sl-jetson
c47ac41573
- Dockerfile: L4T R32.6.1 (JetPack 4.6) base + ROS2 Humble + SLAM stack (slam_toolbox, Nav2, rplidar_ros, realsense2_camera, robot_localization) - docker-compose.yml: multi-service stack (ROS2, RPLIDAR A1M8, D435i, STM32 bridge) with device passthrough, host networking for DDS, persistent map volume - docs/pinout.md: full GPIO/I2C/UART pinout for STM32F722 bridge (USB CDC + UART fallback), RealSense D435i (USB3), RPLIDAR A1M8, udev rules - docs/power-budget.md: 10W envelope analysis with per-component breakdown, mitigation strategies (RPLIDAR gating, D435i 640p, nvpmodel modes) - scripts/setup-jetson.sh: host one-shot setup (Docker, nvidia-container-runtime, udev rules, MAXN power mode, swap) - scripts/build-and-run.sh: build/up/down/shell/slam/status helper Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
5.1 KiB
5.1 KiB
Jetson Nano Power Budget Analysis
Self-Balancing Robot — 10W Envelope
Last updated: 2026-02-28 Target: Operate within 10W SoC power envelope (MAXN 10W mode)
Power Modes
Jetson Nano supports two NVPModel power modes:
| Mode | GPU | CPU cores | CPU freq | Memory freq | TDP |
|---|---|---|---|---|---|
| MAXN (Mode 0) | 128 core | 4 | 1.43GHz | 1600MHz | 10W |
| 5W (Mode 1) | 128 core | 2 | 0.92GHz | 1600MHz | 5W |
For this robot, we target MAXN 10W mode with careful peripheral management.
# Check current mode
sudo nvpmodel -q
# Set 10W MAXN mode
sudo nvpmodel -m 0
# Set 5W mode (thermal/battery save)
sudo nvpmodel -m 1
# Monitor power in real time
sudo tegrastats
Component Power Budget
SoC (Jetson Nano Module)
| Component | Idle (W) | Load (W) | Peak (W) | Notes |
|---|---|---|---|---|
| CPU (4× Cortex-A57) | 1.0 | 3.5 | 4.0 | ROS2 + SLAM compute |
| GPU (128-core Maxwell) | 0.5 | 2.5 | 3.0 | Depth processing, ML inference |
| DDR4 RAM (4GB) | 0.3 | 0.6 | 0.8 | |
| eMMC / SD | 0.1 | 0.2 | 0.3 | |
| SoC Subtotal | 1.9 | 6.8 | 8.1 |
Peripherals (USB / GPIO)
| Peripheral | Idle (W) | Active (W) | Peak (W) | Interface | Notes |
|---|---|---|---|---|---|
| RealSense D435i | 0.3 | 1.5 | 3.5 | USB 3.1 | Peak during boot/init |
| RPLIDAR A1M8 | 0.4 | 2.6 | 3.0 | USB (UART adapter) | Motor spinning |
| STM32F722 bridge | 0.3 | 0.5 | 0.8 | USB CDC | Powered from Jetson USB |
| Peripheral Subtotal | 1.0 | 4.6 | 7.3 |
Total System (from Jetson 5V barrel jack)
| Scenario | SoC (W) | Peripherals (W) | Total (W) | Margin |
|---|---|---|---|---|
| Idle | 1.9 | 1.0 | 2.9 | +7.1W |
| Nominal (SLAM running) | 6.8 | 4.6 | 11.4 | -1.4W ⚠️ |
| Peak (all active, ML) | 8.1 | 7.3 | 15.4 | -5.4W ❌ |
Budget Compliance Strategy
The nominal load of 11.4W exceeds the 10W envelope — mitigation required:
Mitigation 1: RPLIDAR Power Gating
The RPLIDAR motor can be stopped when not scanning. The ROS2 driver handles this via DTR line.
| Mode | Savings |
|---|---|
| RPLIDAR motor off | −2.2W |
| RPLIDAR idle | 0.4W vs 2.6W |
Mitigation 2: RealSense Resolution Reduction
Lower RGB-D resolution reduces USB bandwidth and D435i processing:
| Profile | Power |
|---|---|
| 1280×720 @ 30fps | 1.5W |
| 640×480 @ 30fps | 1.1W ← Recommended |
| 424×240 @ 30fps | 0.8W |
Mitigation 3: Jetson GPU Workload Scheduling
Avoid running depth inference and SLAM simultaneously at full throttle:
# Cap GPU frequency (reduce from max 921.6MHz)
sudo jetson_clocks --show
# Set conservative clocks
echo 614400000 | sudo tee /sys/devices/17000000.gp10b/devfreq/17000000.gp10b/min_freq
Mitigation 4: STM32 Self-Powered
Power STM32 from robot's 5V bus (separate from Jetson USB rail):
| Option | Jetson USB load |
|---|---|
| STM32 powered from Jetson USB | 0.5W |
| STM32 powered from robot 5V | 0W (data only via USB) |
Revised Budget with Mitigations
Applying: 640×480 D435i + RPLIDAR gating + STM32 self-powered:
| Component | Power (W) |
|---|---|
| CPU (SLAM, 4 cores) | 3.5 |
| GPU (depth processing) | 2.0 |
| RAM + misc SoC | 1.0 |
| RealSense D435i (640×480) | 1.1 |
| RPLIDAR A1M8 (active) | 2.6 |
| STM32 bridge (self-powered) | 0.0 |
| Total | 10.2W |
Near-compliant at 10.2W. Further savings achievable by:
- Enabling RPLIDAR standby between scan cycles (−0.5W avg)
- Using 5W nvpmodel during motor-heavy phases
Input Power Requirements
Jetson Nano Power Input
| Spec | Value |
|---|---|
| Input connector | 5.5mm / 2.1mm barrel jack |
| Input voltage | 5V DC |
| Recommended current | ≥4A (20W supply for headroom) |
| Absolute max | 5.25V |
Use a 5V 4A supply minimum. A 2A supply will brownout under load.
Robot Power Architecture (Recommended)
LiPo 3S (12.6V max)
│
├─► DC-DC Buck → 5V 5A ──► Jetson Nano barrel jack
│ (e.g., XL4016)
│
├─► DC-DC Buck → 5V 3A ──► STM32 + logic 5V rail
│
└─► Hoverboard ESC ──► Hub motors (48V loop)
This isolates the Jetson 5V supply from motor switching noise.
Real-Time Monitoring
# Live power telemetry
sudo tegrastats --interval 500
# Key fields:
# POM_5V_IN X/Y — total input power (current W / average W)
# POM_5V_GPU X/Y — GPU power
# POM_5V_CPU X/Y — CPU power
# Log to file
sudo tegrastats --interval 1000 --logfile /tmp/power_log.txt &
# Parse log
grep "POM_5V_IN" /tmp/power_log.txt | \
awk '{for(i=1;i<=NF;i++) if($i=="POM_5V_IN") print $(i+1)}' | \
awk -F'/' '{sum+=$1; count++} END {print "Avg:", sum/count, "mW"}'
Summary
| Metric | Value |
|---|---|
| Target envelope | 10W |
| Nominal (no mitigation) | 11.4W |
| Nominal (with mitigations) | ~10.2W |
| Compliant scenario | RPLIDAR standby + 640p D435i |
| Recommended PSU | 5V 4A (20W) |
| Power mode | nvpmodel MAXN (Mode 0) |