jetson/ros2_ws/src/saltybot_mode_switch/.gitignore

@@ -0,0 +1,4 @@
+__pycache__/
+*.pyc
+*.egg-info/
+.pytest_cache/

jetson/ros2_ws/src/saltybot_mode_switch/config/mode_switch_params.yaml

@@ -0,0 +1,58 @@
+# mode_switch_params.yaml — saltybot_mode_switch
+#
+# Launch with:
+#   ros2 launch saltybot_mode_switch mode_switch.launch.py
+#
+# Topic wiring:
+#   /rc/joy                          → mode_switch_node (CRSF channels)
+#   /saltybot/balance_state          → mode_switch_node (STM32 state)
+#   /slam_toolbox/pose_with_covariance_stamped → mode_switch_node (SLAM fix)
+#   /saltybot/control_mode           ← mode_switch_node (JSON mode + alpha)
+#   /saltybot/led_pattern            ← mode_switch_node (LED name)
+#
+#   /cmd_vel_auto                    → cmd_vel_mux_node  (from Nav2/follower)
+#   /saltybot/control_mode           → cmd_vel_mux_node
+#   /cmd_vel                         ← cmd_vel_mux_node  (to bridge)
+#
+# IMPORTANT: remap Nav2 and person_follower to publish /cmd_vel_auto:
+#   ros2 launch saltybot_bringup nav2.launch.py cmd_vel_topic:=/cmd_vel_auto
+
+# ── CRSF channel mapping ──────────────────────────────────────────────────────
+# Joy axes index for the AUX2 / CH6 mode-toggle switch.
+# Typical CRSF→Joy mapping: CH1=axes[0] … CH6=axes[5].
+ch6_axis: 5
+
+# Space-separated Joy axes indices for the four control sticks (roll/pitch/throttle/yaw).
+# These are monitored for the 10% override threshold.
+stick_axes: "0 1 2 3"
+
+# ── Override / return-to-auto ──────────────────────────────────────────────────
+# Stick magnitude (0–1) that triggers instant RC override while in AUTO.
+# 10% (0.10) per spec — change only if stick noise causes false triggers.
+override_threshold: 0.10
+
+# Seconds sticks must remain neutral before AUTO re-engages after an override.
+return_delay_s: 2.0
+
+# ── Ramp ──────────────────────────────────────────────────────────────────────
+# RC↔AUTO blend ramp duration (seconds). cmd_vel_mux scales by blend_alpha
+# so autonomous commands are eased in/out over this window.
+ramp_duration_s: 0.5
+
+# ── RC link health ────────────────────────────────────────────────────────────
+# Seconds of Joy silence before RC link is declared lost → force RC mode.
+rc_link_timeout_s: 0.5
+
+# ── SLAM fix ──────────────────────────────────────────────────────────────────
+# ROS2 topic providing the SLAM localisation pose. Any message received within
+# slam_fix_timeout_s counts as a valid fix.
+slam_fix_topic: "/slam_toolbox/pose_with_covariance_stamped"
+
+# Seconds after the last SLAM pose before fix is considered lost.
+slam_fix_timeout_s: 3.0
+
+# ── Control loop ──────────────────────────────────────────────────────────────
+control_rate: 20.0    # Hz
+
+# ── cmd_vel_mux ───────────────────────────────────────────────────────────────
+cmd_vel_timeout_s: 0.5   # deadman: zero output if /cmd_vel_auto goes silent

jetson/ros2_ws/src/saltybot_mode_switch/launch/mode_switch.launch.py

@@ -0,0 +1,59 @@
+"""mode_switch.launch.py — launch mode_switch_node + cmd_vel_mux_node."""
+
+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_mode_switch")
+    default_cfg  = os.path.join(pkg_share, "config", "mode_switch_params.yaml")
+
+    return LaunchDescription([
+        DeclareLaunchArgument(
+            "params_file",
+            default_value=default_cfg,
+            description="Path to mode_switch_params.yaml",
+        ),
+        DeclareLaunchArgument(
+            "ramp_duration_s",
+            default_value="0.5",
+            description="RC↔AUTO blend ramp duration (seconds)",
+        ),
+        DeclareLaunchArgument(
+            "return_delay_s",
+            default_value="2.0",
+            description="Sticks-neutral duration before auto re-engage (seconds)",
+        ),
+        DeclareLaunchArgument(
+            "slam_fix_topic",
+            default_value="/slam_toolbox/pose_with_covariance_stamped",
+            description="SLAM fix pose topic",
+        ),
+
+        Node(
+            package="saltybot_mode_switch",
+            executable="mode_switch_node",
+            name="mode_switch",
+            output="screen",
+            parameters=[
+                LaunchConfiguration("params_file"),
+                {
+                    "ramp_duration_s": LaunchConfiguration("ramp_duration_s"),
+                    "return_delay_s":  LaunchConfiguration("return_delay_s"),
+                    "slam_fix_topic":  LaunchConfiguration("slam_fix_topic"),
+                },
+            ],
+        ),
+
+        Node(
+            package="saltybot_mode_switch",
+            executable="cmd_vel_mux_node",
+            name="cmd_vel_mux",
+            output="screen",
+            parameters=[LaunchConfiguration("params_file")],
+        ),
+    ])

jetson/ros2_ws/src/saltybot_mode_switch/package.xml

@@ -0,0 +1,31 @@
+<?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_mode_switch</name>
+  <version>0.1.0</version>
+  <description>
+    Autonomous/RC mode switch for saltybot (Issue #104).
+    Monitors CRSF CH6 toggle, RC stick override, SLAM fix status, and RC link
+    health to manage RC↔AUTO transitions with a 500 ms blend ramp.
+    Publishes /saltybot/control_mode (JSON) and /saltybot/led_pattern.
+    Includes cmd_vel_mux_node that scales autonomous cmd_vel by blend_alpha.
+  </description>
+  <maintainer email="sl-controls@saltylab.local">sl-controls</maintainer>
+  <license>MIT</license>
+
+  <depend>rclpy</depend>
+  <depend>geometry_msgs</depend>
+  <depend>sensor_msgs</depend>
+  <depend>std_msgs</depend>
+
+  <buildtool_depend>ament_python</buildtool_depend>
+
+  <test_depend>ament_copyright</test_depend>
+  <test_depend>ament_flake8</test_depend>
+  <test_depend>ament_pep257</test_depend>
+  <test_depend>python3-pytest</test_depend>
+
+  <export>
+    <build_type>ament_python</build_type>
+  </export>
+</package>

jetson/ros2_ws/src/saltybot_mode_switch/saltybot_mode_switch/cmd_vel_mux_node.py

@@ -0,0 +1,133 @@
+"""
+cmd_vel_mux_node.py — Blend-alpha cmd_vel multiplexer (Issue #104).
+
+Sits between the Nav2 / person-follower stack and the cmd_vel_bridge_node,
+scaling the autonomous cmd_vel output by the blend_alpha published by
+mode_switch_node so that the RC → AUTO handoff is smooth over 500 ms.
+
+Topic graph
+-----------
+  Nav2 / follower → /cmd_vel_auto ──┐
+                                     ├── cmd_vel_mux → /cmd_vel → bridge
+  mode_switch → /saltybot/control_mode (blend_alpha)
+
+In RC mode (blend_alpha ≈ 0) the node publishes Twist(0,0) so the bridge
+receives zeros — this is harmless because the bridge's mode gate already
+prevents autonomous commands when the STM32 is in RC_MANUAL.
+
+The bridge's existing ESC ramp handles hardware-level smoothing;
+the blend_alpha here provides the higher-level cmd_vel policy ramp.
+
+Parameters
+----------
+  cmd_vel_timeout_s  Deadman: zero output if /cmd_vel_auto goes silent  (0.5 s)
+  control_rate       Output publish rate (Hz)                           (20.0)
+"""
+
+import json
+import time
+
+import rclpy
+from rclpy.node import Node
+from rclpy.qos import HistoryPolicy, QoSProfile, ReliabilityPolicy
+
+from geometry_msgs.msg import Twist
+from std_msgs.msg import String
+
+from saltybot_mode_switch.mode_logic import blend_twist
+
+_BIG = 1e9
+
+
+class CmdVelMuxNode(Node):
+
+    def __init__(self):
+        super().__init__("cmd_vel_mux")
+
+        self.declare_parameter("cmd_vel_timeout_s", 0.5)
+        self.declare_parameter("control_rate",      20.0)
+
+        self._timeout      = self.get_parameter("cmd_vel_timeout_s").value
+        self._rate         = self.get_parameter("control_rate").value
+
+        reliable = QoSProfile(
+            reliability=ReliabilityPolicy.RELIABLE,
+            history=HistoryPolicy.KEEP_LAST,
+            depth=10,
+        )
+
+        # ── State ─────────────────────────────────────────────────────────────
+        self._auto_linear_x:  float = 0.0
+        self._auto_angular_z: float = 0.0
+        self._blend_alpha:    float = 0.0
+        self._last_auto_t:    float = 0.0
+
+        # ── Subscriptions ─────────────────────────────────────────────────────
+        self.create_subscription(
+            Twist, "/cmd_vel_auto", self._auto_cb, reliable)
+        self.create_subscription(
+            String, "/saltybot/control_mode", self._mode_cb, reliable)
+
+        # ── Publisher ─────────────────────────────────────────────────────────
+        self._pub = self.create_publisher(Twist, "/cmd_vel", reliable)
+
+        # ── Timer ─────────────────────────────────────────────────────────────
+        self._timer = self.create_timer(1.0 / self._rate, self._publish_cb)
+
+        self.get_logger().info(
+            f"CmdVelMuxNode ready  rate={self._rate:.0f}Hz  "
+            f"timeout={self._timeout:.1f}s"
+        )
+
+    # ── Callbacks ─────────────────────────────────────────────────────────────
+
+    def _auto_cb(self, msg: Twist) -> None:
+        self._auto_linear_x  = msg.linear.x
+        self._auto_angular_z = msg.angular.z
+        self._last_auto_t    = time.monotonic()
+
+    def _mode_cb(self, msg: String) -> None:
+        try:
+            data = json.loads(msg.data)
+            self._blend_alpha = float(data.get("blend_alpha", 0.0))
+        except (json.JSONDecodeError, TypeError, ValueError):
+            self._blend_alpha = 0.0
+
+    # ── 20 Hz publish ─────────────────────────────────────────────────────────
+
+    def _publish_cb(self) -> None:
+        now = time.monotonic()
+
+        # Deadman: zero if autonomous source went silent
+        auto_age = (now - self._last_auto_t) if self._last_auto_t > 0.0 else _BIG
+        if auto_age > self._timeout:
+            linear_x, angular_z = 0.0, 0.0
+        else:
+            linear_x, angular_z = blend_twist(
+                self._auto_linear_x,
+                self._auto_angular_z,
+                self._blend_alpha,
+            )
+
+        twist          = Twist()
+        twist.linear.x  = linear_x
+        twist.angular.z = angular_z
+        self._pub.publish(twist)
+
+
+# ── Entry point ───────────────────────────────────────────────────────────────
+
+def main(args=None):
+    rclpy.init(args=args)
+    node = CmdVelMuxNode()
+    try:
+        rclpy.spin(node)
+    except KeyboardInterrupt:
+        pass
+    finally:
+        node.destroy_node()
+        rclpy.try_shutdown()
+
+
+if __name__ == "__main__":
+    main()

jetson/ros2_ws/src/saltybot_mode_switch/saltybot_mode_switch/mode_logic.py

@@ -0,0 +1,241 @@
+"""
+mode_logic.py — Pure RC/Autonomous mode-switch logic for saltybot.
+
+All functions are stateless or operate on plain Python types so the full
+state machine can be exercised in unit tests without a ROS2 runtime.
+
+Mode vocabulary
+---------------
+  "RC"            — STM32 executing RC pilot commands; Jetson cmd_vel blocked.
+  "RAMP_TO_AUTO"  — Transitioning RC→AUTO; blend_alpha 0.0→1.0 over ramp_s.
+  "AUTO"          — STM32 executing Jetson cmd_vel; RC sticks idle.
+  "RAMP_TO_RC"    — Transitioning AUTO→RC; blend_alpha 1.0→0.0 over ramp_s.
+
+Blend alpha
+-----------
+  The node publishes blend_alpha ∈ [0, 1] in /saltybot/control_mode.
+  The cmd_vel_mux_node scales autonomous cmd_vel by alpha before publishing
+  to /cmd_vel (the bridge input).  alpha=0 ⟹ no autonomous motion.
+
+RC override
+-----------
+  Any stick magnitude > override_threshold while in AUTO instantly forces RC
+  (alpha → 0 immediately; no ramp).  _override_active is set so the
+  return-to-auto path requires sticks to be neutral for return_delay seconds.
+
+CH6 / toggle
+------------
+  ch6_auto=True  means the pilot has requested AUTO (switch up/high).
+  ch6_auto=False means the pilot has requested RC   (switch down/low).
+  Pure CH6 transitions (no stick activity) always ramp gracefully.
+
+Safety interlocks
+-----------------
+  • No AUTO without slam_ok (valid SLAM fix received recently).
+  • Instant drop to RC when rc_link_ok goes False (RC link lost).
+  These checks are applied at every control tick.
+"""
+
+import math
+
+# ─── Constants ──────────────────────────────────────────────────────────────
+
+MODES = ("RC", "RAMP_TO_AUTO", "AUTO", "RAMP_TO_RC")
+
+LED_PATTERNS = {
+    "rc_normal":      "solid_yellow",
+    "rc_no_slam":     "solid_yellow",       # no slam needed in RC
+    "auto_ok":        "solid_green",
+    "auto_no_slam":   "blink_orange_fast",  # shouldn't happen (interlock), but defensive
+    "ramp_to_auto":   "blink_green_slow",
+    "ramp_to_rc":     "blink_yellow_slow",
+    "no_rc_link":     "blink_red_fast",
+}
+
+# ─── Stick / CH6 helpers ────────────────────────────────────────────────────
+
+def ch6_wants_auto(axes: list, ch6_idx: int, threshold: float = 0.5) -> bool:
+    """
+    Return True when CH6 (aux switch) is in the AUTO position.
+
+    Typical CRSF→Joy mapping: AUX2 = axes[5].
+    Switch up (high) = +1.0 → AUTO; switch down (low) = -1.0 → RC.
+    threshold default 0.5 gives hysteresis against mid-travel noise.
+    """
+    if ch6_idx < 0 or ch6_idx >= len(axes):
+        return False
+    return float(axes[ch6_idx]) > threshold
+
+
+def sticks_active(
+    axes: list,
+    stick_indices: tuple,
+    threshold: float = 0.10,
+) -> bool:
+    """
+    Return True when any monitored stick axis exceeds threshold magnitude.
+
+    stick_indices: tuple of Joy axes indices for the four sticks
+                   (roll, pitch, throttle, yaw).  Default CRSF mapping:
+                   (0, 1, 2, 3).
+    threshold: 10 % of full deflection (issue spec).
+    """
+    return any(
+        abs(float(axes[i])) > threshold
+        for i in stick_indices
+        if i < len(axes)
+    )
+
+
+# ─── Alpha / ramp helpers ───────────────────────────────────────────────────
+
+def ramp_alpha(elapsed_s: float, ramp_s: float) -> float:
+    """
+    Compute forward blend alpha [0, 1] linearly over ramp_s seconds.
+
+    elapsed_s : seconds since ramp start.
+    ramp_s    : total ramp duration (seconds).
+    """
+    if ramp_s <= 0.0:
+        return 1.0
+    return min(1.0, elapsed_s / ramp_s)
+
+
+def alpha_for_mode(mode: str, elapsed_s: float, ramp_s: float) -> float:
+    """
+    Return the blend alpha appropriate for `mode`.
+
+      RC           → 0.0
+      RAMP_TO_AUTO → 0.0 … 1.0  (linear with elapsed)
+      AUTO         → 1.0
+      RAMP_TO_RC   → 1.0 … 0.0  (linear with elapsed)
+    """
+    if mode == "RC":
+        return 0.0
+    if mode == "AUTO":
+        return 1.0
+    if mode == "RAMP_TO_AUTO":
+        return ramp_alpha(elapsed_s, ramp_s)
+    if mode == "RAMP_TO_RC":
+        return 1.0 - ramp_alpha(elapsed_s, ramp_s)
+    return 0.0
+
+
+# ─── LED pattern ────────────────────────────────────────────────────────────
+
+def led_pattern(mode: str, slam_ok: bool, rc_link_ok: bool) -> str:
+    """Return the LED pattern name for the current operating state."""
+    if not rc_link_ok:
+        return LED_PATTERNS["no_rc_link"]
+    if mode in ("RC", "RAMP_TO_RC"):
+        return LED_PATTERNS["rc_normal"]
+    if mode == "RAMP_TO_AUTO":
+        return LED_PATTERNS["ramp_to_auto"]
+    if mode == "AUTO":
+        return LED_PATTERNS["auto_ok"] if slam_ok else LED_PATTERNS["auto_no_slam"]
+    return LED_PATTERNS["no_rc_link"]
+
+
+# ─── State machine step ──────────────────────────────────────────────────────
+
+def step_state(
+    current_mode: str,
+    ramp_elapsed_s: float,
+    ramp_s: float,
+    ch6_auto: bool,
+    sticks_now_active: bool,
+    sticks_neutral_elapsed_s: float,
+    return_delay_s: float,
+    slam_ok: bool,
+    rc_link_ok: bool,
+    override_active: bool,
+) -> tuple:
+    """
+    Compute (next_mode, next_override_active, instant_to_rc).
+
+    Parameters
+    ----------
+    current_mode            : current mode string
+    ramp_elapsed_s          : seconds since last ramp start
+    ramp_s                  : configured ramp duration
+    ch6_auto                : True = pilot switch requests AUTO
+    sticks_now_active       : True = at least one stick > override_threshold
+    sticks_neutral_elapsed_s: seconds since all sticks went neutral (0 if still active)
+    return_delay_s          : seconds of stick-neutral required to re-engage AUTO
+    slam_ok                 : True = SLAM fix is valid
+    rc_link_ok              : True = CRSF link alive
+    override_active         : True = currently in stick-override RC
+
+    Returns
+    -------
+    next_mode           : new mode string
+    next_override_active: updated override flag
+    instant_to_rc       : True when dropping to RC without ramp (safety / stick override)
+    """
+    instant_to_rc = False
+
+    # ── Safety: RC link lost → instant RC ────────────────────────────────────
+    if not rc_link_ok:
+        return "RC", False, True
+
+    # ── Compute override transitions ──────────────────────────────────────────
+    next_override = override_active
+    if sticks_now_active and current_mode in ("AUTO", "RAMP_TO_AUTO", "RAMP_TO_RC"):
+        # Stick input while AUTO or any transition → instant RC override
+        next_override = True
+        return "RC", True, True
+
+    if not sticks_now_active and sticks_neutral_elapsed_s >= return_delay_s:
+        # Sticks have been neutral long enough — clear override
+        next_override = False
+
+    # ── Ramp completion ───────────────────────────────────────────────────────
+    ramp_done = ramp_elapsed_s >= ramp_s - 1e-9   # epsilon guards float accumulation
+
+    # ── Transitions per state ─────────────────────────────────────────────────
+    if current_mode == "RC":
+        if ch6_auto and slam_ok and not next_override:
+            return "RAMP_TO_AUTO", next_override, False
+        return "RC", next_override, False
+
+    if current_mode == "RAMP_TO_AUTO":
+        if not ch6_auto:
+            return "RAMP_TO_RC", next_override, False
+        if not slam_ok:
+            return "RAMP_TO_RC", next_override, False
+        if ramp_done:
+            return "AUTO", next_override, False
+        return "RAMP_TO_AUTO", next_override, False
+
+    if current_mode == "AUTO":
+        if not ch6_auto:
+            return "RAMP_TO_RC", next_override, False
+        if not slam_ok:
+            return "RAMP_TO_RC", next_override, False
+        return "AUTO", next_override, False
+
+    if current_mode == "RAMP_TO_RC":
+        if ch6_auto and slam_ok and not next_override:
+            return "RAMP_TO_AUTO", next_override, False
+        if ramp_done:
+            return "RC", next_override, False
+        return "RAMP_TO_RC", next_override, False
+
+    return "RC", next_override, True   # unknown state → safe fallback
+
+
+# ─── cmd_vel blending ───────────────────────────────────────────────────────
+
+def blend_twist(
+    auto_linear_x: float,
+    auto_angular_z: float,
+    alpha: float,
+) -> tuple:
+    """
+    Scale autonomous cmd_vel by blend alpha.
+
+    Returns (linear_x, angular_z) for the blended output.
+    alpha=0 → zeros; alpha=1 → full autonomous.
+    """
+    a = max(0.0, min(1.0, alpha))
+    return auto_linear_x * a, auto_angular_z * a

jetson/ros2_ws/src/saltybot_mode_switch/saltybot_mode_switch/mode_switch_node.py

@@ -0,0 +1,307 @@
+"""
+mode_switch_node.py — Autonomous/RC mode switch for saltybot (Issue #104).
+
+Inputs
+------
+  /rc/joy                  (sensor_msgs/Joy)
+      CRSF channel axes:
+        axes[ch6_axis]        AUX2 — mode toggle  (+1=AUTO, -1=RC)
+        axes[stick_axes...]   Roll/Pitch/Throttle/Yaw — override detection
+
+  /saltybot/balance_state  (std_msgs/String JSON)
+      Parsed for RC link health (field "rc_link") and STM32 mode.
+
+  <slam_fix_topic>         (geometry_msgs/PoseWithCovarianceStamped)
+      Any message received within slam_fix_timeout_s → SLAM fix valid.
+
+Outputs
+-------
+  /saltybot/control_mode   (std_msgs/String JSON)
+      {
+        "mode":           "RC" | "RAMP_TO_AUTO" | "AUTO" | "RAMP_TO_RC",
+        "blend_alpha":    0.0 … 1.0,
+        "slam_ok":        bool,
+        "rc_link_ok":     bool,
+        "override_active": bool
+      }
+
+  /saltybot/led_pattern    (std_msgs/String)
+      LED pattern name consumed by the LED controller node.
+
+State machine
+-------------
+  See mode_logic.py for the full transition table.
+
+  RC           — pilot in control; blend_alpha = 0.0
+  RAMP_TO_AUTO — transitioning; blend_alpha 0→1 over ramp_duration_s (500 ms)
+  AUTO         — Jetson in control; blend_alpha = 1.0
+  RAMP_TO_RC   — transitioning; blend_alpha 1→0 over ramp_duration_s (500 ms)
+
+Safety interlocks
+-----------------
+  • Any stick > override_threshold while AUTO/RAMP_TO_AUTO → instant RC.
+  • rc_link lost (Joy silent > rc_link_timeout_s) → instant RC.
+  • AUTO not entered without valid SLAM fix.
+  • SLAM fix lost while AUTO → RAMP_TO_RC.
+
+Parameters
+----------
+  ch6_axis            Joy axes index for the mode-toggle switch   (default 5)
+  stick_axes          Space-separated Joy indices for 4 sticks    (default "0 1 2 3")
+  override_threshold  Stick magnitude for instant RC override      (default 0.10)
+  return_delay_s      Sticks-neutral duration before re-engage    (default 2.0 s)
+  ramp_duration_s     RC↔AUTO blend ramp duration                 (default 0.5 s)
+  rc_link_timeout_s   Joy silence threshold → RC link lost        (default 0.5 s)
+  slam_fix_topic      Pose topic for SLAM fix detection           (default /slam_toolbox/pose_with_covariance_stamped)
+  slam_fix_timeout_s  Pose silence threshold → SLAM fix lost      (default 3.0 s)
+  control_rate        State machine tick rate (Hz)                (default 20.0)
+"""
+
+import json
+import time
+
+import rclpy
+from rclpy.node import Node
+from rclpy.qos import HistoryPolicy, QoSProfile, ReliabilityPolicy, DurabilityPolicy
+
+from geometry_msgs.msg import PoseWithCovarianceStamped
+from sensor_msgs.msg import Joy
+from std_msgs.msg import String
+
+from saltybot_mode_switch.mode_logic import (
+    alpha_for_mode,
+    ch6_wants_auto,
+    led_pattern,
+    step_state,
+    sticks_active,
+)
+
+_BIG = 1e9   # sentinel: "never received"
+
+
+class ModeSwitchNode(Node):
+
+    def __init__(self):
+        super().__init__("mode_switch")
+
+        # ── Parameters ────────────────────────────────────────────────────────
+        self.declare_parameter("ch6_axis",           5)
+        self.declare_parameter("stick_axes",         "0 1 2 3")
+        self.declare_parameter("override_threshold",  0.10)
+        self.declare_parameter("return_delay_s",      2.0)
+        self.declare_parameter("ramp_duration_s",     0.5)
+        self.declare_parameter("rc_link_timeout_s",   0.5)
+        self.declare_parameter("slam_fix_topic",
+                               "/slam_toolbox/pose_with_covariance_stamped")
+        self.declare_parameter("slam_fix_timeout_s",  3.0)
+        self.declare_parameter("control_rate",       20.0)
+
+        self._reload_params()
+
+        # ── State ─────────────────────────────────────────────────────────────
+        self._mode: str           = "RC"
+        self._override_active: bool = False
+        self._ramp_start: float   = 0.0      # monotonic time ramp began
+        self._sticks_neutral_since: float | None = None
+        self._last_joy_t: float   = 0.0      # monotonic; 0 = never
+        self._last_slam_t: float  = 0.0
+        self._joy_axes: list      = []
+        self._stm32_mode: int     = 0        # from balance_state JSON
+
+        # ── QoS ───────────────────────────────────────────────────────────────
+        best_effort = QoSProfile(
+            reliability=ReliabilityPolicy.BEST_EFFORT,
+            history=HistoryPolicy.KEEP_LAST,
+            depth=5,
+        )
+        reliable = QoSProfile(
+            reliability=ReliabilityPolicy.RELIABLE,
+            history=HistoryPolicy.KEEP_LAST,
+            depth=10,
+        )
+        # SLAM toolbox uses transient_local for its pose topic
+        transient = QoSProfile(
+            reliability=ReliabilityPolicy.RELIABLE,
+            history=HistoryPolicy.KEEP_LAST,
+            durability=DurabilityPolicy.TRANSIENT_LOCAL,
+            depth=1,
+        )
+
+        # ── Subscriptions ─────────────────────────────────────────────────────
+        self.create_subscription(Joy, "/rc/joy", self._joy_cb, best_effort)
+        self.create_subscription(
+            String, "/saltybot/balance_state", self._balance_cb, reliable)
+
+        slam_topic = self.get_parameter("slam_fix_topic").value
+        # Try transient_local first (SLAM Toolbox); fall back gracefully if not
+        try:
+            self.create_subscription(
+                PoseWithCovarianceStamped, slam_topic,
+                self._slam_cb, transient)
+        except Exception:
+            self.create_subscription(
+                PoseWithCovarianceStamped, slam_topic,
+                self._slam_cb, reliable)
+
+        # ── Publishers ────────────────────────────────────────────────────────
+        self._mode_pub = self.create_publisher(
+            String, "/saltybot/control_mode", reliable)
+        self._led_pub  = self.create_publisher(
+            String, "/saltybot/led_pattern", reliable)
+
+        # ── Control timer ─────────────────────────────────────────────────────
+        self._timer = self.create_timer(
+            1.0 / self._p["control_rate"], self._tick)
+
+        self.get_logger().info(
+            f"ModeSwitchNode ready  "
+            f"ramp={self._p['ramp_duration_s']:.1f}s  "
+            f"override_thresh={self._p['override_threshold']:.0%}  "
+            f"return_delay={self._p['return_delay_s']:.1f}s  "
+            f"slam_topic={slam_topic}"
+        )
+
+    # ── Parameter reload ──────────────────────────────────────────────────────
+
+    def _reload_params(self):
+        raw_stick_axes = self.get_parameter("stick_axes").value
+        self._p = {
+            "ch6_axis":           int(self.get_parameter("ch6_axis").value),
+            "stick_indices":      tuple(int(x) for x in raw_stick_axes.split()),
+            "override_threshold": float(self.get_parameter("override_threshold").value),
+            "return_delay_s":     float(self.get_parameter("return_delay_s").value),
+            "ramp_duration_s":    float(self.get_parameter("ramp_duration_s").value),
+            "rc_link_timeout_s":  float(self.get_parameter("rc_link_timeout_s").value),
+            "slam_fix_timeout_s": float(self.get_parameter("slam_fix_timeout_s").value),
+            "control_rate":       float(self.get_parameter("control_rate").value),
+        }
+
+    # ── Callbacks ─────────────────────────────────────────────────────────────
+
+    def _joy_cb(self, msg: Joy) -> None:
+        self._joy_axes  = list(msg.axes)
+        self._last_joy_t = time.monotonic()
+
+    def _balance_cb(self, msg: String) -> None:
+        try:
+            data = json.loads(msg.data)
+            # "mode" is a label string; map back to int for reference
+            mode_label = data.get("mode", "RC_MANUAL")
+            self._stm32_mode = {"RC_MANUAL": 0, "RC_ASSISTED": 1,
+                                "AUTONOMOUS": 2}.get(mode_label, 0)
+        except (json.JSONDecodeError, TypeError):
+            pass
+
+    def _slam_cb(self, msg: PoseWithCovarianceStamped) -> None:
+        self._last_slam_t = time.monotonic()
+
+    # ── 20 Hz control tick ────────────────────────────────────────────────────
+
+    def _tick(self) -> None:
+        self._reload_params()
+        now = time.monotonic()
+        p   = self._p
+
+        # ── Derive inputs ─────────────────────────────────────────────────────
+        joy_age    = (now - self._last_joy_t) if self._last_joy_t > 0.0 else _BIG
+        slam_age   = (now - self._last_slam_t) if self._last_slam_t > 0.0 else _BIG
+        rc_link_ok = joy_age <= p["rc_link_timeout_s"]
+        slam_ok    = slam_age <= p["slam_fix_timeout_s"]
+
+        axes = self._joy_axes
+        ch6  = ch6_wants_auto(axes, p["ch6_axis"])
+        active = sticks_active(axes, p["stick_indices"], p["override_threshold"])
+
+        # Track neutral timer
+        if active:
+            self._sticks_neutral_since = None
+        elif self._sticks_neutral_since is None:
+            self._sticks_neutral_since = now
+
+        neutral_elapsed = (
+            (now - self._sticks_neutral_since)
+            if self._sticks_neutral_since is not None else 0.0
+        )
+
+        ramp_elapsed = now - self._ramp_start
+
+        # ── State machine step ────────────────────────────────────────────────
+        prev_mode = self._mode
+        next_mode, next_override, instant = step_state(
+            current_mode=self._mode,
+            ramp_elapsed_s=ramp_elapsed,
+            ramp_s=p["ramp_duration_s"],
+            ch6_auto=ch6,
+            sticks_now_active=active,
+            sticks_neutral_elapsed_s=neutral_elapsed,
+            return_delay_s=p["return_delay_s"],
+            slam_ok=slam_ok,
+            rc_link_ok=rc_link_ok,
+            override_active=self._override_active,
+        )
+
+        if next_mode != prev_mode:
+            # Log transitions; reset ramp timer on ramp entry
+            self.get_logger().info(
+                f"Mode: {prev_mode} → {next_mode}"
+                f"  slam_ok={slam_ok} rc_link={rc_link_ok}"
+                f"  override={next_override}"
+            )
+            if next_mode in ("RAMP_TO_AUTO", "RAMP_TO_RC"):
+                self._ramp_start = now
+                ramp_elapsed = 0.0
+
+        self._mode            = next_mode
+        self._override_active = next_override
+
+        # ── Compute blend alpha ───────────────────────────────────────────────
+        current_ramp_elapsed = now - self._ramp_start
+        alpha = alpha_for_mode(next_mode, current_ramp_elapsed, p["ramp_duration_s"])
+        if instant:
+            alpha = 0.0
+
+        # ── Log steady warnings ───────────────────────────────────────────────
+        if next_mode == "AUTO" and not slam_ok:
+            self.get_logger().warn(
+                "In AUTO but SLAM fix lost — will exit AUTO",
+                throttle_duration_sec=2.0)
+        if not rc_link_ok:
+            self.get_logger().warn(
+                "RC link lost — forced RC mode",
+                throttle_duration_sec=1.0)
+
+        # ── Publish /saltybot/control_mode ────────────────────────────────────
+        payload = {
+            "mode":            next_mode,
+            "blend_alpha":     round(alpha, 4),
+            "slam_ok":         slam_ok,
+            "rc_link_ok":      rc_link_ok,
+            "override_active": next_override,
+        }
+        mode_msg      = String()
+        mode_msg.data = json.dumps(payload)
+        self._mode_pub.publish(mode_msg)
+
+        # ── Publish /saltybot/led_pattern ─────────────────────────────────────
+        pattern      = led_pattern(next_mode, slam_ok, rc_link_ok)
+        led_msg      = String()
+        led_msg.data = pattern
+        self._led_pub.publish(led_msg)
+
+
+# ── Entry point ───────────────────────────────────────────────────────────────
+
+def main(args=None):
+    rclpy.init(args=args)
+    node = ModeSwitchNode()
+    try:
+        rclpy.spin(node)
+    except KeyboardInterrupt:
+        pass
+    finally:
+        node.destroy_node()
+        rclpy.try_shutdown()
+
+
+if __name__ == "__main__":
+    main()

jetson/ros2_ws/src/saltybot_mode_switch/setup.cfg

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

jetson/ros2_ws/src/saltybot_mode_switch/setup.py

@@ -0,0 +1,33 @@
+from setuptools import setup, find_packages
+import os
+from glob import glob
+
+package_name = "saltybot_mode_switch"
+
+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="Autonomous/RC mode switch with 500 ms blend ramp (Issue #104)",
+    license="MIT",
+    tests_require=["pytest"],
+    entry_points={
+        "console_scripts": [
+            f"mode_switch_node = {package_name}.mode_switch_node:main",
+            f"cmd_vel_mux_node = {package_name}.cmd_vel_mux_node:main",
+        ],
+    },
+)

jetson/ros2_ws/src/saltybot_mode_switch/test/test_mode_logic.py

@@ -0,0 +1,481 @@
+"""
+test_mode_logic.py — Unit tests for saltybot_mode_switch pure logic.
+
+Covers:
+  - ch6_wants_auto / sticks_active helpers
+  - ramp_alpha / alpha_for_mode
+  - led_pattern
+  - step_state state machine (all transitions)
+  - blend_twist
+
+No ROS2 runtime required.
+"""
+
+import sys
+import os
+import math
+
+sys.path.insert(0, os.path.join(os.path.dirname(__file__), ".."))
+
+import pytest
+
+from saltybot_mode_switch.mode_logic import (
+    alpha_for_mode,
+    blend_twist,
+    ch6_wants_auto,
+    led_pattern,
+    ramp_alpha,
+    step_state,
+    sticks_active,
+)
+
+# ─────────────────────────────────────────────────────────────────────────────
+# Helpers
+# ─────────────────────────────────────────────────────────────────────────────
+
+def _step(
+    mode="RC",
+    ramp_elapsed=0.0,
+    ramp_s=0.5,
+    ch6_auto=False,
+    sticks_now_active=False,
+    neutral_elapsed=0.0,
+    return_delay=2.0,
+    slam_ok=True,
+    rc_link_ok=True,
+    override_active=False,
+):
+    return step_state(
+        current_mode=mode,
+        ramp_elapsed_s=ramp_elapsed,
+        ramp_s=ramp_s,
+        ch6_auto=ch6_auto,
+        sticks_now_active=sticks_now_active,
+        sticks_neutral_elapsed_s=neutral_elapsed,
+        return_delay_s=return_delay,
+        slam_ok=slam_ok,
+        rc_link_ok=rc_link_ok,
+        override_active=override_active,
+    )
+
+
+# ─────────────────────────────────────────────────────────────────────────────
+# ch6_wants_auto
+# ─────────────────────────────────────────────────────────────────────────────
+
+class TestCh6WantsAuto:
+
+    def test_high_value_returns_true(self):
+        assert ch6_wants_auto([0.0, 0.0, 0.0, 0.0, 0.0, 1.0], ch6_idx=5) is True
+
+    def test_low_value_returns_false(self):
+        assert ch6_wants_auto([0.0, 0.0, 0.0, 0.0, 0.0, -1.0], ch6_idx=5) is False
+
+    def test_mid_value_below_threshold(self):
+        assert ch6_wants_auto([0.0, 0.0, 0.0, 0.0, 0.0, 0.4], ch6_idx=5) is False
+
+    def test_exactly_at_threshold(self):
+        # 0.5 > 0.5 is False; require strictly above
+        assert ch6_wants_auto([0.0, 0.0, 0.0, 0.0, 0.0, 0.5], ch6_idx=5) is False
+
+    def test_just_above_threshold(self):
+        assert ch6_wants_auto([0.0, 0.0, 0.0, 0.0, 0.0, 0.51], ch6_idx=5) is True
+
+    def test_out_of_range_index(self):
+        assert ch6_wants_auto([1.0], ch6_idx=10) is False
+
+    def test_empty_axes(self):
+        assert ch6_wants_auto([], ch6_idx=0) is False
+
+    def test_custom_threshold(self):
+        assert ch6_wants_auto([0.0, 0.3], ch6_idx=1, threshold=0.2) is True
+        assert ch6_wants_auto([0.0, 0.3], ch6_idx=1, threshold=0.5) is False
+
+
+# ─────────────────────────────────────────────────────────────────────────────
+# sticks_active
+# ─────────────────────────────────────────────────────────────────────────────
+
+class TestSticksActive:
+
+    def test_all_zero_not_active(self):
+        assert sticks_active([0.0, 0.0, 0.0, 0.0], (0, 1, 2, 3)) is False
+
+    def test_one_stick_above_threshold(self):
+        assert sticks_active([0.0, 0.15, 0.0, 0.0], (0, 1, 2, 3)) is True
+
+    def test_all_exactly_at_threshold_not_active(self):
+        axes = [0.10, 0.10, 0.10, 0.10]
+        assert sticks_active(axes, (0, 1, 2, 3), threshold=0.10) is False
+
+    def test_just_above_threshold_active(self):
+        assert sticks_active([0.0, 0.0, 0.0, 0.11], (0, 1, 2, 3)) is True
+
+    def test_negative_deflection_detected(self):
+        assert sticks_active([-0.5, 0.0, 0.0, 0.0], (0,)) is True
+
+    def test_out_of_range_index_skipped(self):
+        # index 5 is out of range for a 4-element list — should not crash
+        assert sticks_active([0.0, 0.0, 0.0, 0.0], (0, 1, 5)) is False
+
+    def test_only_monitored_axes_checked(self):
+        # axes[4] = 0.9 but only (0,1,2,3) are monitored
+        assert sticks_active([0.0, 0.0, 0.0, 0.0, 0.9], (0, 1, 2, 3)) is False
+
+
+# ─────────────────────────────────────────────────────────────────────────────
+# ramp_alpha / alpha_for_mode
+# ─────────────────────────────────────────────────────────────────────────────
+
+class TestRampAlpha:
+
+    def test_zero_elapsed_returns_zero(self):
+        assert ramp_alpha(0.0, 0.5) == pytest.approx(0.0)
+
+    def test_full_elapsed_returns_one(self):
+        assert ramp_alpha(0.5, 0.5) == pytest.approx(1.0)
+
+    def test_beyond_duration_clamped_to_one(self):
+        assert ramp_alpha(1.0, 0.5) == pytest.approx(1.0)
+
+    def test_halfway(self):
+        assert ramp_alpha(0.25, 0.5) == pytest.approx(0.5)
+
+    def test_zero_duration_returns_one(self):
+        assert ramp_alpha(0.0, 0.0) == pytest.approx(1.0)
+
+
+class TestAlphaForMode:
+
+    def test_rc_returns_zero(self):
+        assert alpha_for_mode("RC", 0.3, 0.5) == 0.0
+
+    def test_auto_returns_one(self):
+        assert alpha_for_mode("AUTO", 0.3, 0.5) == 1.0
+
+    def test_ramp_to_auto_mid(self):
+        assert alpha_for_mode("RAMP_TO_AUTO", 0.25, 0.5) == pytest.approx(0.5)
+
+    def test_ramp_to_rc_mid(self):
+        assert alpha_for_mode("RAMP_TO_RC", 0.25, 0.5) == pytest.approx(0.5)
+
+    def test_ramp_to_auto_start(self):
+        assert alpha_for_mode("RAMP_TO_AUTO", 0.0, 0.5) == pytest.approx(0.0)
+
+    def test_ramp_to_rc_start(self):
+        assert alpha_for_mode("RAMP_TO_RC", 0.0, 0.5) == pytest.approx(1.0)
+
+    def test_ramp_to_auto_complete(self):
+        assert alpha_for_mode("RAMP_TO_AUTO", 0.5, 0.5) == pytest.approx(1.0)
+
+    def test_ramp_to_rc_complete(self):
+        assert alpha_for_mode("RAMP_TO_RC", 0.5, 0.5) == pytest.approx(0.0)
+
+    def test_unknown_mode_returns_zero(self):
+        assert alpha_for_mode("UNKNOWN", 0.3, 0.5) == 0.0
+
+
+# ─────────────────────────────────────────────────────────────────────────────
+# led_pattern
+# ─────────────────────────────────────────────────────────────────────────────
+
+class TestLedPattern:
+
+    def test_no_rc_link_overrides_all(self):
+        for mode in ("RC", "RAMP_TO_AUTO", "AUTO", "RAMP_TO_RC"):
+            assert led_pattern(mode, slam_ok=True, rc_link_ok=False) == "blink_red_fast"
+
+    def test_rc_mode_yellow(self):
+        assert led_pattern("RC", slam_ok=True, rc_link_ok=True) == "solid_yellow"
+
+    def test_ramp_to_rc_yellow(self):
+        assert led_pattern("RAMP_TO_RC", slam_ok=True, rc_link_ok=True) == "solid_yellow"
+
+    def test_auto_ok_green(self):
+        assert led_pattern("AUTO", slam_ok=True, rc_link_ok=True) == "solid_green"
+
+    def test_auto_no_slam_orange(self):
+        assert led_pattern("AUTO", slam_ok=False, rc_link_ok=True) == "blink_orange_fast"
+
+    def test_ramp_to_auto_blink_green(self):
+        assert led_pattern("RAMP_TO_AUTO", slam_ok=True, rc_link_ok=True) == "blink_green_slow"
+
+    def test_rc_slam_state_irrelevant(self):
+        p1 = led_pattern("RC", slam_ok=True,  rc_link_ok=True)
+        p2 = led_pattern("RC", slam_ok=False, rc_link_ok=True)
+        assert p1 == p2  # SLAM doesn't affect RC LED
+
+
+# ─────────────────────────────────────────────────────────────────────────────
+# step_state — RC transitions
+# ─────────────────────────────────────────────────────────────────────────────
+
+class TestStepStateFromRC:
+
+    def test_rc_stays_rc_by_default(self):
+        mode, override, instant = _step(mode="RC")
+        assert mode == "RC"
+        assert not instant
+
+    def test_rc_to_ramp_when_ch6_auto(self):
+        mode, override, instant = _step(mode="RC", ch6_auto=True)
+        assert mode == "RAMP_TO_AUTO"
+        assert not instant
+
+    def test_rc_stays_rc_when_ch6_auto_but_no_slam(self):
+        mode, _, _ = _step(mode="RC", ch6_auto=True, slam_ok=False)
+        assert mode == "RC"
+
+    def test_rc_stays_rc_when_ch6_auto_but_no_rc_link(self):
+        mode, _, instant = _step(mode="RC", ch6_auto=True, rc_link_ok=False)
+        assert mode == "RC"
+        assert instant  # RC link lost → instant
+
+    def test_rc_stays_rc_when_override_active_and_sticks_not_neutral(self):
+        mode, _, _ = _step(
+            mode="RC", ch6_auto=True, slam_ok=True,
+            override_active=True, neutral_elapsed=1.0, return_delay=2.0)
+        assert mode == "RC"
+
+    def test_rc_to_ramp_after_override_clears(self):
+        # override clears after sticks neutral >= return_delay
+        mode, override, _ = _step(
+            mode="RC", ch6_auto=True, slam_ok=True,
+            override_active=True, neutral_elapsed=2.5, return_delay=2.0)
+        assert mode == "RAMP_TO_AUTO"
+        assert not override
+
+
+# ─────────────────────────────────────────────────────────────────────────────
+# step_state — RAMP_TO_AUTO transitions
+# ─────────────────────────────────────────────────────────────────────────────
+
+class TestStepStateFromRampToAuto:
+
+    def test_continues_ramping(self):
+        mode, _, _ = _step(mode="RAMP_TO_AUTO", ramp_elapsed=0.2, ramp_s=0.5,
+                           ch6_auto=True)
+        assert mode == "RAMP_TO_AUTO"
+
+    def test_completes_to_auto(self):
+        mode, _, _ = _step(mode="RAMP_TO_AUTO", ramp_elapsed=0.5, ramp_s=0.5,
+                           ch6_auto=True)
+        assert mode == "AUTO"
+
+    def test_stick_override_instant_rc(self):
+        mode, override, instant = _step(
+            mode="RAMP_TO_AUTO", ch6_auto=True,
+            sticks_now_active=True)
+        assert mode == "RC"
+        assert override is True
+        assert instant is True
+
+    def test_ch6_low_aborts_to_ramp_rc(self):
+        mode, _, _ = _step(mode="RAMP_TO_AUTO", ch6_auto=False, ramp_elapsed=0.1)
+        assert mode == "RAMP_TO_RC"
+
+    def test_slam_lost_aborts_to_ramp_rc(self):
+        mode, _, _ = _step(mode="RAMP_TO_AUTO", ch6_auto=True, slam_ok=False)
+        assert mode == "RAMP_TO_RC"
+
+    def test_rc_link_lost_instant_rc(self):
+        mode, _, instant = _step(mode="RAMP_TO_AUTO", ch6_auto=True, rc_link_ok=False)
+        assert mode == "RC"
+        assert instant is True
+
+
+# ─────────────────────────────────────────────────────────────────────────────
+# step_state — AUTO transitions
+# ─────────────────────────────────────────────────────────────────────────────
+
+class TestStepStateFromAuto:
+
+    def test_auto_stays_auto(self):
+        mode, _, _ = _step(mode="AUTO", ch6_auto=True)
+        assert mode == "AUTO"
+
+    def test_stick_override_instant_rc(self):
+        mode, override, instant = _step(
+            mode="AUTO", ch6_auto=True, sticks_now_active=True)
+        assert mode == "RC"
+        assert override is True
+        assert instant is True
+
+    def test_ch6_low_ramp_to_rc(self):
+        mode, _, instant = _step(mode="AUTO", ch6_auto=False)
+        assert mode == "RAMP_TO_RC"
+        assert not instant
+
+    def test_slam_lost_ramp_to_rc(self):
+        mode, _, instant = _step(mode="AUTO", ch6_auto=True, slam_ok=False)
+        assert mode == "RAMP_TO_RC"
+        assert not instant
+
+    def test_rc_link_lost_instant_rc(self):
+        mode, _, instant = _step(mode="AUTO", ch6_auto=True, rc_link_ok=False)
+        assert mode == "RC"
+        assert instant is True
+
+
+# ─────────────────────────────────────────────────────────────────────────────
+# step_state — RAMP_TO_RC transitions
+# ─────────────────────────────────────────────────────────────────────────────
+
+class TestStepStateFromRampToRC:
+
+    def test_continues_ramping(self):
+        mode, _, _ = _step(mode="RAMP_TO_RC", ramp_elapsed=0.2, ramp_s=0.5)
+        assert mode == "RAMP_TO_RC"
+
+    def test_completes_to_rc(self):
+        mode, _, _ = _step(mode="RAMP_TO_RC", ramp_elapsed=0.5, ramp_s=0.5)
+        assert mode == "RC"
+
+    def test_stick_override_instant_rc(self):
+        mode, override, instant = _step(
+            mode="RAMP_TO_RC", sticks_now_active=True, ch6_auto=True)
+        assert mode == "RC"
+        assert instant is True
+
+    def test_ch6_still_auto_no_override_cancels_back(self):
+        # CH6 still says AUTO, no override → cancel to RAMP_TO_AUTO
+        mode, _, _ = _step(mode="RAMP_TO_RC", ch6_auto=True, slam_ok=True,
+                           override_active=False, ramp_elapsed=0.1)
+        assert mode == "RAMP_TO_AUTO"
+
+    def test_rc_link_lost_instant_rc(self):
+        _, _, instant = _step(mode="RAMP_TO_RC", rc_link_ok=False)
+        assert instant is True
+
+
+# ─────────────────────────────────────────────────────────────────────────────
+# blend_twist
+# ─────────────────────────────────────────────────────────────────────────────
+
+class TestBlendTwist:
+
+    def test_alpha_zero_zeros_output(self):
+        lx, az = blend_twist(1.0, 0.5, 0.0)
+        assert lx == pytest.approx(0.0)
+        assert az == pytest.approx(0.0)
+
+    def test_alpha_one_passes_through(self):
+        lx, az = blend_twist(0.4, -0.7, 1.0)
+        assert lx == pytest.approx(0.4)
+        assert az == pytest.approx(-0.7)
+
+    def test_alpha_half_scales(self):
+        lx, az = blend_twist(1.0, 1.0, 0.5)
+        assert lx == pytest.approx(0.5)
+        assert az == pytest.approx(0.5)
+
+    def test_alpha_above_one_clamped(self):
+        lx, az = blend_twist(1.0, 1.0, 1.5)
+        assert lx == pytest.approx(1.0)
+        assert az == pytest.approx(1.0)
+
+    def test_alpha_below_zero_clamped(self):
+        lx, az = blend_twist(1.0, 1.0, -0.5)
+        assert lx == pytest.approx(0.0)
+        assert az == pytest.approx(0.0)
+
+    def test_negative_cmd_vel_preserved(self):
+        lx, az = blend_twist(-0.5, -1.0, 0.5)
+        assert lx == pytest.approx(-0.25)
+        assert az == pytest.approx(-0.5)
+
+
+# ─────────────────────────────────────────────────────────────────────────────
+# Integration scenarios
+# ─────────────────────────────────────────────────────────────────────────────
+
+class TestIntegrationScenarios:
+
+    def test_full_rc_to_auto_ramp_cycle(self):
+        """Simulate RC→RAMP_TO_AUTO→AUTO with 0.5s ramp at 20Hz."""
+        mode      = "RC"
+        override  = False
+        ramp_s    = 0.5
+        dt        = 1.0 / 20.0
+        t_ramp    = 0.0
+
+        # CH6 flipped to AUTO, SLAM ok
+        mode, override, instant = step_state(
+            mode, 0.0, ramp_s, True, False, 5.0, 2.0, True, True, override)
+        assert mode == "RAMP_TO_AUTO"
+        t_ramp = 0.0
+
+        # Simulate 10 ticks (0.5s) → should complete
+        for tick in range(10):
+            t_ramp += dt
+            mode, override, instant = step_state(
+                mode, t_ramp, ramp_s, True, False, 5.0, 2.0, True, True, override)
+
+        assert mode == "AUTO"
+        assert alpha_for_mode("AUTO", t_ramp, ramp_s) == 1.0
+
+    def test_stick_override_then_return_to_auto(self):
+        """Stick override while AUTO → RC → neutral 2s → back to AUTO."""
+        # Start in AUTO
+        mode, override, instant = _step(
+            mode="AUTO", ch6_auto=True, sticks_now_active=True)
+        assert mode == "RC"
+        assert override is True
+        assert instant is True
+
+        # Sticks neutral but < 2s — stays RC
+        mode, override, _ = _step(
+            mode="RC", ch6_auto=True, slam_ok=True,
+            override_active=True, neutral_elapsed=1.0, return_delay=2.0)
+        assert mode == "RC"
+
+        # Sticks neutral >= 2s — override clears, enters ramp
+        mode, override, _ = _step(
+            mode="RC", ch6_auto=True, slam_ok=True,
+            override_active=True, neutral_elapsed=2.0, return_delay=2.0)
+        assert mode == "RAMP_TO_AUTO"
+        assert override is False
+
+    def test_slam_lost_during_auto_graceful_exit(self):
+        """SLAM fix lost during AUTO → graceful RAMP_TO_RC (not instant)."""
+        mode, _, instant = _step(mode="AUTO", ch6_auto=True, slam_ok=False)
+        assert mode == "RAMP_TO_RC"
+        assert not instant
+
+    def test_rc_link_lost_always_instant(self):
+        """RC link lost from any state → instant RC."""
+        for start_mode in ("AUTO", "RAMP_TO_AUTO", "RAMP_TO_RC"):
+            mode, _, instant = _step(mode=start_mode, ch6_auto=True, rc_link_ok=False)
+            assert mode == "RC", f"from {start_mode}"
+            assert instant is True, f"from {start_mode}"
+
+    def test_blend_alpha_sequence_through_ramp(self):
+        """Blend alpha is monotonically increasing during RAMP_TO_AUTO."""
+        ramp_s = 0.5
+        alphas = [
+            alpha_for_mode("RAMP_TO_AUTO", t * 0.05, ramp_s)
+            for t in range(11)   # 0.0 to 0.5 s in 0.05 s steps
+        ]
+        for a, b in zip(alphas, alphas[1:]):
+            assert b >= a
+
+    def test_ramp_to_rc_alpha_decreasing(self):
+        """Blend alpha is monotonically decreasing during RAMP_TO_RC."""
+        ramp_s = 0.5
+        alphas = [
+            alpha_for_mode("RAMP_TO_RC", t * 0.05, ramp_s)
+            for t in range(11)
+        ]
+        for a, b in zip(alphas, alphas[1:]):
+            assert b <= a
+
+    def test_no_auto_without_slam(self):
+        """From RC, ch6=AUTO but slam=False → cannot enter AUTO path."""
+        mode, _, _ = _step(mode="RC", ch6_auto=True, slam_ok=False)
+        assert mode == "RC"
+
+    def test_no_auto_without_rc_link(self):
+        """From RC, ch6=AUTO but rc_link=False → stays RC (instant)."""
+        mode, _, instant = _step(mode="RC", ch6_auto=True, rc_link_ok=False)
+        assert mode == "RC"
+        assert instant is True