feat(social): multi-language support — Whisper LID + per-lang Piper TTS (Issue #167)

- Add SpeechTranscript.language (BCP-47), ConversationResponse.language fields
- speech_pipeline_node: whisper_language param (""=auto-detect via Whisper LID);
  detected language published in every transcript
- conversation_node: track per-speaker language; inject "[Please respond in X.]"
  hint for non-English speakers; propagate language to ConversationResponse.
  _LANG_NAMES: 24 BCP-47 codes -> English names. Also adds Issue #161 emotion
  context plumbing (co-located in same branch for clean merge)
- tts_node: voice_map_json param (JSON BCP-47->ONNX path); lazy voice loading
  per language; playback queue now carries (text, lang) tuples for voice routing
- speech_params.yaml, tts_params.yaml: new language params with docs
- 47/47 tests pass (test_multilang.py)

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

include/config.h

@@ -189,6 +189,18 @@
 /* Full blend transition time: MANUAL→AUTO takes this many ms */
 #define MODE_BLEND_MS           500
 
+// --- Power Management (STOP mode, Issue #178) ---
+#define PM_IDLE_TIMEOUT_MS      30000u  // 30s no activity → PM_SLEEP_PENDING
+#define PM_FADE_MS               3000u  // LED fade-out duration before STOP entry
+#define PM_LED_PERIOD_MS         2000u  // sleep-pending triangle-wave period (ms)
+// Estimated per-subsystem currents (mA) — used for JLINK_TLM_POWER telemetry
+#define PM_CURRENT_BASE_MA         30   // SPI1(IMU)+UART4(CRSF)+USART1(JLink)+core
+#define PM_CURRENT_AUDIO_MA         8   // I2S3 + amplifier quiescent
+#define PM_CURRENT_OSD_MA           5   // SPI2 OSD (MAX7456)
+#define PM_CURRENT_DEBUG_MA         1   // UART5 + USART6
+#define PM_CURRENT_STOP_MA          1   // MCU in STOP mode (< 1 mA)
+#define PM_TLM_HZ                   1   // JLINK_TLM_POWER transmit rate (Hz)
+
 // --- Audio Amplifier (I2S3, Issue #143) ---
 // SPI3 repurposed as I2S3; blackbox flash unused on balance bot
 #define AUDIO_BCLK_PORT         GPIOC

include/jlink.h

@@ -54,9 +54,11 @@
 #define JLINK_CMD_DFU_ENTER     0x06u
 #define JLINK_CMD_ESTOP         0x07u
 #define JLINK_CMD_AUDIO         0x08u  /* PCM audio chunk: int16 samples, up to 126 */
+#define JLINK_CMD_SLEEP         0x09u  /* no payload; request STOP-mode sleep */
 
 /* ---- Telemetry IDs (STM32 → Jetson) ---- */
 #define JLINK_TLM_STATUS        0x80u
+#define JLINK_TLM_POWER         0x81u  /* jlink_tlm_power_t (11 bytes) */
 
 /* ---- Telemetry STATUS payload (20 bytes, packed) ---- */
 typedef struct __attribute__((packed)) {
@@ -77,6 +79,15 @@ typedef struct __attribute__((packed)) {
     uint8_t  fw_patch;
 } jlink_tlm_status_t;  /* 20 bytes */
 
+/* ---- Telemetry POWER payload (11 bytes, packed) ---- */
+typedef struct __attribute__((packed)) {
+    uint8_t  power_state;   /* PowerState: 0=ACTIVE,1=SLEEP_PENDING,2=SLEEPING,3=WAKING */
+    uint16_t est_total_ma;  /* estimated total current draw (mA) */
+    uint16_t est_audio_ma;  /* estimated I2S3+amp current (mA); 0 if gated */
+    uint16_t est_osd_ma;    /* estimated OSD SPI2 current (mA); 0 if gated */
+    uint32_t idle_ms;       /* ms since last cmd_vel activity */
+} jlink_tlm_power_t;  /* 11 bytes */
+
 /* ---- Volatile state (read from main loop) ---- */
 typedef struct {
     /* Drive command — updated on JLINK_CMD_DRIVE */
@@ -99,6 +110,8 @@ typedef struct {
 
     /* DFU reboot request — set by parser, cleared by main loop */
     volatile uint8_t  dfu_req;
+    /* Sleep request — set by JLINK_CMD_SLEEP, cleared by main loop */
+    volatile uint8_t  sleep_req;
 } JLinkState;
 
 extern volatile JLinkState jlink_state;
@@ -130,4 +143,10 @@ void jlink_send_telemetry(const jlink_tlm_status_t *status);
  */
 void jlink_process(void);
 
+/*
+ * jlink_send_power_telemetry(power) — build and transmit a JLINK_TLM_POWER
+ * frame (17 bytes) at PM_TLM_HZ. Call from main loop when not in STOP mode.
+ */
+void jlink_send_power_telemetry(const jlink_tlm_power_t *power);
+
 #endif /* JLINK_H */

include/power_mgmt.h

@@ -0,0 +1,96 @@
+#ifndef POWER_MGMT_H
+#define POWER_MGMT_H
+
+#include <stdint.h>
+#include <stdbool.h>
+
+/*
+ * power_mgmt — STM32F7 STOP-mode sleep/wake manager (Issue #178).
+ *
+ * State machine:
+ *   PM_ACTIVE ──(idle ≥ PM_IDLE_TIMEOUT_MS or sleep cmd)──► PM_SLEEP_PENDING
+ *   PM_SLEEP_PENDING ──(fade complete, ≥ PM_FADE_MS)──► PM_SLEEPING (WFI)
+ *   PM_SLEEPING ──(EXTI wake)──► PM_WAKING ──(clocks restored)──► PM_ACTIVE
+ *
+ *   Any call to power_mgmt_activity() during SLEEP_PENDING or SLEEPING
+ *   immediately transitions back toward PM_ACTIVE.
+ *
+ * Wake sources (EXTI, falling edge on UART idle-high RX pin or IMU INT):
+ *   EXTI1  PA1  UART4_RX   — CRSF/ELRS start bit
+ *   EXTI7  PB7  USART1_RX  — JLink start bit
+ *   EXTI4  PC4  MPU6000 INT — IMU motion (handler owned by mpu6000.c)
+ *
+ * Peripheral gating on sleep entry (clock disable, state preserved):
+ *   Disabled: SPI3/I2S3 (audio amp), SPI2 (OSD), USART6, UART5 (debug)
+ *   Active:   SPI1 (IMU), UART4 (CRSF), USART1 (JLink), I2C1 (baro/mag)
+ *
+ * Sleep LED (LED1, active-low PC15):
+ *   PM_SLEEP_PENDING: triangle-wave pulse, period PM_LED_PERIOD_MS
+ *   All other states: 0 (caller uses normal LED logic)
+ *
+ * IWDG:
+ *   Fed immediately before WFI. STOP wakeup <10 ms typical — well within
+ *   WATCHDOG_TIMEOUT_MS (50 ms).
+ *
+ * Safety interlock:
+ *   Caller MUST NOT call power_mgmt_tick() while armed; call
+ *   power_mgmt_activity() instead to keep the idle timer reset.
+ *
+ * JLink integration:
+ *   JLINK_CMD_SLEEP (0x09) → power_mgmt_request_sleep()
+ *   Any valid JLink frame → power_mgmt_activity()  (handled in main loop)
+ */
+
+typedef enum {
+    PM_ACTIVE        = 0,  /* Normal, all peripherals running */
+    PM_SLEEP_PENDING = 1,  /* Idle timeout reached; LED fade-out in progress */
+    PM_SLEEPING      = 2,  /* In STOP mode (WFI); execution blocked in tick() */
+    PM_WAKING        = 3,  /* Transitional; clocks/peripherals being restored */
+} PowerState;
+
+/* ---- API ---- */
+
+/*
+ * power_mgmt_init() — configure wake EXTI lines (EXTI1, EXTI7).
+ * Call after crsf_init() and jlink_init().
+ */
+void power_mgmt_init(void);
+
+/*
+ * power_mgmt_activity() — record cmd_vel event (CRSF frame, JLink frame).
+ * Resets idle timer; aborts any pending/active sleep.
+ */
+void power_mgmt_activity(void);
+
+/*
+ * power_mgmt_request_sleep() — force sleep regardless of idle timer
+ * (called on JLINK_CMD_SLEEP). Next tick() enters PM_SLEEP_PENDING.
+ */
+void power_mgmt_request_sleep(void);
+
+/*
+ * power_mgmt_tick(now_ms) — drive state machine. May block in WFI during
+ * STOP mode. Returns state after this tick.
+ * MUST NOT be called while balance_state == BALANCE_ARMED.
+ */
+PowerState power_mgmt_tick(uint32_t now_ms);
+
+/* power_mgmt_state() — non-blocking read of current state. */
+PowerState power_mgmt_state(void);
+
+/*
+ * power_mgmt_led_brightness() — 0-255 brightness for sleep-pending pulse.
+ * Returns 0 when not in PM_SLEEP_PENDING; caller uses normal LED logic.
+ */
+uint8_t power_mgmt_led_brightness(void);
+
+/*
+ * power_mgmt_current_ma() — estimated total current draw (mA) based on
+ * gating state; populated in JLINK_TLM_POWER telemetry.
+ */
+uint16_t power_mgmt_current_ma(void);
+
+/* power_mgmt_idle_ms() — ms elapsed since last power_mgmt_activity() call. */
+uint32_t power_mgmt_idle_ms(void);
+
+#endif /* POWER_MGMT_H */

jetson/ros2_ws/src/saltybot_bringup/config/rosbridge_params.yaml

@@ -40,6 +40,11 @@ rosbridge_websocket:
        "/person/target",
        "/person/detections",
        "/camera/*/image_raw/compressed",
+       "/camera/depth/image_rect_raw/compressed",
+       "/camera/panoramic/compressed",
+       "/social/faces/detections",
+       "/social/gestures",
+       "/social/scene/objects",
        "/scan",
        "/cmd_vel",
        "/saltybot/imu",

jetson/ros2_ws/src/saltybot_bringup/launch/rosbridge.launch.py

@@ -94,4 +94,33 @@ def generate_launch_description():
         for name in _CAMERAS
     ]
 
-    return LaunchDescription([rosbridge] + republishers)
+    # ── D435i colour republisher (Issue #177) ────────────────────────────────
+    d435i_color = Node(
+        package='image_transport',
+        executable='republish',
+        name='compress_d435i_color',
+        arguments=['raw', 'compressed'],
+        remappings=[
+            ('in',             '/camera/color/image_raw'),
+            ('out/compressed', '/camera/color/image_raw/compressed'),
+        ],
+        parameters=[{'compressed.jpeg_quality': _JPEG_QUALITY}],
+        output='screen',
+    )
+
+    # ── D435i depth republisher (Issue #177) ─────────────────────────────────
+    # Depth stream as compressedDepth (PNG16) — preserves uint16 depth values.
+    # Browser displays as greyscale PNG (darker = closer).
+    d435i_depth = Node(
+        package='image_transport',
+        executable='republish',
+        name='compress_d435i_depth',
+        arguments=['raw', 'compressedDepth'],
+        remappings=[
+            ('in',                  '/camera/depth/image_rect_raw'),
+            ('out/compressedDepth', '/camera/depth/image_rect_raw/compressed'),
+        ],
+        output='screen',
+    )
+
+    return LaunchDescription([rosbridge] + republishers + [d435i_color, d435i_depth])

jetson/ros2_ws/src/saltybot_dynamic_obs_msgs/CMakeLists.txt

@@ -0,0 +1,16 @@
+cmake_minimum_required(VERSION 3.8)
+project(saltybot_dynamic_obs_msgs)
+
+find_package(ament_cmake REQUIRED)
+find_package(rosidl_default_generators REQUIRED)
+find_package(std_msgs REQUIRED)
+find_package(geometry_msgs REQUIRED)
+
+rosidl_generate_interfaces(${PROJECT_NAME}
+  "msg/TrackedObject.msg"
+  "msg/MovingObjectArray.msg"
+  DEPENDENCIES std_msgs geometry_msgs
+)
+
+ament_export_dependencies(rosidl_default_runtime)
+ament_package()

jetson/ros2_ws/src/saltybot_dynamic_obs_msgs/msg/MovingObjectArray.msg

@@ -0,0 +1,12 @@
+# MovingObjectArray — all currently tracked moving obstacles.
+#
+# Published at ~10 Hz on /saltybot/moving_objects.
+# Only confirmed tracks (hits >= confirm_frames) appear here.
+
+std_msgs/Header header
+
+saltybot_dynamic_obs_msgs/TrackedObject[] objects
+
+uint32 active_count        # number of confirmed tracks
+uint32 tentative_count     # tracks not yet confirmed
+float32 detector_latency_ms # pipeline latency hint

jetson/ros2_ws/src/saltybot_dynamic_obs_msgs/msg/TrackedObject.msg

@@ -0,0 +1,21 @@
+# TrackedObject — a single tracked moving obstacle.
+#
+# predicted_path[i] is the estimated pose at predicted_times[i] seconds from now.
+# All poses are in the same frame as header.frame_id (typically 'odom').
+
+std_msgs/Header header
+
+uint32 object_id               # stable ID across frames (monotonically increasing)
+
+geometry_msgs/PoseWithCovariance pose    # current best-estimate pose (x, y, yaw)
+geometry_msgs/Vector3            velocity # vx, vy in m/s (vz = 0 for ground objects)
+
+geometry_msgs/Pose[] predicted_path   # future positions at predicted_times
+float32[]            predicted_times  # seconds from header.stamp for each pose
+
+float32 speed_mps    # scalar |v|
+float32 confidence   # 0.0–1.0 (higher after more confirmed frames)
+
+uint32 age_frames    # frames since first detection
+uint32 hits          # number of successful associations
+bool   is_valid      # false if in tentative / just-created state

jetson/ros2_ws/src/saltybot_dynamic_obs_msgs/package.xml

@@ -0,0 +1,23 @@
+<?xml version="1.0"?>
+<?xml-model href="http://download.ros.org/schema/package_format3.xsd" schematypens="http://www.w3.org/2001/XMLSchema"?>
+<package format="3">
+  <name>saltybot_dynamic_obs_msgs</name>
+  <version>0.1.0</version>
+  <description>Custom message types for dynamic obstacle tracking.</description>
+  <maintainer email="robot@saltylab.local">SaltyLab</maintainer>
+  <license>MIT</license>
+
+  <buildtool_depend>ament_cmake</buildtool_depend>
+  <buildtool_depend>rosidl_default_generators</buildtool_depend>
+
+  <depend>std_msgs</depend>
+  <depend>geometry_msgs</depend>
+
+  <exec_depend>rosidl_default_runtime</exec_depend>
+
+  <member_of_group>rosidl_interface_packages</member_of_group>
+
+  <export>
+    <build_type>ament_cmake</build_type>
+  </export>
+</package>

jetson/ros2_ws/src/saltybot_dynamic_obstacles/config/dynamic_obstacles_params.yaml

@@ -0,0 +1,52 @@
+# saltybot_dynamic_obstacles — runtime parameters
+#
+# Requires:
+#   /scan  (sensor_msgs/LaserScan)  — RPLIDAR A1M8 at ~5.5 Hz
+#
+# LIDAR scan is published by rplidar_ros node.
+# Make sure RPLIDAR is running before starting this stack.
+
+dynamic_obs_tracker:
+  ros__parameters:
+    max_tracks:          20      # max simultaneous tracked objects
+    confirm_frames:       3      # hits before a track is published
+    max_missed_frames:    6      # missed frames before track deletion
+    assoc_dist_m:         1.5    # max assignment distance (Hungarian)
+    prediction_hz:       10.0    # tracker update + publish rate
+    horizon_s:            2.5    # prediction look-ahead
+    pred_step_s:          0.5    # time between predicted waypoints
+    odom_frame:          'odom'
+    min_speed_mps:        0.05   # suppress near-stationary tracks
+    max_range_m:          8.0    # ignore detections beyond this
+
+dynamic_obs_costmap:
+  ros__parameters:
+    inflation_radius_m:   0.35   # safety bubble around each predicted point
+    ring_points:          8      # polygon points for inflation circle
+    clear_on_empty:       true   # push empty cloud to clear stale Nav2 markings
+
+
+# ── Nav2 costmap integration ───────────────────────────────────────────────────
+# In your nav2_params.yaml, under local_costmap or global_costmap > plugins, add
+# an ObstacleLayer with:
+#
+#   obstacle_layer:
+#     plugin: "nav2_costmap_2d::ObstacleLayer"
+#     enabled: true
+#     observation_sources: static_scan dynamic_obs
+#     static_scan:
+#       topic: /scan
+#       data_type: LaserScan
+#       ...
+#     dynamic_obs:
+#       topic: /saltybot/dynamic_obs_cloud
+#       data_type: PointCloud2
+#       sensor_frame: odom
+#       obstacle_max_range: 10.0
+#       raytrace_max_range: 10.0
+#       marking: true
+#       clearing: false
+#
+# This feeds the predicted trajectory smear directly into Nav2's obstacle
+# inflation, forcing the planner to route around the predicted future path
+# of every tracked moving object.

jetson/ros2_ws/src/saltybot_dynamic_obstacles/launch/dynamic_obstacles.launch.py

@@ -0,0 +1,75 @@
+"""
+dynamic_obstacles.launch.py — Dynamic obstacle tracking + Nav2 costmap layer.
+
+Starts:
+  dynamic_obs_tracker  — LIDAR motion detection + Kalman tracking @10 Hz
+  dynamic_obs_costmap  — Predicted-trajectory → PointCloud2 for Nav2
+
+Launch args:
+  max_tracks         int    '20'
+  assoc_dist_m       float  '1.5'
+  horizon_s          float  '2.5'
+  inflation_radius_m float  '0.35'
+
+Verify:
+  ros2 topic hz /saltybot/moving_objects        # ~10 Hz
+  ros2 topic echo /saltybot/moving_objects      # TrackedObject list
+  ros2 topic hz /saltybot/dynamic_obs_cloud     # ~10 Hz (when objects present)
+  rviz2 → add MarkerArray → /saltybot/moving_objects_viz
+
+Nav2 costmap integration:
+  In your costmap_params.yaml ObstacleLayer observation_sources, add:
+    dynamic_obs:
+      topic: /saltybot/dynamic_obs_cloud
+      data_type: PointCloud2
+      marking: true
+      clearing: false
+"""
+
+from launch import LaunchDescription
+from launch.actions import DeclareLaunchArgument
+from launch.substitutions import LaunchConfiguration
+from launch_ros.actions import Node
+
+
+def generate_launch_description():
+    args = [
+        DeclareLaunchArgument('max_tracks',         default_value='20'),
+        DeclareLaunchArgument('assoc_dist_m',        default_value='1.5'),
+        DeclareLaunchArgument('horizon_s',           default_value='2.5'),
+        DeclareLaunchArgument('inflation_radius_m',  default_value='0.35'),
+        DeclareLaunchArgument('min_speed_mps',       default_value='0.05'),
+    ]
+
+    tracker = Node(
+        package='saltybot_dynamic_obstacles',
+        executable='dynamic_obs_tracker',
+        name='dynamic_obs_tracker',
+        output='screen',
+        parameters=[{
+            'max_tracks':       LaunchConfiguration('max_tracks'),
+            'assoc_dist_m':     LaunchConfiguration('assoc_dist_m'),
+            'horizon_s':        LaunchConfiguration('horizon_s'),
+            'min_speed_mps':    LaunchConfiguration('min_speed_mps'),
+            'prediction_hz':    10.0,
+            'confirm_frames':   3,
+            'max_missed_frames': 6,
+            'pred_step_s':      0.5,
+            'odom_frame':       'odom',
+            'max_range_m':      8.0,
+        }],
+    )
+
+    costmap = Node(
+        package='saltybot_dynamic_obstacles',
+        executable='dynamic_obs_costmap',
+        name='dynamic_obs_costmap',
+        output='screen',
+        parameters=[{
+            'inflation_radius_m': LaunchConfiguration('inflation_radius_m'),
+            'ring_points':        8,
+            'clear_on_empty':     True,
+        }],
+    )
+
+    return LaunchDescription(args + [tracker, costmap])

jetson/ros2_ws/src/saltybot_dynamic_obstacles/package.xml

@@ -0,0 +1,29 @@
+<?xml version="1.0"?>
+<?xml-model href="http://download.ros.org/schema/package_format3.xsd" schematypens="http://www.w3.org/2001/XMLSchema"?>
+<package format="3">
+  <name>saltybot_dynamic_obstacles</name>
+  <version>0.1.0</version>
+  <description>
+    Dynamic obstacle detection, multi-object Kalman tracking, trajectory
+    prediction, and Nav2 costmap layer integration for SaltyBot.
+  </description>
+  <maintainer email="robot@saltylab.local">SaltyLab</maintainer>
+  <license>MIT</license>
+
+  <depend>rclpy</depend>
+  <depend>std_msgs</depend>
+  <depend>sensor_msgs</depend>
+  <depend>geometry_msgs</depend>
+  <depend>nav_msgs</depend>
+  <depend>visualization_msgs</depend>
+  <depend>saltybot_dynamic_obs_msgs</depend>
+
+  <exec_depend>python3-numpy</exec_depend>
+  <exec_depend>python3-scipy</exec_depend>
+
+  <test_depend>pytest</test_depend>
+
+  <export>
+    <build_type>ament_python</build_type>
+  </export>
+</package>

jetson/ros2_ws/src/saltybot_dynamic_obstacles/saltybot_dynamic_obstacles/costmap_layer_node.py

@@ -0,0 +1,178 @@
+"""
+costmap_layer_node.py — Nav2 costmap integration for dynamic obstacles.
+
+Converts predicted trajectories from /saltybot/moving_objects into a
+PointCloud2 fed into Nav2's ObstacleLayer.  Each predicted future position
+is added as a point, creating a "smeared" dynamic obstacle zone that
+covers the full 2-3 s prediction horizon.
+
+Nav2 ObstacleLayer config (in costmap_params.yaml):
+  obstacle_layer:
+    enabled: true
+    observation_sources: dynamic_obs
+    dynamic_obs:
+      topic: /saltybot/dynamic_obs_cloud
+      sensor_frame: odom
+      data_type: PointCloud2
+      obstacle_max_range: 12.0
+      obstacle_min_range: 0.0
+      raytrace_max_range: 12.0
+      marking: true
+      clearing: false   # let the tracker handle clearing
+
+The node also clears old obstacle points when tracks are dropped, by
+publishing a clearing cloud to /saltybot/dynamic_obs_clear.
+
+Subscribes:
+  /saltybot/moving_objects     saltybot_dynamic_obs_msgs/MovingObjectArray
+
+Publishes:
+  /saltybot/dynamic_obs_cloud  sensor_msgs/PointCloud2   marking cloud
+  /saltybot/dynamic_obs_clear  sensor_msgs/PointCloud2   clearing cloud
+
+Parameters:
+  inflation_radius_m  float  0.35   (each predicted point inflated by this radius)
+  ring_points         int    8      (polygon approximation of inflation circle)
+  clear_on_empty      bool   true   (publish clear cloud when no objects tracked)
+"""
+
+from __future__ import annotations
+
+import math
+import struct
+from typing import List
+
+import rclpy
+from rclpy.node import Node
+from rclpy.qos import QoSProfile, ReliabilityPolicy, HistoryPolicy
+
+from sensor_msgs.msg import PointCloud2, PointField
+from std_msgs.msg import Header
+
+try:
+    from saltybot_dynamic_obs_msgs.msg import MovingObjectArray
+    _MSGS_AVAILABLE = True
+except ImportError:
+    _MSGS_AVAILABLE = False
+
+
+_RELIABLE_QOS = QoSProfile(
+    reliability=ReliabilityPolicy.RELIABLE,
+    history=HistoryPolicy.KEEP_LAST,
+    depth=10,
+)
+
+
+def _make_pc2(header: Header, points_xyz: List[tuple]) -> PointCloud2:
+    """Pack a list of (x, y, z) into a PointCloud2 message."""
+    fields = [
+        PointField(name='x', offset=0,  datatype=PointField.FLOAT32, count=1),
+        PointField(name='y', offset=4,  datatype=PointField.FLOAT32, count=1),
+        PointField(name='z', offset=8,  datatype=PointField.FLOAT32, count=1),
+    ]
+    point_step = 12   # 3 × float32
+    data = bytearray(len(points_xyz) * point_step)
+    for i, (x, y, z) in enumerate(points_xyz):
+        struct.pack_into('<fff', data, i * point_step, x, y, z)
+
+    pc = PointCloud2()
+    pc.header      = header
+    pc.height      = 1
+    pc.width       = len(points_xyz)
+    pc.fields      = fields
+    pc.is_bigendian = False
+    pc.point_step  = point_step
+    pc.row_step    = point_step * len(points_xyz)
+    pc.data        = bytes(data)
+    pc.is_dense    = True
+    return pc
+
+
+class CostmapLayerNode(Node):
+
+    def __init__(self):
+        super().__init__('dynamic_obs_costmap')
+
+        self.declare_parameter('inflation_radius_m', 0.35)
+        self.declare_parameter('ring_points',         8)
+        self.declare_parameter('clear_on_empty',      True)
+
+        self._infl_r      = self.get_parameter('inflation_radius_m').value
+        self._ring_n      = self.get_parameter('ring_points').value
+        self._clear_empty = self.get_parameter('clear_on_empty').value
+
+        # Pre-compute ring offsets for inflation
+        self._ring_offsets = [
+            (self._infl_r * math.cos(2 * math.pi * i / self._ring_n),
+             self._infl_r * math.sin(2 * math.pi * i / self._ring_n))
+            for i in range(self._ring_n)
+        ]
+
+        if _MSGS_AVAILABLE:
+            self.create_subscription(
+                MovingObjectArray,
+                '/saltybot/moving_objects',
+                self._on_objects,
+                _RELIABLE_QOS,
+            )
+        else:
+            self.get_logger().warning(
+                '[costmap_layer] saltybot_dynamic_obs_msgs not built — '
+                'will not subscribe to MovingObjectArray'
+            )
+
+        self._mark_pub  = self.create_publisher(
+            PointCloud2, '/saltybot/dynamic_obs_cloud', 10
+        )
+        self._clear_pub = self.create_publisher(
+            PointCloud2, '/saltybot/dynamic_obs_clear', 10
+        )
+
+        self.get_logger().info(
+            f'dynamic_obs_costmap ready — '
+            f'inflation={self._infl_r}m ring_pts={self._ring_n}'
+        )
+
+    # ── Callback ──────────────────────────────────────────────────────────────
+
+    def _on_objects(self, msg: 'MovingObjectArray') -> None:
+        hdr = msg.header
+        mark_pts: List[tuple] = []
+
+        for obj in msg.objects:
+            if not obj.is_valid:
+                continue
+
+            # Current position
+            self._add_inflated(obj.pose.pose.position.x,
+                               obj.pose.pose.position.y, mark_pts)
+
+            # Predicted future positions
+            for pose in obj.predicted_path:
+                self._add_inflated(pose.position.x, pose.position.y, mark_pts)
+
+        if mark_pts:
+            self._mark_pub.publish(_make_pc2(hdr, mark_pts))
+        elif self._clear_empty:
+            # Publish tiny clear cloud so Nav2 clears stale markings
+            self._clear_pub.publish(_make_pc2(hdr, []))
+
+    def _add_inflated(self, cx: float, cy: float, pts: List[tuple]) -> None:
+        """Add the centre + ring of inflation points at height 0.5 m."""
+        pts.append((cx, cy, 0.5))
+        for ox, oy in self._ring_offsets:
+            pts.append((cx + ox, cy + oy, 0.5))
+
+
+def main(args=None):
+    rclpy.init(args=args)
+    node = CostmapLayerNode()
+    try:
+        rclpy.spin(node)
+    finally:
+        node.destroy_node()
+        rclpy.shutdown()
+
+
+if __name__ == '__main__':
+    main()

jetson/ros2_ws/src/saltybot_dynamic_obstacles/saltybot_dynamic_obstacles/dynamic_obs_node.py

@@ -0,0 +1,319 @@
+"""
+dynamic_obs_node.py — ROS2 node: LIDAR moving-object detection + Kalman tracking.
+
+Pipeline:
+  1. Subscribe /scan (RPLIDAR LaserScan, ~5.5 Hz).
+  2. ObjectDetector performs background subtraction → moving blobs.
+  3. TrackerManager runs Hungarian assignment + Kalman predict/update at 10 Hz.
+  4. Publish /saltybot/moving_objects (MovingObjectArray).
+  5. Publish /saltybot/moving_objects_viz (MarkerArray) for RViz.
+
+The 10 Hz timer drives the tracker regardless of scan rate, so prediction
+continues between scans (pure-predict steps).
+
+Subscribes:
+  /scan                    sensor_msgs/LaserScan   RPLIDAR A1M8
+
+Publishes:
+  /saltybot/moving_objects          saltybot_dynamic_obs_msgs/MovingObjectArray
+  /saltybot/moving_objects_viz      visualization_msgs/MarkerArray
+
+Parameters:
+  max_tracks          int    20
+  confirm_frames      int    3
+  max_missed_frames   int    6
+  assoc_dist_m        float  1.5
+  prediction_hz       float  10.0     (tracker + publish rate)
+  horizon_s           float  2.5
+  pred_step_s         float  0.5
+  odom_frame          str    'odom'
+  min_speed_mps       float  0.05     (suppress near-zero velocity tracks)
+  max_range_m         float  8.0
+"""
+
+import time
+import math
+
+import numpy as np
+import rclpy
+from rclpy.node import Node
+from rclpy.qos import QoSProfile, ReliabilityPolicy, HistoryPolicy
+
+from sensor_msgs.msg import LaserScan
+from geometry_msgs.msg import Pose, Point, Quaternion, Vector3
+from std_msgs.msg import Header, ColorRGBA
+from visualization_msgs.msg import Marker, MarkerArray
+
+try:
+    from saltybot_dynamic_obs_msgs.msg import TrackedObject, MovingObjectArray
+    _MSGS_AVAILABLE = True
+except ImportError:
+    _MSGS_AVAILABLE = False
+
+from .object_detector import ObjectDetector
+from .tracker_manager import TrackerManager, Track
+
+
+_SENSOR_QOS = QoSProfile(
+    reliability=ReliabilityPolicy.BEST_EFFORT,
+    history=HistoryPolicy.KEEP_LAST,
+    depth=5,
+)
+
+
+def _yaw_quat(yaw: float) -> Quaternion:
+    q = Quaternion()
+    q.w = math.cos(yaw * 0.5)
+    q.z = math.sin(yaw * 0.5)
+    return q
+
+
+class DynamicObsNode(Node):
+
+    def __init__(self):
+        super().__init__('dynamic_obs_tracker')
+
+        # ── Parameters ──────────────────────────────────────────────────────
+        self.declare_parameter('max_tracks',         20)
+        self.declare_parameter('confirm_frames',      3)
+        self.declare_parameter('max_missed_frames',   6)
+        self.declare_parameter('assoc_dist_m',        1.5)
+        self.declare_parameter('prediction_hz',      10.0)
+        self.declare_parameter('horizon_s',           2.5)
+        self.declare_parameter('pred_step_s',         0.5)
+        self.declare_parameter('odom_frame',         'odom')
+        self.declare_parameter('min_speed_mps',       0.05)
+        self.declare_parameter('max_range_m',         8.0)
+
+        max_tracks    = self.get_parameter('max_tracks').value
+        confirm_f     = self.get_parameter('confirm_frames').value
+        max_missed    = self.get_parameter('max_missed_frames').value
+        assoc_dist    = self.get_parameter('assoc_dist_m').value
+        pred_hz       = self.get_parameter('prediction_hz').value
+        horizon_s     = self.get_parameter('horizon_s').value
+        pred_step     = self.get_parameter('pred_step_s').value
+        self._frame   = self.get_parameter('odom_frame').value
+        self._min_spd = self.get_parameter('min_speed_mps').value
+        self._max_rng = self.get_parameter('max_range_m').value
+
+        # ── Core modules ────────────────────────────────────────────────────
+        self._detector = ObjectDetector(
+            grid_radius_m=min(self._max_rng + 2.0, 12.0),
+            max_cluster=int((self._max_rng / 0.1) ** 2 * 0.5),
+        )
+        self._tracker = TrackerManager(
+            max_tracks=max_tracks,
+            confirm_frames=confirm_f,
+            max_missed=max_missed,
+            assoc_dist_m=assoc_dist,
+            horizon_s=horizon_s,
+            pred_step_s=pred_step,
+        )
+        self._horizon_s  = horizon_s
+        self._pred_step  = pred_step
+
+        # ── State ────────────────────────────────────────────────────────────
+        self._latest_scan: LaserScan | None = None
+        self._last_track_t: float = time.monotonic()
+        self._scan_processed_stamp: float | None = None
+
+        # ── Subscriptions ────────────────────────────────────────────────────
+        self.create_subscription(LaserScan, '/scan', self._on_scan, _SENSOR_QOS)
+
+        # ── Publishers ───────────────────────────────────────────────────────
+        if _MSGS_AVAILABLE:
+            self._obj_pub = self.create_publisher(
+                MovingObjectArray, '/saltybot/moving_objects', 10
+            )
+        else:
+            self._obj_pub = None
+            self.get_logger().warning(
+                '[dyn_obs] saltybot_dynamic_obs_msgs not built — '
+                'MovingObjectArray will not be published'
+            )
+
+        self._viz_pub = self.create_publisher(
+            MarkerArray, '/saltybot/moving_objects_viz', 10
+        )
+
+        # ── Timer ────────────────────────────────────────────────────────────
+        self.create_timer(1.0 / pred_hz, self._track_tick)
+
+        self.get_logger().info(
+            f'dynamic_obs_tracker ready — '
+            f'max_tracks={max_tracks} horizon={horizon_s}s assoc={assoc_dist}m'
+        )
+
+    # ── Scan callback ─────────────────────────────────────────────────────────
+
+    def _on_scan(self, msg: LaserScan) -> None:
+        self._latest_scan = msg
+
+    # ── 10 Hz tracker tick ────────────────────────────────────────────────────
+
+    def _track_tick(self) -> None:
+        t0 = time.monotonic()
+
+        now_mono = t0
+        dt = now_mono - self._last_track_t
+        dt = max(1e-3, min(dt, 0.5))
+        self._last_track_t = now_mono
+
+        scan = self._latest_scan
+        detections = []
+
+        if scan is not None:
+            stamp_sec = scan.header.stamp.sec + scan.header.stamp.nanosec * 1e-9
+            if stamp_sec != self._scan_processed_stamp:
+                self._scan_processed_stamp = stamp_sec
+                ranges = np.asarray(scan.ranges, dtype=np.float32)
+                detections = self._detector.update(
+                    ranges,
+                    scan.angle_min,
+                    scan.angle_increment,
+                    min(scan.range_max, self._max_rng),
+                )
+
+        confirmed = self._tracker.update(detections, dt)
+
+        latency_ms = (time.monotonic() - t0) * 1000.0
+        stamp = self.get_clock().now().to_msg()
+
+        if _MSGS_AVAILABLE and self._obj_pub is not None:
+            self._publish_objects(confirmed, stamp, latency_ms)
+        self._publish_viz(confirmed, stamp)
+
+    # ── Publish helpers ───────────────────────────────────────────────────────
+
+    def _publish_objects(self, confirmed: list, stamp, latency_ms: float) -> None:
+        arr = MovingObjectArray()
+        arr.header.stamp    = stamp
+        arr.header.frame_id = self._frame
+        arr.active_count    = len(confirmed)
+        arr.tentative_count = self._tracker.tentative_count
+        arr.detector_latency_ms = float(latency_ms)
+
+        for tr in confirmed:
+            px, py   = tr.kalman.position
+            vx, vy   = tr.kalman.velocity
+            speed    = tr.kalman.speed
+            if speed < self._min_spd:
+                continue
+
+            obj = TrackedObject()
+            obj.header          = arr.header
+            obj.object_id       = tr.track_id
+            obj.pose.pose.position.x = px
+            obj.pose.pose.position.y = py
+            obj.pose.pose.orientation = _yaw_quat(math.atan2(vy, vx))
+            cov = tr.kalman.covariance_2x2
+            obj.pose.covariance[0]  = float(cov[0, 0])
+            obj.pose.covariance[1]  = float(cov[0, 1])
+            obj.pose.covariance[6]  = float(cov[1, 0])
+            obj.pose.covariance[7]  = float(cov[1, 1])
+            obj.velocity.x  = vx
+            obj.velocity.y  = vy
+            obj.speed_mps   = speed
+            obj.confidence  = min(1.0, tr.hits / (self._tracker._confirm_frames * 3))
+            obj.age_frames  = tr.age
+            obj.hits        = tr.hits
+            obj.is_valid    = True
+
+            # Predicted path
+            for px_f, py_f, t_f in tr.kalman.predict_horizon(
+                self._horizon_s, self._pred_step
+            ):
+                p = Pose()
+                p.position.x = px_f
+                p.position.y = py_f
+                p.orientation.w = 1.0
+                obj.predicted_path.append(p)
+                obj.predicted_times.append(float(t_f))
+
+            arr.objects.append(obj)
+
+        self._obj_pub.publish(arr)
+
+    def _publish_viz(self, confirmed: list, stamp) -> None:
+        markers = MarkerArray()
+
+        # Delete old markers
+        del_marker = Marker()
+        del_marker.header.stamp    = stamp
+        del_marker.header.frame_id = self._frame
+        del_marker.action = Marker.DELETEALL
+        markers.markers.append(del_marker)
+
+        for tr in confirmed:
+            px, py = tr.kalman.position
+            vx, vy = tr.kalman.velocity
+            speed  = tr.kalman.speed
+            if speed < self._min_spd:
+                continue
+
+            # Cylinder at current position
+            m = Marker()
+            m.header.stamp    = stamp
+            m.header.frame_id = self._frame
+            m.ns     = 'dyn_obs'
+            m.id     = tr.track_id
+            m.type   = Marker.CYLINDER
+            m.action = Marker.ADD
+            m.pose.position.x = px
+            m.pose.position.y = py
+            m.pose.position.z = 0.5
+            m.pose.orientation.w = 1.0
+            m.scale.x = 0.4
+            m.scale.y = 0.4
+            m.scale.z = 1.0
+            m.color   = ColorRGBA(r=1.0, g=0.2, b=0.0, a=0.7)
+            m.lifetime.sec = 1
+            markers.markers.append(m)
+
+            # Arrow for velocity
+            vel_m = Marker()
+            vel_m.header  = m.header
+            vel_m.ns      = 'dyn_obs_vel'
+            vel_m.id      = tr.track_id
+            vel_m.type    = Marker.ARROW
+            vel_m.action  = Marker.ADD
+            from geometry_msgs.msg import Point as GPoint
+            p_start = GPoint(x=px, y=py, z=1.0)
+            p_end   = GPoint(x=px + vx, y=py + vy, z=1.0)
+            vel_m.points  = [p_start, p_end]
+            vel_m.scale.x = 0.05
+            vel_m.scale.y = 0.10
+            vel_m.color   = ColorRGBA(r=1.0, g=1.0, b=0.0, a=0.9)
+            vel_m.lifetime.sec = 1
+            markers.markers.append(vel_m)
+
+            # Line strip for predicted path
+            path_m = Marker()
+            path_m.header  = m.header
+            path_m.ns      = 'dyn_obs_path'
+            path_m.id      = tr.track_id
+            path_m.type    = Marker.LINE_STRIP
+            path_m.action  = Marker.ADD
+            path_m.scale.x = 0.04
+            path_m.color   = ColorRGBA(r=1.0, g=0.5, b=0.0, a=0.5)
+            path_m.lifetime.sec = 1
+            path_m.points.append(GPoint(x=px, y=py, z=0.5))
+            for fx, fy, _ in tr.kalman.predict_horizon(self._horizon_s, self._pred_step):
+                path_m.points.append(GPoint(x=fx, y=fy, z=0.5))
+            markers.markers.append(path_m)
+
+        self._viz_pub.publish(markers)
+
+
+def main(args=None):
+    rclpy.init(args=args)
+    node = DynamicObsNode()
+    try:
+        rclpy.spin(node)
+    finally:
+        node.destroy_node()
+        rclpy.shutdown()
+
+
+if __name__ == '__main__':
+    main()

jetson/ros2_ws/src/saltybot_dynamic_obstacles/saltybot_dynamic_obstacles/kalman_tracker.py

@@ -0,0 +1,132 @@
+"""
+kalman_tracker.py — Single-object Kalman filter for 2-D ground-plane tracking.
+
+State vector:  x = [px, py, vx, vy]   (position + velocity)
+Motion model:  constant-velocity (CV) with process noise on acceleration
+
+Predict step:
+    F = | 1 0 dt  0 |      x_{k|k-1} = F @ x_{k-1|k-1}
+        | 0 1  0 dt |      P_{k|k-1} = F @ P @ F^T + Q
+        | 0 0  1  0 |
+        | 0 0  0  1 |
+
+Update step (position observation only):
+    H = | 1 0 0 0 |      y   = z - H @ x
+        | 0 1 0 0 |      S   = H @ P @ H^T + R
+                          K   = P @ H^T @ inv(S)
+                          x   = x + K @ y
+                          P   = (I - K @ H) @ P   (Joseph form for stability)
+
+Trajectory prediction: unrolls the CV model forward at fixed time steps.
+"""
+
+from __future__ import annotations
+
+from typing import List, Tuple
+
+import numpy as np
+
+
+# ── Default noise matrices ────────────────────────────────────────────────────
+# Process noise: models uncertainty in acceleration between frames
+_Q_BASE = np.diag([0.02, 0.02, 0.8, 0.8]).astype(np.float64)
+
+# Measurement noise: LIDAR centroid uncertainty (~0.15 m std)
+_R = np.diag([0.025, 0.025]).astype(np.float64)   # 0.16 m sigma each axis
+
+# Observation matrix
+_H = np.array([[1, 0, 0, 0],
+               [0, 1, 0, 0]], dtype=np.float64)
+_I4 = np.eye(4, dtype=np.float64)
+
+
+class KalmanTracker:
+    """
+    Kalman filter tracking one object.
+
+    Parameters
+    ----------
+    x0, y0         : initial position (metres, odom frame)
+    process_noise  : scalar multiplier on _Q_BASE
+    """
+
+    def __init__(self, x0: float, y0: float, process_noise: float = 1.0):
+        self._x = np.array([x0, y0, 0.0, 0.0], dtype=np.float64)
+        self._P = np.eye(4, dtype=np.float64) * 0.5
+        self._Q = _Q_BASE * process_noise
+
+    # ── Core filter ───────────────────────────────────────────────────────────
+
+    def predict(self, dt: float) -> None:
+        """Propagate state by dt seconds."""
+        F = np.array([
+            [1, 0, dt, 0],
+            [0, 1,  0, dt],
+            [0, 0,  1,  0],
+            [0, 0,  0,  1],
+        ], dtype=np.float64)
+        self._x = F @ self._x
+        self._P = F @ self._P @ F.T + self._Q
+
+    def update(self, z: np.ndarray) -> None:
+        """
+        Incorporate a position measurement z = [x, y].
+        Uses Joseph-form covariance update for numerical stability.
+        """
+        y = z.astype(np.float64) - _H @ self._x
+        S = _H @ self._P @ _H.T + _R
+        K = self._P @ _H.T @ np.linalg.inv(S)
+        self._x = self._x + K @ y
+        IKH = _I4 - K @ _H
+        # Joseph form: (I-KH) P (I-KH)^T + K R K^T
+        self._P = IKH @ self._P @ IKH.T + K @ _R @ K.T
+
+    # ── Prediction horizon ────────────────────────────────────────────────────
+
+    def predict_horizon(
+        self,
+        horizon_s: float = 2.5,
+        step_s: float = 0.5,
+    ) -> List[Tuple[float, float, float]]:
+        """
+        Return [(x, y, t), ...] at equally-spaced future times.
+
+        Does NOT modify internal filter state.
+        """
+        predictions: List[Tuple[float, float, float]] = []
+        state = self._x.copy()
+        t = 0.0
+        F_step = np.array([
+            [1, 0, step_s,      0],
+            [0, 1,      0, step_s],
+            [0, 0,      1,      0],
+            [0, 0,      0,      1],
+        ], dtype=np.float64)
+        while t < horizon_s - 1e-6:
+            state = F_step @ state
+            t += step_s
+            predictions.append((float(state[0]), float(state[1]), t))
+        return predictions
+
+    # ── Properties ────────────────────────────────────────────────────────────
+
+    @property
+    def position(self) -> Tuple[float, float]:
+        return float(self._x[0]), float(self._x[1])
+
+    @property
+    def velocity(self) -> Tuple[float, float]:
+        return float(self._x[2]), float(self._x[3])
+
+    @property
+    def speed(self) -> float:
+        return float(np.hypot(self._x[2], self._x[3]))
+
+    @property
+    def covariance_2x2(self) -> np.ndarray:
+        """Position covariance (top-left 2×2 of P)."""
+        return self._P[:2, :2].copy()
+
+    @property
+    def state(self) -> np.ndarray:
+        return self._x.copy()

jetson/ros2_ws/src/saltybot_dynamic_obstacles/saltybot_dynamic_obstacles/object_detector.py

@@ -0,0 +1,168 @@
+"""
+object_detector.py — LIDAR-based moving object detector.
+
+Algorithm:
+  1. Convert each LaserScan to a 2-D occupancy grid (robot-centred, fixed size).
+  2. Maintain a background model via exponential moving average (EMA):
+       bg_t = α * current + (1-α) * bg_{t-1}   (only for non-moving cells)
+  3. Foreground = cells whose occupancy significantly exceeds the background.
+  4. Cluster foreground cells with scipy connected-components → Detection list.
+
+The grid is robot-relative (origin at robot centre) so it naturally tracks
+the robot's motion without needing TF at this stage.  The caller is responsible
+for transforming detections into a stable frame (odom) before passing to the
+tracker.
+"""
+
+from __future__ import annotations
+
+from dataclasses import dataclass, field
+from typing import List, Optional
+
+import numpy as np
+from scipy import ndimage
+
+
+@dataclass
+class Detection:
+    """A clustered moving foreground blob from one scan."""
+    x: float          # centroid x in sensor frame (m)
+    y: float          # centroid y in sensor frame (m)
+    size_m2: float    # approximate area of the cluster (m²)
+    range_m: float    # distance from robot (m)
+
+
+class ObjectDetector:
+    """
+    Detects moving objects in consecutive 2-D LIDAR scans.
+
+    Parameters
+    ----------
+    grid_radius_m : half-size of the occupancy grid (grid covers ±radius)
+    resolution    : metres per cell
+    bg_alpha      : EMA update rate for background (small = slow forgetting)
+    motion_thr    : occupancy delta above background to count as moving
+    min_cluster   : minimum cells to keep a cluster
+    max_cluster   : maximum cells before a cluster is considered static wall
+    """
+
+    def __init__(
+        self,
+        grid_radius_m: float = 10.0,
+        resolution: float = 0.10,
+        bg_alpha: float = 0.04,
+        motion_thr: float = 0.45,
+        min_cluster: int = 3,
+        max_cluster: int = 200,
+    ):
+        cells = int(2 * grid_radius_m / resolution)
+        self._cells     = cells
+        self._res       = resolution
+        self._origin    = -grid_radius_m        # world x/y at grid index 0
+        self._bg_alpha  = bg_alpha
+        self._motion_thr = motion_thr
+        self._min_clust = min_cluster
+        self._max_clust = max_cluster
+
+        self._bg = np.zeros((cells, cells), dtype=np.float32)
+        self._initialized = False
+
+    # ── Public API ────────────────────────────────────────────────────────────
+
+    def update(
+        self,
+        ranges: np.ndarray,
+        angle_min: float,
+        angle_increment: float,
+        range_max: float,
+    ) -> List[Detection]:
+        """
+        Process one LaserScan and return detected moving blobs.
+
+        Parameters
+        ----------
+        ranges          : 1-D array of range readings (metres)
+        angle_min       : angle of first beam (radians)
+        angle_increment : angular step between beams (radians)
+        range_max       : maximum valid range (metres)
+        """
+        curr = self._scan_to_grid(ranges, angle_min, angle_increment, range_max)
+
+        if not self._initialized:
+            self._bg = curr.copy()
+            self._initialized = True
+            return []
+
+        # Foreground mask
+        motion_mask = (curr - self._bg) > self._motion_thr
+
+        # Update background only on non-moving cells
+        static = ~motion_mask
+        self._bg[static] = (
+            self._bg[static] * (1.0 - self._bg_alpha)
+            + curr[static] * self._bg_alpha
+        )
+
+        return self._cluster(motion_mask)
+
+    def reset(self) -> None:
+        self._bg[:] = 0.0
+        self._initialized = False
+
+    # ── Internals ─────────────────────────────────────────────────────────────
+
+    def _scan_to_grid(
+        self,
+        ranges: np.ndarray,
+        angle_min: float,
+        angle_increment: float,
+        range_max: float,
+    ) -> np.ndarray:
+        grid = np.zeros((self._cells, self._cells), dtype=np.float32)
+        n = len(ranges)
+        angles = angle_min + np.arange(n) * angle_increment
+        r = np.asarray(ranges, dtype=np.float32)
+
+        valid = (r > 0.05) & (r < range_max) & np.isfinite(r)
+        r, a = r[valid], angles[valid]
+
+        x = r * np.cos(a)
+        y = r * np.sin(a)
+
+        ix = np.clip(
+            ((x - self._origin) / self._res).astype(np.int32), 0, self._cells - 1
+        )
+        iy = np.clip(
+            ((y - self._origin) / self._res).astype(np.int32), 0, self._cells - 1
+        )
+        grid[iy, ix] = 1.0
+        return grid
+
+    def _cluster(self, mask: np.ndarray) -> List[Detection]:
+        # Dilate slightly to connect nearby hit cells into one blob
+        struct = ndimage.generate_binary_structure(2, 2)
+        dilated = ndimage.binary_dilation(mask, structure=struct, iterations=1)
+
+        labeled, n_labels = ndimage.label(dilated)
+        detections: List[Detection] = []
+
+        for label_id in range(1, n_labels + 1):
+            coords = np.argwhere(labeled == label_id)
+            n_cells = len(coords)
+            if n_cells < self._min_clust or n_cells > self._max_clust:
+                continue
+
+            ys, xs = coords[:, 0], coords[:, 1]
+            cx_grid = float(np.mean(xs))
+            cy_grid = float(np.mean(ys))
+            cx = cx_grid * self._res + self._origin
+            cy = cy_grid * self._res + self._origin
+
+            detections.append(Detection(
+                x=cx,
+                y=cy,
+                size_m2=n_cells * self._res ** 2,
+                range_m=float(np.hypot(cx, cy)),
+            ))
+
+        return detections

jetson/ros2_ws/src/saltybot_dynamic_obstacles/saltybot_dynamic_obstacles/tracker_manager.py

@@ -0,0 +1,206 @@
+"""
+tracker_manager.py — Multi-object tracker with Hungarian data association.
+
+Track lifecycle:
+  TENTATIVE → confirmed after `confirm_frames` consecutive hits
+  CONFIRMED → normal tracked state
+  LOST      → missed for 1..max_missed frames (still predicts, not published)
+  DEAD      → missed > max_missed → removed
+
+Association:
+  Uses scipy.optimize.linear_sum_assignment (Hungarian algorithm) on a cost
+  matrix of Euclidean distances between predicted track positions and new
+  detections.  Assignments with cost > assoc_dist_m are rejected.
+
+Up to `max_tracks` simultaneous live tracks (tentative + confirmed).
+"""
+
+from __future__ import annotations
+
+from dataclasses import dataclass, field
+from enum import IntEnum
+from typing import Dict, List, Optional, Tuple
+
+import numpy as np
+from scipy.optimize import linear_sum_assignment
+
+from .kalman_tracker import KalmanTracker
+from .object_detector import Detection
+
+
+class TrackState(IntEnum):
+    TENTATIVE = 0
+    CONFIRMED = 1
+    LOST      = 2
+
+
+@dataclass
+class Track:
+    track_id:   int
+    kalman:     KalmanTracker
+    state:      TrackState = TrackState.TENTATIVE
+    hits:       int = 1
+    age:        int = 1      # frames since creation
+    missed:     int = 0      # consecutive missed frames
+
+
+class TrackerManager:
+    """
+    Manages a pool of Kalman tracks.
+
+    Parameters
+    ----------
+    max_tracks      : hard cap on simultaneously alive tracks
+    confirm_frames  : hits needed before a track is CONFIRMED
+    max_missed      : consecutive missed frames before a track is DEAD
+    assoc_dist_m    : max allowed distance (m) for a valid assignment
+    horizon_s       : prediction horizon for trajectory output (seconds)
+    pred_step_s     : time step between predicted waypoints
+    process_noise   : KalmanTracker process-noise multiplier
+    """
+
+    def __init__(
+        self,
+        max_tracks:     int   = 20,
+        confirm_frames: int   = 3,
+        max_missed:     int   = 6,
+        assoc_dist_m:   float = 1.5,
+        horizon_s:      float = 2.5,
+        pred_step_s:    float = 0.5,
+        process_noise:  float = 1.0,
+    ):
+        self._max_tracks     = max_tracks
+        self._confirm_frames = confirm_frames
+        self._max_missed     = max_missed
+        self._assoc_dist     = assoc_dist_m
+        self._horizon_s      = horizon_s
+        self._pred_step      = pred_step_s
+        self._proc_noise     = process_noise
+
+        self._tracks:  Dict[int, Track] = {}
+        self._next_id: int = 1
+
+    # ── Public API ────────────────────────────────────────────────────────────
+
+    def update(self, detections: List[Detection], dt: float) -> List[Track]:
+        """
+        Process one frame of detections.
+
+        1. Predict all tracks by dt.
+        2. Hungarian assignment of predictions → detections.
+        3. Update matched tracks; mark unmatched tracks as LOST.
+        4. Promote tracks crossing `confirm_frames`.
+        5. Create new tracks for unmatched detections (if room).
+        6. Remove DEAD tracks.
+
+        Returns confirmed tracks only.
+        """
+        # 1. Predict
+        for tr in self._tracks.values():
+            tr.kalman.predict(dt)
+            tr.age += 1
+
+        # 2. Assign
+        matched, unmatched_tracks, unmatched_dets = self._assign(detections)
+
+        # 3a. Update matched
+        for tid, did in matched:
+            tr = self._tracks[tid]
+            det = detections[did]
+            tr.kalman.update(np.array([det.x, det.y]))
+            tr.hits   += 1
+            tr.missed  = 0
+            if tr.state == TrackState.LOST:
+                tr.state = TrackState.CONFIRMED
+            elif tr.state == TrackState.TENTATIVE and tr.hits >= self._confirm_frames:
+                tr.state = TrackState.CONFIRMED
+
+        # 3b. Unmatched tracks
+        for tid in unmatched_tracks:
+            tr = self._tracks[tid]
+            tr.missed += 1
+            if tr.missed > 1:
+                tr.state = TrackState.LOST
+
+        # 4. New tracks for unmatched detections
+        live = sum(1 for t in self._tracks.values() if t.state != TrackState.LOST
+                   or t.missed <= self._max_missed)
+        for did in unmatched_dets:
+            if live >= self._max_tracks:
+                break
+            det = detections[did]
+            init_state = (TrackState.CONFIRMED
+                          if self._confirm_frames <= 1
+                          else TrackState.TENTATIVE)
+            new_tr = Track(
+                track_id=self._next_id,
+                kalman=KalmanTracker(det.x, det.y, self._proc_noise),
+                state=init_state,
+            )
+            self._tracks[self._next_id] = new_tr
+            self._next_id += 1
+            live += 1
+
+        # 5. Prune dead
+        dead = [tid for tid, t in self._tracks.items() if t.missed > self._max_missed]
+        for tid in dead:
+            del self._tracks[tid]
+
+        return [t for t in self._tracks.values() if t.state == TrackState.CONFIRMED]
+
+    @property
+    def all_tracks(self) -> List[Track]:
+        return list(self._tracks.values())
+
+    @property
+    def tentative_count(self) -> int:
+        return sum(1 for t in self._tracks.values()
+                   if t.state == TrackState.TENTATIVE)
+
+    def reset(self) -> None:
+        self._tracks.clear()
+        self._next_id = 1
+
+    # ── Hungarian assignment ──────────────────────────────────────────────────
+
+    def _assign(
+        self,
+        detections: List[Detection],
+    ) -> Tuple[List[Tuple[int, int]], List[int], List[int]]:
+        """
+        Returns:
+          matched        — list of (track_id, det_index)
+          unmatched_tids — track IDs with no detection assigned
+          unmatched_dids — detection indices with no track assigned
+        """
+        track_ids = list(self._tracks.keys())
+
+        if not track_ids or not detections:
+            return [], track_ids, list(range(len(detections)))
+
+        # Build cost matrix: rows=tracks, cols=detections
+        cost = np.full((len(track_ids), len(detections)), fill_value=np.inf)
+        for r, tid in enumerate(track_ids):
+            tx, ty = self._tracks[tid].kalman.position
+            for c, det in enumerate(detections):
+                cost[r, c] = np.hypot(tx - det.x, ty - det.y)
+
+        row_ind, col_ind = linear_sum_assignment(cost)
+
+        matched:          List[Tuple[int, int]] = []
+        matched_track_idx: set = set()
+        matched_det_idx:   set = set()
+
+        for r, c in zip(row_ind, col_ind):
+            if cost[r, c] > self._assoc_dist:
+                continue
+            matched.append((track_ids[r], c))
+            matched_track_idx.add(r)
+            matched_det_idx.add(c)
+
+        unmatched_tids = [track_ids[r] for r in range(len(track_ids))
+                          if r not in matched_track_idx]
+        unmatched_dids = [c for c in range(len(detections))
+                          if c not in matched_det_idx]
+
+        return matched, unmatched_tids, unmatched_dids

jetson/ros2_ws/src/saltybot_dynamic_obstacles/setup.cfg

@@ -0,0 +1,4 @@
+[develop]
+script_dir=$base/lib/saltybot_dynamic_obstacles
+[install]
+install_scripts=$base/lib/saltybot_dynamic_obstacles

jetson/ros2_ws/src/saltybot_dynamic_obstacles/setup.py

@@ -0,0 +1,32 @@
+from setuptools import setup, find_packages
+from glob import glob
+
+package_name = 'saltybot_dynamic_obstacles'
+
+setup(
+    name=package_name,
+    version='0.1.0',
+    packages=find_packages(exclude=['test']),
+    data_files=[
+        ('share/ament_index/resource_index/packages',
+            ['resource/' + package_name]),
+        ('share/' + package_name, ['package.xml']),
+        ('share/' + package_name + '/launch',
+            glob('launch/*.launch.py')),
+        ('share/' + package_name + '/config',
+            glob('config/*.yaml')),
+    ],
+    install_requires=['setuptools'],
+    zip_safe=True,
+    maintainer='SaltyLab',
+    maintainer_email='robot@saltylab.local',
+    description='Dynamic obstacle tracking: LIDAR motion detection, Kalman tracking, Nav2 costmap',
+    license='MIT',
+    tests_require=['pytest'],
+    entry_points={
+        'console_scripts': [
+            'dynamic_obs_tracker = saltybot_dynamic_obstacles.dynamic_obs_node:main',
+            'dynamic_obs_costmap = saltybot_dynamic_obstacles.costmap_layer_node:main',
+        ],
+    },
+)

jetson/ros2_ws/src/saltybot_dynamic_obstacles/test/test_dynamic_obstacles.py

@@ -0,0 +1,262 @@
+"""
+test_dynamic_obstacles.py — Unit tests for KalmanTracker, TrackerManager,
+and ObjectDetector.
+
+Runs without ROS2 / hardware (no rclpy imports).
+"""
+
+from __future__ import annotations
+
+import math
+import sys
+import os
+
+import numpy as np
+import pytest
+
+sys.path.insert(0, os.path.join(os.path.dirname(__file__), '..'))
+
+from saltybot_dynamic_obstacles.kalman_tracker import KalmanTracker
+from saltybot_dynamic_obstacles.tracker_manager import TrackerManager, TrackState
+from saltybot_dynamic_obstacles.object_detector import ObjectDetector, Detection
+
+
+# ── KalmanTracker ─────────────────────────────────────────────────────────────
+
+class TestKalmanTracker:
+
+    def test_initial_position(self):
+        kt = KalmanTracker(3.0, 4.0)
+        px, py = kt.position
+        assert px == pytest.approx(3.0)
+        assert py == pytest.approx(4.0)
+
+    def test_initial_velocity_zero(self):
+        kt = KalmanTracker(0.0, 0.0)
+        vx, vy = kt.velocity
+        assert vx == pytest.approx(0.0)
+        assert vy == pytest.approx(0.0)
+
+    def test_predict_moves_position(self):
+        kt = KalmanTracker(0.0, 0.0)
+        # Give it some velocity via update sequence
+        kt.update(np.array([0.1, 0.0]))
+        kt.update(np.array([0.2, 0.0]))
+        kt.predict(0.1)
+        px, _ = kt.position
+        assert px > 0.0   # should have moved forward
+
+    def test_pure_predict_constant_velocity(self):
+        """After velocity is established, predict() should move linearly."""
+        kt = KalmanTracker(0.0, 0.0)
+        # Force velocity by repeated updates
+        for i in range(10):
+            kt.update(np.array([i * 0.1, 0.0]))
+            kt.predict(0.1)
+        vx, _ = kt.velocity
+        px0, _ = kt.position
+        kt.predict(1.0)
+        px1, _ = kt.position
+        # Should advance roughly vx * 1.0 metres
+        assert px1 == pytest.approx(px0 + vx * 1.0, abs=0.3)
+
+    def test_update_corrects_position(self):
+        kt = KalmanTracker(0.0, 0.0)
+        # Predict way off
+        kt.predict(10.0)
+        # Then update to ground truth
+        kt.update(np.array([1.0, 2.0]))
+        px, py = kt.position
+        # Should move toward (1, 2)
+        assert px == pytest.approx(1.0, abs=0.5)
+        assert py == pytest.approx(2.0, abs=0.5)
+
+    def test_predict_horizon_length(self):
+        kt = KalmanTracker(0.0, 0.0)
+        preds = kt.predict_horizon(horizon_s=2.5, step_s=0.5)
+        assert len(preds) == 5   # 0.5, 1.0, 1.5, 2.0, 2.5
+
+    def test_predict_horizon_times(self):
+        kt = KalmanTracker(0.0, 0.0)
+        preds = kt.predict_horizon(horizon_s=2.0, step_s=0.5)
+        times = [t for _, _, t in preds]
+        assert times == pytest.approx([0.5, 1.0, 1.5, 2.0], abs=0.01)
+
+    def test_predict_horizon_does_not_mutate_state(self):
+        kt = KalmanTracker(1.0, 2.0)
+        kt.predict_horizon(horizon_s=3.0, step_s=0.5)
+        px, py = kt.position
+        assert px == pytest.approx(1.0)
+        assert py == pytest.approx(2.0)
+
+    def test_speed_zero_at_init(self):
+        kt = KalmanTracker(5.0, 5.0)
+        assert kt.speed == pytest.approx(0.0)
+
+    def test_covariance_shape(self):
+        kt = KalmanTracker(0.0, 0.0)
+        cov = kt.covariance_2x2
+        assert cov.shape == (2, 2)
+
+    def test_covariance_positive_definite(self):
+        kt = KalmanTracker(0.0, 0.0)
+        for _ in range(5):
+            kt.predict(0.1)
+            kt.update(np.array([0.1, 0.0]))
+        eigvals = np.linalg.eigvalsh(kt.covariance_2x2)
+        assert np.all(eigvals > 0)
+
+    def test_joseph_form_stays_symmetric(self):
+        """Covariance should remain symmetric after many updates."""
+        kt = KalmanTracker(0.0, 0.0)
+        for i in range(50):
+            kt.predict(0.1)
+            kt.update(np.array([i * 0.01, 0.0]))
+        P = kt._P
+        assert np.allclose(P, P.T, atol=1e-10)
+
+
+# ── TrackerManager ────────────────────────────────────────────────────────────
+
+class TestTrackerManager:
+
+    def _det(self, x, y):
+        return Detection(x=x, y=y, size_m2=0.1, range_m=math.hypot(x, y))
+
+    def test_empty_detections_no_tracks(self):
+        tm = TrackerManager()
+        confirmed = tm.update([], 0.1)
+        assert confirmed == []
+
+    def test_track_created_on_detection(self):
+        tm = TrackerManager(confirm_frames=1)
+        confirmed = tm.update([self._det(1.0, 0.0)], 0.1)
+        assert len(confirmed) == 1
+
+    def test_track_tentative_before_confirm(self):
+        tm = TrackerManager(confirm_frames=3)
+        tm.update([self._det(1.0, 0.0)], 0.1)
+        # Only 1 hit — should still be tentative
+        assert tm.tentative_count == 1
+
+    def test_track_confirmed_after_N_hits(self):
+        tm = TrackerManager(confirm_frames=3, assoc_dist_m=2.0)
+        for _ in range(4):
+            confirmed = tm.update([self._det(1.0, 0.0)], 0.1)
+        assert len(confirmed) == 1
+
+    def test_track_deleted_after_max_missed(self):
+        tm = TrackerManager(confirm_frames=1, max_missed=3, assoc_dist_m=2.0)
+        tm.update([self._det(1.0, 0.0)], 0.1)   # create
+        for _ in range(5):
+            tm.update([], 0.1)   # no detections → missed++
+        assert len(tm.all_tracks) == 0
+
+    def test_max_tracks_cap(self):
+        tm = TrackerManager(max_tracks=5, confirm_frames=1)
+        dets = [self._det(float(i), 0.0) for i in range(10)]
+        tm.update(dets, 0.1)
+        assert len(tm.all_tracks) <= 5
+
+    def test_consistent_track_id(self):
+        tm = TrackerManager(confirm_frames=3, assoc_dist_m=1.5)
+        for i in range(5):
+            confirmed = tm.update([self._det(1.0 + i * 0.01, 0.0)], 0.1)
+        assert len(confirmed) == 1
+        track_id = confirmed[0].track_id
+        # One more tick — ID should be stable
+        confirmed2 = tm.update([self._det(1.06, 0.0)], 0.1)
+        assert confirmed2[0].track_id == track_id
+
+    def test_two_independent_tracks(self):
+        tm = TrackerManager(confirm_frames=3, assoc_dist_m=0.8)
+        for _ in range(5):
+            confirmed = tm.update([self._det(1.0, 0.0), self._det(5.0, 0.0)], 0.1)
+        assert len(confirmed) == 2
+
+    def test_reset_clears_all(self):
+        tm = TrackerManager(confirm_frames=1)
+        tm.update([self._det(1.0, 0.0)], 0.1)
+        tm.reset()
+        assert len(tm.all_tracks) == 0
+
+    def test_far_detection_not_assigned(self):
+        tm = TrackerManager(confirm_frames=1, assoc_dist_m=0.5)
+        tm.update([self._det(1.0, 0.0)], 0.1)   # create track at (1,0)
+        # Detection 3 m away → new track, not update
+        tm.update([self._det(4.0, 0.0)], 0.1)
+        assert len(tm.all_tracks) == 2
+
+
+# ── ObjectDetector ────────────────────────────────────────────────────────────
+
+class TestObjectDetector:
+
+    def _empty_scan(self, n=360, rmax=8.0) -> tuple:
+        """All readings at max range (static background)."""
+        ranges = np.full(n, rmax - 0.1, dtype=np.float32)
+        return ranges, -math.pi, 2 * math.pi / n, rmax
+
+    def _scan_with_blob(self, blob_r=2.0, blob_theta=0.0, n=360, rmax=8.0) -> tuple:
+        """Background scan + a short-range cluster at blob_theta."""
+        ranges = np.full(n, rmax - 0.1, dtype=np.float32)
+        angle_inc = 2 * math.pi / n
+        angle_min = -math.pi
+        # Put a cluster of ~10 beams at blob_r
+        center_idx = int((blob_theta - angle_min) / angle_inc) % n
+        for di in range(-5, 6):
+            idx = (center_idx + di) % n
+            ranges[idx] = blob_r
+        return ranges, angle_min, angle_inc, rmax
+
+    def test_empty_scan_no_detections_after_warmup(self):
+        od = ObjectDetector()
+        r, a_min, a_inc, rmax = self._empty_scan()
+        od.update(r, a_min, a_inc, rmax)   # init background
+        for _ in range(3):
+            dets = od.update(r, a_min, a_inc, rmax)
+        assert len(dets) == 0
+
+    def test_moving_blob_detected(self):
+        od = ObjectDetector()
+        bg_r, a_min, a_inc, rmax = self._empty_scan()
+        od.update(bg_r, a_min, a_inc, rmax)   # seed background
+        for _ in range(5):
+            od.update(bg_r, a_min, a_inc, rmax)
+        # Now inject a foreground blob
+        fg_r, _, _, _ = self._scan_with_blob(blob_r=2.0, blob_theta=0.0)
+        dets = od.update(fg_r, a_min, a_inc, rmax)
+        assert len(dets) >= 1
+
+    def test_detection_centroid_approximate(self):
+        od = ObjectDetector()
+        bg_r, a_min, a_inc, rmax = self._empty_scan()
+        for _ in range(8):
+            od.update(bg_r, a_min, a_inc, rmax)
+        fg_r, _, _, _ = self._scan_with_blob(blob_r=3.0, blob_theta=0.0)
+        dets = od.update(fg_r, a_min, a_inc, rmax)
+        assert len(dets) >= 1
+        # Blob is at ~3 m along x-axis (theta=0)
+        cx = dets[0].x
+        cy = dets[0].y
+        assert abs(cx - 3.0) < 0.5
+        assert abs(cy) < 0.5
+
+    def test_reset_clears_background(self):
+        od = ObjectDetector()
+        bg_r, a_min, a_inc, rmax = self._empty_scan()
+        for _ in range(5):
+            od.update(bg_r, a_min, a_inc, rmax)
+        od.reset()
+        assert not od._initialized
+
+    def test_no_inf_nan_ranges(self):
+        od = ObjectDetector()
+        r = np.array([np.inf, np.nan, 5.0, -1.0, 0.0] * 72, dtype=np.float32)
+        a_min = -math.pi
+        a_inc = 2 * math.pi / len(r)
+        od.update(r, a_min, a_inc, 8.0)   # should not raise
+
+
+if __name__ == '__main__':
+    pytest.main([__file__, '-v'])

jetson/ros2_ws/src/saltybot_emergency/config/emergency_params.yaml

@@ -0,0 +1,43 @@
+/**:
+  ros__parameters:
+    # Control loop rate (Hz)
+    control_rate: 20.0
+
+    # Odometry topic for stuck detection
+    odom_topic: "/saltybot/rover_odom"
+
+    # ── LaserScan forward sector ───────────────────────────────────────────────
+    forward_scan_angle_rad: 0.785   # ±45° forward sector
+
+    # ── Obstacle proximity ────────────────────────────────────────────────────
+    stop_distance_m:     0.30       # MAJOR threshold (spec: <30 cm)
+    critical_distance_m: 0.10       # CRITICAL threshold
+    min_cmd_speed_ms:    0.05       # ignore obstacle when nearly stopped
+
+    # ── Fall detection (IMU tilt) ─────────────────────────────────────────────
+    minor_tilt_rad:    0.20         # advisory
+    major_tilt_rad:    0.35         # stop + recover
+    critical_tilt_rad: 0.52         # ~30° — full shutdown
+    floor_drop_m:      0.15         # depth discontinuity triggering MAJOR
+
+    # ── Stuck detection ───────────────────────────────────────────────────────
+    stuck_timeout_s: 3.0            # (spec: 3 s wheel stall)
+
+    # ── Bump / jerk detection ─────────────────────────────────────────────────
+    jerk_threshold_ms3:          8.0
+    critical_jerk_threshold_ms3: 25.0
+
+    # ── FSM / recovery ────────────────────────────────────────────────────────
+    stopped_ms:            0.03     # speed below which robot is "stopped" (m/s)
+    major_count_threshold: 3        # MAJOR alerts before escalation to CRITICAL
+    escalation_window_s:   10.0     # sliding window for escalation counter (s)
+    suppression_s:         1.0      # de-bounce period for duplicate alerts (s)
+
+    # Recovery sequence
+    reverse_speed_ms:   -0.15       # back-up speed (m/s; must be negative)
+    reverse_distance_m:  0.30       # distance to reverse each cycle (m)
+    angular_speed_rads:  0.60       # turn speed (rad/s)
+    turn_angle_rad:      1.5708     # ~90° turn (rad)
+    retry_timeout_s:     3.0        # time in RETRYING per attempt (s)
+    clear_hold_s:        0.50       # consecutive clear time to declare success (s)
+    max_retries:         3          # maximum reverse+turn attempts before GAVE_UP

jetson/ros2_ws/src/saltybot_emergency/launch/emergency.launch.py

@@ -0,0 +1,53 @@
+"""
+emergency.launch.py — Launch the emergency behavior system (Issue #169).
+
+Usage
+-----
+  ros2 launch saltybot_emergency emergency.launch.py
+  ros2 launch saltybot_emergency emergency.launch.py \
+      stop_distance_m:=0.30 max_retries:=3
+"""
+
+import os
+
+from ament_index_python.packages import get_package_share_directory
+from launch import LaunchDescription
+from launch.actions import DeclareLaunchArgument
+from launch.substitutions import LaunchConfiguration
+from launch_ros.actions import Node
+
+
+def generate_launch_description():
+    pkg_share = get_package_share_directory("saltybot_emergency")
+    default_params = os.path.join(pkg_share, "config", "emergency_params.yaml")
+
+    return LaunchDescription([
+        DeclareLaunchArgument(
+            "params_file",
+            default_value=default_params,
+            description="Path to emergency_params.yaml",
+        ),
+        DeclareLaunchArgument(
+            "stop_distance_m",
+            default_value="0.30",
+            description="Obstacle distance triggering MAJOR stop (m)",
+        ),
+        DeclareLaunchArgument(
+            "max_retries",
+            default_value="3",
+            description="Maximum recovery cycles before ESCALATED",
+        ),
+        Node(
+            package="saltybot_emergency",
+            executable="emergency_node",
+            name="emergency",
+            output="screen",
+            parameters=[
+                LaunchConfiguration("params_file"),
+                {
+                    "stop_distance_m": LaunchConfiguration("stop_distance_m"),
+                    "max_retries":     LaunchConfiguration("max_retries"),
+                },
+            ],
+        ),
+    ])

jetson/ros2_ws/src/saltybot_emergency/package.xml

@@ -0,0 +1,24 @@
+<?xml version="1.0"?>
+<?xml-model href="http://download.ros.org/schema/package_format3.xsd" schematypens="http://www.w3.org/2001/XMLSchema"?>
+<package format="3">
+  <name>saltybot_emergency</name>
+  <version>0.1.0</version>
+  <description>Emergency behavior system — collision avoidance, fall prevention, stuck detection, recovery (Issue #169)</description>
+  <maintainer email="sl-controls@saltylab.local">sl-controls</maintainer>
+  <license>MIT</license>
+
+  <depend>rclpy</depend>
+  <depend>sensor_msgs</depend>
+  <depend>nav_msgs</depend>
+  <depend>geometry_msgs</depend>
+  <depend>std_msgs</depend>
+
+  <test_depend>ament_copyright</test_depend>
+  <test_depend>ament_flake8</test_depend>
+  <test_depend>ament_pep257</test_depend>
+  <test_depend>pytest</test_depend>
+
+  <export>
+    <build_type>ament_python</build_type>
+  </export>
+</package>

jetson/ros2_ws/src/saltybot_emergency/saltybot_emergency/alert_manager.py

@@ -0,0 +1,139 @@
+"""
+alert_manager.py — Alert severity escalation for emergency behavior (Issue #169).
+
+Alert levels
+────────────
+  NONE     : no action
+  MINOR    : advisory beep  → publish to /saltybot/alert_beep
+  MAJOR    : stop + LED flash → publish to /saltybot/alert_flash; cmd_vel override
+  CRITICAL : full shutdown + MQTT → publish to /saltybot/e_stop + /saltybot/critical_alert
+
+Escalation
+──────────
+  If major_count_threshold MAJOR alerts occur within escalation_window_s, the
+  next MAJOR is promoted to CRITICAL.  This catches persistent stuck / repeated
+  collision scenarios.
+
+Suppression
+───────────
+  Identical (type, level) alerts are suppressed within suppression_s to avoid
+  flooding downstream topics.
+
+Pure module — no ROS2 dependency.
+"""
+
+from __future__ import annotations
+
+from collections import deque
+from dataclasses import dataclass
+from enum import Enum
+from typing import Optional
+
+from saltybot_emergency.threat_detector import ThreatEvent, ThreatLevel, ThreatType
+
+
+# ── Alert level ───────────────────────────────────────────────────────────────
+
+class AlertLevel(Enum):
+    NONE     = 0
+    MINOR    = 1   # beep
+    MAJOR    = 2   # stop + flash
+    CRITICAL = 3   # shutdown + MQTT
+
+
+# ── Alert ─────────────────────────────────────────────────────────────────────
+
+@dataclass
+class Alert:
+    level:       AlertLevel
+    source:      str     # ThreatType value string
+    message:     str
+    timestamp_s: float
+
+
+# ── AlertManager ─────────────────────────────────────────────────────────────
+
+class AlertManager:
+    """
+    Converts ThreatEvents to Alerts with escalation and suppression logic.
+
+    Parameters
+    ----------
+    major_count_threshold : number of MAJOR alerts within window to escalate
+    escalation_window_s   : sliding window for escalation counting (s)
+    suppression_s         : suppress duplicate (type, level) alerts within this period
+    """
+
+    def __init__(
+        self,
+        major_count_threshold: int   = 3,
+        escalation_window_s:   float = 10.0,
+        suppression_s:         float = 1.0,
+    ):
+        self._major_threshold  = max(1, int(major_count_threshold))
+        self._esc_window       = float(escalation_window_s)
+        self._suppress         = float(suppression_s)
+        self._major_times:     deque  = deque()            # timestamps of MAJOR alerts
+        self._last_seen:       dict   = {}                 # (type, level) → timestamp
+
+    # ── Update ────────────────────────────────────────────────────────────────
+
+    def update(self, threat: ThreatEvent) -> Optional[Alert]:
+        """
+        Convert one ThreatEvent to an Alert, applying escalation and suppression.
+
+        Returns None if threat is CLEAR or the alert is suppressed.
+        """
+        if threat.level == ThreatLevel.CLEAR:
+            return None
+
+        now = threat.timestamp_s
+        alert_level = _threat_to_alert(threat.level)
+
+        # ── Suppression ───────────────────────────────────────────────────────
+        key = (threat.threat_type, threat.level)
+        last = self._last_seen.get(key)
+        if last is not None and (now - last) < self._suppress:
+            return None
+        self._last_seen[key] = now
+
+        # ── Escalation ────────────────────────────────────────────────────────
+        if alert_level == AlertLevel.MAJOR:
+            # Prune old timestamps
+            while self._major_times and (now - self._major_times[0]) > self._esc_window:
+                self._major_times.popleft()
+            self._major_times.append(now)
+            if len(self._major_times) >= self._major_threshold:
+                alert_level = AlertLevel.CRITICAL
+
+        msg = _build_message(alert_level, threat)
+        return Alert(
+            level=alert_level,
+            source=threat.threat_type.value,
+            message=msg,
+            timestamp_s=now,
+        )
+
+    def reset(self) -> None:
+        """Clear escalation history and suppression state."""
+        self._major_times.clear()
+        self._last_seen.clear()
+
+
+# ── Helpers ───────────────────────────────────────────────────────────────────
+
+def _threat_to_alert(level: ThreatLevel) -> AlertLevel:
+    return {
+        ThreatLevel.MINOR:    AlertLevel.MINOR,
+        ThreatLevel.MAJOR:    AlertLevel.MAJOR,
+        ThreatLevel.CRITICAL: AlertLevel.CRITICAL,
+    }.get(level, AlertLevel.NONE)
+
+
+def _build_message(level: AlertLevel, threat: ThreatEvent) -> str:
+    prefix = {
+        AlertLevel.MINOR:    "[MINOR]",
+        AlertLevel.MAJOR:    "[MAJOR]",
+        AlertLevel.CRITICAL: "[CRITICAL]",
+    }.get(level, "[?]")
+    return f"{prefix} {threat.threat_type.value}: {threat.detail}"

jetson/ros2_ws/src/saltybot_emergency/saltybot_emergency/emergency_fsm.py

@@ -0,0 +1,232 @@
+"""
+emergency_fsm.py — Master emergency FSM integrating all detectors (Issue #169).
+
+States
+──────
+  NOMINAL    : normal operation; minor alerts pass through; major/critical → STOPPING.
+  STOPPING   : commanding zero velocity until robot speed drops below stopped_ms.
+               Critical threats skip RECOVERING → ESCALATED immediately.
+  RECOVERING : RecoverySequencer executing reverse+turn sequence.
+               Success → NOMINAL; gave-up → ESCALATED.
+  ESCALATED  : full stop; critical alert emitted once; stays until acknowledge.
+
+Alert actions produced by state
+────────────────────────────────
+  NOMINAL    : emit MINOR alert (beep only); no velocity override.
+  STOPPING   : suppress nav, publish zero; emit MAJOR alert once.
+  RECOVERING : suppress nav, publish recovery cmds; no new alerts.
+  ESCALATED  : suppress nav, publish zero; emit CRITICAL alert once per entry.
+
+Pure module — no ROS2 dependency.
+"""
+
+from __future__ import annotations
+
+from dataclasses import dataclass, field
+from enum import Enum
+from typing import Optional
+
+from saltybot_emergency.alert_manager import Alert, AlertLevel, AlertManager
+from saltybot_emergency.recovery_sequencer import RecoveryInputs, RecoverySequencer
+from saltybot_emergency.threat_detector import ThreatEvent, ThreatLevel
+
+
+# ── States ────────────────────────────────────────────────────────────────────
+
+class EmergencyState(Enum):
+    NOMINAL    = "NOMINAL"
+    STOPPING   = "STOPPING"
+    RECOVERING = "RECOVERING"
+    ESCALATED  = "ESCALATED"
+
+
+# ── I/O ───────────────────────────────────────────────────────────────────────
+
+@dataclass
+class EmergencyInputs:
+    threat:         ThreatEvent  # highest-severity threat this tick
+    robot_speed_ms: float = 0.0  # actual speed from odometry (m/s)
+    acknowledge:    bool  = False # operator cleared the escalation
+
+
+@dataclass
+class EmergencyOutputs:
+    state:        EmergencyState    = EmergencyState.NOMINAL
+    cmd_override: bool              = False   # True = emergency owns cmd_vel
+    cmd_linear:   float             = 0.0
+    cmd_angular:  float             = 0.0
+    alert:        Optional[Alert]   = None
+    e_stop:       bool              = False   # assert /saltybot/e_stop
+    state_changed: bool             = False
+    recovery_progress: float        = 0.0
+    recovery_retry_count: int       = 0
+
+
+# ── EmergencyFSM ──────────────────────────────────────────────────────────────
+
+class EmergencyFSM:
+    """
+    Master emergency FSM.
+
+    Owns an AlertManager and a RecoverySequencer; coordinates them each tick.
+
+    Parameters
+    ----------
+    stopped_ms              : speed below which robot is considered stopped (m/s)
+    major_count_threshold   : MAJOR events within window before escalation
+    escalation_window_s     : sliding window for escalation (s)
+    suppression_s           : alert de-bounce period (s)
+    reverse_speed_ms        : reverse speed during recovery (m/s)
+    reverse_distance_m      : reverse travel per cycle (m)
+    angular_speed_rads      : turn speed during recovery (rad/s)
+    turn_angle_rad          : turn per cycle (rad)
+    retry_timeout_s         : time in RETRYING before next cycle (s)
+    clear_hold_s            : clear duration required to declare success (s)
+    max_retries             : recovery cycles before GAVE_UP
+    """
+
+    def __init__(
+        self,
+        stopped_ms:            float = 0.03,
+        major_count_threshold: int   = 3,
+        escalation_window_s:   float = 10.0,
+        suppression_s:         float = 1.0,
+        reverse_speed_ms:      float = -0.15,
+        reverse_distance_m:    float = 0.30,
+        angular_speed_rads:    float = 0.60,
+        turn_angle_rad:        float = 1.5708,
+        retry_timeout_s:       float = 3.0,
+        clear_hold_s:          float = 0.5,
+        max_retries:           int   = 3,
+    ):
+        self._stopped_ms = max(0.0, stopped_ms)
+        self._alert_mgr  = AlertManager(
+            major_count_threshold=major_count_threshold,
+            escalation_window_s=escalation_window_s,
+            suppression_s=suppression_s,
+        )
+        self._recovery = RecoverySequencer(
+            reverse_speed_ms=reverse_speed_ms,
+            reverse_distance_m=reverse_distance_m,
+            angular_speed_rads=angular_speed_rads,
+            turn_angle_rad=turn_angle_rad,
+            retry_timeout_s=retry_timeout_s,
+            clear_hold_s=clear_hold_s,
+            max_retries=max_retries,
+        )
+        self._state            = EmergencyState.NOMINAL
+        self._critical_pending = False   # STOPPING → ESCALATED (not RECOVERING)
+        self._escalation_alerted = False # CRITICAL alert emitted once per ESCALATED entry
+
+    # ── Public API ────────────────────────────────────────────────────────────
+
+    @property
+    def state(self) -> EmergencyState:
+        return self._state
+
+    def reset(self) -> None:
+        self._state              = EmergencyState.NOMINAL
+        self._critical_pending   = False
+        self._escalation_alerted = False
+        self._alert_mgr.reset()
+        self._recovery.reset()
+
+    def tick(self, inputs: EmergencyInputs) -> EmergencyOutputs:
+        prev = self._state
+        out  = self._step(inputs)
+        out.state         = self._state
+        out.state_changed = (self._state != prev)
+        return out
+
+    # ── Step ─────────────────────────────────────────────────────────────────
+
+    def _step(self, inp: EmergencyInputs) -> EmergencyOutputs:
+        out = EmergencyOutputs(state=self._state)
+
+        # Run alert manager for this threat
+        alert = self._alert_mgr.update(inp.threat)
+
+        # ── NOMINAL ───────────────────────────────────────────────────────────
+        if self._state == EmergencyState.NOMINAL:
+            if inp.threat.level == ThreatLevel.CRITICAL:
+                self._state            = EmergencyState.STOPPING
+                self._critical_pending = True
+                out.alert              = alert
+                out.cmd_override       = True   # start overriding on entry tick
+            elif inp.threat.level == ThreatLevel.MAJOR:
+                self._state            = EmergencyState.STOPPING
+                self._critical_pending = False
+                out.alert              = alert
+                out.cmd_override       = True   # start overriding on entry tick
+            elif inp.threat.level == ThreatLevel.MINOR:
+                # Advisory only — no override
+                out.alert = alert
+
+        # ── STOPPING ──────────────────────────────────────────────────────────
+        elif self._state == EmergencyState.STOPPING:
+            out.cmd_override = True
+            out.cmd_linear   = 0.0
+            out.cmd_angular  = 0.0
+            # Upgrade to critical if new critical arrives
+            if inp.threat.level == ThreatLevel.CRITICAL:
+                self._critical_pending = True
+            if abs(inp.robot_speed_ms) <= self._stopped_ms:
+                if self._critical_pending:
+                    self._state              = EmergencyState.ESCALATED
+                    self._escalation_alerted = False
+                else:
+                    self._state = EmergencyState.RECOVERING
+                    self._recovery.reset()
+                    self._recovery.tick(RecoveryInputs(trigger=True, dt=0.0))
+
+        # ── RECOVERING ────────────────────────────────────────────────────────
+        elif self._state == EmergencyState.RECOVERING:
+            threat_cleared = inp.threat.level == ThreatLevel.CLEAR
+            rec_inp = RecoveryInputs(
+                trigger=False,
+                threat_cleared=threat_cleared,
+                dt=0.02,   # nominal dt; node should pass actual dt
+            )
+            rec_out = self._recovery.tick(rec_inp)
+
+            out.cmd_override          = True
+            out.cmd_linear            = rec_out.cmd_linear
+            out.cmd_angular           = rec_out.cmd_angular
+            out.recovery_progress     = rec_out.progress
+            out.recovery_retry_count  = rec_out.retry_count
+
+            if rec_out.gave_up:
+                self._state              = EmergencyState.ESCALATED
+                self._escalation_alerted = False
+            elif rec_out.state.value == "IDLE" and not inp.trigger if hasattr(inp, "trigger") else True:
+                # RecoverySequencer returned to IDLE = success
+                from saltybot_emergency.recovery_sequencer import RecoveryState
+                if self._recovery.state == RecoveryState.IDLE and not rec_out.gave_up:
+                    self._state = EmergencyState.NOMINAL
+                    self._recovery.reset()
+
+        # ── ESCALATED ─────────────────────────────────────────────────────────
+        elif self._state == EmergencyState.ESCALATED:
+            out.cmd_override = True
+            out.cmd_linear   = 0.0
+            out.cmd_angular  = 0.0
+            out.e_stop       = True
+            if not self._escalation_alerted:
+                # Force a CRITICAL alert regardless of suppression
+                from saltybot_emergency.alert_manager import Alert, AlertLevel
+                out.alert = Alert(
+                    level=AlertLevel.CRITICAL,
+                    source=inp.threat.threat_type.value,
+                    message=f"[CRITICAL] ESCALATED: {inp.threat.detail or 'Recovery gave up'}",
+                    timestamp_s=inp.threat.timestamp_s,
+                )
+                self._escalation_alerted = True
+            if inp.acknowledge:
+                self._state              = EmergencyState.NOMINAL
+                self._critical_pending   = False
+                self._escalation_alerted = False
+                out.e_stop               = False
+                self._alert_mgr.reset()
+                self._recovery.reset()
+
+        return out

jetson/ros2_ws/src/saltybot_emergency/saltybot_emergency/emergency_node.py

@@ -0,0 +1,383 @@
+"""
+emergency_node.py — Emergency behavior system orchestration (Issue #169).
+
+Overview
+────────
+  Aggregates threats from four independent detectors and drives the
+  EmergencyFSM.  Overrides /cmd_vel when an emergency is active.  Escalates
+  via /saltybot/e_stop and /saltybot/critical_alert for CRITICAL events.
+
+Pipeline (20 Hz)
+────────────────
+  1. LaserScan callback → ObstacleDetector → ThreatEvent
+  2. IMU callback       → FallDetector + BumpDetector → ThreatEvent (×2)
+  3. Odom callback      → StuckDetector (fed in timer) → ThreatEvent
+  4. 20 Hz timer        → highest_threat() → EmergencyFSM.tick()
+                        → publish overriding cmd_vel or pass-through
+                        → publish /saltybot/emergency + /saltybot/recovery_action
+
+Subscribes
+──────────
+  /scan                    sensor_msgs/LaserScan  — obstacle detection
+  /saltybot/imu            sensor_msgs/Imu        — fall + bump detection
+  <odom_topic>             nav_msgs/Odometry      — stuck + speed tracking
+  /cmd_vel                 geometry_msgs/Twist    — nav commands (pass-through)
+
+Publishes
+─────────
+  /saltybot/cmd_vel_out    geometry_msgs/Twist    — muxed cmd_vel (to drive nodes)
+  /saltybot/e_stop         std_msgs/Bool          — emergency stop flag
+  /saltybot/alert_beep     std_msgs/Empty         — beep trigger (MINOR)
+  /saltybot/alert_flash    std_msgs/Empty         — LED flash trigger (MAJOR)
+  /saltybot/critical_alert std_msgs/String (JSON) — CRITICAL event for MQTT bridge
+  /saltybot/emergency      saltybot_emergency_msgs/EmergencyEvent
+  /saltybot/recovery_action saltybot_emergency_msgs/RecoveryAction
+
+Parameters
+──────────
+  control_rate             20.0    Hz
+  odom_topic               /saltybot/rover_odom
+  forward_scan_angle_rad   0.785   rad  (±45° forward sector for obstacle check)
+  stop_distance_m          0.30    m
+  critical_distance_m      0.10    m
+  min_cmd_speed_ms         0.05    m/s
+  minor_tilt_rad           0.20    rad
+  major_tilt_rad           0.35    rad
+  critical_tilt_rad        0.52    rad
+  floor_drop_m             0.15    m
+  stuck_timeout_s          3.0     s
+  jerk_threshold_ms3       8.0     m/s³
+  critical_jerk_threshold_ms3  25.0 m/s³
+  stopped_ms               0.03    m/s
+  major_count_threshold    3
+  escalation_window_s      10.0    s
+  suppression_s            1.0     s
+  reverse_speed_ms        -0.15    m/s
+  reverse_distance_m       0.30    m
+  angular_speed_rads       0.60    rad/s
+  turn_angle_rad           1.5708  rad
+  retry_timeout_s          3.0     s
+  clear_hold_s             0.50    s
+  max_retries              3
+"""
+
+import json
+import math
+import time
+
+import rclpy
+from rclpy.node import Node
+from rclpy.qos import HistoryPolicy, QoSProfile, ReliabilityPolicy
+
+from geometry_msgs.msg import Twist
+from nav_msgs.msg import Odometry
+from sensor_msgs.msg import Imu, LaserScan
+from std_msgs.msg import Bool, Empty, String
+
+from saltybot_emergency.alert_manager import AlertLevel
+from saltybot_emergency.emergency_fsm import EmergencyFSM, EmergencyInputs, EmergencyState
+from saltybot_emergency.threat_detector import (
+    BumpDetector,
+    FallDetector,
+    ObstacleDetector,
+    StuckDetector,
+    ThreatEvent,
+    ThreatType,
+    highest_threat,
+)
+
+try:
+    from saltybot_emergency_msgs.msg import EmergencyEvent, RecoveryAction
+    _MSGS_OK = True
+except ImportError:
+    _MSGS_OK = False
+
+
+def _quaternion_to_pitch_roll(qx, qy, qz, qw):
+    pitch = math.asin(max(-1.0, min(1.0, 2.0 * (qw * qy - qz * qx))))
+    roll  = math.atan2(2.0 * (qw * qx + qy * qz), 1.0 - 2.0 * (qx * qx + qy * qy))
+    return pitch, roll
+
+
+class EmergencyNode(Node):
+
+    def __init__(self):
+        super().__init__("emergency")
+
+        self._declare_params()
+        p = self._load_params()
+
+        # ── Detectors ────────────────────────────────────────────────────────
+        self._obstacle = ObstacleDetector(
+            stop_distance_m=p["stop_distance_m"],
+            critical_distance_m=p["critical_distance_m"],
+            min_speed_ms=p["min_cmd_speed_ms"],
+        )
+        self._fall = FallDetector(
+            minor_tilt_rad=p["minor_tilt_rad"],
+            major_tilt_rad=p["major_tilt_rad"],
+            critical_tilt_rad=p["critical_tilt_rad"],
+            floor_drop_m=p["floor_drop_m"],
+        )
+        self._stuck = StuckDetector(
+            stuck_timeout_s=p["stuck_timeout_s"],
+            min_cmd_ms=p["min_cmd_speed_ms"],
+        )
+        self._bump = BumpDetector(
+            jerk_threshold_ms3=p["jerk_threshold_ms3"],
+            critical_jerk_threshold_ms3=p["critical_jerk_threshold_ms3"],
+        )
+        self._fsm = EmergencyFSM(
+            stopped_ms=p["stopped_ms"],
+            major_count_threshold=p["major_count_threshold"],
+            escalation_window_s=p["escalation_window_s"],
+            suppression_s=p["suppression_s"],
+            reverse_speed_ms=p["reverse_speed_ms"],
+            reverse_distance_m=p["reverse_distance_m"],
+            angular_speed_rads=p["angular_speed_rads"],
+            turn_angle_rad=p["turn_angle_rad"],
+            retry_timeout_s=p["retry_timeout_s"],
+            clear_hold_s=p["clear_hold_s"],
+            max_retries=p["max_retries"],
+        )
+
+        # ── State ────────────────────────────────────────────────────────────
+        self._latest_obstacle_threat = ThreatEvent()
+        self._latest_fall_threat     = ThreatEvent()
+        self._latest_bump_threat     = ThreatEvent()
+        self._cmd_speed_ms           = 0.0
+        self._actual_speed_ms        = 0.0
+        self._last_ctrl_t            = time.monotonic()
+        self._scan_forward_angle     = p["forward_scan_angle_rad"]
+        self._acknowledge_flag       = False
+
+        # ── QoS ──────────────────────────────────────────────────────────────
+        reliable = QoSProfile(
+            reliability=ReliabilityPolicy.RELIABLE,
+            history=HistoryPolicy.KEEP_LAST,
+            depth=10,
+        )
+        best_effort = QoSProfile(
+            reliability=ReliabilityPolicy.BEST_EFFORT,
+            history=HistoryPolicy.KEEP_LAST,
+            depth=1,
+        )
+
+        # ── Subscriptions ────────────────────────────────────────────────────
+        self.create_subscription(LaserScan, "/scan",              self._scan_cb,   best_effort)
+        self.create_subscription(Imu,       "/saltybot/imu",      self._imu_cb,    best_effort)
+        self.create_subscription(Odometry,  p["odom_topic"],      self._odom_cb,   reliable)
+        self.create_subscription(Twist,     "/cmd_vel",           self._cmd_vel_cb, reliable)
+        self.create_subscription(Bool,      "/saltybot/emergency_ack", self._ack_cb, reliable)
+
+        # ── Publishers ───────────────────────────────────────────────────────
+        self._cmd_out_pub    = self.create_publisher(Twist,  "/saltybot/cmd_vel_out", reliable)
+        self._estop_pub      = self.create_publisher(Bool,   "/saltybot/e_stop",      reliable)
+        self._beep_pub       = self.create_publisher(Empty,  "/saltybot/alert_beep",  reliable)
+        self._flash_pub      = self.create_publisher(Empty,  "/saltybot/alert_flash", reliable)
+        self._critical_pub   = self.create_publisher(String, "/saltybot/critical_alert", reliable)
+
+        self._event_pub    = None
+        self._recovery_pub = None
+        if _MSGS_OK:
+            self._event_pub    = self.create_publisher(EmergencyEvent,  "/saltybot/emergency",       reliable)
+            self._recovery_pub = self.create_publisher(RecoveryAction, "/saltybot/recovery_action", reliable)
+
+        # ── Timer ────────────────────────────────────────────────────────────
+        rate = p["control_rate"]
+        self._timer = self.create_timer(1.0 / rate, self._control_cb)
+        self.get_logger().info(f"EmergencyNode ready  rate={rate}Hz")
+
+    # ── Parameters ────────────────────────────────────────────────────────────
+
+    def _declare_params(self) -> None:
+        self.declare_parameter("control_rate",               20.0)
+        self.declare_parameter("odom_topic",                 "/saltybot/rover_odom")
+        self.declare_parameter("forward_scan_angle_rad",     0.785)
+        self.declare_parameter("stop_distance_m",            0.30)
+        self.declare_parameter("critical_distance_m",        0.10)
+        self.declare_parameter("min_cmd_speed_ms",           0.05)
+        self.declare_parameter("minor_tilt_rad",             0.20)
+        self.declare_parameter("major_tilt_rad",             0.35)
+        self.declare_parameter("critical_tilt_rad",          0.52)
+        self.declare_parameter("floor_drop_m",               0.15)
+        self.declare_parameter("stuck_timeout_s",            3.0)
+        self.declare_parameter("jerk_threshold_ms3",         8.0)
+        self.declare_parameter("critical_jerk_threshold_ms3", 25.0)
+        self.declare_parameter("stopped_ms",                 0.03)
+        self.declare_parameter("major_count_threshold",      3)
+        self.declare_parameter("escalation_window_s",        10.0)
+        self.declare_parameter("suppression_s",              1.0)
+        self.declare_parameter("reverse_speed_ms",          -0.15)
+        self.declare_parameter("reverse_distance_m",         0.30)
+        self.declare_parameter("angular_speed_rads",         0.60)
+        self.declare_parameter("turn_angle_rad",             1.5708)
+        self.declare_parameter("retry_timeout_s",            3.0)
+        self.declare_parameter("clear_hold_s",               0.50)
+        self.declare_parameter("max_retries",                3)
+
+    def _load_params(self) -> dict:
+        g = self.get_parameter
+        return {k: g(k).value for k in [
+            "control_rate", "odom_topic",
+            "forward_scan_angle_rad",
+            "stop_distance_m", "critical_distance_m", "min_cmd_speed_ms",
+            "minor_tilt_rad", "major_tilt_rad", "critical_tilt_rad", "floor_drop_m",
+            "stuck_timeout_s", "jerk_threshold_ms3", "critical_jerk_threshold_ms3",
+            "stopped_ms",
+            "major_count_threshold", "escalation_window_s", "suppression_s",
+            "reverse_speed_ms", "reverse_distance_m",
+            "angular_speed_rads", "turn_angle_rad",
+            "retry_timeout_s", "clear_hold_s", "max_retries",
+        ]}
+
+    # ── Callbacks ─────────────────────────────────────────────────────────────
+
+    def _scan_cb(self, msg: LaserScan) -> None:
+        # Extract minimum range within forward sector (±forward_scan_angle_rad)
+        half = self._scan_forward_angle
+        ranges = []
+        for i, r in enumerate(msg.ranges):
+            angle = msg.angle_min + i * msg.angle_increment
+            if abs(angle) <= half and msg.range_min < r < msg.range_max:
+                ranges.append(r)
+        min_r = min(ranges) if ranges else float("inf")
+        self._latest_obstacle_threat = self._obstacle.update(
+            min_r, self._cmd_speed_ms, time.monotonic()
+        )
+
+    def _imu_cb(self, msg: Imu) -> None:
+        now = time.monotonic()
+        ax = msg.linear_acceleration.x
+        ay = msg.linear_acceleration.y
+        az = msg.linear_acceleration.z
+        pitch, roll = _quaternion_to_pitch_roll(
+            msg.orientation.x, msg.orientation.y,
+            msg.orientation.z, msg.orientation.w,
+        )
+        # dt for jerk is approximated; bump detector handles None on first call
+        dt = 0.02  # nominal 20 Hz
+        self._latest_fall_threat = self._fall.update(pitch, roll, 0.0, now)
+        self._latest_bump_threat = self._bump.update(ax, ay, az, dt, now)
+
+    def _odom_cb(self, msg: Odometry) -> None:
+        self._actual_speed_ms = msg.twist.twist.linear.x
+
+    def _cmd_vel_cb(self, msg: Twist) -> None:
+        self._cmd_speed_ms = msg.linear.x
+
+    def _ack_cb(self, msg: Bool) -> None:
+        if msg.data:
+            self._acknowledge_flag = True
+
+    # ── 20 Hz control loop ────────────────────────────────────────────────────
+
+    def _control_cb(self) -> None:
+        now = time.monotonic()
+        dt  = now - self._last_ctrl_t
+        self._last_ctrl_t = now
+
+        stuck_threat = self._stuck.update(
+            self._cmd_speed_ms, self._actual_speed_ms, dt, now
+        )
+
+        threat = highest_threat([
+            self._latest_obstacle_threat,
+            self._latest_fall_threat,
+            self._latest_bump_threat,
+            stuck_threat,
+        ])
+
+        inp = EmergencyInputs(
+            threat=threat,
+            robot_speed_ms=self._actual_speed_ms,
+            acknowledge=self._acknowledge_flag,
+        )
+        self._acknowledge_flag = False
+
+        out = self._fsm.tick(inp)
+
+        if out.state_changed:
+            self.get_logger().info(f"Emergency FSM → {out.state.value}")
+
+        # ── Alert dispatch ────────────────────────────────────────────────────
+        if out.alert is not None:
+            lvl = out.alert.level
+            self.get_logger().warn(out.alert.message)
+            if lvl == AlertLevel.MINOR:
+                self._beep_pub.publish(Empty())
+            elif lvl == AlertLevel.MAJOR:
+                self._flash_pub.publish(Empty())
+            elif lvl == AlertLevel.CRITICAL:
+                self._flash_pub.publish(Empty())
+                self._publish_critical_alert(out.alert.message, threat)
+
+        # ── E-stop ───────────────────────────────────────────────────────────
+        estop_msg = Bool()
+        estop_msg.data = out.e_stop
+        self._estop_pub.publish(estop_msg)
+
+        # ── cmd_vel mux ───────────────────────────────────────────────────────
+        twist = Twist()
+        if out.cmd_override:
+            twist.linear.x  = out.cmd_linear
+            twist.angular.z = out.cmd_angular
+        else:
+            twist.linear.x  = self._cmd_speed_ms
+        self._cmd_out_pub.publish(twist)
+
+        # ── Status topics ─────────────────────────────────────────────────────
+        if self._event_pub is not None:
+            self._publish_event(out, threat)
+        if self._recovery_pub is not None:
+            self._publish_recovery(out)
+
+    # ── Publishers ────────────────────────────────────────────────────────────
+
+    def _publish_critical_alert(self, message: str, threat: ThreatEvent) -> None:
+        msg = String()
+        msg.data = json.dumps({
+            "severity":   "CRITICAL",
+            "threat":     threat.threat_type.value,
+            "value":      round(threat.value, 3),
+            "detail":     threat.detail,
+            "message":    message,
+        })
+        self._critical_pub.publish(msg)
+
+    def _publish_event(self, out, threat: ThreatEvent) -> None:
+        msg = EmergencyEvent()
+        msg.stamp        = self.get_clock().now().to_msg()
+        msg.state        = out.state.value
+        msg.threat_type  = threat.threat_type.value
+        msg.severity     = threat.level.name
+        msg.threat_value = float(threat.value)
+        msg.detail       = threat.detail
+        msg.cmd_override = out.cmd_override
+        self._event_pub.publish(msg)
+
+    def _publish_recovery(self, out) -> None:
+        msg = RecoveryAction()
+        msg.stamp        = self.get_clock().now().to_msg()
+        msg.action       = self._fsm._recovery.state.value
+        msg.retry_count  = out.recovery_retry_count
+        msg.progress     = float(out.recovery_progress)
+        self._recovery_pub.publish(msg)
+
+
+# ── Entry point ───────────────────────────────────────────────────────────────
+
+def main(args=None):
+    rclpy.init(args=args)
+    node = EmergencyNode()
+    try:
+        rclpy.spin(node)
+    except KeyboardInterrupt:
+        pass
+    finally:
+        node.destroy_node()
+        rclpy.try_shutdown()
+
+
+if __name__ == "__main__":
+    main()

jetson/ros2_ws/src/saltybot_emergency/saltybot_emergency/recovery_sequencer.py

@@ -0,0 +1,193 @@
+"""
+recovery_sequencer.py — Reverse + turn recovery FSM for emergency behavior (Issue #169).
+
+Sequence
+────────
+  IDLE → REVERSING → TURNING → RETRYING → (IDLE on success)
+                                         → (REVERSING on re-threat, retry loop)
+                                         → (GAVE_UP after max_retries)
+
+  REVERSING : command reverse at reverse_speed_ms until reverse_distance_m covered.
+  TURNING   : command angular_speed_rads until turn_angle_rad covered (90°).
+  RETRYING  : zero velocity; wait up to retry_timeout_s for threat to clear.
+              If threat clears within clear_hold_s → back to IDLE (success).
+              If timeout without clearance → start another REVERSING cycle.
+              If retry_count >= max_retries → GAVE_UP.
+
+Pure module — no ROS2 dependency.
+"""
+
+from __future__ import annotations
+
+from dataclasses import dataclass
+from enum import Enum
+
+
+# ── States ────────────────────────────────────────────────────────────────────
+
+class RecoveryState(Enum):
+    IDLE      = "IDLE"
+    REVERSING = "REVERSING"
+    TURNING   = "TURNING"
+    RETRYING  = "RETRYING"
+    GAVE_UP   = "GAVE_UP"
+
+
+# ── I/O ───────────────────────────────────────────────────────────────────────
+
+@dataclass
+class RecoveryInputs:
+    trigger:        bool  = False   # True to start (or restart) recovery
+    threat_cleared: bool  = False   # True when all threats are CLEAR
+    dt:             float = 0.02    # time step (s)
+
+
+@dataclass
+class RecoveryOutputs:
+    state:         RecoveryState = RecoveryState.IDLE
+    cmd_linear:    float         = 0.0   # m/s
+    cmd_angular:   float         = 0.0   # rad/s
+    progress:      float         = 0.0   # [0, 1] completion of current phase
+    retry_count:   int           = 0
+    gave_up:       bool          = False
+    state_changed: bool          = False
+
+
+# ── RecoverySequencer ────────────────────────────────────────────────────────
+
+class RecoverySequencer:
+    """
+    Tick-based FSM for executing reverse + turn recovery sequences.
+
+    Parameters
+    ----------
+    reverse_speed_ms  : backward speed during REVERSING (m/s; stored as negative)
+    reverse_distance_m: total reverse travel before turning (m)
+    angular_speed_rads: yaw rate during TURNING (rad/s; positive = left)
+    turn_angle_rad    : total turn angle before RETRYING (rad; default π/2)
+    retry_timeout_s   : max RETRYING time per attempt before next reverse cycle
+    clear_hold_s      : consecutive clear time needed to declare success
+    max_retries       : maximum reverse+turn attempts before GAVE_UP
+    """
+
+    def __init__(
+        self,
+        reverse_speed_ms:   float = -0.15,
+        reverse_distance_m: float = 0.30,
+        angular_speed_rads: float = 0.60,
+        turn_angle_rad:     float = 1.5708,    # π/2
+        retry_timeout_s:    float = 3.0,
+        clear_hold_s:       float = 0.5,
+        max_retries:        int   = 3,
+    ):
+        self._rev_speed  = min(0.0, float(reverse_speed_ms))   # ensure negative
+        self._rev_dist   = max(0.01, float(reverse_distance_m))
+        self._ang_speed  = abs(float(angular_speed_rads))
+        self._turn_angle = max(0.01, float(turn_angle_rad))
+        self._retry_tout = max(0.1, float(retry_timeout_s))
+        self._clear_hold = max(0.0, float(clear_hold_s))
+        self._max_retry  = max(1, int(max_retries))
+
+        self._state        = RecoveryState.IDLE
+        self._rev_done     = 0.0   # distance reversed so far
+        self._turn_done    = 0.0   # angle turned so far
+        self._retry_count  = 0
+        self._retry_timer  = 0.0   # time spent in RETRYING
+        self._clear_timer  = 0.0   # consecutive clear time in RETRYING
+
+    # ── Public API ────────────────────────────────────────────────────────────
+
+    @property
+    def state(self) -> RecoveryState:
+        return self._state
+
+    @property
+    def retry_count(self) -> int:
+        return self._retry_count
+
+    def reset(self) -> None:
+        """Return to IDLE and clear all counters."""
+        self._state       = RecoveryState.IDLE
+        self._rev_done    = 0.0
+        self._turn_done   = 0.0
+        self._retry_count = 0
+        self._retry_timer = 0.0
+        self._clear_timer = 0.0
+
+    def tick(self, inputs: RecoveryInputs) -> RecoveryOutputs:
+        prev = self._state
+        out  = self._step(inputs)
+        out.state         = self._state
+        out.retry_count   = self._retry_count
+        out.state_changed = (self._state != prev)
+        if out.state_changed:
+            self._on_enter(self._state)
+        return out
+
+    # ── Internal step ─────────────────────────────────────────────────────────
+
+    def _step(self, inp: RecoveryInputs) -> RecoveryOutputs:
+        out = RecoveryOutputs(state=self._state)
+        dt  = max(0.0, inp.dt)
+
+        # ── IDLE ──────────────────────────────────────────────────────────────
+        if self._state == RecoveryState.IDLE:
+            if inp.trigger:
+                self._state = RecoveryState.REVERSING
+
+        # ── REVERSING ─────────────────────────────────────────────────────────
+        elif self._state == RecoveryState.REVERSING:
+            step = abs(self._rev_speed) * dt
+            self._rev_done += step
+            out.cmd_linear = self._rev_speed
+            out.progress   = min(1.0, self._rev_done / self._rev_dist)
+            if self._rev_done >= self._rev_dist:
+                self._state = RecoveryState.TURNING
+
+        # ── TURNING ───────────────────────────────────────────────────────────
+        elif self._state == RecoveryState.TURNING:
+            step = self._ang_speed * dt
+            self._turn_done += step
+            out.cmd_angular = self._ang_speed
+            out.progress    = min(1.0, self._turn_done / self._turn_angle)
+            if self._turn_done >= self._turn_angle:
+                self._retry_count += 1
+                self._state = RecoveryState.RETRYING
+
+        # ── RETRYING ──────────────────────────────────────────────────────────
+        elif self._state == RecoveryState.RETRYING:
+            self._retry_timer += dt
+
+            if inp.threat_cleared:
+                self._clear_timer += dt
+                if self._clear_timer >= self._clear_hold:
+                    # Success — threat has cleared
+                    self._state = RecoveryState.IDLE
+                    return out
+            else:
+                self._clear_timer = 0.0
+
+            if self._retry_timer >= self._retry_tout:
+                if self._retry_count >= self._max_retry:
+                    self._state  = RecoveryState.GAVE_UP
+                    out.gave_up  = True
+                else:
+                    self._state = RecoveryState.REVERSING
+
+        # ── GAVE_UP ───────────────────────────────────────────────────────────
+        elif self._state == RecoveryState.GAVE_UP:
+            # Stay in GAVE_UP until external reset()
+            out.gave_up = True
+
+        return out
+
+    # ── Entry actions ─────────────────────────────────────────────────────────
+
+    def _on_enter(self, state: RecoveryState) -> None:
+        if state == RecoveryState.REVERSING:
+            self._rev_done  = 0.0
+        elif state == RecoveryState.TURNING:
+            self._turn_done = 0.0
+        elif state == RecoveryState.RETRYING:
+            self._retry_timer = 0.0
+            self._clear_timer = 0.0

jetson/ros2_ws/src/saltybot_emergency/saltybot_emergency/threat_detector.py

@@ -0,0 +1,354 @@
+"""
+threat_detector.py — Multi-source threat detection for emergency behavior (Issue #169).
+
+Detectors
+─────────
+  ObstacleDetector  : Forward-sector minimum range < stop thresholds at speed.
+                      Inputs: min_range_m (pre-filtered from LaserScan forward
+                      sector), cmd_speed_ms.
+
+  FallDetector      : Extreme pitch/roll from IMU, or depth floor-drop ahead.
+                      Inputs: pitch_rad, roll_rad, floor_drop_m (depth-derived;
+                      0.0 if depth unavailable).
+
+  StuckDetector     : Commanded speed vs actual speed mismatch sustained for
+                      stuck_timeout_s.  Tracks elapsed time with dt argument.
+
+  BumpDetector      : IMU acceleration jerk (|Δ|a||/dt) above threshold.
+                      MAJOR at jerk_threshold_ms3, CRITICAL at
+                      critical_jerk_threshold_ms3.
+
+ThreatLevel
+───────────
+  CLEAR    : no threat; normal operation
+  MINOR    : advisory; log/beep only
+  MAJOR    : stop and execute recovery
+  CRITICAL : full shutdown + MQTT escalation
+
+Pure module — no ROS2 dependency.
+"""
+
+from __future__ import annotations
+
+import math
+import time
+from dataclasses import dataclass
+from enum import Enum
+from typing import Optional
+
+
+# ── Enumerations ──────────────────────────────────────────────────────────────
+
+class ThreatLevel(Enum):
+    CLEAR    = 0
+    MINOR    = 1
+    MAJOR    = 2
+    CRITICAL = 3
+
+
+class ThreatType(Enum):
+    NONE                = "NONE"
+    OBSTACLE_PROXIMITY  = "OBSTACLE_PROXIMITY"
+    FALL_RISK           = "FALL_RISK"
+    WHEEL_STUCK         = "WHEEL_STUCK"
+    BUMP                = "BUMP"
+
+
+# ── ThreatEvent ───────────────────────────────────────────────────────────────
+
+@dataclass
+class ThreatEvent:
+    """Snapshot of a single detected threat."""
+    threat_type:  ThreatType  = ThreatType.NONE
+    level:        ThreatLevel = ThreatLevel.CLEAR
+    value:        float       = 0.0    # triggering metric
+    detail:       str         = ""
+    timestamp_s:  float       = 0.0
+
+    @staticmethod
+    def clear(timestamp_s: float = 0.0) -> "ThreatEvent":
+        return ThreatEvent(timestamp_s=timestamp_s)
+
+
+_CLEAR = ThreatEvent()
+
+
+# ── ObstacleDetector ─────────────────────────────────────────────────────────
+
+class ObstacleDetector:
+    """
+    Obstacle proximity threat from forward-sector minimum range.
+
+    Parameters
+    ----------
+    stop_distance_m      : range below which MAJOR is raised (default 0.30 m)
+    critical_distance_m  : range below which CRITICAL is raised (default 0.10 m)
+    min_speed_ms         : only active above this commanded speed (default 0.05 m/s)
+    """
+
+    def __init__(
+        self,
+        stop_distance_m:     float = 0.30,
+        critical_distance_m: float = 0.10,
+        min_speed_ms:        float = 0.05,
+    ):
+        self._stop     = max(1e-3, stop_distance_m)
+        self._critical = max(1e-3, min(self._stop, critical_distance_m))
+        self._min_spd  = abs(min_speed_ms)
+
+    def update(
+        self,
+        min_range_m:   float,
+        cmd_speed_ms:  float,
+        timestamp_s:   float = 0.0,
+    ) -> ThreatEvent:
+        """
+        Parameters
+        ----------
+        min_range_m   : minimum obstacle range in forward sector (m)
+        cmd_speed_ms  : signed commanded forward speed (m/s)
+        """
+        if abs(cmd_speed_ms) < self._min_spd:
+            return ThreatEvent.clear(timestamp_s)
+
+        if min_range_m <= self._critical:
+            return ThreatEvent(
+                threat_type=ThreatType.OBSTACLE_PROXIMITY,
+                level=ThreatLevel.CRITICAL,
+                value=min_range_m,
+                detail=f"Obstacle {min_range_m:.2f} m (critical zone)",
+                timestamp_s=timestamp_s,
+            )
+        if min_range_m <= self._stop:
+            return ThreatEvent(
+                threat_type=ThreatType.OBSTACLE_PROXIMITY,
+                level=ThreatLevel.MAJOR,
+                value=min_range_m,
+                detail=f"Obstacle {min_range_m:.2f} m ahead",
+                timestamp_s=timestamp_s,
+            )
+        return ThreatEvent.clear(timestamp_s)
+
+
+# ── FallDetector ──────────────────────────────────────────────────────────────
+
+class FallDetector:
+    """
+    Fall / tipping risk from IMU pitch/roll and optional depth floor-drop.
+
+    Parameters
+    ----------
+    minor_tilt_rad    : |pitch| or |roll| above which MINOR fires (default 0.20 rad)
+    major_tilt_rad    : above which MAJOR fires (default 0.35 rad)
+    critical_tilt_rad : above which CRITICAL fires (default 0.52 rad ≈ 30°)
+    floor_drop_m      : depth discontinuity (m) triggering MAJOR (default 0.15 m)
+    """
+
+    def __init__(
+        self,
+        minor_tilt_rad:    float = 0.20,
+        major_tilt_rad:    float = 0.35,
+        critical_tilt_rad: float = 0.52,
+        floor_drop_m:      float = 0.15,
+    ):
+        self._minor    = float(minor_tilt_rad)
+        self._major    = float(major_tilt_rad)
+        self._critical = float(critical_tilt_rad)
+        self._drop     = float(floor_drop_m)
+
+    def update(
+        self,
+        pitch_rad:    float,
+        roll_rad:     float,
+        floor_drop_m: float = 0.0,
+        timestamp_s:  float = 0.0,
+    ) -> ThreatEvent:
+        """
+        Parameters
+        ----------
+        pitch_rad    : forward tilt (rad); +ve = nose up
+        roll_rad     : lateral tilt (rad); +ve = left side up
+        floor_drop_m : depth discontinuity ahead of robot (m); 0 = not measured
+        """
+        tilt = max(abs(pitch_rad), abs(roll_rad))
+
+        if tilt >= self._critical:
+            return ThreatEvent(
+                threat_type=ThreatType.FALL_RISK,
+                level=ThreatLevel.CRITICAL,
+                value=tilt,
+                detail=f"Critical tilt {math.degrees(tilt):.1f}°",
+                timestamp_s=timestamp_s,
+            )
+        if tilt >= self._major or floor_drop_m >= self._drop:
+            value   = max(tilt, floor_drop_m)
+            detail  = (
+                f"Floor drop {floor_drop_m:.2f} m" if floor_drop_m >= self._drop
+                else f"Major tilt {math.degrees(tilt):.1f}°"
+            )
+            return ThreatEvent(
+                threat_type=ThreatType.FALL_RISK,
+                level=ThreatLevel.MAJOR,
+                value=value,
+                detail=detail,
+                timestamp_s=timestamp_s,
+            )
+        if tilt >= self._minor:
+            return ThreatEvent(
+                threat_type=ThreatType.FALL_RISK,
+                level=ThreatLevel.MINOR,
+                value=tilt,
+                detail=f"Tilt advisory {math.degrees(tilt):.1f}°",
+                timestamp_s=timestamp_s,
+            )
+        return ThreatEvent.clear(timestamp_s)
+
+
+# ── StuckDetector ─────────────────────────────────────────────────────────────
+
+class StuckDetector:
+    """
+    Wheel stall / stuck detection from cmd_vel vs odometry mismatch.
+
+    Accumulates stuck time while |cmd| > min_cmd_ms AND |actual| < moving_ms.
+    Resets when motion resumes or commanded speed drops below min_cmd_ms.
+
+    Parameters
+    ----------
+    stuck_timeout_s  : accumulated stuck time before MAJOR fires (default 3.0 s)
+    min_cmd_ms       : minimum commanded speed to consider stalling (0.05 m/s)
+    moving_threshold_ms : actual speed above which robot is considered moving
+    """
+
+    def __init__(
+        self,
+        stuck_timeout_s:     float = 3.0,
+        min_cmd_ms:          float = 0.05,
+        moving_threshold_ms: float = 0.05,
+    ):
+        self._timeout   = max(0.1, stuck_timeout_s)
+        self._min_cmd   = abs(min_cmd_ms)
+        self._moving    = abs(moving_threshold_ms)
+        self._stuck_time: float = 0.0
+
+    @property
+    def stuck_time(self) -> float:
+        """Accumulated stuck duration (s)."""
+        return self._stuck_time
+
+    def update(
+        self,
+        cmd_speed_ms:    float,
+        actual_speed_ms: float,
+        dt:              float,
+        timestamp_s:     float = 0.0,
+    ) -> ThreatEvent:
+        """
+        Parameters
+        ----------
+        cmd_speed_ms    : commanded forward speed (m/s)
+        actual_speed_ms : measured forward speed from odometry (m/s)
+        dt              : elapsed time since last call (s)
+        """
+        commanding = abs(cmd_speed_ms) >= self._min_cmd
+        moving     = abs(actual_speed_ms) >= self._moving
+
+        if not commanding or moving:
+            self._stuck_time = 0.0
+            return ThreatEvent.clear(timestamp_s)
+
+        self._stuck_time += max(0.0, dt)
+
+        if self._stuck_time >= self._timeout:
+            return ThreatEvent(
+                threat_type=ThreatType.WHEEL_STUCK,
+                level=ThreatLevel.MAJOR,
+                value=self._stuck_time,
+                detail=f"Wheels stuck for {self._stuck_time:.1f} s",
+                timestamp_s=timestamp_s,
+            )
+        return ThreatEvent.clear(timestamp_s)
+
+    def reset(self) -> None:
+        self._stuck_time = 0.0
+
+
+# ── BumpDetector ─────────────────────────────────────────────────────────────
+
+class BumpDetector:
+    """
+    Collision / bump detection via IMU acceleration jerk.
+
+    Jerk = |Δ|a|| / dt   where |a| = sqrt(ax²+ay²+az²) − g  (gravity removed)
+
+    Parameters
+    ----------
+    jerk_threshold_ms3          : MAJOR at jerk above this (m/s³, default 8.0)
+    critical_jerk_threshold_ms3 : CRITICAL at jerk above this (m/s³, default 25.0)
+    gravity_ms2                 : gravity magnitude to subtract (default 9.81)
+    """
+
+    def __init__(
+        self,
+        jerk_threshold_ms3:          float = 8.0,
+        critical_jerk_threshold_ms3: float = 25.0,
+        gravity_ms2:                 float = 9.81,
+    ):
+        self._jerk_major    = float(jerk_threshold_ms3)
+        self._jerk_critical = float(critical_jerk_threshold_ms3)
+        self._gravity       = float(gravity_ms2)
+        self._prev_dyn_mag: Optional[float] = None   # previous |a_dynamic|
+
+    def update(
+        self,
+        ax: float,
+        ay: float,
+        az: float,
+        dt: float,
+        timestamp_s: float = 0.0,
+    ) -> ThreatEvent:
+        """
+        Parameters
+        ----------
+        ax, ay, az : IMU linear acceleration (m/s²)
+        dt         : elapsed time since last call (s)
+        """
+        raw_mag = math.sqrt(ax * ax + ay * ay + az * az)
+        dyn_mag = abs(raw_mag - self._gravity)   # remove gravity component
+
+        if self._prev_dyn_mag is None or dt <= 0.0:
+            self._prev_dyn_mag = dyn_mag
+            return ThreatEvent.clear(timestamp_s)
+
+        jerk = abs(dyn_mag - self._prev_dyn_mag) / dt
+        self._prev_dyn_mag = dyn_mag
+
+        if jerk >= self._jerk_critical:
+            return ThreatEvent(
+                threat_type=ThreatType.BUMP,
+                level=ThreatLevel.CRITICAL,
+                value=jerk,
+                detail=f"Critical impact: jerk {jerk:.1f} m/s³",
+                timestamp_s=timestamp_s,
+            )
+        if jerk >= self._jerk_major:
+            return ThreatEvent(
+                threat_type=ThreatType.BUMP,
+                level=ThreatLevel.MAJOR,
+                value=jerk,
+                detail=f"Bump detected: jerk {jerk:.1f} m/s³",
+                timestamp_s=timestamp_s,
+            )
+        return ThreatEvent.clear(timestamp_s)
+
+    def reset(self) -> None:
+        self._prev_dyn_mag = None
+
+
+# ── Utility: pick highest-severity threat ─────────────────────────────────────
+
+def highest_threat(events: list[ThreatEvent]) -> ThreatEvent:
+    """Return the ThreatEvent with the highest ThreatLevel from a list."""
+    if not events:
+        return _CLEAR
+    return max(events, key=lambda e: e.level.value)

jetson/ros2_ws/src/saltybot_emergency/setup.cfg

@@ -0,0 +1,5 @@
+[develop]
+script_dir=$base/lib/saltybot_emergency
+
+[install]
+install_scripts=$base/lib/saltybot_emergency

jetson/ros2_ws/src/saltybot_emergency/setup.py

@@ -0,0 +1,32 @@
+from setuptools import setup, find_packages
+import os
+from glob import glob
+
+package_name = "saltybot_emergency"
+
+setup(
+    name=package_name,
+    version="0.1.0",
+    packages=find_packages(exclude=["test"]),
+    data_files=[
+        ("share/ament_index/resource_index/packages",
+         [f"resource/{package_name}"]),
+        (f"share/{package_name}", ["package.xml"]),
+        (os.path.join("share", package_name, "config"),
+         glob("config/*.yaml")),
+        (os.path.join("share", package_name, "launch"),
+         glob("launch/*.py")),
+    ],
+    install_requires=["setuptools"],
+    zip_safe=True,
+    maintainer="sl-controls",
+    maintainer_email="sl-controls@saltylab.local",
+    description="Emergency behavior system — collision avoidance, fall prevention, stuck detection, recovery",
+    license="MIT",
+    tests_require=["pytest"],
+    entry_points={
+        "console_scripts": [
+            f"emergency_node = {package_name}.emergency_node:main",
+        ],
+    },
+)

jetson/ros2_ws/src/saltybot_emergency/test/test_emergency.py

@@ -0,0 +1,560 @@
+"""
+test_emergency.py — Unit tests for Issue #169 emergency behavior modules.
+
+Covers:
+  ObstacleDetector   — proximity thresholds, speed gate
+  FallDetector       — tilt levels, floor drop
+  StuckDetector      — timeout accumulation, reset on motion
+  BumpDetector       — jerk thresholds, first-call safety
+  AlertManager       — severity mapping, escalation, suppression
+  RecoverySequencer  — full sequence, retry, give-up
+  EmergencyFSM       — all state transitions and guard conditions
+"""
+
+import math
+import sys
+import os
+
+sys.path.insert(0, os.path.join(os.path.dirname(__file__), ".."))
+
+import pytest
+
+from saltybot_emergency.threat_detector import (
+    BumpDetector,
+    FallDetector,
+    ObstacleDetector,
+    StuckDetector,
+    ThreatEvent,
+    ThreatLevel,
+    ThreatType,
+    highest_threat,
+)
+from saltybot_emergency.alert_manager import Alert, AlertLevel, AlertManager
+from saltybot_emergency.recovery_sequencer import (
+    RecoveryInputs,
+    RecoverySequencer,
+    RecoveryState,
+)
+from saltybot_emergency.emergency_fsm import (
+    EmergencyFSM,
+    EmergencyInputs,
+    EmergencyState,
+)
+
+
+# ── Helpers ───────────────────────────────────────────────────────────────────
+
+def _obs(**kw):
+    d = dict(stop_distance_m=0.30, critical_distance_m=0.10, min_speed_ms=0.05)
+    d.update(kw)
+    return ObstacleDetector(**d)
+
+def _fall(**kw):
+    d = dict(minor_tilt_rad=0.20, major_tilt_rad=0.35,
+             critical_tilt_rad=0.52, floor_drop_m=0.15)
+    d.update(kw)
+    return FallDetector(**d)
+
+def _stuck(**kw):
+    d = dict(stuck_timeout_s=3.0, min_cmd_ms=0.05, moving_threshold_ms=0.05)
+    d.update(kw)
+    return StuckDetector(**d)
+
+def _bump(**kw):
+    d = dict(jerk_threshold_ms3=8.0, critical_jerk_threshold_ms3=25.0)
+    d.update(kw)
+    return BumpDetector(**d)
+
+def _alert_mgr(**kw):
+    d = dict(major_count_threshold=3, escalation_window_s=10.0, suppression_s=1.0)
+    d.update(kw)
+    return AlertManager(**d)
+
+def _seq(**kw):
+    d = dict(
+        reverse_speed_ms=-0.15, reverse_distance_m=0.30,
+        angular_speed_rads=0.60, turn_angle_rad=1.5708,
+        retry_timeout_s=3.0, clear_hold_s=0.5, max_retries=3,
+    )
+    d.update(kw)
+    return RecoverySequencer(**d)
+
+def _fsm(**kw):
+    d = dict(
+        stopped_ms=0.03, major_count_threshold=3, escalation_window_s=10.0,
+        suppression_s=0.0,  # disable suppression for cleaner tests
+        reverse_speed_ms=-0.15, reverse_distance_m=0.30,
+        angular_speed_rads=0.60, turn_angle_rad=1.5708,
+        retry_timeout_s=3.0, clear_hold_s=0.5, max_retries=3,
+    )
+    d.update(kw)
+    return EmergencyFSM(**d)
+
+def _major_threat(threat_type=ThreatType.OBSTACLE_PROXIMITY, ts=0.0):
+    return ThreatEvent(threat_type=threat_type, level=ThreatLevel.MAJOR,
+                       value=1.0, detail="test", timestamp_s=ts)
+
+def _critical_threat(ts=0.0):
+    return ThreatEvent(threat_type=ThreatType.OBSTACLE_PROXIMITY,
+                       level=ThreatLevel.CRITICAL, value=0.05,
+                       detail="critical test", timestamp_s=ts)
+
+def _minor_threat(ts=0.0):
+    return ThreatEvent(threat_type=ThreatType.FALL_RISK, level=ThreatLevel.MINOR,
+                       value=0.21, detail="tilt", timestamp_s=ts)
+
+def _clear_threat():
+    return ThreatEvent.clear()
+
+def _inp(threat=None, speed=0.0, ack=False):
+    return EmergencyInputs(
+        threat=threat or _clear_threat(),
+        robot_speed_ms=speed,
+        acknowledge=ack,
+    )
+
+
+# ══════════════════════════════════════════════════════════════════════════════
+# ObstacleDetector
+# ══════════════════════════════════════════════════════════════════════════════
+
+class TestObstacleDetector:
+
+    def test_clear_when_far(self):
+        ev = _obs().update(0.5, 0.3)
+        assert ev.level == ThreatLevel.CLEAR
+
+    def test_major_within_stop_distance(self):
+        ev = _obs(stop_distance_m=0.30).update(0.25, 0.3)
+        assert ev.level == ThreatLevel.MAJOR
+        assert ev.threat_type == ThreatType.OBSTACLE_PROXIMITY
+
+    def test_critical_within_critical_distance(self):
+        ev = _obs(critical_distance_m=0.10).update(0.05, 0.3)
+        assert ev.level == ThreatLevel.CRITICAL
+
+    def test_clear_when_stopped(self):
+        """Obstacle detection suppressed when not moving."""
+        ev = _obs(min_speed_ms=0.05).update(0.05, 0.01)
+        assert ev.level == ThreatLevel.CLEAR
+
+    def test_active_at_min_speed(self):
+        ev = _obs(min_speed_ms=0.05).update(0.20, 0.06)
+        assert ev.level == ThreatLevel.MAJOR
+
+    def test_value_is_distance(self):
+        ev = _obs().update(0.20, 0.3)
+        assert ev.value == pytest.approx(0.20, abs=1e-9)
+
+
+# ══════════════════════════════════════════════════════════════════════════════
+# FallDetector
+# ══════════════════════════════════════════════════════════════════════════════
+
+class TestFallDetector:
+
+    def test_clear_on_flat(self):
+        ev = _fall().update(0.0, 0.0)
+        assert ev.level == ThreatLevel.CLEAR
+
+    def test_minor_moderate_tilt(self):
+        ev = _fall(minor_tilt_rad=0.20, major_tilt_rad=0.35).update(0.25, 0.0)
+        assert ev.level == ThreatLevel.MINOR
+
+    def test_major_high_tilt(self):
+        ev = _fall(major_tilt_rad=0.35, critical_tilt_rad=0.52).update(0.40, 0.0)
+        assert ev.level == ThreatLevel.MAJOR
+
+    def test_critical_extreme_tilt(self):
+        ev = _fall(critical_tilt_rad=0.52).update(0.60, 0.0)
+        assert ev.level == ThreatLevel.CRITICAL
+
+    def test_major_on_floor_drop(self):
+        ev = _fall(floor_drop_m=0.15).update(0.0, 0.0, floor_drop_m=0.20)
+        assert ev.level == ThreatLevel.MAJOR
+        assert "drop" in ev.detail.lower()
+
+    def test_roll_triggers_same_as_pitch(self):
+        """Roll beyond minor threshold also fires."""
+        ev = _fall(minor_tilt_rad=0.20).update(0.0, 0.25)
+        assert ev.level == ThreatLevel.MINOR
+
+
+# ══════════════════════════════════════════════════════════════════════════════
+# StuckDetector
+# ══════════════════════════════════════════════════════════════════════════════
+
+class TestStuckDetector:
+
+    def test_clear_when_not_commanded(self):
+        s = _stuck(stuck_timeout_s=1.0, min_cmd_ms=0.05)
+        ev = s.update(0.01, 0.0, dt=1.0)  # cmd below threshold
+        assert ev.level == ThreatLevel.CLEAR
+
+    def test_clear_when_moving(self):
+        s = _stuck(stuck_timeout_s=1.0)
+        ev = s.update(0.2, 0.2, dt=1.0)   # actually moving
+        assert ev.level == ThreatLevel.CLEAR
+
+    def test_major_after_timeout(self):
+        s = _stuck(stuck_timeout_s=3.0, min_cmd_ms=0.05, moving_threshold_ms=0.05)
+        for _ in range(6):
+            ev = s.update(0.2, 0.0, dt=0.5)   # cmd=0.2, actual=0 → stuck
+        assert ev.level == ThreatLevel.MAJOR
+
+    def test_no_major_before_timeout(self):
+        s = _stuck(stuck_timeout_s=3.0)
+        ev = s.update(0.2, 0.0, dt=1.0)   # only 1s — not yet
+        assert ev.level == ThreatLevel.CLEAR
+
+    def test_reset_on_motion_resume(self):
+        s = _stuck(stuck_timeout_s=1.0)
+        s.update(0.2, 0.0, dt=0.8)        # accumulate stuck time
+        s.update(0.2, 0.3, dt=0.1)        # motion resumes → reset
+        ev = s.update(0.2, 0.0, dt=0.3)   # only 0.3s since reset → still clear
+        assert ev.level == ThreatLevel.CLEAR
+
+    def test_stuck_time_property(self):
+        s = _stuck(stuck_timeout_s=5.0)
+        s.update(0.2, 0.0, dt=1.0)
+        s.update(0.2, 0.0, dt=1.0)
+        assert s.stuck_time == pytest.approx(2.0, abs=1e-6)
+
+
+# ══════════════════════════════════════════════════════════════════════════════
+# BumpDetector
+# ══════════════════════════════════════════════════════════════════════════════
+
+class TestBumpDetector:
+
+    def test_clear_on_first_call(self):
+        """No jerk on first sample (no previous value)."""
+        ev = _bump().update(0.0, 0.0, 9.81, dt=0.05)
+        assert ev.level == ThreatLevel.CLEAR
+
+    def test_major_on_jerk(self):
+        b = _bump(jerk_threshold_ms3=5.0, critical_jerk_threshold_ms3=20.0)
+        b.update(0.0, 0.0, 9.81, dt=0.05)    # seed → dyn_mag = 0
+        # ax=4.5: raw≈10.79, dyn≈0.98, jerk≈9.8 m/s³ → MAJOR (5.0 < 9.8 < 20.0)
+        ev = b.update(4.5, 0.0, 9.81, dt=0.1)
+        assert ev.level == ThreatLevel.MAJOR
+
+    def test_critical_on_severe_jerk(self):
+        b = _bump(jerk_threshold_ms3=5.0, critical_jerk_threshold_ms3=20.0)
+        b.update(0.0, 0.0, 9.81, dt=0.05)
+        # Very large spike
+        ev = b.update(50.0, 0.0, 9.81, dt=0.1)
+        assert ev.level == ThreatLevel.CRITICAL
+
+    def test_clear_on_gentle_acceleration(self):
+        b = _bump(jerk_threshold_ms3=8.0)
+        b.update(0.0, 0.0, 9.81, dt=0.05)
+        ev = b.update(0.1, 0.0, 9.81, dt=0.05)   # tiny change
+        assert ev.level == ThreatLevel.CLEAR
+
+
+# ══════════════════════════════════════════════════════════════════════════════
+# highest_threat
+# ══════════════════════════════════════════════════════════════════════════════
+
+class TestHighestThreat:
+
+    def test_empty_returns_clear(self):
+        assert highest_threat([]).level == ThreatLevel.CLEAR
+
+    def test_picks_highest(self):
+        a = ThreatEvent(level=ThreatLevel.MINOR)
+        b = ThreatEvent(level=ThreatLevel.CRITICAL)
+        c = ThreatEvent(level=ThreatLevel.MAJOR)
+        assert highest_threat([a, b, c]).level == ThreatLevel.CRITICAL
+
+    def test_single_item(self):
+        ev = ThreatEvent(level=ThreatLevel.MAJOR)
+        assert highest_threat([ev]) is ev
+
+
+# ══════════════════════════════════════════════════════════════════════════════
+# AlertManager
+# ══════════════════════════════════════════════════════════════════════════════
+
+class TestAlertManager:
+
+    def test_clear_returns_none(self):
+        am = _alert_mgr()
+        assert am.update(_clear_threat()) is None
+
+    def test_minor_threat_gives_minor_alert(self):
+        am = _alert_mgr(suppression_s=0.0)
+        alert = am.update(_minor_threat(ts=0.0))
+        assert alert is not None
+        assert alert.level == AlertLevel.MINOR
+
+    def test_major_threat_gives_major_alert(self):
+        am = _alert_mgr(suppression_s=0.0)
+        alert = am.update(_major_threat(ts=0.0))
+        assert alert is not None
+        assert alert.level == AlertLevel.MAJOR
+
+    def test_critical_threat_gives_critical_alert(self):
+        am = _alert_mgr(suppression_s=0.0)
+        alert = am.update(_critical_threat(ts=0.0))
+        assert alert is not None
+        assert alert.level == AlertLevel.CRITICAL
+
+    def test_suppression_blocks_duplicate(self):
+        am = _alert_mgr(suppression_s=5.0)
+        am.update(_major_threat(ts=0.0))
+        alert = am.update(_major_threat(ts=1.0))   # within 5s window
+        assert alert is None
+
+    def test_suppression_expires(self):
+        am = _alert_mgr(suppression_s=2.0)
+        am.update(_major_threat(ts=0.0))
+        alert = am.update(_major_threat(ts=3.0))   # after 2s window
+        assert alert is not None
+
+    def test_escalation_major_to_critical(self):
+        """After major_count_threshold major alerts, next one becomes CRITICAL."""
+        am = _alert_mgr(major_count_threshold=3, escalation_window_s=60.0,
+                        suppression_s=0.0)
+        for i in range(3):
+            am.update(_major_threat(ts=float(i)))
+        # 4th should be escalated
+        alert = am.update(_major_threat(ts=4.0))
+        assert alert is not None
+        assert alert.level == AlertLevel.CRITICAL
+
+    def test_escalation_resets_after_window(self):
+        """Major alerts outside the window don't contribute to escalation."""
+        am = _alert_mgr(major_count_threshold=3, escalation_window_s=5.0,
+                        suppression_s=0.0)
+        am.update(_major_threat(ts=0.0))
+        am.update(_major_threat(ts=1.0))
+        am.update(_major_threat(ts=2.0))
+        # All 3 are old; new one at t=10 is outside window
+        alert = am.update(_major_threat(ts=10.0))
+        assert alert is not None
+        assert alert.level == AlertLevel.MAJOR   # not escalated
+
+    def test_reset_clears_escalation_state(self):
+        am = _alert_mgr(major_count_threshold=2, suppression_s=0.0)
+        am.update(_major_threat(ts=0.0))
+        am.update(_major_threat(ts=1.0))  # now at threshold
+        am.reset()
+        alert = am.update(_major_threat(ts=2.0))
+        assert alert.level == AlertLevel.MAJOR  # back to major after reset
+
+
+# ══════════════════════════════════════════════════════════════════════════════
+# RecoverySequencer
+# ══════════════════════════════════════════════════════════════════════════════
+
+class TestRecoverySequencer:
+
+    def _trigger(self, seq):
+        return seq.tick(RecoveryInputs(trigger=True, dt=0.02))
+
+    def test_idle_on_init(self):
+        seq = _seq()
+        assert seq.state == RecoveryState.IDLE
+
+    def test_trigger_starts_reversing(self):
+        seq = _seq()
+        out = self._trigger(seq)
+        assert seq.state == RecoveryState.REVERSING
+
+    def test_reversing_backward_velocity(self):
+        seq = _seq(reverse_speed_ms=-0.15)
+        self._trigger(seq)
+        out = seq.tick(RecoveryInputs(dt=0.02))
+        assert out.cmd_linear < 0.0
+
+    def test_reversing_completes_to_turning(self):
+        seq = _seq(reverse_speed_ms=-1.0, reverse_distance_m=0.5)
+        self._trigger(seq)
+        for _ in range(30):
+            out = seq.tick(RecoveryInputs(dt=0.02))
+        assert seq.state == RecoveryState.TURNING
+
+    def test_turning_positive_angular(self):
+        seq = _seq(reverse_speed_ms=-1.0, reverse_distance_m=0.1,
+                   angular_speed_rads=1.0)
+        self._trigger(seq)
+        # Skip through reversing quickly
+        for _ in range(20):
+            seq.tick(RecoveryInputs(dt=0.02))
+        if seq.state == RecoveryState.TURNING:
+            out = seq.tick(RecoveryInputs(dt=0.02))
+            assert out.cmd_angular > 0.0
+
+    def test_retrying_increments_count(self):
+        seq = _seq(reverse_speed_ms=-1.0, reverse_distance_m=0.05,
+                   angular_speed_rads=10.0, turn_angle_rad=0.1)
+        self._trigger(seq)
+        for _ in range(100):
+            seq.tick(RecoveryInputs(dt=0.02))
+        assert seq.state == RecoveryState.RETRYING
+        assert seq.retry_count == 1
+
+    def test_threat_cleared_returns_idle(self):
+        seq = _seq(reverse_speed_ms=-1.0, reverse_distance_m=0.05,
+                   angular_speed_rads=10.0, turn_angle_rad=0.1,
+                   clear_hold_s=0.1)
+        self._trigger(seq)
+        # Fast-forward to RETRYING
+        for _ in range(100):
+            seq.tick(RecoveryInputs(dt=0.02))
+        assert seq.state == RecoveryState.RETRYING
+        # Feed cleared ticks until clear_hold met
+        for _ in range(20):
+            seq.tick(RecoveryInputs(threat_cleared=True, dt=0.02))
+        assert seq.state == RecoveryState.IDLE
+
+    def test_max_retries_gives_up(self):
+        seq = _seq(reverse_speed_ms=-1.0, reverse_distance_m=0.05,
+                   angular_speed_rads=10.0, turn_angle_rad=0.1,
+                   retry_timeout_s=0.1, max_retries=2)
+        self._trigger(seq)
+        for _ in range(500):
+            out = seq.tick(RecoveryInputs(threat_cleared=False, dt=0.05))
+            if seq.state == RecoveryState.GAVE_UP:
+                break
+        assert seq.state == RecoveryState.GAVE_UP
+
+    def test_reset_returns_to_idle(self):
+        seq = _seq()
+        self._trigger(seq)
+        seq.reset()
+        assert seq.state == RecoveryState.IDLE
+        assert seq.retry_count == 0
+
+
+# ══════════════════════════════════════════════════════════════════════════════
+# EmergencyFSM
+# ══════════════════════════════════════════════════════════════════════════════
+
+class TestEmergencyFSMBasic:
+
+    def test_initial_state_nominal(self):
+        fsm = _fsm()
+        assert fsm.state == EmergencyState.NOMINAL
+
+    def test_nominal_stays_on_clear(self):
+        fsm = _fsm()
+        out = fsm.tick(_inp())
+        assert fsm.state == EmergencyState.NOMINAL
+        assert out.cmd_override is False
+
+    def test_minor_alert_no_override(self):
+        fsm = _fsm()
+        out = fsm.tick(_inp(_minor_threat(ts=0.0)))
+        assert fsm.state == EmergencyState.NOMINAL
+        assert out.cmd_override is False
+        assert out.alert is not None
+        assert out.alert.level == AlertLevel.MINOR
+
+    def test_major_threat_enters_stopping(self):
+        fsm = _fsm()
+        out = fsm.tick(_inp(_major_threat()))
+        assert fsm.state == EmergencyState.STOPPING
+        assert out.cmd_override is True
+
+    def test_critical_threat_enters_stopping_critical_pending(self):
+        fsm = _fsm()
+        fsm.tick(_inp(_critical_threat()))
+        assert fsm.state == EmergencyState.STOPPING
+        assert fsm._critical_pending is True
+
+
+class TestEmergencyFSMStopping:
+
+    def test_stopping_commands_zero(self):
+        fsm = _fsm()
+        fsm.tick(_inp(_major_threat()))
+        out = fsm.tick(_inp(_major_threat(), speed=0.5))
+        assert out.cmd_linear  == pytest.approx(0.0, abs=1e-9)
+        assert out.cmd_angular == pytest.approx(0.0, abs=1e-9)
+
+    def test_stopped_enters_recovering(self):
+        fsm = _fsm(stopped_ms=0.03)
+        fsm.tick(_inp(_major_threat()))
+        out = fsm.tick(_inp(_major_threat(), speed=0.01))  # below stopped_ms
+        assert fsm.state == EmergencyState.RECOVERING
+
+    def test_critical_pending_enters_escalated(self):
+        fsm = _fsm(stopped_ms=0.03)
+        fsm.tick(_inp(_critical_threat()))
+        fsm.tick(_inp(_critical_threat(), speed=0.01))  # stopped → ESCALATED
+        assert fsm.state == EmergencyState.ESCALATED
+
+
+class TestEmergencyFSMRecovering:
+
+    def _reach_recovering(self, fsm):
+        fsm.tick(_inp(_major_threat()))
+        fsm.tick(_inp(_major_threat(), speed=0.0))  # stopped → RECOVERING
+        assert fsm.state == EmergencyState.RECOVERING
+
+    def test_recovering_has_cmd_override(self):
+        fsm = _fsm()
+        self._reach_recovering(fsm)
+        out = fsm.tick(_inp(_clear_threat()))
+        assert out.cmd_override is True
+
+    def test_recovering_gave_up_escalates(self):
+        fsm = _fsm(max_retries=1, retry_timeout_s=0.05)
+        self._reach_recovering(fsm)
+        # Drive recovery to GAVE_UP by feeding many non-clearing ticks
+        for _ in range(500):
+            out = fsm.tick(_inp(_major_threat()))
+            if fsm.state == EmergencyState.ESCALATED:
+                break
+        assert fsm.state == EmergencyState.ESCALATED
+
+
+class TestEmergencyFSMEscalated:
+
+    def _reach_escalated(self, fsm):
+        fsm.tick(_inp(_critical_threat()))
+        fsm.tick(_inp(_critical_threat(), speed=0.0))
+        assert fsm.state == EmergencyState.ESCALATED
+
+    def test_escalated_emits_critical_alert_once(self):
+        fsm = _fsm()
+        self._reach_escalated(fsm)
+        out1 = fsm.tick(_inp())
+        out2 = fsm.tick(_inp())
+        assert out1.alert is not None
+        assert out1.alert.level == AlertLevel.CRITICAL
+        assert out2.alert is None   # suppressed after first emission
+
+    def test_escalated_e_stop_asserted(self):
+        fsm = _fsm()
+        self._reach_escalated(fsm)
+        out = fsm.tick(_inp())
+        assert out.e_stop is True
+
+    def test_escalated_stays_without_ack(self):
+        fsm = _fsm()
+        self._reach_escalated(fsm)
+        for _ in range(5):
+            fsm.tick(_inp())
+        assert fsm.state == EmergencyState.ESCALATED
+
+    def test_acknowledge_returns_to_nominal(self):
+        fsm = _fsm()
+        self._reach_escalated(fsm)
+        fsm.tick(_inp(ack=True))
+        assert fsm.state == EmergencyState.NOMINAL
+
+    def test_reset_returns_to_nominal(self):
+        fsm = _fsm()
+        self._reach_escalated(fsm)
+        fsm.reset()
+        assert fsm.state == EmergencyState.NOMINAL
+
+    def test_e_stop_cleared_on_ack(self):
+        fsm = _fsm()
+        self._reach_escalated(fsm)
+        out = fsm.tick(_inp(ack=True))
+        assert out.e_stop is False

jetson/ros2_ws/src/saltybot_emergency_msgs/CMakeLists.txt

@@ -0,0 +1,15 @@
+cmake_minimum_required(VERSION 3.8)
+project(saltybot_emergency_msgs)
+
+find_package(ament_cmake REQUIRED)
+find_package(rosidl_default_generators REQUIRED)
+find_package(builtin_interfaces REQUIRED)
+
+rosidl_generate_interfaces(${PROJECT_NAME}
+  "msg/EmergencyEvent.msg"
+  "msg/RecoveryAction.msg"
+  DEPENDENCIES builtin_interfaces
+)
+
+ament_export_dependencies(rosidl_default_runtime)
+ament_package()

jetson/ros2_ws/src/saltybot_emergency_msgs/msg/EmergencyEvent.msg

@@ -0,0 +1,25 @@
+# EmergencyEvent.msg — Real-time emergency system state snapshot (Issue #169)
+# Published by: /saltybot/emergency_node
+# Topic: /saltybot/emergency
+
+builtin_interfaces/Time stamp
+
+# Overall FSM state
+# Values: "NOMINAL" | "STOPPING" | "RECOVERING" | "ESCALATED"
+string state
+
+# Active threat (highest severity across all detectors)
+# threat_type values: "NONE" | "OBSTACLE_PROXIMITY" | "FALL_RISK" | "WHEEL_STUCK" | "BUMP"
+string threat_type
+
+# Severity: "CLEAR" | "MINOR" | "MAJOR" | "CRITICAL"
+string severity
+
+# Triggering metric value (e.g. distance in m, jerk in m/s³, stuck seconds)
+float32 threat_value
+
+# Human-readable description of the active threat
+string detail
+
+# True when emergency system is overriding normal cmd_vel with its own commands
+bool cmd_override

jetson/ros2_ws/src/saltybot_emergency_msgs/msg/RecoveryAction.msg

@@ -0,0 +1,15 @@
+# RecoveryAction.msg — Recovery sequencer state (Issue #169)
+# Published by: /saltybot/emergency_node
+# Topic: /saltybot/recovery_action
+
+builtin_interfaces/Time stamp
+
+# Current recovery action
+# Values: "IDLE" | "REVERSING" | "TURNING" | "RETRYING" | "GAVE_UP"
+string action
+
+# Number of reverse+turn attempts completed so far
+int32 retry_count
+
+# Progress through current phase [0.0 – 1.0]
+float32 progress

jetson/ros2_ws/src/saltybot_emergency_msgs/package.xml

@@ -0,0 +1,22 @@
+<?xml version="1.0"?>
+<?xml-model href="http://download.ros.org/schema/package_format3.xsd" schematypens="http://www.w3.org/2001/XMLSchema"?>
+<package format="3">
+  <name>saltybot_emergency_msgs</name>
+  <version>0.1.0</version>
+  <description>Emergency behavior message definitions for SaltyBot (Issue #169)</description>
+  <maintainer email="sl-controls@saltylab.local">sl-controls</maintainer>
+  <license>MIT</license>
+
+  <buildtool_depend>ament_cmake</buildtool_depend>
+  <buildtool_depend>rosidl_default_generators</buildtool_depend>
+
+  <depend>builtin_interfaces</depend>
+
+  <exec_depend>rosidl_default_runtime</exec_depend>
+
+  <member_of_group>rosidl_interface_packages</member_of_group>
+
+  <export>
+    <build_type>ament_cmake</build_type>
+  </export>
+</package>

jetson/ros2_ws/src/saltybot_social/config/emotion_params.yaml

@@ -0,0 +1,38 @@
+emotion_node:
+  ros__parameters:
+    # Path to the TRT FP16 emotion engine (built from emotion_cnn.onnx).
+    # Build with:
+    #   trtexec --onnx=emotion_cnn.onnx --fp16 --saveEngine=emotion_fp16.trt
+    # Leave empty to use landmark heuristic only (no GPU required).
+    engine_path: "/models/emotion_fp16.trt"
+
+    # Minimum smoothed confidence to publish an expression.
+    # Lower = more detections but more noise; 0.40 is a good production default.
+    min_confidence: 0.40
+
+    # EMA smoothing weight applied to each new observation.
+    #   0.0 = frozen (never updates)   1.0 = no smoothing (raw output)
+    # 0.30 gives stable ~3-frame moving average at 10 Hz face detection rate.
+    smoothing_alpha: 0.30
+
+    # Comma-separated person_ids that have opted out of emotion tracking.
+    # Empty string = everyone is tracked by default.
+    # Example: "1,3,7"
+    opt_out_persons: ""
+
+    # Skip face crops smaller than this side length (pixels).
+    # Very small crops produce unreliable emotion predictions.
+    face_min_size: 24
+
+    # If true and TRT engine is unavailable, fall back to the 5-point
+    # landmark heuristic.  Set false to suppress output entirely without TRT.
+    landmark_fallback: true
+
+    # Camera names matching saltybot_cameras topic naming convention.
+    camera_names: "front,left,rear,right"
+
+    # Number of active CSI camera streams to subscribe to.
+    n_cameras: 4
+
+    # Publish /social/emotion/context (JSON string) for conversation_node.
+    publish_context: true

jetson/ros2_ws/src/saltybot_social/config/speech_params.yaml

@@ -8,6 +8,7 @@ speech_pipeline_node:
     use_silero_vad: true
     whisper_model: "small"        # small (~500ms), medium (better quality, ~900ms)
     whisper_compute_type: "float16"
+    whisper_language: ""          # "" = auto-detect; set e.g. "fr" to force
     speaker_threshold: 0.65
     speaker_db_path: "/social_db/speaker_embeddings.json"
     publish_partial: true

jetson/ros2_ws/src/saltybot_social/config/tts_params.yaml

@@ -1,6 +1,8 @@
 tts_node:
   ros__parameters:
     voice_path: "/models/piper/en_US-lessac-medium.onnx"
+    voice_map_json: "{}"
+    default_language: "en"
     sample_rate: 22050
     volume: 1.0
     audio_device: ""              # "" = system default; set to device name if needed

jetson/ros2_ws/src/saltybot_social/launch/emotion.launch.py

@@ -0,0 +1,67 @@
+"""emotion.launch.py — Launch emotion_node for facial expression recognition (Issue #161)."""
+
+from launch import LaunchDescription
+from launch.actions import DeclareLaunchArgument
+from launch.substitutions import LaunchConfiguration, PathJoinSubstitution
+from launch_ros.actions import Node
+from launch_ros.substitutions import FindPackageShare
+
+
+def generate_launch_description() -> LaunchDescription:
+    pkg = FindPackageShare("saltybot_social")
+    params_file = PathJoinSubstitution([pkg, "config", "emotion_params.yaml"])
+
+    return LaunchDescription([
+        DeclareLaunchArgument(
+            "params_file",
+            default_value=params_file,
+            description="Path to emotion_node parameter YAML",
+        ),
+        DeclareLaunchArgument(
+            "engine_path",
+            default_value="/models/emotion_fp16.trt",
+            description="Path to TensorRT FP16 emotion engine (empty = landmark heuristic)",
+        ),
+        DeclareLaunchArgument(
+            "min_confidence",
+            default_value="0.40",
+            description="Minimum detection confidence to publish an expression",
+        ),
+        DeclareLaunchArgument(
+            "smoothing_alpha",
+            default_value="0.30",
+            description="EMA smoothing weight (0=frozen, 1=no smoothing)",
+        ),
+        DeclareLaunchArgument(
+            "opt_out_persons",
+            default_value="",
+            description="Comma-separated person_ids that opted out",
+        ),
+        DeclareLaunchArgument(
+            "landmark_fallback",
+            default_value="true",
+            description="Use landmark heuristic when TRT engine unavailable",
+        ),
+        DeclareLaunchArgument(
+            "publish_context",
+            default_value="true",
+            description="Publish /social/emotion/context JSON for LLM context",
+        ),
+        Node(
+            package="saltybot_social",
+            executable="emotion_node",
+            name="emotion_node",
+            output="screen",
+            parameters=[
+                LaunchConfiguration("params_file"),
+                {
+                    "engine_path":      LaunchConfiguration("engine_path"),
+                    "min_confidence":   LaunchConfiguration("min_confidence"),
+                    "smoothing_alpha":  LaunchConfiguration("smoothing_alpha"),
+                    "opt_out_persons":  LaunchConfiguration("opt_out_persons"),
+                    "landmark_fallback": LaunchConfiguration("landmark_fallback"),
+                    "publish_context":  LaunchConfiguration("publish_context"),
+                },
+            ],
+        ),
+    ])

jetson/ros2_ws/src/saltybot_social/saltybot_social/conversation_node.py

@@ -1,54 +1,30 @@
 """conversation_node.py — Local LLM conversation engine with per-person context.
-
+Issue #83/#161/#167
-Issue #83: Conversation engine for social-bot.
-
-Stack: Phi-3-mini or Llama-3.2-3B GGUF Q4_K_M via llama-cpp-python (CUDA).
-Subscribes /social/speech/transcript → builds per-person prompt → streams
-token output → publishes /social/conversation/response.
-
-Streaming: publishes partial=true tokens as they arrive, then final=false
-at end of generation. TTS node can begin synthesis on first sentence boundary.
-
-ROS2 topics:
-  Subscribe:  /social/speech/transcript (saltybot_social_msgs/SpeechTranscript)
-  Publish:    /social/conversation/response (saltybot_social_msgs/ConversationResponse)
-
-Parameters:
-  model_path       (str, "/models/gguf/phi-3-mini-4k-instruct-q4_k_m.gguf")
-  n_ctx            (int, 4096)
-  n_gpu_layers     (int, 20)    — GPU offload layers (increase for more VRAM usage)
-  max_tokens       (int, 200)
-  temperature      (float, 0.7)
-  top_p            (float, 0.9)
-  soul_path        (str, "/soul/SOUL.md")
-  context_db_path  (str, "/social_db/conversation_context.json")
-  save_interval_s  (float, 30.0) — how often to persist context to disk
-  stream           (bool, true)
 """
-
 from __future__ import annotations
-
+import json, threading, time
-import threading
+from typing import Dict, Optional
-import time
-from typing import Optional
-
 import rclpy
 from rclpy.node import Node
 from rclpy.qos import QoSProfile
-
+from std_msgs.msg import String
 from saltybot_social_msgs.msg import SpeechTranscript, ConversationResponse
 from .llm_context import ContextStore, build_llama_prompt, load_system_prompt, needs_summary_prompt
 
+_LANG_NAMES: Dict[str, str] = {
+    "en": "English", "fr": "French", "es": "Spanish", "de": "German",
+    "it": "Italian", "pt": "Portuguese", "ja": "Japanese", "zh": "Chinese",
+    "ko": "Korean", "ar": "Arabic", "ru": "Russian", "nl": "Dutch",
+    "pl": "Polish", "sv": "Swedish", "da": "Danish", "fi": "Finnish",
+    "no": "Norwegian", "tr": "Turkish", "hi": "Hindi", "uk": "Ukrainian",
+    "cs": "Czech", "ro": "Romanian", "hu": "Hungarian", "el": "Greek",
+}
 
 class ConversationNode(Node):
     """Local LLM inference node with per-person conversation memory."""
-
     def __init__(self) -> None:
         super().__init__("conversation_node")
-
+        self.declare_parameter("model_path", "/models/gguf/phi-3-mini-4k-instruct-q4_k_m.gguf")
-        # ── Parameters ──────────────────────────────────────────────────────
-        self.declare_parameter("model_path",
-                               "/models/gguf/phi-3-mini-4k-instruct-q4_k_m.gguf")
         self.declare_parameter("n_ctx", 4096)
         self.declare_parameter("n_gpu_layers", 20)
         self.declare_parameter("max_tokens", 200)
@@ -58,7 +34,6 @@ class ConversationNode(Node):
         self.declare_parameter("context_db_path", "/social_db/conversation_context.json")
         self.declare_parameter("save_interval_s", 30.0)
         self.declare_parameter("stream", True)
-
         self._model_path = self.get_parameter("model_path").value
         self._n_ctx = self.get_parameter("n_ctx").value
         self._n_gpu = self.get_parameter("n_gpu_layers").value
@@ -69,18 +44,9 @@ class ConversationNode(Node):
         self._db_path = self.get_parameter("context_db_path").value
         self._save_interval = self.get_parameter("save_interval_s").value
         self._stream = self.get_parameter("stream").value
-
-        # ── Publishers / Subscribers ─────────────────────────────────────────
         qos = QoSProfile(depth=10)
-        self._resp_pub = self.create_publisher(
+        self._resp_pub = self.create_publisher(ConversationResponse, "/social/conversation/response", qos)
-            ConversationResponse, "/social/conversation/response", qos
+        self._transcript_sub = self.create_subscription(SpeechTranscript, "/social/speech/transcript", self._on_transcript, qos)
-        )
-        self._transcript_sub = self.create_subscription(
-            SpeechTranscript, "/social/speech/transcript",
-            self._on_transcript, qos
-        )
-
-        # ── State ────────────────────────────────────────────────────────────
         self._llm = None
         self._system_prompt = load_system_prompt(self._soul_path)
         self._ctx_store = ContextStore(self._db_path)
@@ -88,180 +54,114 @@ class ConversationNode(Node):
         self._turn_counter = 0
         self._generating = False
         self._last_save = time.time()
-
+        self._speaker_lang: Dict[str, str] = {}
-        # ── Load LLM in background ────────────────────────────────────────────
+        self._emotions: Dict[str, str] = {}
+        self.create_subscription(String, "/social/emotion/context", self._on_emotion_context, 10)
         threading.Thread(target=self._load_llm, daemon=True).start()
-
-        # ── Periodic context save ────────────────────────────────────────────
         self._save_timer = self.create_timer(self._save_interval, self._save_context)
-
         self.get_logger().info(
             f"ConversationNode init (model={self._model_path}, "
             f"gpu_layers={self._n_gpu}, ctx={self._n_ctx})"
         )
 
-    # ── Model loading ─────────────────────────────────────────────────────────
-
     def _load_llm(self) -> None:
         t0 = time.time()
-        self.get_logger().info(f"Loading LLM: {self._model_path}")
         try:
             from llama_cpp import Llama
-            self._llm = Llama(
+            self._llm = Llama(model_path=self._model_path, n_ctx=self._n_ctx, n_gpu_layers=self._n_gpu, n_threads=4, verbose=False)
-                model_path=self._model_path,
+            self.get_logger().info(f"LLM ready ({time.time()-t0:.1f}s)")
-                n_ctx=self._n_ctx,
-                n_gpu_layers=self._n_gpu,
-                n_threads=4,
-                verbose=False,
-            )
-            self.get_logger().info(
-                f"LLM ready ({time.time()-t0:.1f}s). "
-                f"Context: {self._n_ctx} tokens, GPU layers: {self._n_gpu}"
-            )
         except Exception as e:
             self.get_logger().error(f"LLM load failed: {e}")
 
-    # ── Transcript callback ───────────────────────────────────────────────────
-
     def _on_transcript(self, msg: SpeechTranscript) -> None:
-        """Handle final transcripts only (skip streaming partials)."""
+        if msg.is_partial or not msg.text.strip():
-        if msg.is_partial:
             return
-        if not msg.text.strip():
+        if msg.language:
-            return
+            self._speaker_lang[msg.speaker_id] = msg.language
-
+        self.get_logger().info(f"Transcript [{msg.speaker_id}/{msg.language or '?'}]: '{msg.text}'")
-        self.get_logger().info(
+        threading.Thread(target=self._generate_response, args=(msg.text.strip(), msg.speaker_id), daemon=True).start()
-            f"Transcript [{msg.speaker_id}]: '{msg.text}'"
-        )
-
-        threading.Thread(
-            target=self._generate_response,
-            args=(msg.text.strip(), msg.speaker_id),
-            daemon=True,
-        ).start()
-
-    # ── LLM inference ─────────────────────────────────────────────────────────
 
     def _generate_response(self, user_text: str, speaker_id: str) -> None:
-        """Generate LLM response with streaming. Runs in thread."""
         if self._llm is None:
-            self.get_logger().warn("LLM not loaded yet, dropping utterance")
+            self.get_logger().warn("LLM not loaded yet, dropping utterance"); return
-            return
-
         with self._lock:
             if self._generating:
-                self.get_logger().warn("LLM busy, dropping utterance")
+                self.get_logger().warn("LLM busy, dropping utterance"); return
-                return
             self._generating = True
             self._turn_counter += 1
             turn_id = self._turn_counter
-
+        lang = self._speaker_lang.get(speaker_id, "en")
         try:
             ctx = self._ctx_store.get(speaker_id)
-
-            # Summary compression if context is long
             if ctx.needs_compression():
                 self._compress_context(ctx)
-
+            emotion_hint = self._emotion_hint(speaker_id)
-            ctx.add_user(user_text)
+            lang_hint = self._language_hint(speaker_id)
-
+            hints = " ".join(h for h in (emotion_hint, lang_hint) if h)
-            prompt = build_llama_prompt(
+            annotated = f"{user_text}  {hints}".rstrip() if hints else user_text
-                ctx, user_text, self._system_prompt
+            ctx.add_user(annotated)
-            )
+            prompt = build_llama_prompt(ctx, annotated, self._system_prompt)
-
             t0 = time.perf_counter()
             full_response = ""
-
             if self._stream:
-                output = self._llm(
+                output = self._llm(prompt, max_tokens=self._max_tokens, temperature=self._temperature, top_p=self._top_p, stream=True, stop=["<|user|>", "<|system|>", "\n\n\n"])
-                    prompt,
-                    max_tokens=self._max_tokens,
-                    temperature=self._temperature,
-                    top_p=self._top_p,
-                    stream=True,
-                    stop=["<|user|>", "<|system|>", "\n\n\n"],
-                )
                 for chunk in output:
                     token = chunk["choices"][0]["text"]
                     full_response += token
-                    # Publish partial after each sentence boundary for low TTS latency
                     if token.endswith((".", "!", "?", "\n")):
-                        self._publish_response(
+                        self._publish_response(full_response.strip(), speaker_id, turn_id, language=lang, is_partial=True)
-                            full_response.strip(), speaker_id, turn_id, is_partial=True
-                        )
             else:
-                output = self._llm(
+                output = self._llm(prompt, max_tokens=self._max_tokens, temperature=self._temperature, top_p=self._top_p, stream=False, stop=["<|user|>", "<|system|>"])
-                    prompt,
-                    max_tokens=self._max_tokens,
-                    temperature=self._temperature,
-                    top_p=self._top_p,
-                    stream=False,
-                    stop=["<|user|>", "<|system|>"],
-                )
                 full_response = output["choices"][0]["text"]
-
             full_response = full_response.strip()
-            latency_ms = (time.perf_counter() - t0) * 1000
+            self.get_logger().info(f"LLM [{speaker_id}/{lang}] ({(time.perf_counter()-t0)*1000:.0f}ms): '{full_response[:80]}'")
-            self.get_logger().info(
-                f"LLM [{speaker_id}] ({latency_ms:.0f}ms): '{full_response[:80]}'"
-            )
-
             ctx.add_assistant(full_response)
-            self._publish_response(full_response, speaker_id, turn_id, is_partial=False)
+            self._publish_response(full_response, speaker_id, turn_id, language=lang, is_partial=False)
-
         except Exception as e:
             self.get_logger().error(f"LLM inference error: {e}")
         finally:
-            with self._lock:
+            with self._lock: self._generating = False
-                self._generating = False
 
     def _compress_context(self, ctx) -> None:
-        """Ask LLM to summarize old turns for context compression."""
+        if self._llm is None: ctx.compress("(history omitted)"); return
-        if self._llm is None:
-            ctx.compress("(history omitted)")
-            return
         try:
-            summary_prompt = needs_summary_prompt(ctx)
+            result = self._llm(needs_summary_prompt(ctx), max_tokens=80, temperature=0.3, stream=False)
-            result = self._llm(summary_prompt, max_tokens=80, temperature=0.3, stream=False)
+            ctx.compress(result["choices"][0]["text"].strip())
-            summary = result["choices"][0]["text"].strip()
+        except Exception: ctx.compress("(history omitted)")
-            ctx.compress(summary)
-            self.get_logger().debug(
-                f"Context compressed for {ctx.person_id}: '{summary[:60]}'"
-            )
-        except Exception:
-            ctx.compress("(history omitted)")
 
-    # ── Publish ───────────────────────────────────────────────────────────────
+    def _language_hint(self, speaker_id: str) -> str:
+        lang = self._speaker_lang.get(speaker_id, "en")
+        if lang and lang != "en":
+            return f"[Please respond in {_LANG_NAMES.get(lang, lang)}.]"
+        return ""
 
-    def _publish_response(
+    def _on_emotion_context(self, msg: String) -> None:
-        self, text: str, speaker_id: str, turn_id: int, is_partial: bool
+        try:
-    ) -> None:
+            for k, v in json.loads(msg.data).get("emotions", {}).items():
+                self._emotions[k] = v
+        except Exception: pass
+
+    def _emotion_hint(self, speaker_id: str) -> str:
+        emo = self._emotions.get(speaker_id, "")
+        return f"[The person seems {emo} right now.]" if emo and emo != "neutral" else ""
+
+    def _publish_response(self, text: str, speaker_id: str, turn_id: int, language: str = "en", is_partial: bool = False) -> None:
         msg = ConversationResponse()
         msg.header.stamp = self.get_clock().now().to_msg()
-        msg.text = text
+        msg.text = text; msg.speaker_id = speaker_id; msg.is_partial = is_partial
-        msg.speaker_id = speaker_id
+        msg.turn_id = turn_id; msg.language = language
-        msg.is_partial = is_partial
-        msg.turn_id = turn_id
         self._resp_pub.publish(msg)
 
     def _save_context(self) -> None:
-        try:
+        try: self._ctx_store.save()
-            self._ctx_store.save()
+        except Exception as e: self.get_logger().error(f"Context save error: {e}")
-        except Exception as e:
-            self.get_logger().error(f"Context save error: {e}")
 
     def destroy_node(self) -> None:
-        self._save_context()
+        self._save_context(); super().destroy_node()
-        super().destroy_node()
-
 
 def main(args=None) -> None:
     rclpy.init(args=args)
     node = ConversationNode()
-    try:
+    try: rclpy.spin(node)
-        rclpy.spin(node)
+    except KeyboardInterrupt: pass
-    except KeyboardInterrupt:
+    finally: node.destroy_node(); rclpy.shutdown()
-        pass
-    finally:
-        node.destroy_node()
-        rclpy.shutdown()

jetson/ros2_ws/src/saltybot_social/saltybot_social/emotion_classifier.py

@@ -0,0 +1,414 @@
+"""emotion_classifier.py — 7-class facial expression classifier (Issue #161).
+
+Pure Python, no ROS2 / TensorRT / OpenCV dependencies.
+Wraps a TensorRT FP16 emotion CNN and provides:
+  - On-device TRT inference on 48×48 grayscale face crops
+  - Heuristic fallback from 5-point SCRFD facial landmarks
+  - Per-person EMA temporal smoothing for stable outputs
+  - Per-person opt-out registry
+
+Emotion classes (index order matches CNN output layer)
+------------------------------------------------------
+  0 = happy
+  1 = sad
+  2 = angry
+  3 = surprised
+  4 = fearful
+  5 = disgusted
+  6 = neutral
+
+Coordinate convention
+---------------------
+  Face crop: BGR uint8 ndarray, any size (resized to INPUT_SIZE internally).
+  Landmarks (lm10): 10 floats from FaceDetection.landmarks —
+    [left_eye_x, left_eye_y, right_eye_x, right_eye_y,
+     nose_x, nose_y, left_mouth_x, left_mouth_y,
+     right_mouth_x, right_mouth_y]
+  All coordinates are normalised image-space (0.0–1.0).
+"""
+
+from __future__ import annotations
+
+import math
+from dataclasses import dataclass, field
+from typing import Dict, List, Optional, Tuple
+
+# ── Constants ─────────────────────────────────────────────────────────────────
+
+EMOTIONS: List[str] = [
+    "happy", "sad", "angry", "surprised", "fearful", "disgusted", "neutral"
+]
+N_CLASSES: int = len(EMOTIONS)
+
+# Landmark slot indices within the 10-value array
+_LEX, _LEY = 0, 1   # left eye
+_REX, _REY = 2, 3   # right eye
+_NX,  _NY  = 4, 5   # nose
+_LMX, _LMY = 6, 7   # left mouth corner
+_RMX, _RMY = 8, 9   # right mouth corner
+
+# CNN input size
+INPUT_SIZE: int = 48
+
+
+# ── Result type ───────────────────────────────────────────────────────────────
+
+
+@dataclass
+class ClassifiedEmotion:
+    emotion: str            # dominant emotion label
+    confidence: float       # smoothed softmax probability (0.0–1.0)
+    scores: List[float]     # per-class scores, len == N_CLASSES
+    source: str = "cnn_trt" # "cnn_trt" | "landmark_heuristic" | "opt_out"
+
+
+# ── Math helpers ──────────────────────────────────────────────────────────────
+
+
+def _softmax(logits: List[float]) -> List[float]:
+    """Numerically stable softmax."""
+    m = max(logits)
+    exps = [math.exp(x - m) for x in logits]
+    total = sum(exps)
+    return [e / total for e in exps]
+
+
+def _argmax(values: List[float]) -> int:
+    best = 0
+    for i in range(1, len(values)):
+        if values[i] > values[best]:
+            best = i
+    return best
+
+
+def _uniform_scores() -> List[float]:
+    return [1.0 / N_CLASSES] * N_CLASSES
+
+
+# ── Per-person temporal smoother + opt-out registry ───────────────────────────
+
+
+class PersonEmotionTracker:
+    """Exponential moving-average smoother + opt-out registry per person.
+
+    Args:
+        alpha:   EMA weight for the newest observation (0.0 = frozen, 1.0 = no smooth).
+        max_age: Frames without update before the EMA is expired and reset.
+    """
+
+    def __init__(self, alpha: float = 0.3, max_age: int = 16) -> None:
+        self._alpha = alpha
+        self._max_age = max_age
+        self._ema: Dict[int, List[float]] = {}      # person_id → EMA scores
+        self._age: Dict[int, int] = {}              # person_id → frames since last update
+        self._opt_out: set = set()                  # person_ids that opted out
+
+    # ── Opt-out management ────────────────────────────────────────────────────
+
+    def set_opt_out(self, person_id: int, value: bool) -> None:
+        if value:
+            self._opt_out.add(person_id)
+            self._ema.pop(person_id, None)
+            self._age.pop(person_id, None)
+        else:
+            self._opt_out.discard(person_id)
+
+    def is_opt_out(self, person_id: int) -> bool:
+        return person_id in self._opt_out
+
+    # ── Smoothing ─────────────────────────────────────────────────────────────
+
+    def smooth(self, person_id: int, scores: List[float]) -> List[float]:
+        """Apply EMA to raw scores.  Returns smoothed scores (length N_CLASSES)."""
+        if person_id < 0:
+            return scores[:]
+
+        # Reset stale entries
+        if person_id in self._age and self._age[person_id] > self._max_age:
+            self.reset(person_id)
+
+        if person_id not in self._ema:
+            self._ema[person_id] = scores[:]
+            self._age[person_id] = 0
+            return scores[:]
+
+        ema = self._ema[person_id]
+        a = self._alpha
+        smoothed = [a * s + (1.0 - a) * e for s, e in zip(scores, ema)]
+        self._ema[person_id] = smoothed
+        self._age[person_id] = 0
+        return smoothed
+
+    def tick(self) -> None:
+        """Advance age counter for all tracked persons (call once per frame)."""
+        for pid in list(self._age.keys()):
+            self._age[pid] += 1
+
+    def reset(self, person_id: int) -> None:
+        self._ema.pop(person_id, None)
+        self._age.pop(person_id, None)
+
+    def reset_all(self) -> None:
+        self._ema.clear()
+        self._age.clear()
+
+    @property
+    def tracked_ids(self) -> List[int]:
+        return list(self._ema.keys())
+
+
+# ── TensorRT engine loader ────────────────────────────────────────────────────
+
+
+class _TrtEngine:
+    """Thin wrapper around a TensorRT FP16 emotion CNN engine.
+
+    Expected engine:
+      - Input binding:  name "input"  shape (1, 1, 48, 48)  float32
+      - Output binding: name "output" shape (1, 7)           float32 (softmax logits)
+
+    The engine is built offline from a MobileNetV2-based 48×48 grayscale CNN
+    (FER+ or AffectNet trained) via:
+        trtexec --onnx=emotion_cnn.onnx --fp16 --saveEngine=emotion_fp16.trt
+    """
+
+    def __init__(self) -> None:
+        self._context = None
+        self._h_input = None
+        self._h_output = None
+        self._d_input = None
+        self._d_output = None
+        self._stream = None
+        self._bindings: List = []
+
+    def load(self, engine_path: str) -> bool:
+        """Load TRT engine from file.  Returns True on success."""
+        try:
+            import tensorrt as trt              # type: ignore
+            import pycuda.autoinit             # type: ignore  # noqa: F401
+            import pycuda.driver as cuda       # type: ignore
+            import numpy as np
+
+            TRT_LOGGER = trt.Logger(trt.Logger.WARNING)
+            with open(engine_path, "rb") as f:
+                engine_data = f.read()
+
+            runtime = trt.Runtime(TRT_LOGGER)
+            engine = runtime.deserialize_cuda_engine(engine_data)
+            self._context = engine.create_execution_context()
+
+            # Allocate host + device buffers
+            self._h_input = cuda.pagelocked_empty((1, 1, INPUT_SIZE, INPUT_SIZE),
+                                                  dtype=np.float32)
+            self._h_output = cuda.pagelocked_empty((1, N_CLASSES), dtype=np.float32)
+            self._d_input = cuda.mem_alloc(self._h_input.nbytes)
+            self._d_output = cuda.mem_alloc(self._h_output.nbytes)
+            self._stream = cuda.Stream()
+            self._bindings = [int(self._d_input), int(self._d_output)]
+            return True
+
+        except Exception:
+            self._context = None
+            return False
+
+    @property
+    def ready(self) -> bool:
+        return self._context is not None
+
+    def infer(self, face_bgr) -> List[float]:
+        """Run TRT inference on a BGR face crop.  Returns 7 softmax scores."""
+        import pycuda.driver as cuda   # type: ignore
+        import numpy as np
+
+        img = _preprocess(face_bgr)    # (1, 1, 48, 48) float32
+        np.copyto(self._h_input, img)
+        cuda.memcpy_htod_async(self._d_input, self._h_input, self._stream)
+        self._context.execute_async_v2(self._bindings, self._stream.handle, None)
+        cuda.memcpy_dtoh_async(self._h_output, self._d_output, self._stream)
+        self._stream.synchronize()
+        logits = list(self._h_output[0])
+        return _softmax(logits)
+
+
+# ── Image pre-processing helper (importable without cv2 in test mode) ─────────
+
+
+def _preprocess(face_bgr) -> "np.ndarray":  # type: ignore
+    """Resize BGR crop to 48×48 grayscale, normalise to [-1, 1].
+
+    Returns ndarray shape (1, 1, 48, 48) float32.
+    """
+    import numpy as np  # type: ignore
+    import cv2          # type: ignore
+
+    gray = cv2.cvtColor(face_bgr, cv2.COLOR_BGR2GRAY)
+    resized = cv2.resize(gray, (INPUT_SIZE, INPUT_SIZE),
+                         interpolation=cv2.INTER_LINEAR)
+    norm = resized.astype(np.float32) / 127.5 - 1.0
+    return norm.reshape(1, 1, INPUT_SIZE, INPUT_SIZE)
+
+
+# ── Landmark heuristic ────────────────────────────────────────────────────────
+
+
+def classify_from_landmarks(lm10: List[float]) -> ClassifiedEmotion:
+    """Estimate emotion from 5-point SCRFD landmarks (10 floats).
+
+    Uses geometric ratios between eyes, nose, and mouth corners.
+    Accuracy is limited — treats this as a soft prior, not a definitive label.
+
+    Returns ClassifiedEmotion with source="landmark_heuristic".
+    """
+    if len(lm10) < 10:
+        scores = _uniform_scores()
+        scores[6] = 0.5           # bias neutral
+        scores = _renorm(scores)
+        return ClassifiedEmotion("neutral", scores[6], scores, "landmark_heuristic")
+
+    eye_y = (lm10[_LEY] + lm10[_REY]) / 2.0
+    nose_y = lm10[_NY]
+    mouth_y = (lm10[_LMY] + lm10[_RMY]) / 2.0
+    eye_span = max(abs(lm10[_REX] - lm10[_LEX]), 1e-4)
+    mouth_width = abs(lm10[_RMX] - lm10[_LMX])
+    mouth_asymm = abs(lm10[_LMY] - lm10[_RMY])
+
+    face_h = max(mouth_y - eye_y, 1e-4)
+
+    # Ratio of mouth span to interocular distance
+    width_ratio = mouth_width / eye_span          # >1.0 = wide open / happy smile
+    # How far mouth is below nose, relative to face height
+    mouth_below_nose = (mouth_y - nose_y) / face_h  # ~0.3–0.6 typical
+    # Relative asymmetry of mouth corners
+    asym_ratio = mouth_asymm / face_h             # >0.05 = notable asymmetry
+
+    # Build soft scores
+    scores = _uniform_scores()  # start uniform
+
+    if width_ratio > 0.85 and mouth_below_nose > 0.35:
+        # Wide mouth, normal vertical position → happy
+        scores[0] = 0.55 + 0.25 * min(1.0, (width_ratio - 0.85) / 0.5)
+    elif mouth_below_nose < 0.20 and width_ratio < 0.7:
+        # Tight, compressed mouth high up → surprised OR angry
+        scores[3] = 0.35   # surprised
+        scores[2] = 0.30   # angry
+    elif asym_ratio > 0.06:
+        # Asymmetric mouth → disgust or sadness
+        scores[5] = 0.30   # disgusted
+        scores[1] = 0.25   # sad
+    elif width_ratio < 0.65 and mouth_below_nose < 0.30:
+        # Tight and compressed → sad/angry
+        scores[1] = 0.35   # sad
+        scores[2] = 0.25   # angry
+    else:
+        # Default to neutral
+        scores[6] = 0.45
+
+    scores = _renorm(scores)
+    top_idx = _argmax(scores)
+    return ClassifiedEmotion(
+        emotion=EMOTIONS[top_idx],
+        confidence=round(scores[top_idx], 3),
+        scores=[round(s, 4) for s in scores],
+        source="landmark_heuristic",
+    )
+
+
+def _renorm(scores: List[float]) -> List[float]:
+    """Re-normalise scores so they sum to 1.0."""
+    total = sum(scores)
+    if total <= 0:
+        return _uniform_scores()
+    return [s / total for s in scores]
+
+
+# ── Public classifier ─────────────────────────────────────────────────────────
+
+
+class EmotionClassifier:
+    """Facade combining TRT inference + landmark fallback + per-person smoothing.
+
+    Usage
+    -----
+    >>> clf = EmotionClassifier(engine_path="/models/emotion_fp16.trt")
+    >>> clf.load()
+    True
+    >>> result = clf.classify_crop(face_bgr, person_id=42, tracker=tracker)
+    >>> result.emotion, result.confidence
+    ('happy', 0.87)
+    """
+
+    def __init__(
+        self,
+        engine_path: str = "",
+        alpha: float = 0.3,
+    ) -> None:
+        self._engine_path = engine_path
+        self._alpha = alpha
+        self._engine = _TrtEngine()
+
+    def load(self) -> bool:
+        """Load TRT engine.  Returns True if engine is ready."""
+        if not self._engine_path:
+            return False
+        return self._engine.load(self._engine_path)
+
+    @property
+    def ready(self) -> bool:
+        return self._engine.ready
+
+    def classify_crop(
+        self,
+        face_bgr,
+        person_id: int = -1,
+        tracker: Optional[PersonEmotionTracker] = None,
+    ) -> ClassifiedEmotion:
+        """Classify a BGR face crop.
+
+        Args:
+            face_bgr:  BGR ndarray from cv_bridge / direct crop.
+            person_id: For temporal smoothing.  -1 = no smoothing.
+            tracker:   Optional PersonEmotionTracker for EMA smoothing.
+
+        Returns:
+            ClassifiedEmotion with smoothed scores.
+        """
+        if self._engine.ready:
+            raw = self._engine.infer(face_bgr)
+            source = "cnn_trt"
+        else:
+            # Uniform fallback — no inference without engine
+            raw = _uniform_scores()
+            raw[6] = 0.40          # mild neutral bias
+            raw = _renorm(raw)
+            source = "landmark_heuristic"
+
+        smoothed = raw
+        if tracker is not None and person_id >= 0:
+            smoothed = tracker.smooth(person_id, raw)
+
+        top_idx = _argmax(smoothed)
+        return ClassifiedEmotion(
+            emotion=EMOTIONS[top_idx],
+            confidence=round(smoothed[top_idx], 3),
+            scores=[round(s, 4) for s in smoothed],
+            source=source,
+        )
+
+    def classify_from_landmarks(
+        self,
+        lm10: List[float],
+        person_id: int = -1,
+        tracker: Optional[PersonEmotionTracker] = None,
+    ) -> ClassifiedEmotion:
+        """Classify using landmark geometry only (no crop required)."""
+        result = classify_from_landmarks(lm10)
+        if tracker is not None and person_id >= 0:
+            smoothed = tracker.smooth(person_id, result.scores)
+            top_idx = _argmax(smoothed)
+            result = ClassifiedEmotion(
+                emotion=EMOTIONS[top_idx],
+                confidence=round(smoothed[top_idx], 3),
+                scores=[round(s, 4) for s in smoothed],
+                source=result.source,
+            )
+        return result

jetson/ros2_ws/src/saltybot_social/saltybot_social/emotion_node.py

@@ -0,0 +1,380 @@
+"""emotion_node.py — Facial expression recognition node (Issue #161).
+
+Piggybacks on the face detection pipeline: subscribes to
+/social/faces/detections (FaceDetectionArray), extracts face crops from the
+latest camera frames, runs a TensorRT FP16 emotion CNN, applies per-person
+EMA temporal smoothing, and publishes results on /social/faces/expressions.
+
+Architecture
+------------
+  FaceDetectionArray → crop extraction → TRT FP16 inference (< 5 ms) →
+  EMA smoothing → ExpressionArray publish
+
+If TRT engine is not available the node falls back to the 5-point landmark
+heuristic (classify_from_landmarks) which requires no GPU and adds < 0.1 ms.
+
+ROS2 topics
+-----------
+  Subscribe:
+    /social/faces/detections       (saltybot_social_msgs/FaceDetectionArray)
+    /camera/{name}/image_raw       (sensor_msgs/Image) × 4
+    /social/persons                (saltybot_social_msgs/PersonStateArray)
+  Publish:
+    /social/faces/expressions      (saltybot_social_msgs/ExpressionArray)
+    /social/emotion/context        (std_msgs/String)  — JSON for LLM context
+
+Parameters
+----------
+  engine_path          (str)   ""        — path to emotion_fp16.trt; empty = landmark only
+  min_confidence       (float) 0.40      — suppress results below this threshold
+  smoothing_alpha      (float) 0.30      — EMA weight (higher = faster, less stable)
+  opt_out_persons      (str)   ""        — comma-separated person_ids that opted out
+  face_min_size        (int)   24        — skip faces whose bbox is smaller (px side)
+  landmark_fallback    (bool)  true      — use landmark heuristic when TRT unavailable
+  camera_names         (str)   "front,left,rear,right"
+  n_cameras            (int)   4
+  publish_context      (bool)  true      — publish /social/emotion/context JSON
+"""
+
+from __future__ import annotations
+
+import json
+import threading
+import time
+from typing import Dict, List, Optional, Tuple
+
+import rclpy
+from rclpy.node import Node
+from rclpy.qos import QoSProfile, ReliabilityPolicy, HistoryPolicy
+
+from std_msgs.msg import String
+from sensor_msgs.msg import Image
+
+from saltybot_social_msgs.msg import (
+    FaceDetectionArray,
+    PersonStateArray,
+    ExpressionArray,
+    Expression,
+)
+from saltybot_social.emotion_classifier import (
+    EmotionClassifier,
+    PersonEmotionTracker,
+    EMOTIONS,
+)
+
+try:
+    from cv_bridge import CvBridge
+    _CV_BRIDGE_OK = True
+except ImportError:
+    _CV_BRIDGE_OK = False
+
+try:
+    import cv2
+    _CV2_OK = True
+except ImportError:
+    _CV2_OK = False
+
+
+# ── Per-camera latest-frame buffer ────────────────────────────────────────────
+
+
+class _FrameBuffer:
+    """Thread-safe store of the most recent image per camera."""
+
+    def __init__(self) -> None:
+        self._lock = threading.Lock()
+        self._frames: Dict[int, object] = {}   # camera_id → cv2 BGR image
+
+    def put(self, camera_id: int, img) -> None:
+        with self._lock:
+            self._frames[camera_id] = img
+
+    def get(self, camera_id: int):
+        with self._lock:
+            return self._frames.get(camera_id)
+
+
+# ── ROS2 Node ─────────────────────────────────────────────────────────────────
+
+
+class EmotionNode(Node):
+    """Facial expression recognition — TRT FP16 + landmark fallback."""
+
+    def __init__(self) -> None:
+        super().__init__("emotion_node")
+
+        # ── Parameters ────────────────────────────────────────────────────────
+        self.declare_parameter("engine_path",      "")
+        self.declare_parameter("min_confidence",   0.40)
+        self.declare_parameter("smoothing_alpha",  0.30)
+        self.declare_parameter("opt_out_persons",  "")
+        self.declare_parameter("face_min_size",    24)
+        self.declare_parameter("landmark_fallback", True)
+        self.declare_parameter("camera_names",     "front,left,rear,right")
+        self.declare_parameter("n_cameras",        4)
+        self.declare_parameter("publish_context",  True)
+
+        engine_path   = self.get_parameter("engine_path").value
+        self._min_conf = self.get_parameter("min_confidence").value
+        alpha          = self.get_parameter("smoothing_alpha").value
+        opt_out_str    = self.get_parameter("opt_out_persons").value
+        self._face_min = self.get_parameter("face_min_size").value
+        self._lm_fallback = self.get_parameter("landmark_fallback").value
+        cam_names_str  = self.get_parameter("camera_names").value
+        n_cameras      = self.get_parameter("n_cameras").value
+        self._pub_ctx  = self.get_parameter("publish_context").value
+
+        # ── Classifier + tracker ───────────────────────────────────────────
+        self._classifier = EmotionClassifier(engine_path=engine_path, alpha=alpha)
+        self._tracker    = PersonEmotionTracker(alpha=alpha)
+
+        # Parse opt-out list
+        for pid_str in opt_out_str.split(","):
+            pid_str = pid_str.strip()
+            if pid_str.isdigit():
+                self._tracker.set_opt_out(int(pid_str), True)
+
+        # ── Camera frame buffer + cv_bridge ───────────────────────────────
+        self._frame_buf = _FrameBuffer()
+        self._bridge    = CvBridge() if _CV_BRIDGE_OK else None
+        self._cam_names = [n.strip() for n in cam_names_str.split(",")]
+        self._cam_id_map: Dict[str, int] = {
+            name: idx for idx, name in enumerate(self._cam_names)
+        }
+
+        # ── Latest persons for person_id ↔ face_id correlation ────────────
+        self._persons_lock = threading.Lock()
+        self._face_to_person: Dict[int, int] = {}   # face_id → person_id
+
+        # ── Context for LLM (latest emotion per person) ───────────────────
+        self._emotion_context: Dict[int, str] = {}  # person_id → emotion label
+
+        # ── QoS profiles ──────────────────────────────────────────────────
+        be_qos = QoSProfile(
+            reliability=ReliabilityPolicy.BEST_EFFORT,
+            history=HistoryPolicy.KEEP_LAST,
+            depth=4,
+        )
+
+        # ── Subscriptions ──────────────────────────────────────────────────
+        self.create_subscription(
+            FaceDetectionArray,
+            "/social/faces/detections",
+            self._on_faces,
+            be_qos,
+        )
+        self.create_subscription(
+            PersonStateArray,
+            "/social/persons",
+            self._on_persons,
+            be_qos,
+        )
+        for name in self._cam_names[:n_cameras]:
+            cam_id = self._cam_id_map.get(name, 0)
+            self.create_subscription(
+                Image,
+                f"/camera/{name}/image_raw",
+                lambda msg, cid=cam_id: self._on_image(msg, cid),
+                be_qos,
+            )
+
+        # ── Publishers ─────────────────────────────────────────────────────
+        self._expr_pub = self.create_publisher(
+            ExpressionArray, "/social/faces/expressions", be_qos
+        )
+        if self._pub_ctx:
+            self._ctx_pub = self.create_publisher(
+                String, "/social/emotion/context", 10
+            )
+        else:
+            self._ctx_pub = None
+
+        # ── Load TRT engine in background ──────────────────────────────────
+        threading.Thread(target=self._load_engine, daemon=True).start()
+
+        self.get_logger().info(
+            f"EmotionNode ready — engine='{engine_path}', "
+            f"alpha={alpha:.2f}, min_conf={self._min_conf:.2f}, "
+            f"cameras={self._cam_names[:n_cameras]}"
+        )
+
+    # ── Engine loading ────────────────────────────────────────────────────────
+
+    def _load_engine(self) -> None:
+        if not self._classifier._engine_path:
+            if self._lm_fallback:
+                self.get_logger().info(
+                    "No TRT engine path — using landmark heuristic fallback"
+                )
+            else:
+                self.get_logger().warn(
+                    "No TRT engine and landmark_fallback=false — "
+                    "emotion_node will not classify"
+                )
+            return
+
+        ok = self._classifier.load()
+        if ok:
+            self.get_logger().info("TRT emotion engine loaded ✓")
+        else:
+            self.get_logger().warn(
+                "TRT engine load failed — falling back to landmark heuristic"
+            )
+
+    # ── Camera frame ingestion ────────────────────────────────────────────────
+
+    def _on_image(self, msg: Image, camera_id: int) -> None:
+        if not _CV_BRIDGE_OK or self._bridge is None:
+            return
+        try:
+            bgr = self._bridge.imgmsg_to_cv2(msg, desired_encoding="bgr8")
+            self._frame_buf.put(camera_id, bgr)
+        except Exception as exc:
+            self.get_logger().debug(f"cv_bridge error cam{camera_id}: {exc}")
+
+    # ── Person-state for face_id → person_id mapping ─────────────────────────
+
+    def _on_persons(self, msg: PersonStateArray) -> None:
+        mapping: Dict[int, int] = {}
+        for ps in msg.persons:
+            if ps.face_id >= 0:
+                mapping[ps.face_id] = ps.person_id
+        with self._persons_lock:
+            self._face_to_person = mapping
+
+    def _get_person_id(self, face_id: int) -> int:
+        with self._persons_lock:
+            return self._face_to_person.get(face_id, -1)
+
+    # ── Main face-detection callback ──────────────────────────────────────────
+
+    def _on_faces(self, msg: FaceDetectionArray) -> None:
+        if not msg.faces:
+            return
+
+        expressions: List[Expression] = []
+
+        for face in msg.faces:
+            person_id = self._get_person_id(face.face_id)
+
+            # Opt-out check
+            if person_id >= 0 and self._tracker.is_opt_out(person_id):
+                expr = Expression()
+                expr.header = msg.header
+                expr.person_id = person_id
+                expr.face_id   = face.face_id
+                expr.emotion   = ""
+                expr.confidence = 0.0
+                expr.scores    = [0.0] * 7
+                expr.is_opt_out = True
+                expr.source     = "opt_out"
+                expressions.append(expr)
+                continue
+
+            # Try TRT crop classification first
+            result = None
+            if self._classifier.ready and _CV2_OK:
+                result = self._classify_crop(face, person_id)
+
+            # Fallback to landmark heuristic
+            if result is None and self._lm_fallback:
+                result = self._classifier.classify_from_landmarks(
+                    list(face.landmarks),
+                    person_id=person_id,
+                    tracker=self._tracker,
+                )
+
+            if result is None:
+                continue
+
+            if result.confidence < self._min_conf:
+                continue
+
+            # Build ROS message
+            expr = Expression()
+            expr.header     = msg.header
+            expr.person_id  = person_id
+            expr.face_id    = face.face_id
+            expr.emotion    = result.emotion
+            expr.confidence = result.confidence
+            # Pad/trim scores to exactly 7
+            sc = result.scores
+            expr.scores = (sc + [0.0] * 7)[:7]
+            expr.is_opt_out = False
+            expr.source     = result.source
+            expressions.append(expr)
+
+            # Update LLM context cache
+            if person_id >= 0:
+                self._emotion_context[person_id] = result.emotion
+
+        if not expressions:
+            return
+
+        # Publish ExpressionArray
+        arr = ExpressionArray()
+        arr.header = msg.header
+        arr.expressions = expressions
+        self._expr_pub.publish(arr)
+
+        # Publish JSON context for conversation node
+        if self._ctx_pub is not None:
+            self._publish_context()
+
+        self._tracker.tick()
+
+    def _classify_crop(self, face, person_id: int):
+        """Extract crop from frame buffer and run TRT inference."""
+        # Resolve camera_id from face header frame_id
+        frame_id = face.header.frame_id if face.header.frame_id else "front"
+        cam_name = frame_id.split("/")[-1].split("_")[0]   # "front", "left", etc.
+        camera_id = self._cam_id_map.get(cam_name, 0)
+
+        frame = self._frame_buf.get(camera_id)
+        if frame is None:
+            return None
+
+        h, w = frame.shape[:2]
+        x1 = max(0, int(face.bbox_x * w))
+        y1 = max(0, int(face.bbox_y * h))
+        x2 = min(w, int((face.bbox_x + face.bbox_w) * w))
+        y2 = min(h, int((face.bbox_y + face.bbox_h) * h))
+
+        if (x2 - x1) < self._face_min or (y2 - y1) < self._face_min:
+            return None
+
+        crop = frame[y1:y2, x1:x2]
+        if crop.size == 0:
+            return None
+
+        return self._classifier.classify_crop(
+            crop, person_id=person_id, tracker=self._tracker
+        )
+
+    # ── LLM context publisher ─────────────────────────────────────────────────
+
+    def _publish_context(self) -> None:
+        """Publish latest-emotions dict as JSON for conversation_node."""
+        ctx = {str(pid): emo for pid, emo in self._emotion_context.items()}
+        msg = String()
+        msg.data = json.dumps({"emotions": ctx, "ts": time.time()})
+        self._ctx_pub.publish(msg)
+
+
+# ── Entry point ───────────────────────────────────────────────────────────────
+
+
+def main(args=None) -> None:
+    rclpy.init(args=args)
+    node = EmotionNode()
+    try:
+        rclpy.spin(node)
+    except KeyboardInterrupt:
+        pass
+    finally:
+        node.destroy_node()
+        rclpy.try_shutdown()
+
+
+if __name__ == "__main__":
+    main()

jetson/ros2_ws/src/saltybot_social/saltybot_social/speech_pipeline_node.py

@@ -66,6 +66,7 @@ class SpeechPipelineNode(Node):
         self.declare_parameter("use_silero_vad", True)
         self.declare_parameter("whisper_model", "small")
         self.declare_parameter("whisper_compute_type", "float16")
+        self.declare_parameter("whisper_language", "")
         self.declare_parameter("speaker_threshold", 0.65)
         self.declare_parameter("speaker_db_path", "/social_db/speaker_embeddings.json")
         self.declare_parameter("publish_partial", True)
@@ -78,6 +79,7 @@ class SpeechPipelineNode(Node):
         self._use_silero = self.get_parameter("use_silero_vad").value
         self._whisper_model_name = self.get_parameter("whisper_model").value
         self._compute_type = self.get_parameter("whisper_compute_type").value
+        self._whisper_language = self.get_parameter("whisper_language").value or None
         self._speaker_thresh = self.get_parameter("speaker_threshold").value
         self._speaker_db = self.get_parameter("speaker_db_path").value
         self._publish_partial = self.get_parameter("publish_partial").value
@@ -315,20 +317,24 @@ class SpeechPipelineNode(Node):
             except Exception as e:
                 self.get_logger().debug(f"Speaker ID error: {e}")
 
-        # Streaming Whisper transcription
+        # Streaming Whisper transcription with language detection
         partial_text = ""
+        detected_lang = self._whisper_language or "en"
         try:
-            segments_gen, _info = self._whisper.transcribe(
+            segments_gen, info = self._whisper.transcribe(
                 audio_np,
-                language="en",
+                language=self._whisper_language,  # None = auto-detect
                 beam_size=3,
                 vad_filter=False,
             )
+            if hasattr(info, "language") and info.language:
+                detected_lang = info.language
             for seg in segments_gen:
                 partial_text += seg.text.strip() + " "
                 if self._publish_partial:
                     self._publish_transcript(
-                        partial_text.strip(), speaker_id, 0.0, duration, is_partial=True
+                        partial_text.strip(), speaker_id, 0.0, duration,
+                        language=detected_lang, is_partial=True,
                     )
         except Exception as e:
             self.get_logger().error(f"Whisper error: {e}")
@@ -340,15 +346,19 @@ class SpeechPipelineNode(Node):
 
         latency_ms = (time.perf_counter() - t0) * 1000
         self.get_logger().info(
-            f"STT [{speaker_id}] ({duration:.1f}s, {latency_ms:.0f}ms): '{final_text}'"
+            f"STT [{speaker_id}/{detected_lang}] ({duration:.1f}s, {latency_ms:.0f}ms): "
+            f"'{final_text}'"
+        )
+        self._publish_transcript(
+            final_text, speaker_id, 0.9, duration,
+            language=detected_lang, is_partial=False,
         )
-        self._publish_transcript(final_text, speaker_id, 0.9, duration, is_partial=False)
 
     # ── Publishers ────────────────────────────────────────────────────────────
 
     def _publish_transcript(
         self, text: str, speaker_id: str, confidence: float,
-        duration: float, is_partial: bool
+        duration: float, language: str = "en", is_partial: bool = False,
     ) -> None:
         msg = SpeechTranscript()
         msg.header.stamp = self.get_clock().now().to_msg()
@@ -356,6 +366,7 @@ class SpeechPipelineNode(Node):
         msg.speaker_id = speaker_id
         msg.confidence = confidence
         msg.audio_duration = duration
+        msg.language = language
         msg.is_partial = is_partial
         self._transcript_pub.publish(msg)
 

jetson/ros2_ws/src/saltybot_social/saltybot_social/tts_node.py

@@ -1,228 +1,136 @@
-"""tts_node.py — Streaming TTS with Piper / first-chunk streaming.
+"""tts_node.py -- Streaming TTS with Piper / first-chunk streaming.
-
+Issue #85/#167
-Issue #85: Streaming TTS — Piper/XTTS integration with first-chunk streaming.
-
-Pipeline:
-  /social/conversation/response (ConversationResponse)
-    → sentence split → Piper ONNX synthesis (sentence by sentence)
-      → PCM16 chunks → USB speaker (sounddevice) + publish /social/tts/audio
-
-First-chunk strategy:
-  - On partial=true ConversationResponse, extract first sentence and synthesize
-    immediately → audio starts before LLM finishes generating
-  - On final=false, synthesize remaining sentences
-
-Latency target: <200ms to first audio chunk.
-
-ROS2 topics:
-  Subscribe: /social/conversation/response (saltybot_social_msgs/ConversationResponse)
-  Publish:   /social/tts/audio (audio_msgs/Audio or std_msgs/UInt8MultiArray fallback)
-
-Parameters:
-  voice_path      (str, "/models/piper/en_US-lessac-medium.onnx")
-  sample_rate     (int, 22050)
-  volume          (float, 1.0)
-  audio_device    (str, "")  — sounddevice device name; "" = system default
-  playback_enabled (bool, true)
-  publish_audio   (bool, false)  — publish PCM to ROS2 topic
-  sentence_streaming (bool, true) — synthesize sentence-by-sentence
 """
-
 from __future__ import annotations
-
+import json, queue, threading, time
-import queue
+from typing import Any, Dict, Optional
-import threading
-import time
-from typing import Optional
-
 import rclpy
 from rclpy.node import Node
 from rclpy.qos import QoSProfile
 from std_msgs.msg import UInt8MultiArray
-
 from saltybot_social_msgs.msg import ConversationResponse
 from .tts_utils import split_sentences, strip_ssml, apply_volume, chunk_pcm, estimate_duration_ms
 
-
 class TtsNode(Node):
-    """Streaming TTS node using Piper ONNX."""
+    """Streaming TTS node using Piper ONNX with per-language voice switching."""
-
     def __init__(self) -> None:
         super().__init__("tts_node")
-
-        # ── Parameters ──────────────────────────────────────────────────────
         self.declare_parameter("voice_path", "/models/piper/en_US-lessac-medium.onnx")
+        self.declare_parameter("voice_map_json", "{}")
+        self.declare_parameter("default_language", "en")
         self.declare_parameter("sample_rate", 22050)
         self.declare_parameter("volume", 1.0)
         self.declare_parameter("audio_device", "")
         self.declare_parameter("playback_enabled", True)
         self.declare_parameter("publish_audio", False)
         self.declare_parameter("sentence_streaming", True)
-
         self._voice_path = self.get_parameter("voice_path").value
+        self._voice_map_json = self.get_parameter("voice_map_json").value
+        self._default_language = self.get_parameter("default_language").value or "en"
         self._sample_rate = self.get_parameter("sample_rate").value
         self._volume = self.get_parameter("volume").value
         self._audio_device = self.get_parameter("audio_device").value or None
         self._playback = self.get_parameter("playback_enabled").value
         self._publish_audio = self.get_parameter("publish_audio").value
         self._sentence_streaming = self.get_parameter("sentence_streaming").value
-
+        try:
-        # ── Publishers / Subscribers ─────────────────────────────────────────
+            extra: Dict[str, str] = json.loads(self._voice_map_json) if self._voice_map_json.strip() not in ("{}","") else {}
+        except Exception as e:
+            self.get_logger().warn(f"voice_map_json parse error: {e}"); extra = {}
+        self._voice_paths: Dict[str, str] = {self._default_language: self._voice_path}
+        self._voice_paths.update(extra)
         qos = QoSProfile(depth=10)
-        self._resp_sub = self.create_subscription(
+        self._resp_sub = self.create_subscription(ConversationResponse, "/social/conversation/response", self._on_response, qos)
-            ConversationResponse, "/social/conversation/response",
-            self._on_response, qos
-        )
         if self._publish_audio:
-            self._audio_pub = self.create_publisher(
+            self._audio_pub = self.create_publisher(UInt8MultiArray, "/social/tts/audio", qos)
-                UInt8MultiArray, "/social/tts/audio", qos
+        self._voices: Dict[str, Any] = {}
-            )
+        self._voices_lock = threading.Lock()
-
-        # ── TTS engine ────────────────────────────────────────────────────────
-        self._voice = None
         self._playback_queue: queue.Queue = queue.Queue(maxsize=16)
         self._current_turn = -1
-        self._synthesized_turns: set = set()   # turn_ids already synthesized
+        self._synthesized_turns: set = set()
         self._lock = threading.Lock()
-
+        threading.Thread(target=self._load_voice_for_lang, args=(self._default_language,), daemon=True).start()
-        threading.Thread(target=self._load_voice, daemon=True).start()
         threading.Thread(target=self._playback_worker, daemon=True).start()
+        self.get_logger().info(f"TtsNode init (langs={list(self._voice_paths.keys())})")
 
-        self.get_logger().info(
+    def _load_voice_for_lang(self, lang: str) -> None:
-            f"TtsNode init (voice={self._voice_path}, "
+        path = self._voice_paths.get(lang)
-            f"streaming={self._sentence_streaming})"
+        if not path:
-        )
+            self.get_logger().warn(f"No voice for '{lang}', fallback to '{self._default_language}'"); return
-
+        with self._voices_lock:
-    # ── Voice loading ─────────────────────────────────────────────────────────
+            if lang in self._voices: return
-
-    def _load_voice(self) -> None:
-        t0 = time.time()
-        self.get_logger().info(f"Loading Piper voice: {self._voice_path}")
         try:
             from piper import PiperVoice
-            self._voice = PiperVoice.load(self._voice_path)
+            voice = PiperVoice.load(path)
-            # Warmup synthesis to pre-JIT ONNX graph
+            list(voice.synthesize_stream_raw("Hello."))
-            warmup_text = "Hello."
+            with self._voices_lock: self._voices[lang] = voice
-            list(self._voice.synthesize_stream_raw(warmup_text))
+            self.get_logger().info(f"Piper [{lang}] ready")
-            self.get_logger().info(f"Piper voice ready ({time.time()-t0:.1f}s)")
         except Exception as e:
-            self.get_logger().error(f"Piper voice load failed: {e}")
+            self.get_logger().error(f"Piper voice load failed [{lang}]: {e}")
 
-    # ── Response handler ──────────────────────────────────────────────────────
+    def _get_voice(self, lang: str):
+        with self._voices_lock:
+            v = self._voices.get(lang)
+            if v is not None: return v
+            if lang in self._voice_paths:
+                threading.Thread(target=self._load_voice_for_lang, args=(lang,), daemon=True).start()
+            return self._voices.get(self._default_language)
 
     def _on_response(self, msg: ConversationResponse) -> None:
-        """Handle streaming LLM response — synthesize sentence by sentence."""
+        if not msg.text.strip(): return
-        if not msg.text.strip():
+        lang = msg.language if msg.language else self._default_language
-            return
-
         with self._lock:
-            is_new_turn = msg.turn_id != self._current_turn
+            if msg.turn_id != self._current_turn:
-            if is_new_turn:
+                self._current_turn = msg.turn_id; self._synthesized_turns = set()
-                self._current_turn = msg.turn_id
-                # Clear old synthesized sentence cache for this new turn
-                self._synthesized_turns = set()
-
         text = strip_ssml(msg.text)
-
         if self._sentence_streaming:
-            sentences = split_sentences(text)
+            for sentence in split_sentences(text):
-            for sentence in sentences:
-                # Track which sentences we've already queued by content hash
                 key = (msg.turn_id, hash(sentence))
                 with self._lock:
-                    if key in self._synthesized_turns:
+                    if key in self._synthesized_turns: continue
-                        continue
                     self._synthesized_turns.add(key)
-                self._queue_synthesis(sentence)
+                self._queue_synthesis(sentence, lang)
         elif not msg.is_partial:
-            # Non-streaming: synthesize full response at end
+            self._queue_synthesis(text, lang)
-            self._queue_synthesis(text)
 
-    def _queue_synthesis(self, text: str) -> None:
+    def _queue_synthesis(self, text: str, lang: str) -> None:
-        """Queue a text segment for synthesis in the playback worker."""
+        if not text.strip(): return
-        if not text.strip():
+        try: self._playback_queue.put_nowait((text.strip(), lang))
-            return
+        except queue.Full: self.get_logger().warn("TTS queue full")
-        try:
-            self._playback_queue.put_nowait(text.strip())
-        except queue.Full:
-            self.get_logger().warn("TTS playback queue full, dropping segment")
-
-    # ── Playback worker ───────────────────────────────────────────────────────
 
     def _playback_worker(self) -> None:
-        """Consume synthesis queue: synthesize → play → publish."""
         while rclpy.ok():
-            try:
+            try: item = self._playback_queue.get(timeout=0.5)
-                text = self._playback_queue.get(timeout=0.5)
+            except queue.Empty: continue
-            except queue.Empty:
+            text, lang = item
-                continue
+            voice = self._get_voice(lang)
-
+            if voice is None:
-            if self._voice is None:
+                self.get_logger().warn(f"No voice for '{lang}'"); self._playback_queue.task_done(); continue
-                self.get_logger().warn("TTS voice not loaded yet")
-                self._playback_queue.task_done()
-                continue
-
             t0 = time.perf_counter()
-            pcm_data = self._synthesize(text)
+            pcm_data = self._synthesize(text, voice)
-            if pcm_data is None:
+            if pcm_data is None: self._playback_queue.task_done(); continue
-                self._playback_queue.task_done()
+            if self._volume != 1.0: pcm_data = apply_volume(pcm_data, self._volume)
-                continue
+            if self._playback: self._play_audio(pcm_data)
-
+            if self._publish_audio: self._publish_pcm(pcm_data)
-            synth_ms = (time.perf_counter() - t0) * 1000
-            dur_ms = estimate_duration_ms(pcm_data, self._sample_rate)
-            self.get_logger().debug(
-                f"TTS '{text[:40]}' synth={synth_ms:.0f}ms, dur={dur_ms:.0f}ms"
-            )
-
-            if self._volume != 1.0:
-                pcm_data = apply_volume(pcm_data, self._volume)
-
-            if self._playback:
-                self._play_audio(pcm_data)
-
-            if self._publish_audio:
-                self._publish_pcm(pcm_data)
-
             self._playback_queue.task_done()
 
-    def _synthesize(self, text: str) -> Optional[bytes]:
+    def _synthesize(self, text: str, voice) -> Optional[bytes]:
-        """Synthesize text to PCM16 bytes using Piper streaming."""
+        try: return b"".join(voice.synthesize_stream_raw(text))
-        if self._voice is None:
+        except Exception as e: self.get_logger().error(f"TTS error: {e}"); return None
-            return None
-        try:
-            chunks = list(self._voice.synthesize_stream_raw(text))
-            return b"".join(chunks)
-        except Exception as e:
-            self.get_logger().error(f"TTS synthesis error: {e}")
-            return None
 
     def _play_audio(self, pcm_data: bytes) -> None:
-        """Play PCM16 data on USB speaker via sounddevice."""
         try:
-            import sounddevice as sd
+            import sounddevice as sd, numpy as np
-            import numpy as np
+            sd.play(np.frombuffer(pcm_data,dtype=np.int16).astype(np.float32)/32768.0, samplerate=self._sample_rate, device=self._audio_device, blocking=True)
-            samples = np.frombuffer(pcm_data, dtype=np.int16).astype(np.float32) / 32768.0
+        except Exception as e: self.get_logger().error(f"Playback error: {e}")
-            sd.play(samples, samplerate=self._sample_rate, device=self._audio_device,
-                    blocking=True)
-        except Exception as e:
-            self.get_logger().error(f"Audio playback error: {e}")
 
     def _publish_pcm(self, pcm_data: bytes) -> None:
-        """Publish PCM data as UInt8MultiArray."""
+        if not hasattr(self,"_audio_pub"): return
-        if not hasattr(self, "_audio_pub"):
+        msg = UInt8MultiArray(); msg.data = list(pcm_data); self._audio_pub.publish(msg)
-            return
-        msg = UInt8MultiArray()
-        msg.data = list(pcm_data)
-        self._audio_pub.publish(msg)
-
 
 def main(args=None) -> None:
     rclpy.init(args=args)
     node = TtsNode()
-    try:
+    try: rclpy.spin(node)
-        rclpy.spin(node)
+    except KeyboardInterrupt: pass
-    except KeyboardInterrupt:
+    finally: node.destroy_node(); rclpy.shutdown()
-        pass
-    finally:
-        node.destroy_node()
-        rclpy.shutdown()

jetson/ros2_ws/src/saltybot_social/setup.py

@@ -37,6 +37,8 @@ setup(
             'voice_command_node = saltybot_social.voice_command_node:main',
             # Multi-camera gesture recognition (Issue #140)
             'gesture_node = saltybot_social.gesture_node:main',
+            # Facial expression recognition (Issue #161)
+            'emotion_node = saltybot_social.emotion_node:main',
         ],
     },
 )

jetson/ros2_ws/src/saltybot_social/test/test_emotion_classifier.py

@@ -0,0 +1,528 @@
+"""test_emotion_classifier.py — Unit tests for emotion_classifier (Issue #161).
+
+Tests cover:
+  - Emotion constant definitions
+  - Softmax normalisation
+  - PersonEmotionTracker: EMA smoothing, age expiry, opt-out, reset
+  - Landmark heuristic classifier: geometric edge cases
+  - EmotionClassifier: classify_crop fallback, classify_from_landmarks
+  - Score renormalisation, argmax, utility helpers
+  - Integration: full classify pipeline with mock engine
+
+No ROS2, TensorRT, or OpenCV runtime required.
+"""
+
+import math
+import pytest
+
+from saltybot_social.emotion_classifier import (
+    EMOTIONS,
+    N_CLASSES,
+    ClassifiedEmotion,
+    PersonEmotionTracker,
+    EmotionClassifier,
+    classify_from_landmarks,
+    _softmax,
+    _argmax,
+    _uniform_scores,
+    _renorm,
+)
+
+
+# ── Helpers ───────────────────────────────────────────────────────────────────
+
+
+def _lm_neutral() -> list:
+    """5-point SCRFD landmarks for a frontal neutral face (normalised)."""
+    return [
+        0.35, 0.38,   # left_eye
+        0.65, 0.38,   # right_eye
+        0.50, 0.52,   # nose
+        0.38, 0.72,   # left_mouth
+        0.62, 0.72,   # right_mouth
+    ]
+
+
+def _lm_happy() -> list:
+    """Wide mouth, symmetric corners, mouth well below nose."""
+    return [
+        0.35, 0.38,
+        0.65, 0.38,
+        0.50, 0.52,
+        0.30, 0.74,   # wide mouth
+        0.70, 0.74,
+    ]
+
+
+def _lm_sad() -> list:
+    """Compressed mouth, tight width."""
+    return [
+        0.35, 0.38,
+        0.65, 0.38,
+        0.50, 0.52,
+        0.44, 0.62,   # tight, close to nose
+        0.56, 0.62,
+    ]
+
+
+def _lm_asymmetric() -> list:
+    """Asymmetric mouth corners → disgust/sad signal."""
+    return [
+        0.35, 0.38,
+        0.65, 0.38,
+        0.50, 0.52,
+        0.38, 0.65,
+        0.62, 0.74,   # right corner lower → asymmetric
+    ]
+
+
+# ── TestEmotionConstants ──────────────────────────────────────────────────────
+
+
+class TestEmotionConstants:
+    def test_emotions_length(self):
+        assert len(EMOTIONS) == 7
+
+    def test_n_classes(self):
+        assert N_CLASSES == 7
+
+    def test_emotions_labels(self):
+        expected = ["happy", "sad", "angry", "surprised", "fearful", "disgusted", "neutral"]
+        assert EMOTIONS == expected
+
+    def test_happy_index(self):
+        assert EMOTIONS[0] == "happy"
+
+    def test_neutral_index(self):
+        assert EMOTIONS[6] == "neutral"
+
+    def test_all_lowercase(self):
+        for e in EMOTIONS:
+            assert e == e.lower()
+
+
+# ── TestSoftmax ───────────────────────────────────────────────────────────────
+
+
+class TestSoftmax:
+    def test_sums_to_one(self):
+        logits = [1.0, 2.0, 0.5, -1.0, 3.0, 0.0, 1.5]
+        result = _softmax(logits)
+        assert abs(sum(result) - 1.0) < 1e-6
+
+    def test_all_positive(self):
+        logits = [1.0, 2.0, 3.0, 0.0, -1.0, -2.0, 0.5]
+        result = _softmax(logits)
+        assert all(s > 0 for s in result)
+
+    def test_max_logit_gives_max_prob(self):
+        logits = [0.0, 0.0, 10.0, 0.0, 0.0, 0.0, 0.0]
+        result = _softmax(logits)
+        assert _argmax(result) == 2
+
+    def test_uniform_logits_uniform_probs(self):
+        logits = [1.0] * 7
+        result = _softmax(logits)
+        for p in result:
+            assert abs(p - 1.0 / 7.0) < 1e-6
+
+    def test_length_preserved(self):
+        logits = [0.0] * 7
+        assert len(_softmax(logits)) == 7
+
+    def test_numerically_stable_large_values(self):
+        logits = [1000.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0]
+        result = _softmax(logits)
+        assert abs(result[0] - 1.0) < 1e-6
+
+
+# ── TestArgmax ────────────────────────────────────────────────────────────────
+
+
+class TestArgmax:
+    def test_finds_max(self):
+        assert _argmax([0.1, 0.2, 0.9, 0.3, 0.1, 0.1, 0.1]) == 2
+
+    def test_single_element(self):
+        assert _argmax([0.5]) == 0
+
+    def test_first_when_all_equal(self):
+        result = _argmax([0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5])
+        assert result == 0
+
+
+# ── TestUniformScores ─────────────────────────────────────────────────────────
+
+
+class TestUniformScores:
+    def test_length(self):
+        assert len(_uniform_scores()) == N_CLASSES
+
+    def test_sums_to_one(self):
+        assert abs(sum(_uniform_scores()) - 1.0) < 1e-9
+
+    def test_all_equal(self):
+        s = _uniform_scores()
+        assert all(abs(x - s[0]) < 1e-9 for x in s)
+
+
+# ── TestRenorm ────────────────────────────────────────────────────────────────
+
+
+class TestRenorm:
+    def test_sums_to_one(self):
+        s = [0.1, 0.2, 0.3, 0.1, 0.1, 0.1, 0.1]
+        r = _renorm(s)
+        assert abs(sum(r) - 1.0) < 1e-9
+
+    def test_all_zero_returns_uniform(self):
+        r = _renorm([0.0] * 7)
+        assert len(r) == 7
+        assert abs(sum(r) - 1.0) < 1e-9
+
+    def test_preserves_order(self):
+        s = [0.5, 0.3, 0.0, 0.0, 0.0, 0.0, 0.2]
+        r = _renorm(s)
+        assert r[0] > r[1] > r[6]
+
+
+# ── TestPersonEmotionTracker ──────────────────────────────────────────────────
+
+
+class TestPersonEmotionTracker:
+    def _scores(self, idx: int, val: float = 0.8) -> list:
+        s = [0.0] * N_CLASSES
+        s[idx] = val
+        total = val + (N_CLASSES - 1) * 0.02
+        s = [v if i == idx else 0.02 for i, v in enumerate(s)]
+        s[idx] = val
+        t = sum(s)
+        return [x / t for x in s]
+
+    def test_first_call_returns_input(self):
+        tracker = PersonEmotionTracker(alpha=0.5)
+        scores = self._scores(0)
+        result = tracker.smooth(1, scores)
+        assert abs(result[0] - scores[0]) < 1e-9
+
+    def test_ema_converges_toward_new_dominant(self):
+        tracker = PersonEmotionTracker(alpha=0.5)
+        happy = self._scores(0)
+        sad   = self._scores(1)
+        tracker.smooth(1, happy)
+        # Push sad repeatedly
+        prev_sad = tracker.smooth(1, sad)[1]
+        for _ in range(20):
+            result = tracker.smooth(1, sad)
+        assert result[1] > prev_sad   # sad score increased
+
+    def test_alpha_1_no_smoothing(self):
+        tracker = PersonEmotionTracker(alpha=1.0)
+        s1 = self._scores(0)
+        s2 = self._scores(1)
+        tracker.smooth(1, s1)
+        result = tracker.smooth(1, s2)
+        for a, b in zip(result, s2):
+            assert abs(a - b) < 1e-9
+
+    def test_alpha_0_frozen(self):
+        tracker = PersonEmotionTracker(alpha=0.0)
+        s1 = self._scores(0)
+        s2 = self._scores(1)
+        tracker.smooth(1, s1)
+        result = tracker.smooth(1, s2)
+        for a, b in zip(result, s1):
+            assert abs(a - b) < 1e-9
+
+    def test_different_persons_independent(self):
+        tracker = PersonEmotionTracker(alpha=0.5)
+        happy = self._scores(0)
+        sad   = self._scores(1)
+        tracker.smooth(1, happy)
+        tracker.smooth(2, sad)
+        r1 = tracker.smooth(1, happy)
+        r2 = tracker.smooth(2, sad)
+        assert r1[0] > r2[0]   # person 1 more happy
+        assert r2[1] > r1[1]   # person 2 more sad
+
+    def test_negative_person_id_no_tracking(self):
+        tracker = PersonEmotionTracker()
+        scores = self._scores(2)
+        result = tracker.smooth(-1, scores)
+        assert result == scores   # unchanged, not stored
+
+    def test_reset_clears_ema(self):
+        tracker = PersonEmotionTracker(alpha=0.5)
+        s1 = self._scores(0)
+        tracker.smooth(1, s1)
+        tracker.reset(1)
+        assert 1 not in tracker.tracked_ids
+
+    def test_reset_all_clears_all(self):
+        tracker = PersonEmotionTracker(alpha=0.5)
+        for pid in range(5):
+            tracker.smooth(pid, self._scores(pid % N_CLASSES))
+        tracker.reset_all()
+        assert tracker.tracked_ids == []
+
+    def test_tracked_ids_populated(self):
+        tracker = PersonEmotionTracker()
+        tracker.smooth(10, self._scores(0))
+        tracker.smooth(20, self._scores(1))
+        assert set(tracker.tracked_ids) == {10, 20}
+
+    def test_age_expiry_resets_ema(self):
+        tracker = PersonEmotionTracker(alpha=0.5, max_age=3)
+        tracker.smooth(1, self._scores(0))
+        # Advance age beyond max_age
+        for _ in range(4):
+            tracker.tick()
+        # Next smooth after expiry should reset (first call returns input unchanged)
+        fresh_scores = self._scores(1)
+        result = tracker.smooth(1, fresh_scores)
+        # After reset, result should equal fresh_scores exactly
+        for a, b in zip(result, fresh_scores):
+            assert abs(a - b) < 1e-9
+
+
+# ── TestOptOut ────────────────────────────────────────────────────────────────
+
+
+class TestOptOut:
+    def test_set_opt_out_true(self):
+        tracker = PersonEmotionTracker()
+        tracker.set_opt_out(42, True)
+        assert tracker.is_opt_out(42)
+
+    def test_set_opt_out_false(self):
+        tracker = PersonEmotionTracker()
+        tracker.set_opt_out(42, True)
+        tracker.set_opt_out(42, False)
+        assert not tracker.is_opt_out(42)
+
+    def test_opt_out_clears_ema(self):
+        tracker = PersonEmotionTracker()
+        tracker.smooth(42, [0.5, 0.1, 0.1, 0.1, 0.1, 0.0, 0.1])
+        assert 42 in tracker.tracked_ids
+        tracker.set_opt_out(42, True)
+        assert 42 not in tracker.tracked_ids
+
+    def test_unknown_person_not_opted_out(self):
+        tracker = PersonEmotionTracker()
+        assert not tracker.is_opt_out(99)
+
+    def test_multiple_opt_outs(self):
+        tracker = PersonEmotionTracker()
+        for pid in [1, 2, 3]:
+            tracker.set_opt_out(pid, True)
+        for pid in [1, 2, 3]:
+            assert tracker.is_opt_out(pid)
+
+
+# ── TestClassifyFromLandmarks ─────────────────────────────────────────────────
+
+
+class TestClassifyFromLandmarks:
+    def test_returns_classified_emotion(self):
+        result = classify_from_landmarks(_lm_neutral())
+        assert isinstance(result, ClassifiedEmotion)
+
+    def test_emotion_is_valid_label(self):
+        result = classify_from_landmarks(_lm_neutral())
+        assert result.emotion in EMOTIONS
+
+    def test_scores_length(self):
+        result = classify_from_landmarks(_lm_neutral())
+        assert len(result.scores) == N_CLASSES
+
+    def test_scores_sum_to_one(self):
+        result = classify_from_landmarks(_lm_neutral())
+        # Scores are rounded to 4 dp; allow 1e-3 accumulation across 7 terms
+        assert abs(sum(result.scores) - 1.0) < 1e-3
+
+    def test_confidence_matches_top_score(self):
+        result = classify_from_landmarks(_lm_neutral())
+        # confidence is round(score, 3) and max score is round(s, 4) → ≤0.5e-3 diff
+        assert abs(result.confidence - max(result.scores)) < 5e-3
+
+    def test_source_is_landmark_heuristic(self):
+        result = classify_from_landmarks(_lm_neutral())
+        assert result.source == "landmark_heuristic"
+
+    def test_happy_landmarks_boost_happy(self):
+        happy = classify_from_landmarks(_lm_happy())
+        neutral = classify_from_landmarks(_lm_neutral())
+        # Happy landmarks should give relatively higher happy score
+        assert happy.scores[0] >= neutral.scores[0]
+
+    def test_sad_landmarks_suppress_happy(self):
+        sad_result = classify_from_landmarks(_lm_sad())
+        # Happy score should be relatively low for sad landmarks
+        assert sad_result.scores[0] < 0.5
+
+    def test_asymmetric_mouth_non_neutral(self):
+        asym = classify_from_landmarks(_lm_asymmetric())
+        # Asymmetric → disgust or sad should be elevated
+        assert asym.scores[5] > 0.10 or asym.scores[1] > 0.10
+
+    def test_empty_landmarks_returns_neutral(self):
+        result = classify_from_landmarks([])
+        assert result.emotion == "neutral"
+
+    def test_short_landmarks_returns_neutral(self):
+        result = classify_from_landmarks([0.5, 0.5])
+        assert result.emotion == "neutral"
+
+    def test_all_positive_scores(self):
+        result = classify_from_landmarks(_lm_happy())
+        assert all(s >= 0.0 for s in result.scores)
+
+    def test_confidence_in_range(self):
+        result = classify_from_landmarks(_lm_neutral())
+        assert 0.0 <= result.confidence <= 1.0
+
+
+# ── TestEmotionClassifier ─────────────────────────────────────────────────────
+
+
+class TestEmotionClassifier:
+    def test_init_no_engine_not_ready(self):
+        clf = EmotionClassifier(engine_path="")
+        assert not clf.ready
+
+    def test_load_empty_path_returns_false(self):
+        clf = EmotionClassifier(engine_path="")
+        assert clf.load() is False
+
+    def test_load_nonexistent_path_returns_false(self):
+        clf = EmotionClassifier(engine_path="/nonexistent/engine.trt")
+        result = clf.load()
+        assert result is False
+
+    def test_classify_crop_fallback_without_engine(self):
+        """Without TRT engine, classify_crop should return a valid result
+        with landmark heuristic source."""
+        clf = EmotionClassifier(engine_path="")
+        # Build a minimal synthetic 48x48 BGR image
+        try:
+            import numpy as np
+            fake_crop = np.zeros((48, 48, 3), dtype="uint8")
+            result = clf.classify_crop(fake_crop, person_id=-1, tracker=None)
+            assert isinstance(result, ClassifiedEmotion)
+            assert result.emotion in EMOTIONS
+            assert 0.0 <= result.confidence <= 1.0
+            assert len(result.scores) == N_CLASSES
+        except ImportError:
+            pytest.skip("numpy not available")
+
+    def test_classify_from_landmarks_delegates(self):
+        clf = EmotionClassifier(engine_path="")
+        result = clf.classify_from_landmarks(_lm_happy())
+        assert isinstance(result, ClassifiedEmotion)
+        assert result.source == "landmark_heuristic"
+
+    def test_classify_from_landmarks_with_tracker_smooths(self):
+        clf = EmotionClassifier(engine_path="", alpha=0.5)
+        tracker = PersonEmotionTracker(alpha=0.5)
+        r1 = clf.classify_from_landmarks(_lm_happy(), person_id=1, tracker=tracker)
+        r2 = clf.classify_from_landmarks(_lm_sad(), person_id=1, tracker=tracker)
+        # After smoothing, r2's top score should not be identical to raw sad scores
+        # (EMA blends r1 history into r2)
+        raw_sad = classify_from_landmarks(_lm_sad())
+        assert r2.scores != raw_sad.scores
+
+    def test_classify_from_landmarks_no_tracker_no_smooth(self):
+        clf = EmotionClassifier(engine_path="")
+        r1 = clf.classify_from_landmarks(_lm_happy(), person_id=1, tracker=None)
+        r2 = clf.classify_from_landmarks(_lm_happy(), person_id=1, tracker=None)
+        # Without tracker, same input → same output
+        assert r1.scores == r2.scores
+
+    def test_source_fallback_when_no_engine(self):
+        try:
+            import numpy as np
+        except ImportError:
+            pytest.skip("numpy not available")
+        clf = EmotionClassifier(engine_path="")
+        crop = __import__("numpy").zeros((48, 48, 3), dtype="uint8")
+        result = clf.classify_crop(crop)
+        assert result.source == "landmark_heuristic"
+
+    def test_classifier_alpha_propagated_to_tracker(self):
+        clf = EmotionClassifier(engine_path="", alpha=0.1)
+        assert clf._alpha == 0.1
+
+
+# ── TestClassifiedEmotionDataclass ───────────────────────────────────────────
+
+
+class TestClassifiedEmotionDataclass:
+    def test_fields(self):
+        ce = ClassifiedEmotion(
+            emotion="happy",
+            confidence=0.85,
+            scores=[0.85, 0.02, 0.02, 0.03, 0.02, 0.02, 0.04],
+            source="cnn_trt",
+        )
+        assert ce.emotion == "happy"
+        assert ce.confidence == 0.85
+        assert len(ce.scores) == 7
+        assert ce.source == "cnn_trt"
+
+    def test_default_source(self):
+        ce = ClassifiedEmotion("neutral", 0.6, [0.0] * 7)
+        assert ce.source == "cnn_trt"
+
+    def test_emotion_is_mutable(self):
+        ce = ClassifiedEmotion("neutral", 0.6, [0.0] * 7)
+        ce.emotion = "happy"
+        assert ce.emotion == "happy"
+
+
+# ── TestEdgeCases ─────────────────────────────────────────────────────────────
+
+
+class TestEdgeCases:
+    def test_softmax_single_element(self):
+        result = _softmax([5.0])
+        assert abs(result[0] - 1.0) < 1e-9
+
+    def test_tracker_negative_id_no_stored_state(self):
+        tracker = PersonEmotionTracker()
+        scores = [0.1, 0.1, 0.1, 0.1, 0.1, 0.1, 0.4]
+        tracker.smooth(-1, scores)
+        assert tracker.tracked_ids == []
+
+    def test_tick_increments_age(self):
+        tracker = PersonEmotionTracker(max_age=2)
+        tracker.smooth(1, [0.1] * 7)
+        tracker.tick()
+        tracker.tick()
+        tracker.tick()
+        # Age should exceed max_age → next smooth resets
+        fresh = [0.0, 0.0, 1.0, 0.0, 0.0, 0.0, 0.0]
+        result = tracker.smooth(1, fresh)
+        assert abs(result[2] - 1.0) < 1e-9
+
+    def test_renorm_negative_scores_safe(self):
+        # Negative scores shouldn't crash (though unusual in practice)
+        scores = [0.1, 0.2, 0.0, 0.0, 0.0, 0.0, 0.3]
+        r = _renorm(scores)
+        assert abs(sum(r) - 1.0) < 1e-9
+
+    def test_landmark_heuristic_very_close_eye_mouth(self):
+        # Degenerate face where everything is at same y → should not crash
+        lm = [0.3, 0.5, 0.7, 0.5, 0.5, 0.5, 0.4, 0.5, 0.6, 0.5]
+        result = classify_from_landmarks(lm)
+        assert result.emotion in EMOTIONS
+
+    def test_opt_out_then_re_enable(self):
+        tracker = PersonEmotionTracker()
+        tracker.smooth(5, [0.1] * 7)
+        tracker.set_opt_out(5, True)
+        assert tracker.is_opt_out(5)
+        tracker.set_opt_out(5, False)
+        assert not tracker.is_opt_out(5)
+        # Should be able to smooth again after re-enable
+        result = tracker.smooth(5, [0.2] * 7)
+        assert len(result) == N_CLASSES

jetson/ros2_ws/src/saltybot_social/test/test_multilang.py

@@ -0,0 +1,122 @@
+"""test_multilang.py -- Unit tests for Issue #167 multi-language support."""
+
+from __future__ import annotations
+import json, os
+from typing import Any, Dict, Optional
+import pytest
+
+def _pkg_root():
+    return os.path.dirname(os.path.dirname(os.path.abspath(__file__)))
+
+def _read_src(rel_path):
+    with open(os.path.join(_pkg_root(), rel_path)) as f:
+        return f.read()
+
+def _extract_lang_names():
+    import ast
+    src = _read_src("saltybot_social/conversation_node.py")
+    start = src.index("_LANG_NAMES: Dict[str, str] = {")
+    end   = src.index("\n}", start) + 2
+    return ast.literal_eval(src[start:end].split("=",1)[1].strip())
+
+class TestLangNames:
+    @pytest.fixture(scope="class")
+    def ln(self): return _extract_lang_names()
+    def test_english(self, ln): assert ln["en"] == "English"
+    def test_french(self, ln):  assert ln["fr"] == "French"
+    def test_spanish(self, ln): assert ln["es"] == "Spanish"
+    def test_german(self, ln):  assert ln["de"] == "German"
+    def test_japanese(self, ln):assert ln["ja"] == "Japanese"
+    def test_chinese(self, ln): assert ln["zh"] == "Chinese"
+    def test_arabic(self, ln):  assert ln["ar"] == "Arabic"
+    def test_at_least_15(self, ln): assert len(ln) >= 15
+    def test_lowercase_keys(self, ln):
+        for k in ln: assert k == k.lower() and 2 <= len(k) <= 3
+    def test_nonempty_values(self, ln):
+        for k, v in ln.items(): assert v
+
+class TestLanguageHint:
+    def _h(self, sl, sid, ln):
+        lang = sl.get(sid, "en")
+        return f"[Please respond in {ln.get(lang, lang)}.]" if lang and lang != "en" else ""
+    @pytest.fixture(scope="class")
+    def ln(self): return _extract_lang_names()
+    def test_english_no_hint(self, ln): assert self._h({"p": "en"}, "p", ln) == ""
+    def test_unknown_no_hint(self, ln): assert self._h({}, "p", ln) == ""
+    def test_french(self, ln): assert self._h({"p":"fr"},"p",ln) == "[Please respond in French.]"
+    def test_spanish(self, ln): assert self._h({"p":"es"},"p",ln) == "[Please respond in Spanish.]"
+    def test_unknown_code(self, ln): assert "xx" in self._h({"p":"xx"},"p",ln)
+    def test_brackets(self, ln):
+        h = self._h({"p":"de"},"p",ln)
+        assert h.startswith("[") and h.endswith("]")
+
+class TestVoiceMap:
+    def _parse(self, jstr, dl, dp):
+        try: extra = json.loads(jstr) if jstr.strip() not in ("{}","") else {}
+        except: extra = {}
+        r = {dl: dp}; r.update(extra); return r
+    def test_empty(self): assert self._parse("{}","en","/e") == {"en":"/e"}
+    def test_extra(self):
+        vm = self._parse('{"fr":"/f"}', "en", "/e")
+        assert vm["fr"] == "/f"
+    def test_invalid(self): assert self._parse("BAD","en","/e") == {"en":"/e"}
+    def test_multi(self):
+        assert len(self._parse(json.dumps({"fr":"/f","es":"/s"}),"en","/e")) == 3
+
+class TestVoiceSelect:
+    def _s(self, voices, lang, default):
+        return voices.get(lang) or voices.get(default)
+    def test_exact(self): assert self._s({"en":"E","fr":"F"},"fr","en") == "F"
+    def test_fallback(self): assert self._s({"en":"E"},"fr","en") == "E"
+    def test_none(self): assert self._s({},"fr","en") is None
+
+class TestSttFields:
+    @pytest.fixture(scope="class")
+    def src(self): return _read_src("saltybot_social/speech_pipeline_node.py")
+    def test_param(self, src): assert "whisper_language" in src
+    def test_detected_lang(self, src): assert "detected_lang" in src
+    def test_msg_language(self, src): assert "msg.language = language" in src
+    def test_auto_detect(self, src): assert "language=self._whisper_language" in src
+
+class TestConvFields:
+    @pytest.fixture(scope="class")
+    def src(self): return _read_src("saltybot_social/conversation_node.py")
+    def test_speaker_lang(self, src): assert "_speaker_lang" in src
+    def test_lang_hint_method(self, src): assert "_language_hint" in src
+    def test_msg_language(self, src): assert "msg.language = language" in src
+    def test_lang_names(self, src): assert "_LANG_NAMES" in src
+    def test_please_respond(self, src): assert "Please respond in" in src
+    def test_emotion_coexists(self, src): assert "_emotion_hint" in src
+
+class TestTtsFields:
+    @pytest.fixture(scope="class")
+    def src(self): return _read_src("saltybot_social/tts_node.py")
+    def test_voice_map_json(self, src): assert "voice_map_json" in src
+    def test_default_lang(self, src): assert "default_language" in src
+    def test_voices_dict(self, src): assert "_voices" in src
+    def test_get_voice(self, src): assert "_get_voice" in src
+    def test_load_voice_for_lang(self, src): assert "_load_voice_for_lang" in src
+    def test_queue_tuple(self, src): assert "(text.strip(), lang)" in src
+    def test_synthesize_voice_arg(self, src): assert "_synthesize(text, voice)" in src
+
+class TestMsgDefs:
+    @pytest.fixture(scope="class")
+    def tr(self): return _read_src("../saltybot_social_msgs/msg/SpeechTranscript.msg")
+    @pytest.fixture(scope="class")
+    def re(self): return _read_src("../saltybot_social_msgs/msg/ConversationResponse.msg")
+    def test_transcript_lang(self, tr): assert "string  language" in tr
+    def test_transcript_bcp47(self, tr): assert "BCP-47" in tr
+    def test_response_lang(self, re): assert "string  language" in re
+
+class TestEdgeCases:
+    def test_empty_lang_no_hint(self):
+        lang = "" or "en"; assert lang == "en"
+    def test_lang_flows(self):
+        d: Dict[str,str] = {}; d["p1"] = "fr"
+        assert d.get("p1","en") == "fr"
+    def test_multi_speakers(self):
+        d = {"p1":"fr","p2":"es"}
+        assert d["p1"] == "fr" and d["p2"] == "es"
+    def test_voice_map_code_in_tts(self):
+        src = _read_src("saltybot_social/tts_node.py")
+        assert "voice_map_json" in src and "json.loads" in src

jetson/ros2_ws/src/saltybot_social_msgs/CMakeLists.txt

@@ -38,6 +38,9 @@ rosidl_generate_interfaces(${PROJECT_NAME}
   # Issue #140 — gesture recognition
   "msg/Gesture.msg"
   "msg/GestureArray.msg"
+  # Issue #161 — emotion detection
+  "msg/Expression.msg"
+  "msg/ExpressionArray.msg"
   DEPENDENCIES std_msgs geometry_msgs builtin_interfaces
 )
 

jetson/ros2_ws/src/saltybot_social_msgs/msg/ConversationResponse.msg

@@ -7,3 +7,4 @@ string  text            # Full or partial response text
 string  speaker_id      # Who the response is addressed to
 bool    is_partial       # true = streaming token chunk, false = final response
 int32   turn_id          # Conversation turn counter (for deduplication)
+string  language         # BCP-47 language code for TTS voice selection e.g. "en" "fr" "es"

jetson/ros2_ws/src/saltybot_social_msgs/msg/Expression.msg

@@ -0,0 +1,17 @@
+# Expression.msg — Detected facial expression for one person (Issue #161).
+# Published by emotion_node on /social/faces/expressions
+
+std_msgs/Header header
+
+int32   person_id           # -1 = unidentified; matches PersonState.person_id
+int32   face_id             # matches FaceDetection.face_id
+
+string  emotion             # one of: happy sad angry surprised fearful disgusted neutral
+float32 confidence          # smoothed confidence of the top emotion (0.0–1.0)
+
+float32[7] scores           # per-class softmax scores, order:
+                            #   [0]=happy [1]=sad [2]=angry [3]=surprised
+                            #   [4]=fearful [5]=disgusted [6]=neutral
+
+bool    is_opt_out          # true = this person opted out; no emotion data published
+string  source              # "cnn_trt" | "landmark_heuristic" | "opt_out"

jetson/ros2_ws/src/saltybot_social_msgs/msg/ExpressionArray.msg

@@ -0,0 +1,5 @@
+# ExpressionArray.msg — Batch of detected facial expressions (Issue #161).
+# Published by emotion_node on /social/faces/expressions
+
+std_msgs/Header header
+Expression[] expressions

jetson/ros2_ws/src/saltybot_social_msgs/msg/SpeechTranscript.msg

@@ -8,3 +8,4 @@ string  speaker_id      # e.g. "person_42" or "unknown"
 float32 confidence      # ASR confidence 0..1
 float32 audio_duration  # Duration of the utterance in seconds
 bool    is_partial      # true = intermediate streaming result, false = final
+string  language        # BCP-47 detected language code e.g. "en" "fr" "es" (empty = unknown)

src/jlink.c

@@ -176,6 +176,11 @@ static void dispatch(const uint8_t *payload, uint8_t cmd, uint8_t plen)
         }
         break;
 
+    case JLINK_CMD_SLEEP:
+        /* Payload-less; main loop calls power_mgmt_request_sleep() */
+        jlink_state.sleep_req = 1u;
+        break;
+
     default:
         break;
     }
@@ -290,3 +295,28 @@ void jlink_send_telemetry(const jlink_tlm_status_t *status)
 
     HAL_UART_Transmit(&s_uart, frame, sizeof(frame), 5u);
 }
+
+/* ---- jlink_send_power_telemetry() ---- */
+void jlink_send_power_telemetry(const jlink_tlm_power_t *power)
+{
+    /*
+     * Frame: [STX][LEN][0x81][11 bytes POWER][CRC_hi][CRC_lo][ETX]
+     * LEN = 1 (CMD) + 11 (payload) = 12; total = 17 bytes
+     * At 921600 baud: 17×10/921600 ≈ 0.18 ms — safe to block.
+     */
+    static uint8_t frame[17];
+    const uint8_t  plen = (uint8_t)sizeof(jlink_tlm_power_t);  /* 11 */
+    const uint8_t  len  = 1u + plen;                            /* 12 */
+
+    frame[0] = JLINK_STX;
+    frame[1] = len;
+    frame[2] = JLINK_TLM_POWER;
+    memcpy(&frame[3], power, plen);
+
+    uint16_t crc = crc16_xmodem(&frame[2], len);
+    frame[3 + plen]     = (uint8_t)(crc >> 8);
+    frame[3 + plen + 1] = (uint8_t)(crc & 0xFFu);
+    frame[3 + plen + 2] = JLINK_ETX;
+
+    HAL_UART_Transmit(&s_uart, frame, sizeof(frame), 5u);
+}

src/main.c

@@ -19,6 +19,7 @@
 #include "jlink.h"
 #include "ota.h"
 #include "audio.h"
+#include "power_mgmt.h"
 #include "battery.h"
 #include <math.h>
 #include <string.h>
@@ -149,6 +150,9 @@ int main(void) {
     audio_init();
     audio_play_tone(AUDIO_TONE_STARTUP);
 
+    /* Init power management — STOP-mode sleep/wake, wake EXTIs configured */
+    power_mgmt_init();
+
     /* Init mode manager (RC/autonomous blend; CH6 mode switch) */
     mode_manager_t mode;
     mode_manager_init(&mode);
@@ -183,6 +187,8 @@ int main(void) {
     uint32_t esc_tick = 0;
     uint32_t crsf_telem_tick = 0;  /* CRSF uplink telemetry TX timer */
     uint32_t jlink_tlm_tick = 0;   /* Jetson binary telemetry TX timer */
+    uint32_t pm_tlm_tick    = 0;   /* JLINK_TLM_POWER transmit timer */
+    uint8_t  pm_pwm_phase   = 0;   /* Software PWM counter for sleep LED */
     const float dt = 1.0f / PID_LOOP_HZ;  /* 1ms at 1kHz */
 
     while (1) {
@@ -196,6 +202,23 @@ int main(void) {
         /* Advance audio tone sequencer (non-blocking, call every tick) */
         audio_tick(now);
 
+        /* Sleep LED: software PWM on LED1 (active-low PC15) driven by PM brightness.
+         * pm_pwm_phase rolls over each ms; brightness sets duty cycle 0-255. */
+        pm_pwm_phase++;
+        {
+            uint8_t pm_bright = power_mgmt_led_brightness();
+            if (pm_bright > 0u) {
+                bool led_on = (pm_pwm_phase < pm_bright);
+                HAL_GPIO_WritePin(LED1_PORT, LED1_PIN,
+                    led_on ? GPIO_PIN_RESET : GPIO_PIN_SET);
+            }
+        }
+
+        /* Power manager tick — may block in WFI (STOP mode) when disarmed */
+        if (bal.state != BALANCE_ARMED) {
+            power_mgmt_tick(now);
+        }
+
         /* Mode manager: update RC liveness, CH6 mode selection, blend ramp */
         mode_manager_update(&mode, now);
 
@@ -249,6 +272,16 @@ int main(void) {
              * never returns when disarmed — MCU resets into DFU mode. */
             ota_enter_dfu(bal.state == BALANCE_ARMED);
         }
+        if (jlink_state.sleep_req) {
+            jlink_state.sleep_req = 0u;
+            power_mgmt_request_sleep();
+        }
+
+        /* Power management: CRSF/JLink activity or armed state resets idle timer */
+        if (crsf_is_active(now) || jlink_is_active(now) ||
+                bal.state == BALANCE_ARMED) {
+            power_mgmt_activity();
+        }
 
         /* RC CH5 kill switch: disarm immediately if RC is alive and CH5 off.
          * Applies regardless of active mode (CH5 always has kill authority). */
@@ -424,6 +457,18 @@ int main(void) {
             jlink_send_telemetry(&tlm);
         }
 
+        /* JLINK_TLM_POWER telemetry at PM_TLM_HZ (1 Hz) */
+        if (now - pm_tlm_tick >= (1000u / PM_TLM_HZ)) {
+            pm_tlm_tick = now;
+            jlink_tlm_power_t pow;
+            pow.power_state  = (uint8_t)power_mgmt_state();
+            pow.est_total_ma = power_mgmt_current_ma();
+            pow.est_audio_ma = (uint16_t)(power_mgmt_state() == PM_SLEEPING ? 0u : PM_CURRENT_AUDIO_MA);
+            pow.est_osd_ma   = (uint16_t)(power_mgmt_state() == PM_SLEEPING ? 0u : PM_CURRENT_OSD_MA);
+            pow.idle_ms      = power_mgmt_idle_ms();
+            jlink_send_power_telemetry(&pow);
+        }
+
         /* USB telemetry at 50Hz (only when streaming enabled and calibration done) */
         if (cdc_streaming && imu_calibrated() && now - send_tick >= 20) {
             send_tick = now;

src/power_mgmt.c

@@ -0,0 +1,251 @@
+#include "power_mgmt.h"
+#include "config.h"
+#include "stm32f7xx_hal.h"
+#include <string.h>
+
+/* ---- Internal state ---- */
+static PowerState  s_state             = PM_ACTIVE;
+static uint32_t    s_last_active       = 0;
+static uint32_t    s_fade_start        = 0;
+static bool        s_sleep_req         = false;
+static bool        s_peripherals_gated = false;
+
+/* ---- EXTI wake-source configuration ---- */
+/*
+ * EXTI1  → PA1 (UART4_RX / CRSF): falling edge (UART start bit)
+ * EXTI7  → PB7 (USART1_RX / JLink): falling edge
+ * EXTI4  → PC4 (MPU6000 INT): already configured by mpu6000_init();
+ *           we just ensure IMR bit is set.
+ *
+ * GPIO pins remain in their current AF mode; EXTI is pad-level and
+ * fires independently of the AF setting.
+ */
+static void enable_wake_exti(void)
+{
+    __HAL_RCC_SYSCFG_CLK_ENABLE();
+
+    /* EXTI1: PA1 (UART4_RX) — SYSCFG EXTICR1[7:4] = 0000 (PA) */
+    SYSCFG->EXTICR[0] = (SYSCFG->EXTICR[0] & ~(0xFu << 4)) | (0x0u << 4);
+    EXTI->FTSR |=  (1u << 1);
+    EXTI->RTSR &= ~(1u << 1);
+    EXTI->PR    =  (1u << 1);   /* clear pending */
+    EXTI->IMR  |=  (1u << 1);
+    HAL_NVIC_SetPriority(EXTI1_IRQn, 5, 0);
+    HAL_NVIC_EnableIRQ(EXTI1_IRQn);
+
+    /* EXTI7: PB7 (USART1_RX) — SYSCFG EXTICR2[15:12] = 0001 (PB) */
+    SYSCFG->EXTICR[1] = (SYSCFG->EXTICR[1] & ~(0xFu << 12)) | (0x1u << 12);
+    EXTI->FTSR |=  (1u << 7);
+    EXTI->RTSR &= ~(1u << 7);
+    EXTI->PR    =  (1u << 7);
+    EXTI->IMR  |=  (1u << 7);
+    HAL_NVIC_SetPriority(EXTI9_5_IRQn, 5, 0);
+    HAL_NVIC_EnableIRQ(EXTI9_5_IRQn);
+
+    /* EXTI4: PC4 (MPU6000 INT) — handler in mpu6000.c; just ensure IMR set */
+    EXTI->IMR |= (1u << 4);
+}
+
+static void disable_wake_exti(void)
+{
+    /* Mask UART RX wake EXTIs now that UART peripherals handle traffic */
+    EXTI->IMR &= ~(1u << 1);
+    EXTI->IMR &= ~(1u << 7);
+    /* Leave EXTI4 (IMU data-ready) always unmasked */
+}
+
+/* ---- Peripheral clock gating ---- */
+/*
+ * Clock-only gate (no force-reset): peripheral register state is preserved.
+ * On re-enable, DMA circular transfers resume without reinitialisation.
+ */
+static void gate_peripherals(void)
+{
+    if (s_peripherals_gated) return;
+    __HAL_RCC_SPI3_CLK_DISABLE();    /* I2S3 / audio amplifier */
+    __HAL_RCC_SPI2_CLK_DISABLE();    /* OSD MAX7456 */
+    __HAL_RCC_USART6_CLK_DISABLE();  /* legacy Jetson CDC */
+    __HAL_RCC_UART5_CLK_DISABLE();   /* debug UART */
+    s_peripherals_gated = true;
+}
+
+static void ungate_peripherals(void)
+{
+    if (!s_peripherals_gated) return;
+    __HAL_RCC_SPI3_CLK_ENABLE();
+    __HAL_RCC_SPI2_CLK_ENABLE();
+    __HAL_RCC_USART6_CLK_ENABLE();
+    __HAL_RCC_UART5_CLK_ENABLE();
+    s_peripherals_gated = false;
+}
+
+/* ---- PLL clock restore after STOP mode ---- */
+/*
+ * After STOP wakeup SYSCLK = HSI (16 MHz). Re-lock PLL for 216 MHz.
+ * PLLM=8, PLLN=216, PLLP=2, PLLQ=9 — STM32F722 @ 216 MHz, HSI source.
+ *
+ * HAL_RCC_ClockConfig() calls HAL_InitTick() which resets uwTick to 0;
+ * save and restore it so existing timeouts remain valid across sleep.
+ */
+extern volatile uint32_t uwTick;
+
+static void restore_clocks(void)
+{
+    uint32_t saved_tick = uwTick;
+
+    RCC_OscInitTypeDef osc = {0};
+    osc.OscillatorType      = RCC_OSCILLATORTYPE_HSI;
+    osc.HSIState            = RCC_HSI_ON;
+    osc.HSICalibrationValue = RCC_HSICALIBRATION_DEFAULT;
+    osc.PLL.PLLState        = RCC_PLL_ON;
+    osc.PLL.PLLSource       = RCC_PLLSOURCE_HSI;
+    osc.PLL.PLLM            = 8;
+    osc.PLL.PLLN            = 216;
+    osc.PLL.PLLP            = RCC_PLLP_DIV2;
+    osc.PLL.PLLQ            = 9;
+    HAL_RCC_OscConfig(&osc);
+
+    RCC_ClkInitTypeDef clk = {0};
+    clk.ClockType      = RCC_CLOCKTYPE_HCLK | RCC_CLOCKTYPE_SYSCLK |
+                         RCC_CLOCKTYPE_PCLK1 | RCC_CLOCKTYPE_PCLK2;
+    clk.SYSCLKSource   = RCC_SYSCLKSOURCE_PLLCLK;
+    clk.AHBCLKDivider  = RCC_SYSCLK_DIV1;
+    clk.APB1CLKDivider = RCC_HCLK_DIV4;   /* 54 MHz */
+    clk.APB2CLKDivider = RCC_HCLK_DIV2;   /* 108 MHz */
+    HAL_RCC_ClockConfig(&clk, FLASH_LATENCY_7);
+
+    uwTick = saved_tick;  /* restore — HAL_InitTick() reset it to 0 */
+}
+
+/* ---- EXTI IRQ handlers (wake-only: clear pending bit and return) ---- */
+/*
+ * These handlers fire once on wakeup. After restore_clocks() the respective
+ * UART peripherals resume normal DMA/IDLE-interrupt operation.
+ *
+ * NOTE: If EXTI9_5_IRQHandler is already defined elsewhere in the project,
+ * merge that handler with this one.
+ */
+void EXTI1_IRQHandler(void)
+{
+    if (EXTI->PR & (1u << 1)) EXTI->PR = (1u << 1);
+}
+
+void EXTI9_5_IRQHandler(void)
+{
+    /* Clear any pending EXTI5-9 lines (PB7 = EXTI7 is our primary wake) */
+    uint32_t pr = EXTI->PR & 0x3E0u;
+    if (pr) EXTI->PR = pr;
+}
+
+/* ---- LED brightness (integer arithmetic, no float, called from main loop) ---- */
+/*
+ * Triangle wave: 0→255→0 over PM_LED_PERIOD_MS.
+ * Only active during PM_SLEEP_PENDING; returns 0 otherwise.
+ */
+uint8_t power_mgmt_led_brightness(void)
+{
+    if (s_state != PM_SLEEP_PENDING) return 0u;
+
+    uint32_t phase = (HAL_GetTick() - s_fade_start) % PM_LED_PERIOD_MS;
+    uint32_t half  = PM_LED_PERIOD_MS / 2u;
+    if (phase < half)
+        return (uint8_t)(phase * 255u / half);
+    else
+        return (uint8_t)((PM_LED_PERIOD_MS - phase) * 255u / half);
+}
+
+/* ---- Current estimate ---- */
+uint16_t power_mgmt_current_ma(void)
+{
+    if (s_state == PM_SLEEPING)
+        return (uint16_t)PM_CURRENT_STOP_MA;
+    uint16_t ma = (uint16_t)PM_CURRENT_BASE_MA;
+    if (!s_peripherals_gated) {
+        ma += (uint16_t)(PM_CURRENT_AUDIO_MA + PM_CURRENT_OSD_MA +
+                         PM_CURRENT_DEBUG_MA);
+    }
+    return ma;
+}
+
+/* ---- Idle elapsed ---- */
+uint32_t power_mgmt_idle_ms(void)
+{
+    return HAL_GetTick() - s_last_active;
+}
+
+/* ---- Public API ---- */
+void power_mgmt_init(void)
+{
+    s_state             = PM_ACTIVE;
+    s_last_active       = HAL_GetTick();
+    s_fade_start        = 0;
+    s_sleep_req         = false;
+    s_peripherals_gated = false;
+    enable_wake_exti();
+}
+
+void power_mgmt_activity(void)
+{
+    s_last_active = HAL_GetTick();
+    if (s_state != PM_ACTIVE) {
+        s_sleep_req = false;
+        s_state     = PM_WAKING;  /* resolved to PM_ACTIVE on next tick() */
+    }
+}
+
+void power_mgmt_request_sleep(void)
+{
+    s_sleep_req = true;
+}
+
+PowerState power_mgmt_state(void)
+{
+    return s_state;
+}
+
+PowerState power_mgmt_tick(uint32_t now_ms)
+{
+    switch (s_state) {
+
+    case PM_ACTIVE:
+        if (s_sleep_req || (now_ms - s_last_active) >= PM_IDLE_TIMEOUT_MS) {
+            s_sleep_req  = false;
+            s_fade_start = now_ms;
+            s_state      = PM_SLEEP_PENDING;
+        }
+        break;
+
+    case PM_SLEEP_PENDING:
+        if ((now_ms - s_fade_start) >= PM_FADE_MS) {
+            gate_peripherals();
+            enable_wake_exti();
+            s_state = PM_SLEEPING;
+
+            /* Feed IWDG: wakeup <10 ms << WATCHDOG_TIMEOUT_MS (50 ms) */
+            IWDG->KR = 0xAAAAu;
+
+            /* === STOP MODE ENTRY — execution resumes here on EXTI wake === */
+            HAL_PWR_EnterSTOPMode(PWR_LOWPOWERREGULATOR_ON, PWR_STOPENTRY_WFI);
+            /* === WAKEUP POINT (< 10 ms latency) === */
+
+            restore_clocks();
+            ungate_peripherals();
+            disable_wake_exti();
+            s_last_active = HAL_GetTick();
+            s_state       = PM_ACTIVE;
+        }
+        break;
+
+    case PM_SLEEPING:
+        /* Unreachable: WFI is inline in PM_SLEEP_PENDING above */
+        break;
+
+    case PM_WAKING:
+        /* Set by power_mgmt_activity() during SLEEP_PENDING/SLEEPING */
+        ungate_peripherals();
+        s_state = PM_ACTIVE;
+        break;
+    }
+
+    return s_state;
+}

test/test_power_mgmt.py

@@ -0,0 +1,567 @@
+"""
+test_power_mgmt.py — unit tests for Issue #178 power management module.
+
+Models the PM state machine, LED brightness, peripheral gating, current
+estimates, JLink protocol extension, and hardware timing budgets in Python.
+"""
+
+import struct
+import pytest
+
+# ---------------------------------------------------------------------------
+# Constants (mirror config.h / power_mgmt.h)
+# ---------------------------------------------------------------------------
+PM_IDLE_TIMEOUT_MS  = 30_000
+PM_FADE_MS          = 3_000
+PM_LED_PERIOD_MS    = 2_000
+
+PM_CURRENT_BASE_MA  = 30   # SPI1(IMU) + UART4(CRSF) + USART1(JLink) + core
+PM_CURRENT_AUDIO_MA =  8   # I2S3 + amp quiescent
+PM_CURRENT_OSD_MA   =  5   # SPI2 OSD MAX7456
+PM_CURRENT_DEBUG_MA =  1   # UART5 + USART6
+PM_CURRENT_STOP_MA  =  1   # MCU in STOP mode (< 1 mA)
+
+PM_CURRENT_ACTIVE_ALL = (PM_CURRENT_BASE_MA + PM_CURRENT_AUDIO_MA +
+                         PM_CURRENT_OSD_MA + PM_CURRENT_DEBUG_MA)
+
+# JLink additions
+JLINK_STX           = 0x02
+JLINK_ETX           = 0x03
+JLINK_CMD_SLEEP     = 0x09
+JLINK_TLM_STATUS    = 0x80
+JLINK_TLM_POWER     = 0x81
+
+# Power states
+PM_ACTIVE        = 0
+PM_SLEEP_PENDING = 1
+PM_SLEEPING      = 2
+PM_WAKING        = 3
+
+# WATCHDOG_TIMEOUT_MS from config.h
+WATCHDOG_TIMEOUT_MS = 50
+
+
+# ---------------------------------------------------------------------------
+# CRC-16/XModem helper (poly 0x1021, init 0x0000)
+# ---------------------------------------------------------------------------
+def crc16_xmodem(data: bytes) -> int:
+    crc = 0x0000
+    for b in data:
+        crc ^= b << 8
+        for _ in range(8):
+            crc = (crc << 1) ^ 0x1021 if crc & 0x8000 else crc << 1
+        crc &= 0xFFFF
+    return crc
+
+
+def build_frame(cmd: int, payload: bytes = b"") -> bytes:
+    data = bytes([cmd]) + payload
+    length = len(data)
+    crc = crc16_xmodem(data)
+    return bytes([JLINK_STX, length, *data, crc >> 8, crc & 0xFF, JLINK_ETX])
+
+
+# ---------------------------------------------------------------------------
+# Python model of the power_mgmt state machine (mirrors power_mgmt.c)
+# ---------------------------------------------------------------------------
+class PowerMgmtSim:
+    def __init__(self, now: int = 0):
+        self.state              = PM_ACTIVE
+        self.last_active        = now
+        self.fade_start         = 0
+        self.sleep_req          = False
+        self.peripherals_gated  = False
+
+    def activity(self, now: int) -> None:
+        self.last_active = now
+        if self.state != PM_ACTIVE:
+            self.sleep_req = False
+            self.state = PM_WAKING
+
+    def request_sleep(self) -> None:
+        self.sleep_req = True
+
+    def led_brightness(self, now: int) -> int:
+        if self.state != PM_SLEEP_PENDING:
+            return 0
+        phase = (now - self.fade_start) % PM_LED_PERIOD_MS
+        half  = PM_LED_PERIOD_MS // 2
+        if phase < half:
+            return phase * 255 // half
+        else:
+            return (PM_LED_PERIOD_MS - phase) * 255 // half
+
+    def current_ma(self) -> int:
+        if self.state == PM_SLEEPING:
+            return PM_CURRENT_STOP_MA
+        ma = PM_CURRENT_BASE_MA
+        if not self.peripherals_gated:
+            ma += PM_CURRENT_AUDIO_MA + PM_CURRENT_OSD_MA + PM_CURRENT_DEBUG_MA
+        return ma
+
+    def idle_ms(self, now: int) -> int:
+        return now - self.last_active
+
+    def tick(self, now: int) -> int:
+        if self.state == PM_ACTIVE:
+            if self.sleep_req or (now - self.last_active) >= PM_IDLE_TIMEOUT_MS:
+                self.sleep_req  = False
+                self.fade_start = now
+                self.state      = PM_SLEEP_PENDING
+
+        elif self.state == PM_SLEEP_PENDING:
+            if (now - self.fade_start) >= PM_FADE_MS:
+                self.peripherals_gated = True
+                self.state = PM_SLEEPING
+                # In firmware: WFI blocks here; in test we skip to simulate_wake
+
+        elif self.state == PM_WAKING:
+            self.peripherals_gated = False
+            self.state = PM_ACTIVE
+
+        return self.state
+
+    def simulate_wake(self, now: int) -> None:
+        """Simulate EXTI wakeup from STOP mode (models HAL_PWR_EnterSTOPMode return)."""
+        if self.state == PM_SLEEPING:
+            self.peripherals_gated = False
+            self.last_active = now
+            self.state = PM_ACTIVE
+
+
+# ---------------------------------------------------------------------------
+# Tests: Idle timer
+# ---------------------------------------------------------------------------
+class TestIdleTimer:
+    def test_stays_active_before_timeout(self):
+        pm = PowerMgmtSim(now=0)
+        for t in range(0, PM_IDLE_TIMEOUT_MS, 1000):
+            assert pm.tick(t) == PM_ACTIVE
+
+    def test_enters_sleep_pending_at_timeout(self):
+        pm = PowerMgmtSim(now=0)
+        assert pm.tick(PM_IDLE_TIMEOUT_MS - 1) == PM_ACTIVE
+        assert pm.tick(PM_IDLE_TIMEOUT_MS)      == PM_SLEEP_PENDING
+
+    def test_activity_resets_idle_timer(self):
+        pm = PowerMgmtSim(now=0)
+        pm.tick(PM_IDLE_TIMEOUT_MS - 1000)
+        pm.activity(PM_IDLE_TIMEOUT_MS - 1000)   # reset at T=29000
+        assert pm.tick(PM_IDLE_TIMEOUT_MS) == PM_ACTIVE        # 1 s since reset
+        assert pm.tick(PM_IDLE_TIMEOUT_MS - 1000 + PM_IDLE_TIMEOUT_MS) == PM_SLEEP_PENDING
+
+    def test_idle_ms_increases_monotonically(self):
+        pm = PowerMgmtSim(now=0)
+        assert pm.idle_ms(0)     == 0
+        assert pm.idle_ms(5000)  == 5000
+        assert pm.idle_ms(29999) == 29999
+
+    def test_idle_ms_resets_on_activity(self):
+        pm = PowerMgmtSim(now=0)
+        pm.activity(10000)
+        assert pm.idle_ms(10500) == 500
+
+    def test_30s_timeout_matches_spec(self):
+        assert PM_IDLE_TIMEOUT_MS == 30_000
+
+
+# ---------------------------------------------------------------------------
+# Tests: State machine transitions
+# ---------------------------------------------------------------------------
+class TestStateMachine:
+    def test_sleep_req_bypasses_idle_timer(self):
+        pm = PowerMgmtSim(now=0)
+        pm.activity(0)
+        pm.request_sleep()
+        assert pm.tick(500) == PM_SLEEP_PENDING
+
+    def test_fade_complete_enters_sleeping(self):
+        pm = PowerMgmtSim(now=0)
+        pm.tick(PM_IDLE_TIMEOUT_MS)                 # → SLEEP_PENDING
+        assert pm.tick(PM_IDLE_TIMEOUT_MS + PM_FADE_MS) == PM_SLEEPING
+
+    def test_fade_not_complete_stays_pending(self):
+        pm = PowerMgmtSim(now=0)
+        pm.tick(PM_IDLE_TIMEOUT_MS)
+        assert pm.tick(PM_IDLE_TIMEOUT_MS + PM_FADE_MS - 1) == PM_SLEEP_PENDING
+
+    def test_wake_from_stop_returns_active(self):
+        pm = PowerMgmtSim(now=0)
+        pm.tick(PM_IDLE_TIMEOUT_MS)
+        pm.tick(PM_IDLE_TIMEOUT_MS + PM_FADE_MS)    # → SLEEPING
+        pm.simulate_wake(PM_IDLE_TIMEOUT_MS + PM_FADE_MS + 5)
+        assert pm.state == PM_ACTIVE
+
+    def test_activity_during_sleep_pending_aborts(self):
+        pm = PowerMgmtSim(now=0)
+        pm.tick(PM_IDLE_TIMEOUT_MS)                 # → SLEEP_PENDING
+        pm.activity(PM_IDLE_TIMEOUT_MS + 100)       # abort
+        assert pm.state == PM_WAKING
+        pm.tick(PM_IDLE_TIMEOUT_MS + 101)
+        assert pm.state == PM_ACTIVE
+
+    def test_activity_during_sleeping_aborts(self):
+        pm = PowerMgmtSim(now=0)
+        pm.tick(PM_IDLE_TIMEOUT_MS)
+        pm.tick(PM_IDLE_TIMEOUT_MS + PM_FADE_MS)    # → SLEEPING
+        pm.activity(PM_IDLE_TIMEOUT_MS + PM_FADE_MS + 3)
+        assert pm.state == PM_WAKING
+        pm.tick(PM_IDLE_TIMEOUT_MS + PM_FADE_MS + 4)
+        assert pm.state == PM_ACTIVE
+
+    def test_waking_resolves_on_next_tick(self):
+        pm = PowerMgmtSim(now=0)
+        pm.state = PM_WAKING
+        pm.tick(1000)
+        assert pm.state == PM_ACTIVE
+
+    def test_full_sleep_wake_cycle(self):
+        pm = PowerMgmtSim(now=0)
+        # 1. Active
+        assert pm.tick(100) == PM_ACTIVE
+        # 2. Idle → sleep pending
+        assert pm.tick(PM_IDLE_TIMEOUT_MS) == PM_SLEEP_PENDING
+        # 3. Fade → sleeping
+        assert pm.tick(PM_IDLE_TIMEOUT_MS + PM_FADE_MS) == PM_SLEEPING
+        # 4. EXTI wake → active
+        pm.simulate_wake(PM_IDLE_TIMEOUT_MS + PM_FADE_MS + 8)
+        assert pm.state == PM_ACTIVE
+
+    def test_multiple_sleep_wake_cycles(self):
+        pm = PowerMgmtSim(now=0)
+        base = 0
+        for _ in range(3):
+            pm.activity(base)
+            pm.tick(base + PM_IDLE_TIMEOUT_MS)
+            pm.tick(base + PM_IDLE_TIMEOUT_MS + PM_FADE_MS)
+            pm.simulate_wake(base + PM_IDLE_TIMEOUT_MS + PM_FADE_MS + 5)
+            assert pm.state == PM_ACTIVE
+            base += PM_IDLE_TIMEOUT_MS + PM_FADE_MS + 10
+
+
+# ---------------------------------------------------------------------------
+# Tests: Peripheral gating
+# ---------------------------------------------------------------------------
+class TestPeripheralGating:
+    GATED   = {'SPI3_I2S3', 'SPI2_OSD', 'USART6', 'UART5_DEBUG'}
+    ACTIVE  = {'SPI1_IMU', 'UART4_CRSF', 'USART1_JLINK', 'I2C1_BARO'}
+
+    def test_gated_set_has_four_peripherals(self):
+        assert len(self.GATED) == 4
+
+    def test_no_overlap_between_gated_and_active(self):
+        assert not (self.GATED & self.ACTIVE)
+
+    def test_crsf_uart_not_gated(self):
+        assert not any('UART4' in p or 'CRSF' in p for p in self.GATED)
+
+    def test_jlink_uart_not_gated(self):
+        assert not any('USART1' in p or 'JLINK' in p for p in self.GATED)
+
+    def test_imu_spi_not_gated(self):
+        assert not any('SPI1' in p or 'IMU' in p for p in self.GATED)
+
+    def test_peripherals_gated_on_sleep_entry(self):
+        pm = PowerMgmtSim(now=0)
+        assert not pm.peripherals_gated
+        pm.tick(PM_IDLE_TIMEOUT_MS)
+        pm.tick(PM_IDLE_TIMEOUT_MS + PM_FADE_MS)    # → SLEEPING
+        assert pm.peripherals_gated
+
+    def test_peripherals_ungated_on_wake(self):
+        pm = PowerMgmtSim(now=0)
+        pm.tick(PM_IDLE_TIMEOUT_MS)
+        pm.tick(PM_IDLE_TIMEOUT_MS + PM_FADE_MS)
+        pm.simulate_wake(PM_IDLE_TIMEOUT_MS + PM_FADE_MS + 5)
+        assert not pm.peripherals_gated
+
+    def test_peripherals_not_gated_in_sleep_pending(self):
+        pm = PowerMgmtSim(now=0)
+        pm.tick(PM_IDLE_TIMEOUT_MS)                 # → SLEEP_PENDING
+        assert not pm.peripherals_gated
+
+    def test_peripherals_ungated_if_activity_during_pending(self):
+        pm = PowerMgmtSim(now=0)
+        pm.tick(PM_IDLE_TIMEOUT_MS)
+        pm.activity(PM_IDLE_TIMEOUT_MS + 100)
+        pm.tick(PM_IDLE_TIMEOUT_MS + 101)
+        assert not pm.peripherals_gated
+
+
+# ---------------------------------------------------------------------------
+# Tests: LED brightness
+# ---------------------------------------------------------------------------
+class TestLedBrightness:
+    def test_zero_when_active(self):
+        pm = PowerMgmtSim(now=0)
+        assert pm.led_brightness(5000) == 0
+
+    def test_zero_when_sleeping(self):
+        pm = PowerMgmtSim(now=0)
+        pm.tick(PM_IDLE_TIMEOUT_MS)
+        pm.tick(PM_IDLE_TIMEOUT_MS + PM_FADE_MS)    # → SLEEPING
+        assert pm.led_brightness(PM_IDLE_TIMEOUT_MS + PM_FADE_MS + 100) == 0
+
+    def test_zero_when_waking(self):
+        pm = PowerMgmtSim(now=0)
+        pm.state = PM_WAKING
+        assert pm.led_brightness(1000) == 0
+
+    def test_zero_at_phase_start(self):
+        pm = PowerMgmtSim(now=0)
+        pm.tick(PM_IDLE_TIMEOUT_MS)                 # fade_start = PM_IDLE_TIMEOUT_MS
+        assert pm.led_brightness(PM_IDLE_TIMEOUT_MS) == 0
+
+    def test_max_at_half_period(self):
+        pm = PowerMgmtSim(now=0)
+        pm.tick(PM_IDLE_TIMEOUT_MS)
+        t = PM_IDLE_TIMEOUT_MS + PM_LED_PERIOD_MS // 2
+        assert pm.led_brightness(t) == 255
+
+    def test_zero_at_full_period(self):
+        pm = PowerMgmtSim(now=0)
+        pm.tick(PM_IDLE_TIMEOUT_MS)
+        t = PM_IDLE_TIMEOUT_MS + PM_LED_PERIOD_MS
+        assert pm.led_brightness(t) == 0
+
+    def test_symmetric_about_half_period(self):
+        pm = PowerMgmtSim(now=0)
+        pm.tick(PM_IDLE_TIMEOUT_MS)
+        quarter       = PM_LED_PERIOD_MS // 4
+        three_quarter = 3 * PM_LED_PERIOD_MS // 4
+        b1 = pm.led_brightness(PM_IDLE_TIMEOUT_MS + quarter)
+        b2 = pm.led_brightness(PM_IDLE_TIMEOUT_MS + three_quarter)
+        assert abs(b1 - b2) <= 1    # allow 1 LSB for integer division
+
+    def test_range_0_to_255(self):
+        pm = PowerMgmtSim(now=0)
+        pm.tick(PM_IDLE_TIMEOUT_MS)
+        for dt in range(0, PM_LED_PERIOD_MS * 3, 37):
+            b = pm.led_brightness(PM_IDLE_TIMEOUT_MS + dt)
+            assert 0 <= b <= 255
+
+    def test_repeats_over_multiple_periods(self):
+        pm = PowerMgmtSim(now=0)
+        pm.tick(PM_IDLE_TIMEOUT_MS)
+        # Sample at same phase in periods 0, 1, 2 — should be equal
+        phase = PM_LED_PERIOD_MS // 3
+        b0 = pm.led_brightness(PM_IDLE_TIMEOUT_MS + phase)
+        b1 = pm.led_brightness(PM_IDLE_TIMEOUT_MS + PM_LED_PERIOD_MS + phase)
+        b2 = pm.led_brightness(PM_IDLE_TIMEOUT_MS + 2 * PM_LED_PERIOD_MS + phase)
+        assert b0 == b1 == b2
+
+    def test_period_is_2s(self):
+        assert PM_LED_PERIOD_MS == 2000
+
+
+# ---------------------------------------------------------------------------
+# Tests: Power / current estimates
+# ---------------------------------------------------------------------------
+class TestPowerEstimates:
+    def test_active_includes_all_subsystems(self):
+        pm = PowerMgmtSim(now=0)
+        assert pm.current_ma() == PM_CURRENT_ACTIVE_ALL
+
+    def test_sleeping_returns_stop_ma(self):
+        pm = PowerMgmtSim(now=0)
+        pm.tick(PM_IDLE_TIMEOUT_MS)
+        pm.tick(PM_IDLE_TIMEOUT_MS + PM_FADE_MS)    # → SLEEPING
+        assert pm.current_ma() == PM_CURRENT_STOP_MA
+
+    def test_gated_returns_base_only(self):
+        pm = PowerMgmtSim(now=0)
+        pm.peripherals_gated = True
+        assert pm.current_ma() == PM_CURRENT_BASE_MA
+
+    def test_stop_current_less_than_active(self):
+        assert PM_CURRENT_STOP_MA < PM_CURRENT_ACTIVE_ALL
+
+    def test_stop_current_at_most_1ma(self):
+        assert PM_CURRENT_STOP_MA <= 1
+
+    def test_active_current_reasonable(self):
+        # Should be < 100 mA (just MCU + peripherals, no motors)
+        assert PM_CURRENT_ACTIVE_ALL < 100
+
+    def test_audio_subsystem_estimate(self):
+        assert PM_CURRENT_AUDIO_MA > 0
+
+    def test_osd_subsystem_estimate(self):
+        assert PM_CURRENT_OSD_MA > 0
+
+    def test_total_equals_sum_of_parts(self):
+        total = (PM_CURRENT_BASE_MA + PM_CURRENT_AUDIO_MA +
+                 PM_CURRENT_OSD_MA + PM_CURRENT_DEBUG_MA)
+        assert total == PM_CURRENT_ACTIVE_ALL
+
+
+# ---------------------------------------------------------------------------
+# Tests: JLink protocol extension
+# ---------------------------------------------------------------------------
+class TestJlinkProtocol:
+    def test_sleep_cmd_id(self):
+        assert JLINK_CMD_SLEEP == 0x09
+
+    def test_sleep_follows_audio_cmd(self):
+        JLINK_CMD_AUDIO = 0x08
+        assert JLINK_CMD_SLEEP == JLINK_CMD_AUDIO + 1
+
+    def test_power_tlm_id(self):
+        assert JLINK_TLM_POWER == 0x81
+
+    def test_power_tlm_follows_status_tlm(self):
+        assert JLINK_TLM_POWER == JLINK_TLM_STATUS + 1
+
+    def test_sleep_frame_length(self):
+        # SLEEP has no payload: STX(1)+LEN(1)+CMD(1)+CRC(2)+ETX(1) = 6
+        frame = build_frame(JLINK_CMD_SLEEP)
+        assert len(frame) == 6
+
+    def test_sleep_frame_sentinels(self):
+        frame = build_frame(JLINK_CMD_SLEEP)
+        assert frame[0] == JLINK_STX
+        assert frame[-1] == JLINK_ETX
+
+    def test_sleep_frame_len_field(self):
+        frame = build_frame(JLINK_CMD_SLEEP)
+        assert frame[1] == 1    # LEN = 1 (CMD only, no payload)
+
+    def test_sleep_frame_cmd_byte(self):
+        frame = build_frame(JLINK_CMD_SLEEP)
+        assert frame[2] == JLINK_CMD_SLEEP
+
+    def test_sleep_frame_crc_valid(self):
+        frame = build_frame(JLINK_CMD_SLEEP)
+        calc  = crc16_xmodem(bytes([JLINK_CMD_SLEEP]))
+        rx    = (frame[-3] << 8) | frame[-2]
+        assert rx == calc
+
+    def test_power_tlm_frame_length(self):
+        # jlink_tlm_power_t = 11 bytes
+        # Frame: STX(1)+LEN(1)+CMD(1)+payload(11)+CRC(2)+ETX(1) = 17
+        POWER_TLM_PAYLOAD_LEN = 11
+        expected = 1 + 1 + 1 + POWER_TLM_PAYLOAD_LEN + 2 + 1
+        assert expected == 17
+
+    def test_power_tlm_payload_struct(self):
+        """jlink_tlm_power_t: u8 power_state, u16 est_total_ma,
+        u16 est_audio_ma, u16 est_osd_ma, u32 idle_ms = 11 bytes."""
+        fmt  = "<BHHHI"
+        size = struct.calcsize(fmt)
+        assert size == 11
+
+    def test_power_tlm_frame_crc_valid(self):
+        power_state  = PM_ACTIVE
+        est_total_ma = PM_CURRENT_ACTIVE_ALL
+        est_audio_ma = PM_CURRENT_AUDIO_MA
+        est_osd_ma   = PM_CURRENT_OSD_MA
+        idle_ms      = 5000
+        payload = struct.pack("<BHHHI", power_state, est_total_ma,
+                              est_audio_ma, est_osd_ma, idle_ms)
+        frame = build_frame(JLINK_TLM_POWER, payload)
+        assert frame[0] == JLINK_STX
+        assert frame[-1] == JLINK_ETX
+        data_for_crc = bytes([JLINK_TLM_POWER]) + payload
+        expected_crc = crc16_xmodem(data_for_crc)
+        rx_crc = (frame[-3] << 8) | frame[-2]
+        assert rx_crc == expected_crc
+
+
+# ---------------------------------------------------------------------------
+# Tests: Wake latency and IWDG budget
+# ---------------------------------------------------------------------------
+class TestWakeLatencyBudget:
+    # STM32F722 STOP-mode wakeup: HSI ready ~2 ms + PLL lock ~2 ms ≈ 4 ms
+    ESTIMATED_WAKE_MS = 10   # conservative upper bound
+
+    def test_wake_latency_within_50ms(self):
+        assert self.ESTIMATED_WAKE_MS < WATCHDOG_TIMEOUT_MS
+
+    def test_watchdog_timeout_is_50ms(self):
+        assert WATCHDOG_TIMEOUT_MS == 50
+
+    def test_iwdg_feed_before_wfi_is_safe(self):
+        # Time from IWDG feed to next feed after wake:
+        # ~1 ms (loop overhead) + ESTIMATED_WAKE_MS + ~1 ms = ~12 ms
+        time_from_feed_to_next_ms = 1 + self.ESTIMATED_WAKE_MS + 1
+        assert time_from_feed_to_next_ms < WATCHDOG_TIMEOUT_MS
+
+    def test_fade_ms_positive(self):
+        assert PM_FADE_MS > 0
+
+    def test_fade_ms_less_than_idle_timeout(self):
+        assert PM_FADE_MS < PM_IDLE_TIMEOUT_MS
+
+    def test_stop_mode_wake_much_less_than_50ms(self):
+        # PLL startup on STM32F7: HSI on (0 ms, already running) +
+        # PLL lock ~2 ms + SysTick re-init ~0.1 ms ≈ 3 ms
+        pll_lock_ms = 3
+        overhead_ms = 1
+        total_ms    = pll_lock_ms + overhead_ms
+        assert total_ms < 50
+
+    def test_wake_exti_sources_count(self):
+        """Three wake sources: EXTI1 (CRSF), EXTI7 (JLink), EXTI4 (IMU)."""
+        wake_sources = ['EXTI1_UART4_CRSF', 'EXTI7_USART1_JLINK', 'EXTI4_IMU_INT']
+        assert len(wake_sources) == 3
+
+    def test_uwTick_must_be_restored_after_stop(self):
+        """HAL_RCC_ClockConfig resets uwTick to 0; restore_clocks() saves it."""
+        # Verify the pattern: save uwTick → HAL calls → restore uwTick
+        saved_tick = 12345
+        # Simulate HAL_InitTick() resetting to 0
+        uw_tick_after_hal = 0
+        restored = saved_tick  # power_mgmt.c: uwTick = saved_tick
+        assert restored == saved_tick
+        assert uw_tick_after_hal != saved_tick  # HAL reset it
+
+
+# ---------------------------------------------------------------------------
+# Tests: Hardware constants
+# ---------------------------------------------------------------------------
+class TestHardwareConstants:
+    def test_pll_params_for_216mhz(self):
+        """PLLM=8, PLLN=216, PLLP=2 → VCO=216*2=432 MHz, SYSCLK=216 MHz."""
+        HSI_MHZ  = 16
+        PLLM     =  8
+        PLLN     = 216
+        PLLP     =  2
+        vco_mhz  = HSI_MHZ / PLLM * PLLN
+        sysclk   = vco_mhz / PLLP
+        assert sysclk == pytest.approx(216.0, rel=1e-6)
+
+    def test_apb1_54mhz(self):
+        """APB1 = SYSCLK / 4 = 54 MHz."""
+        assert 216 / 4 == 54
+
+    def test_apb2_108mhz(self):
+        """APB2 = SYSCLK / 2 = 108 MHz."""
+        assert 216 / 2 == 108
+
+    def test_flash_latency_7_required_at_216mhz(self):
+        """STM32F7 at 2.7-3.3 V: 7 wait states for 210-216 MHz."""
+        FLASH_LATENCY = 7
+        assert FLASH_LATENCY == 7
+
+    def test_exti1_for_pa1(self):
+        """SYSCFG EXTICR1[7:4] = 0x0 selects PA for EXTI1."""
+        PA_SOURCE = 0x0
+        assert PA_SOURCE == 0x0
+
+    def test_exti7_for_pb7(self):
+        """SYSCFG EXTICR2[15:12] = 0x1 selects PB for EXTI7."""
+        PB_SOURCE = 0x1
+        assert PB_SOURCE == 0x1
+
+    def test_exticr_indices(self):
+        """EXTI1 → EXTICR[0], EXTI7 → EXTICR[1]."""
+        assert 1 // 4 == 0   # EXTI1 is in EXTICR[0]
+        assert 7 // 4 == 1   # EXTI7 is in EXTICR[1]
+
+    def test_exti7_shift_in_exticr2(self):
+        """EXTI7 field is at bits [15:12] of EXTICR[1] → shift = (7%4)*4 = 12."""
+        shift = (7 % 4) * 4
+        assert shift == 12
+
+    def test_idle_timeout_30s(self):
+        assert PM_IDLE_TIMEOUT_MS == 30_000

ui/social-bot/src/App.jsx

@@ -5,13 +5,16 @@
  *   Status | Faces | Conversation | Personality | Navigation
  *
  * Telemetry tabs (issue #126):
- *   IMU | Battery | Motors | Map | Control | Health
+ *   IMU | Battery | Motors | Map | Control | Health | Cameras
  *
  * Fleet tabs (issue #139):
  *   Fleet (self-contained via useFleet)
  *
  * Mission tabs (issue #145):
  *   Missions (waypoint editor, route builder, geofence, schedule, execute)
+ *
+ * Camera viewer (issue #177):
+ *   CSI × 4 + D435i RGB/depth + panoramic, detection overlays, recording
  */
 
 import { useState, useCallback } from 'react';
@@ -41,6 +44,9 @@ import { MissionPlanner } from './components/MissionPlanner.jsx';
 // Settings panel (issue #160)
 import { SettingsPanel } from './components/SettingsPanel.jsx';
 
+// Camera viewer (issue #177)
+import { CameraViewer } from './components/CameraViewer.jsx';
+
 const TAB_GROUPS = [
   {
     label: 'SOCIAL',
@@ -63,6 +69,7 @@ const TAB_GROUPS = [
       { id: 'map',     label: 'Map',     },
       { id: 'control', label: 'Control', },
       { id: 'health',  label: 'Health',  },
+      { id: 'cameras', label: 'Cameras', },
     ],
   },
   {
@@ -206,6 +213,7 @@ export default function App() {
         {activeTab === 'map'     && <MapViewer    subscribe={subscribe} />}
         {activeTab === 'control' && <ControlMode  subscribe={subscribe} />}
         {activeTab === 'health'  && <SystemHealth subscribe={subscribe} />}
+        {activeTab === 'cameras' && <CameraViewer subscribe={subscribe} />}
 
         {activeTab === 'fleet'    && <FleetPanel />}
         {activeTab === 'missions' && <MissionPlanner />}

ui/social-bot/src/components/CameraViewer.jsx

@@ -0,0 +1,671 @@
+/**
+ * CameraViewer.jsx — Live camera stream viewer (Issue #177).
+ *
+ * Features:
+ *   - 7 cameras: front/left/rear/right (CSI), D435i RGB/depth, panoramic
+ *   - Detection overlays: face boxes + names, gesture icons, scene object labels
+ *   - 360° panoramic equirect viewer with mouse drag pan
+ *   - One-click recording (MP4/WebM) + download
+ *   - Snapshot to PNG with annotations + timestamp
+ *   - Picture-in-picture (up to 3 pinned cameras)
+ *   - Per-camera FPS indicator + adaptive quality badge
+ *
+ * Topics consumed:
+ *   /camera/<name>/image_raw/compressed    sensor_msgs/CompressedImage
+ *   /camera/color/image_raw/compressed     sensor_msgs/CompressedImage  (D435i)
+ *   /camera/depth/image_rect_raw/compressed sensor_msgs/CompressedImage  (D435i)
+ *   /camera/panoramic/compressed            sensor_msgs/CompressedImage
+ *   /social/faces/detections               saltybot_social_msgs/FaceDetectionArray
+ *   /social/gestures                       saltybot_social_msgs/GestureArray
+ *   /social/scene/objects                  saltybot_scene_msgs/SceneObjectArray
+ */
+
+import { useEffect, useRef, useState, useCallback } from 'react';
+import { useCamera, CAMERAS, CAMERA_BY_ID, CAMERA_BY_ROS_ID } from '../hooks/useCamera.js';
+
+// ── Constants ─────────────────────────────────────────────────────────────────
+
+const GESTURE_ICONS = {
+  wave:         '👋',
+  point:        '👆',
+  stop_palm:    '✋',
+  thumbs_up:    '👍',
+  thumbs_down:  '👎',
+  come_here:    '🤏',
+  follow:       '☞',
+  arms_up:      '🙌',
+  crouch:       '⬇',
+  arms_spread:  '↔',
+};
+
+const HAZARD_COLORS = {
+  1: '#f59e0b',  // stairs — amber
+  2: '#ef4444',  // drop — red
+  3: '#60a5fa',  // wet floor — blue
+  4: '#a855f7',  // glass door — purple
+  5: '#f97316',  // pet — orange
+};
+
+// ── Detection overlay drawing helpers ─────────────────────────────────────────
+
+function drawFaceBoxes(ctx, faces, scaleX, scaleY) {
+  for (const face of faces) {
+    const x = face.bbox_x * scaleX;
+    const y = face.bbox_y * scaleY;
+    const w = face.bbox_w * scaleX;
+    const h = face.bbox_h * scaleY;
+
+    const isKnown = face.person_name && face.person_name !== 'unknown';
+    ctx.strokeStyle = isKnown ? '#06b6d4' : '#f59e0b';
+    ctx.lineWidth   = 2;
+    ctx.shadowBlur  = 6;
+    ctx.shadowColor = ctx.strokeStyle;
+    ctx.strokeRect(x, y, w, h);
+    ctx.shadowBlur  = 0;
+
+    // Corner accent marks
+    const cLen = 8;
+    ctx.lineWidth = 3;
+    [[x,y,1,1],[x+w,y,-1,1],[x,y+h,1,-1],[x+w,y+h,-1,-1]].forEach(([cx,cy,dx,dy]) => {
+      ctx.beginPath();
+      ctx.moveTo(cx, cy + dy * cLen);
+      ctx.lineTo(cx, cy);
+      ctx.lineTo(cx + dx * cLen, cy);
+      ctx.stroke();
+    });
+
+    // Label
+    const label = isKnown
+      ? `${face.person_name} ${(face.recognition_score * 100).toFixed(0)}%`
+      : `face #${face.face_id}`;
+    ctx.font      = 'bold 11px monospace';
+    const tw      = ctx.measureText(label).width;
+    ctx.fillStyle = isKnown ? 'rgba(6,182,212,0.8)' : 'rgba(245,158,11,0.8)';
+    ctx.fillRect(x, y - 16, tw + 6, 16);
+    ctx.fillStyle = '#000';
+    ctx.fillText(label, x + 3, y - 4);
+  }
+}
+
+function drawGestureIcons(ctx, gestures, activeCamId, scaleX, scaleY) {
+  for (const g of gestures) {
+    // Only show gestures from the currently viewed camera
+    const cam = CAMERA_BY_ROS_ID[g.camera_id];
+    if (!cam || cam.cameraId !== activeCamId) continue;
+
+    const x = g.hand_x * ctx.canvas.width;
+    const y = g.hand_y * ctx.canvas.height;
+    const icon = GESTURE_ICONS[g.gesture_type] ?? '?';
+
+    ctx.font      = '24px serif';
+    ctx.shadowBlur  = 8;
+    ctx.shadowColor = '#f97316';
+    ctx.fillText(icon, x - 12, y + 8);
+    ctx.shadowBlur = 0;
+
+    ctx.font      = 'bold 10px monospace';
+    ctx.fillStyle = '#f97316';
+    const label   = g.gesture_type;
+    ctx.fillText(label, x - ctx.measureText(label).width / 2, y + 22);
+  }
+}
+
+function drawSceneObjects(ctx, objects, scaleX, scaleY) {
+  for (const obj of objects) {
+    // vision_msgs/BoundingBox2D: center_x, center_y, size_x, size_y
+    const bb  = obj.bbox;
+    const cx  = bb?.center?.x ?? bb?.center_x;
+    const cy  = bb?.center?.y ?? bb?.center_y;
+    const sw  = bb?.size_x ?? 0;
+    const sh  = bb?.size_y ?? 0;
+    if (cx == null) continue;
+
+    const x = (cx - sw / 2) * scaleX;
+    const y = (cy - sh / 2) * scaleY;
+    const w = sw * scaleX;
+    const h = sh * scaleY;
+
+    const color = HAZARD_COLORS[obj.hazard_type] ?? '#22c55e';
+    ctx.strokeStyle = color;
+    ctx.lineWidth   = 1.5;
+    ctx.setLineDash([4, 3]);
+    ctx.strokeRect(x, y, w, h);
+    ctx.setLineDash([]);
+
+    const dist  = obj.distance_m > 0 ? ` ${obj.distance_m.toFixed(1)}m` : '';
+    const label = `${obj.class_name}${dist}`;
+    ctx.font      = '10px monospace';
+    const tw      = ctx.measureText(label).width;
+    ctx.fillStyle = `${color}cc`;
+    ctx.fillRect(x, y + h, tw + 4, 14);
+    ctx.fillStyle = '#000';
+    ctx.fillText(label, x + 2, y + h + 11);
+  }
+}
+
+// ── Overlay canvas ─────────────────────────────────────────────────────────────
+
+function OverlayCanvas({ faces, gestures, sceneObjects, activeCam, containerW, containerH }) {
+  const canvasRef = useRef(null);
+
+  useEffect(() => {
+    const canvas = canvasRef.current;
+    if (!canvas) return;
+    const ctx = canvas.getContext('2d');
+    ctx.clearRect(0, 0, canvas.width, canvas.height);
+
+    if (!activeCam) return;
+
+    const scaleX = canvas.width  / (activeCam.width  || 640);
+    const scaleY = canvas.height / (activeCam.height || 480);
+
+    // Draw overlays: only for front camera (face + gesture source)
+    if (activeCam.id === 'front') {
+      drawFaceBoxes(ctx, faces, scaleX, scaleY);
+    }
+    if (!activeCam.isPanoramic) {
+      drawGestureIcons(ctx, gestures, activeCam.cameraId, scaleX, scaleY);
+    }
+    if (activeCam.id === 'color') {
+      drawSceneObjects(ctx, sceneObjects, scaleX, scaleY);
+    }
+  }, [faces, gestures, sceneObjects, activeCam]);
+
+  return (
+    <canvas
+      ref={canvasRef}
+      width={containerW || 640}
+      height={containerH || 480}
+      className="absolute inset-0 w-full h-full pointer-events-none"
+    />
+  );
+}
+
+// ── Panoramic equirect viewer ──────────────────────────────────────────────────
+
+function PanoViewer({ frameUrl }) {
+  const canvasRef = useRef(null);
+  const azRef     = useRef(0);    // 0–1920px offset
+  const dragRef   = useRef(null);
+  const imgRef    = useRef(null);
+
+  const draw = useCallback(() => {
+    const canvas = canvasRef.current;
+    const img    = imgRef.current;
+    if (!canvas || !img || !img.complete) return;
+
+    const ctx  = canvas.getContext('2d');
+    const W    = canvas.width;
+    const H    = canvas.height;
+    const iW   = img.naturalWidth;   // 1920
+    const iH   = img.naturalHeight;  // 960
+    const vW   = iW / 2;             // viewport = 50% of equirect width
+    const vH   = Math.round((H / W) * vW);
+    const vY   = Math.round((iH - vH) / 2);
+    const off  = Math.round(azRef.current) % iW;
+
+    ctx.clearRect(0, 0, W, H);
+
+    // Draw left segment
+    const srcX1 = off;
+    const srcW1 = Math.min(vW, iW - off);
+    const dstW1 = Math.round((srcW1 / vW) * W);
+    if (dstW1 > 0) {
+      ctx.drawImage(img, srcX1, vY, srcW1, vH, 0, 0, dstW1, H);
+    }
+
+    // Draw wrapped right segment (if viewport crosses 0°)
+    if (srcW1 < vW) {
+      const srcX2 = 0;
+      const srcW2 = vW - srcW1;
+      const dstX2 = dstW1;
+      const dstW2 = W - dstW1;
+      ctx.drawImage(img, srcX2, vY, srcW2, vH, dstX2, 0, dstW2, H);
+    }
+
+    // Compass badge
+    const azDeg = Math.round((azRef.current / iW) * 360);
+    ctx.fillStyle = 'rgba(0,0,0,0.5)';
+    ctx.fillRect(W - 58, 6, 52, 18);
+    ctx.fillStyle = '#06b6d4';
+    ctx.font      = 'bold 11px monospace';
+    ctx.fillText(`${azDeg}°`, W - 52, 19);
+  }, []);
+
+  // Load image when URL changes
+  useEffect(() => {
+    if (!frameUrl) return;
+    const img = new Image();
+    img.onload = draw;
+    img.src    = frameUrl;
+    imgRef.current = img;
+  }, [frameUrl, draw]);
+
+  // Re-draw when azimuth changes
+  const onMouseDown = e => { dragRef.current = e.clientX; };
+  const onMouseMove = e => {
+    if (dragRef.current == null) return;
+    const dx = e.clientX - dragRef.current;
+    dragRef.current = e.clientX;
+    azRef.current = ((azRef.current - dx * 2) % 1920 + 1920) % 1920;
+    draw();
+  };
+  const onMouseUp = () => { dragRef.current = null; };
+
+  const onTouchStart = e => { dragRef.current = e.touches[0].clientX; };
+  const onTouchMove  = e => {
+    if (dragRef.current == null) return;
+    const dx = e.touches[0].clientX - dragRef.current;
+    dragRef.current = e.touches[0].clientX;
+    azRef.current = ((azRef.current - dx * 2) % 1920 + 1920) % 1920;
+    draw();
+  };
+
+  return (
+    <canvas
+      ref={canvasRef}
+      width={960}
+      height={240}
+      className="w-full object-contain bg-black cursor-ew-resize rounded"
+      onMouseDown={onMouseDown}
+      onMouseMove={onMouseMove}
+      onMouseUp={onMouseUp}
+      onMouseLeave={onMouseUp}
+      onTouchStart={onTouchStart}
+      onTouchMove={onTouchMove}
+      onTouchEnd={() => { dragRef.current = null; }}
+    />
+  );
+}
+
+// ── PiP mini window ────────────────────────────────────────────────────────────
+
+function PiPWindow({ cam, frameUrl, fps, onClose, index }) {
+  const positions = [
+    'bottom-2 left-2',
+    'bottom-2 left-40',
+    'bottom-2 left-[18rem]',
+  ];
+  return (
+    <div className={`absolute ${positions[index] ?? 'bottom-2 left-2'} w-36 rounded border border-cyan-900 overflow-hidden bg-black shadow-lg shadow-black z-10`}>
+      <div className="flex items-center justify-between px-1.5 py-0.5 bg-gray-950 text-xs">
+        <span className="text-cyan-700 font-bold">{cam.label}</span>
+        <div className="flex items-center gap-1">
+          <span className="text-gray-700">{fps}fps</span>
+          <button onClick={onClose} className="text-gray-600 hover:text-red-400">✕</button>
+        </div>
+      </div>
+      {frameUrl ? (
+        <img src={frameUrl} alt={cam.label} className="w-full aspect-video object-cover block" />
+      ) : (
+        <div className="w-full aspect-video flex items-center justify-center text-gray-800 text-xs">
+          no signal
+        </div>
+      )}
+    </div>
+  );
+}
+
+// ── Camera selector strip ──────────────────────────────────────────────────────
+
+function CameraStrip({ cameras, activeId, pipList, frames, fps, onSelect, onTogglePip }) {
+  return (
+    <div className="flex gap-1.5 flex-wrap">
+      {cameras.map(cam => {
+        const hasFrame = !!frames[cam.id];
+        const camFps   = fps[cam.id] ?? 0;
+        const isActive = activeId === cam.id;
+        const isPip    = pipList.includes(cam.id);
+
+        return (
+          <div key={cam.id} className="relative">
+            <button
+              onClick={() => onSelect(cam.id)}
+              className={`flex flex-col items-start rounded border px-2.5 py-1.5 text-xs font-bold transition-colors ${
+                isActive
+                  ? 'border-cyan-500 bg-cyan-950 bg-opacity-50 text-cyan-300'
+                  : hasFrame
+                  ? 'border-gray-700 bg-gray-900 text-gray-400 hover:border-cyan-800 hover:text-gray-200'
+                  : 'border-gray-800 bg-gray-950 text-gray-700 hover:border-gray-700'
+              }`}
+            >
+              <span>{cam.label.toUpperCase()}</span>
+              <span className={`text-xs font-normal mt-0.5 ${
+                camFps >= 12 ? 'text-green-600' :
+                camFps > 0   ? 'text-amber-600' :
+                               'text-gray-700'
+              }`}>
+                {camFps > 0 ? `${camFps}fps` : 'no signal'}
+              </span>
+            </button>
+
+            {/* PiP pin button — only when NOT the active camera */}
+            {!isActive && (
+              <button
+                onClick={() => onTogglePip(cam.id)}
+                title={isPip ? 'Unpin PiP' : 'Pin PiP'}
+                className={`absolute -top-1.5 -right-1.5 w-4 h-4 rounded-full text-[9px] flex items-center justify-center border transition-colors ${
+                  isPip
+                    ? 'bg-cyan-600 border-cyan-400 text-white'
+                    : 'bg-gray-800 border-gray-700 text-gray-600 hover:border-cyan-700 hover:text-cyan-500'
+                }`}
+              >
+                {isPip ? '×' : '⊕'}
+              </button>
+            )}
+          </div>
+        );
+      })}
+    </div>
+  );
+}
+
+// ── Recording bar ──────────────────────────────────────────────────────────────
+
+function RecordingBar({ recording, recSeconds, onStart, onStop, onSnapshot, overlayRef }) {
+  const fmtTime = s => `${String(Math.floor(s / 60)).padStart(2, '0')}:${String(s % 60).padStart(2, '0')}`;
+
+  return (
+    <div className="flex items-center gap-2 flex-wrap">
+      {!recording ? (
+        <button
+          onClick={onStart}
+          className="flex items-center gap-1.5 px-3 py-1.5 rounded border border-red-900 bg-red-950 text-red-400 hover:bg-red-900 text-xs font-bold transition-colors"
+        >
+          <span className="w-2 h-2 rounded-full bg-red-500" />
+          REC
+        </button>
+      ) : (
+        <button
+          onClick={onStop}
+          className="flex items-center gap-1.5 px-3 py-1.5 rounded border border-red-600 bg-red-900 text-red-300 hover:bg-red-800 text-xs font-bold animate-pulse"
+        >
+          <span className="w-2 h-2 rounded bg-red-400" />
+          STOP  {fmtTime(recSeconds)}
+        </button>
+      )}
+
+      <button
+        onClick={() => onSnapshot(overlayRef?.current)}
+        className="flex items-center gap-1 px-3 py-1.5 rounded border border-gray-700 bg-gray-900 text-gray-400 hover:border-cyan-700 hover:text-cyan-400 text-xs font-bold transition-colors"
+      >
+        📷 SNAP
+      </button>
+
+      {recording && (
+        <span className="text-xs text-red-500 animate-pulse font-mono">
+          ● RECORDING {fmtTime(recSeconds)}
+        </span>
+      )}
+    </div>
+  );
+}
+
+// ── Main component ─────────────────────────────────────────────────────────────
+
+export function CameraViewer({ subscribe }) {
+  const {
+    cameras, frames, fps,
+    activeId, setActiveId,
+    pipList, togglePip,
+    recording, recSeconds,
+    startRecording, stopRecording,
+    takeSnapshot,
+  } = useCamera({ subscribe });
+
+  // ── Detection state ─────────────────────────────────────────────────────────
+  const [faces,        setFaces]        = useState([]);
+  const [gestures,     setGestures]     = useState([]);
+  const [sceneObjects, setSceneObjects] = useState([]);
+
+  const [showOverlay, setShowOverlay] = useState(true);
+  const [overlayMode, setOverlayMode] = useState('all'); // 'all' | 'faces' | 'gestures' | 'objects' | 'off'
+
+  const overlayCanvasRef = useRef(null);
+
+  // Subscribe to detection topics
+  useEffect(() => {
+    if (!subscribe) return;
+    const u1 = subscribe('/social/faces/detections', 'saltybot_social_msgs/FaceDetectionArray', msg => {
+      setFaces(msg.faces ?? []);
+    });
+    const u2 = subscribe('/social/gestures', 'saltybot_social_msgs/GestureArray', msg => {
+      setGestures(msg.gestures ?? []);
+    });
+    const u3 = subscribe('/social/scene/objects', 'saltybot_scene_msgs/SceneObjectArray', msg => {
+      setSceneObjects(msg.objects ?? []);
+    });
+    return () => { u1?.(); u2?.(); u3?.(); };
+  }, [subscribe]);
+
+  const activeCam  = CAMERA_BY_ID[activeId];
+  const activeFrame = frames[activeId];
+
+  // Filter overlay data based on mode
+  const visibleFaces   = (overlayMode === 'all' || overlayMode === 'faces')   ? faces        : [];
+  const visibleGestures = (overlayMode === 'all' || overlayMode === 'gestures') ? gestures    : [];
+  const visibleObjects  = (overlayMode === 'all' || overlayMode === 'objects')  ? sceneObjects : [];
+
+  // ── Container size tracking (for overlay canvas sizing) ────────────────────
+  const containerRef = useRef(null);
+  const [containerSize, setContainerSize] = useState({ w: 640, h: 480 });
+
+  useEffect(() => {
+    if (!containerRef.current) return;
+    const ro = new ResizeObserver(entries => {
+      const e = entries[0];
+      setContainerSize({ w: Math.round(e.contentRect.width), h: Math.round(e.contentRect.height) });
+    });
+    ro.observe(containerRef.current);
+    return () => ro.disconnect();
+  }, []);
+
+  // ── Quality badge ──────────────────────────────────────────────────────────
+  const camFps    = fps[activeId] ?? 0;
+  const quality   = camFps >= 13 ? 'FULL' : camFps >= 8 ? 'GOOD' : camFps > 0 ? 'LOW' : 'NO SIGNAL';
+  const qualColor = camFps >= 13 ? 'text-green-500' : camFps >= 8 ? 'text-amber-500' : camFps > 0 ? 'text-red-500' : 'text-gray-700';
+
+  return (
+    <div className="space-y-3">
+
+      {/* ── Camera strip ── */}
+      <div className="bg-gray-950 rounded-lg border border-cyan-950 p-3 space-y-2">
+        <div className="flex items-center justify-between">
+          <div className="text-cyan-700 text-xs font-bold tracking-widest">CAMERA SELECT</div>
+          <span className={`text-xs font-bold ${qualColor}`}>{quality}  {camFps > 0 ? `${camFps}fps` : ''}</span>
+        </div>
+        <CameraStrip
+          cameras={cameras}
+          activeId={activeId}
+          pipList={pipList}
+          frames={frames}
+          fps={fps}
+          onSelect={setActiveId}
+          onTogglePip={togglePip}
+        />
+      </div>
+
+      {/* ── Main viewer ── */}
+      <div className="bg-gray-950 rounded-lg border border-cyan-950 overflow-hidden">
+
+        {/* Viewer toolbar */}
+        <div className="flex items-center justify-between px-3 py-2 border-b border-cyan-950">
+          <div className="flex items-center gap-2">
+            <span className="text-cyan-400 text-xs font-bold">{activeCam?.label ?? '—'}</span>
+            {activeCam?.isDepth && (
+              <span className="text-xs text-gray-600 border border-gray-800 rounded px-1">DEPTH · greyscale</span>
+            )}
+            {activeCam?.isPanoramic && (
+              <span className="text-xs text-gray-600 border border-gray-800 rounded px-1">360° · drag to pan</span>
+            )}
+          </div>
+
+          {/* Overlay mode selector */}
+          <div className="flex items-center gap-1">
+            {['off','faces','gestures','objects','all'].map(mode => (
+              <button
+                key={mode}
+                onClick={() => setOverlayMode(mode)}
+                className={`px-2 py-0.5 rounded text-xs border transition-colors ${
+                  overlayMode === mode
+                    ? 'border-cyan-600 bg-cyan-950 text-cyan-400'
+                    : 'border-gray-800 text-gray-600 hover:border-gray-700 hover:text-gray-400'
+                }`}
+              >
+                {mode === 'all' ? 'ALL' : mode === 'off' ? 'OFF' : mode.slice(0,3).toUpperCase()}
+              </button>
+            ))}
+          </div>
+        </div>
+
+        {/* Image + overlay */}
+        <div className="relative" ref={containerRef}>
+          {activeCam?.isPanoramic ? (
+            <PanoViewer frameUrl={activeFrame} />
+          ) : activeFrame ? (
+            <img
+              src={activeFrame}
+              alt={activeCam?.label ?? 'camera'}
+              className="w-full object-contain block bg-black"
+              style={{ maxHeight: '480px' }}
+            />
+          ) : (
+            <div className="w-full bg-black flex items-center justify-center text-gray-800 text-sm font-mono"
+                 style={{ height: '360px' }}>
+              <div className="text-center space-y-2">
+                <div className="text-2xl">📷</div>
+                <div>Waiting for {activeCam?.label ?? '—'}…</div>
+                <div className="text-xs text-gray-700">{activeCam?.topic}</div>
+              </div>
+            </div>
+          )}
+
+          {/* Detection overlay canvas */}
+          {overlayMode !== 'off' && !activeCam?.isPanoramic && (
+            <OverlayCanvas
+              ref={overlayCanvasRef}
+              faces={visibleFaces}
+              gestures={visibleGestures}
+              sceneObjects={visibleObjects}
+              activeCam={activeCam}
+              containerW={containerSize.w}
+              containerH={containerSize.h}
+            />
+          )}
+
+          {/* PiP windows */}
+          {pipList.map((id, idx) => {
+            const cam = CAMERA_BY_ID[id];
+            if (!cam) return null;
+            return (
+              <PiPWindow
+                key={id}
+                cam={cam}
+                frameUrl={frames[id]}
+                fps={fps[id] ?? 0}
+                index={idx}
+                onClose={() => togglePip(id)}
+              />
+            );
+          })}
+        </div>
+      </div>
+
+      {/* ── Recording controls ── */}
+      <div className="bg-gray-950 rounded-lg border border-cyan-950 p-3">
+        <div className="flex items-center justify-between mb-2">
+          <div className="text-cyan-700 text-xs font-bold tracking-widest">CAPTURE</div>
+        </div>
+        <RecordingBar
+          recording={recording}
+          recSeconds={recSeconds}
+          onStart={startRecording}
+          onStop={stopRecording}
+          onSnapshot={takeSnapshot}
+          overlayRef={overlayCanvasRef}
+        />
+        <div className="mt-2 text-xs text-gray-700">
+          Recording saves as MP4/WebM to your Downloads.
+          Snapshot includes detection overlay + timestamp.
+        </div>
+      </div>
+
+      {/* ── Detection status ── */}
+      <div className="grid grid-cols-3 gap-2 text-xs">
+        <div className="bg-gray-950 rounded border border-gray-800 p-2">
+          <div className="text-gray-600 mb-1">FACES</div>
+          <div className={`font-bold ${faces.length > 0 ? 'text-cyan-400' : 'text-gray-700'}`}>
+            {faces.length > 0 ? `${faces.length} detected` : 'none'}
+          </div>
+          {faces.slice(0, 2).map((f, i) => (
+            <div key={i} className="text-gray-600 truncate">
+              {f.person_name && f.person_name !== 'unknown'
+                ? `↳ ${f.person_name}`
+                : `↳ unknown #${f.face_id}`}
+            </div>
+          ))}
+        </div>
+
+        <div className="bg-gray-950 rounded border border-gray-800 p-2">
+          <div className="text-gray-600 mb-1">GESTURES</div>
+          <div className={`font-bold ${gestures.length > 0 ? 'text-amber-400' : 'text-gray-700'}`}>
+            {gestures.length > 0 ? `${gestures.length} active` : 'none'}
+          </div>
+          {gestures.slice(0, 2).map((g, i) => {
+            const icon = GESTURE_ICONS[g.gesture_type] ?? '?';
+            return (
+              <div key={i} className="text-gray-600 truncate">
+                {icon} {g.gesture_type} cam{g.camera_id}
+              </div>
+            );
+          })}
+        </div>
+
+        <div className="bg-gray-950 rounded border border-gray-800 p-2">
+          <div className="text-gray-600 mb-1">OBJECTS</div>
+          <div className={`font-bold ${sceneObjects.length > 0 ? 'text-green-400' : 'text-gray-700'}`}>
+            {sceneObjects.length > 0 ? `${sceneObjects.length} objects` : 'none'}
+          </div>
+          {sceneObjects
+            .filter(o => o.hazard_type > 0)
+            .slice(0, 2)
+            .map((o, i) => (
+              <div key={i} className="text-amber-700 truncate">⚠ {o.class_name}</div>
+            ))
+          }
+          {sceneObjects.filter(o => o.hazard_type === 0).slice(0, 2).map((o, i) => (
+            <div key={`ok${i}`} className="text-gray-600 truncate">
+              {o.class_name} {o.distance_m > 0 ? `${o.distance_m.toFixed(1)}m` : ''}
+            </div>
+          ))}
+        </div>
+      </div>
+
+      {/* ── Legend ── */}
+      <div className="flex gap-4 text-xs text-gray-700 flex-wrap">
+        <div className="flex items-center gap-1">
+          <div className="w-3 h-3 rounded-sm border border-cyan-600" />
+          Known face
+        </div>
+        <div className="flex items-center gap-1">
+          <div className="w-3 h-3 rounded-sm border border-amber-600" />
+          Unknown face
+        </div>
+        <div className="flex items-center gap-1">
+          <span>👆</span> Gesture
+        </div>
+        <div className="flex items-center gap-1">
+          <div className="w-3 h-3 rounded-sm border border-green-700 border-dashed" />
+          Object
+        </div>
+        <div className="flex items-center gap-1">
+          <div className="w-3 h-3 rounded-sm border border-amber-600 border-dashed" />
+          Hazard
+        </div>
+        <div className="ml-auto text-gray-800 italic">
+          ⊕ pin = PiP  ·  overlay: {overlayMode}
+        </div>
+      </div>
+    </div>
+  );
+}

ui/social-bot/src/hooks/useCamera.js

@@ -0,0 +1,325 @@
+/**
+ * useCamera.js — Multi-camera stream manager (Issue #177).
+ *
+ * Subscribes to sensor_msgs/CompressedImage topics via rosbridge.
+ * Decodes base64 JPEG/PNG → data URL for <img>/<canvas> display.
+ * Tracks per-camera FPS. Manages MediaRecorder for recording + snapshots.
+ *
+ * Camera sources:
+ *   front / left / rear / right  — 4× CSI IMX219, 640×480
+ *                                  topic: /camera/<name>/image_raw/compressed
+ *   color                        — D435i RGB, 640×480
+ *                                  topic: /camera/color/image_raw/compressed
+ *   depth                        — D435i depth, 640×480 greyscale (PNG16)
+ *                                  topic: /camera/depth/image_rect_raw/compressed
+ *   panoramic                    — equirect stitch 1920×960
+ *                                  topic: /camera/panoramic/compressed
+ */
+
+import { useState, useEffect, useRef, useCallback } from 'react';
+
+// ── Camera catalogue ──────────────────────────────────────────────────────────
+
+export const CAMERAS = [
+  {
+    id: 'front',
+    label: 'Front',
+    shortLabel: 'F',
+    topic: '/camera/front/image_raw/compressed',
+    msgType: 'sensor_msgs/CompressedImage',
+    cameraId: 0,           // matches gesture_node camera_id
+    width: 640, height: 480,
+  },
+  {
+    id: 'left',
+    label: 'Left',
+    shortLabel: 'L',
+    topic: '/camera/left/image_raw/compressed',
+    msgType: 'sensor_msgs/CompressedImage',
+    cameraId: 1,
+    width: 640, height: 480,
+  },
+  {
+    id: 'rear',
+    label: 'Rear',
+    shortLabel: 'R',
+    topic: '/camera/rear/image_raw/compressed',
+    msgType: 'sensor_msgs/CompressedImage',
+    cameraId: 2,
+    width: 640, height: 480,
+  },
+  {
+    id: 'right',
+    label: 'Right',
+    shortLabel: 'Rt',
+    topic: '/camera/right/image_raw/compressed',
+    msgType: 'sensor_msgs/CompressedImage',
+    cameraId: 3,
+    width: 640, height: 480,
+  },
+  {
+    id: 'color',
+    label: 'D435i RGB',
+    shortLabel: 'D',
+    topic: '/camera/color/image_raw/compressed',
+    msgType: 'sensor_msgs/CompressedImage',
+    cameraId: 4,
+    width: 640, height: 480,
+  },
+  {
+    id: 'depth',
+    label: 'Depth',
+    shortLabel: '≋',
+    topic: '/camera/depth/image_rect_raw/compressed',
+    msgType: 'sensor_msgs/CompressedImage',
+    cameraId: 5,
+    width: 640, height: 480,
+    isDepth: true,
+  },
+  {
+    id: 'panoramic',
+    label: 'Panoramic',
+    shortLabel: '360',
+    topic: '/camera/panoramic/compressed',
+    msgType: 'sensor_msgs/CompressedImage',
+    cameraId: -1,
+    width: 1920, height: 960,
+    isPanoramic: true,
+  },
+];
+
+export const CAMERA_BY_ID = Object.fromEntries(CAMERAS.map(c => [c.id, c]));
+export const CAMERA_BY_ROS_ID = Object.fromEntries(
+  CAMERAS.filter(c => c.cameraId >= 0).map(c => [c.cameraId, c])
+);
+
+const TARGET_FPS     = 15;
+const FPS_INTERVAL   = 1000; // ms between FPS counter resets
+
+// ── Hook ──────────────────────────────────────────────────────────────────────
+
+export function useCamera({ subscribe } = {}) {
+  const [frames, setFrames]         = useState(() =>
+    Object.fromEntries(CAMERAS.map(c => [c.id, null]))
+  );
+  const [fps, setFps]               = useState(() =>
+    Object.fromEntries(CAMERAS.map(c => [c.id, 0]))
+  );
+  const [activeId, setActiveId]     = useState('front');
+  const [pipList, setPipList]       = useState([]);   // up to 3 extra camera ids
+  const [recording, setRecording]   = useState(false);
+  const [recSeconds, setRecSeconds] = useState(0);
+
+  // ── Refs (not state — no re-render needed) ─────────────────────────────────
+  const countRef      = useRef(Object.fromEntries(CAMERAS.map(c => [c.id, 0])));
+  const mediaRecRef   = useRef(null);
+  const chunksRef     = useRef([]);
+  const recTimerRef   = useRef(null);
+  const recordCanvas  = useRef(null); // hidden canvas used for recording
+  const recAnimRef    = useRef(null); // rAF handle for record-canvas loop
+  const latestFrameRef = useRef(Object.fromEntries(CAMERAS.map(c => [c.id, null])));
+  const latestTsRef   = useRef(Object.fromEntries(CAMERAS.map(c => [c.id, 0])));
+
+  // ── FPS counter ────────────────────────────────────────────────────────────
+  useEffect(() => {
+    const timer = setInterval(() => {
+      setFps({ ...countRef.current });
+      const reset = Object.fromEntries(CAMERAS.map(c => [c.id, 0]));
+      countRef.current = reset;
+    }, FPS_INTERVAL);
+    return () => clearInterval(timer);
+  }, []);
+
+  // ── Subscribe all camera topics ────────────────────────────────────────────
+  useEffect(() => {
+    if (!subscribe) return;
+
+    const unsubs = CAMERAS.map(cam => {
+      let lastTs = 0;
+      const interval = Math.floor(1000 / TARGET_FPS); // client-side 15fps gate
+
+      return subscribe(cam.topic, cam.msgType, (msg) => {
+        const now = Date.now();
+        if (now - lastTs < interval) return; // drop frames > 15fps
+        lastTs = now;
+
+        const fmt    = msg.format || 'jpeg';
+        const mime   = fmt.includes('png') || fmt.includes('16UC') ? 'image/png' : 'image/jpeg';
+        const dataUrl = `data:${mime};base64,${msg.data}`;
+
+        latestFrameRef.current[cam.id] = dataUrl;
+        latestTsRef.current[cam.id]    = now;
+        countRef.current[cam.id]       = (countRef.current[cam.id] ?? 0) + 1;
+
+        setFrames(prev => ({ ...prev, [cam.id]: dataUrl }));
+      });
+    });
+
+    return () => unsubs.forEach(fn => fn?.());
+  }, [subscribe]);
+
+  // ── Create hidden record canvas ────────────────────────────────────────────
+  useEffect(() => {
+    const c = document.createElement('canvas');
+    c.width  = 640;
+    c.height = 480;
+    c.style.display = 'none';
+    document.body.appendChild(c);
+    recordCanvas.current = c;
+    return () => { c.remove(); };
+  }, []);
+
+  // ── Draw loop for record canvas ────────────────────────────────────────────
+  // Runs at TARGET_FPS when recording — draws active frame to hidden canvas
+  const startRecordLoop = useCallback(() => {
+    const canvas = recordCanvas.current;
+    if (!canvas) return;
+
+    const step = () => {
+      const cam    = CAMERA_BY_ID[activeId];
+      const src    = latestFrameRef.current[activeId];
+      const ctx    = canvas.getContext('2d');
+
+      if (!cam || !src) {
+        recAnimRef.current = requestAnimationFrame(step);
+        return;
+      }
+
+      // Resize canvas to match source
+      if (canvas.width !== cam.width || canvas.height !== cam.height) {
+        canvas.width  = cam.width;
+        canvas.height = cam.height;
+      }
+
+      const img = new Image();
+      img.onload = () => {
+        ctx.drawImage(img, 0, 0, canvas.width, canvas.height);
+      };
+      img.src = src;
+
+      recAnimRef.current = setTimeout(step, Math.floor(1000 / TARGET_FPS));
+    };
+
+    recAnimRef.current = setTimeout(step, 0);
+  }, [activeId]);
+
+  const stopRecordLoop = useCallback(() => {
+    if (recAnimRef.current) {
+      clearTimeout(recAnimRef.current);
+      cancelAnimationFrame(recAnimRef.current);
+      recAnimRef.current = null;
+    }
+  }, []);
+
+  // ── Recording ──────────────────────────────────────────────────────────────
+
+  const startRecording = useCallback(() => {
+    const canvas = recordCanvas.current;
+    if (!canvas || recording) return;
+
+    startRecordLoop();
+
+    const stream   = canvas.captureStream(TARGET_FPS);
+    const mimeType =
+      MediaRecorder.isTypeSupported('video/mp4')                ? 'video/mp4'            :
+      MediaRecorder.isTypeSupported('video/webm;codecs=vp9')    ? 'video/webm;codecs=vp9' :
+      MediaRecorder.isTypeSupported('video/webm;codecs=vp8')    ? 'video/webm;codecs=vp8' :
+                                                                   'video/webm';
+
+    chunksRef.current = [];
+    const mr = new MediaRecorder(stream, { mimeType, videoBitsPerSecond: 2_500_000 });
+    mr.ondataavailable = e => { if (e.data?.size > 0) chunksRef.current.push(e.data); };
+    mr.start(200);
+    mediaRecRef.current = mr;
+    setRecording(true);
+    setRecSeconds(0);
+    recTimerRef.current = setInterval(() => setRecSeconds(s => s + 1), 1000);
+  }, [recording, startRecordLoop]);
+
+  const stopRecording = useCallback(() => {
+    const mr = mediaRecRef.current;
+    if (!mr || mr.state === 'inactive') return;
+
+    mr.onstop = () => {
+      const ext  = mr.mimeType.includes('mp4') ? 'mp4' : 'webm';
+      const blob = new Blob(chunksRef.current, { type: mr.mimeType });
+      const url  = URL.createObjectURL(blob);
+      const a    = document.createElement('a');
+      a.href     = url;
+      a.download = `saltybot-${activeId}-${Date.now()}.${ext}`;
+      a.click();
+      URL.revokeObjectURL(url);
+    };
+
+    mr.stop();
+    stopRecordLoop();
+    clearInterval(recTimerRef.current);
+    setRecording(false);
+  }, [activeId, stopRecordLoop]);
+
+  // ── Snapshot ───────────────────────────────────────────────────────────────
+
+  const takeSnapshot = useCallback((overlayCanvasEl) => {
+    const src = latestFrameRef.current[activeId];
+    if (!src) return;
+
+    const cam    = CAMERA_BY_ID[activeId];
+    const canvas = document.createElement('canvas');
+    canvas.width  = cam.width;
+    canvas.height = cam.height;
+    const ctx = canvas.getContext('2d');
+
+    const img = new Image();
+    img.onload = () => {
+      ctx.drawImage(img, 0, 0, canvas.width, canvas.height);
+
+      // Composite detection overlay if provided
+      if (overlayCanvasEl) {
+        ctx.drawImage(overlayCanvasEl, 0, 0, canvas.width, canvas.height);
+      }
+
+      // Timestamp watermark
+      ctx.fillStyle = 'rgba(0,0,0,0.5)';
+      ctx.fillRect(0, canvas.height - 20, canvas.width, 20);
+      ctx.fillStyle = '#06b6d4';
+      ctx.font = '11px monospace';
+      ctx.fillText(`SALTYBOT  ${cam.label}  ${new Date().toISOString()}`, 8, canvas.height - 6);
+
+      canvas.toBlob(blob => {
+        const url = URL.createObjectURL(blob);
+        const a   = document.createElement('a');
+        a.href    = url;
+        a.download = `saltybot-snap-${activeId}-${Date.now()}.png`;
+        a.click();
+        URL.revokeObjectURL(url);
+      }, 'image/png');
+    };
+    img.src = src;
+  }, [activeId]);
+
+  // ── PiP management ─────────────────────────────────────────────────────────
+
+  const togglePip = useCallback(id => {
+    setPipList(prev => {
+      if (prev.includes(id)) return prev.filter(x => x !== id);
+      const next = [...prev, id].filter(x => x !== activeId);
+      return next.slice(-3); // max 3 PIPs
+    });
+  }, [activeId]);
+
+  // Remove PiP if it becomes the active camera
+  useEffect(() => {
+    setPipList(prev => prev.filter(id => id !== activeId));
+  }, [activeId]);
+
+  return {
+    cameras:      CAMERAS,
+    frames,
+    fps,
+    activeId,     setActiveId,
+    pipList,      togglePip,
+    recording,    recSeconds,
+    startRecording, stopRecording,
+    takeSnapshot,
+  };
+}