Skip to content

Sensor Troubleshooting#

This guide provides solutions to common sensor integration issues in AutoSDV.

Quick Diagnostic Commands#

Before diving into specific issues, run these commands to gather information:

# List all active topics
ros2 topic list

# Check sensor topics specifically
ros2 topic list | grep /sensing

# View TF tree
ros2 run tf2_tools view_frames
evince frames.pdf

# Check node status
ros2 node list

# View logs with debug level
ros2 launch autosdv_launch autosdv.launch.yaml --log-level DEBUG

General Issues#

No Sensor Data#

Symptoms: - Topics exist but ros2 topic echo shows nothing - ros2 topic hz shows 0 Hz - Driver launches but no data published

Diagnostic Steps:

  1. Check physical connections: 

    # USB devices
lsusb
# Should list your sensor

# Serial devices
ls -l /dev/tty*
# Should show device (e.g., /dev/ttyUSB0)

# Network devices
ip addr show
# Verify Ethernet interface up

# Test network sensors
ping <sensor_ip>
# Should respond with <5ms latency

  2. Verify device permissions: 

    # Add user to dialout group (for serial)
sudo usermod -aG dialout $USER

# Add user to video group (for cameras)
sudo usermod -aG video $USER

# Log out and back in for changes to take effect

  3. Check driver logs: 

    # Launch with debug logging
ros2 launch autosdv_launch autosdv.launch.yaml --log-level DEBUG

# Or check specific node
ros2 node info /sensing/lidar/robin_lidar/driver_node

  4. Verify power:

    • Check sensor LED indicators
    • Measure voltage with multimeter
    • Ensure power supply has sufficient current

Solutions by Sensor Type: - LiDAR: Check network settings, IP configuration - Camera: Verify ZED SDK installed, GMSL connection - IMU: Check I2C bus, device address - GNSS: Ensure clear sky view, wait for satellite acquisition

Wrong Coordinate Frame / Sensor in Wrong Location#

Symptoms: - Sensor data visible in RViz but in wrong position - Point cloud upside down or rotated incorrectly - Transform errors in RViz

Diagnostic Steps:

  1. Verify TF transforms: 

    # Generate TF tree
ros2 run tf2_tools view_frames
evince frames.pdf

# Check specific transform
ros2 run tf2_ros tf2_echo base_link sensor_link
# Compare output with calibration values

  2. Check calibration file: 

    # View current calibration
cat src/sensor_kit/autosdv_sensor_kit_description/config/sensor_kit_calibration.yaml

# Look for your sensor's x, y, z, roll, pitch, yaw values

Solutions:

  1. Fix calibration values: 

    # Edit: sensor_kit_calibration.yaml
sensor_kit_base_link:
  your_sensor_link:
    x: 0.15      # Forward/backward (meters)
    y: 0.02      # Left/right (meters)
    z: 0.20      # Up/down (meters)
    roll: 0.0    # Rotation around X-axis (radians)
    pitch: 0.0   # Rotation around Y-axis (radians)
    yaw: 0.0     # Rotation around Z-axis (radians)

  2. Understand coordinate systems:

    • ROS Standard (REP-103): X=forward, Y=left, Z=up
    • Some sensors use different conventions (see sensor guides)
    • Calculate rotation needed to align sensor → ROS standard

  3. Iteratively adjust in RViz:

    • Launch system and view in RViz
    • Adjust roll/pitch/yaw in calibration file
    • Rebuild and relaunch to see changes
    • Repeat until correct

Common Rotations: - Robin-W LiDAR: roll=3.14159, pitch=-1.5708, yaw=0.0 - ZED Camera: roll=0.0, pitch=0.0, yaw=0.0 (standard)

Topic Namespace Issues#

Symptoms: - Topics appear in unexpected namespaces - Cannot find sensor topics in /sensing/ - Topic remapping not working

Diagnostic Steps:

  1. Check actual topic names: 

    ros2 topic list | grep -E '(lidar|camera|imu|gnss)'
# Compare with expected: /sensing/[type]/[name]/[data]

  2. Verify launch file remapping: 

    # Check sensor kit launch files
cat src/sensor_kit/autosdv_sensor_kit_launch/launch/lidar.launch.xml
# Look for <remap> tags

Solutions:

  1. Add or fix topic remapping: 

    <!-- In sensor kit launch file -->
<remap from="/driver_topic_name" to="/sensing/lidar/sensor_name/points_raw"/>

  2. Use absolute topic paths: 

    <!-- Bad (relative) -->
<remap from="points" to="points_raw"/>

<!-- Good (absolute) -->
<remap from="/sensor/points" to="/sensing/lidar/sensor/points_raw"/>

  3. For ZED camera namespace issues: Use composable node loading (see Camera Sensors Guide)

Performance Issues / High CPU Usage#

Symptoms: - System lag or slow response - High CPU usage (check with htop) - Low frame rate from sensors - Dropped messages

Diagnostic Steps:

  1. Check topic bandwidth: 

    # Monitor data rate
ros2 topic bw /sensing/lidar/robin_lidar/points_raw

# Check all sensor topics
ros2 topic bw /sensing/*

  2. Check CPU usage: 

    # Real-time process monitor
htop

# GPU usage (NVIDIA)
nvidia-smi

  3. Verify CycloneDDS configuration: 

    # Check kernel buffers
sysctl net.core.rmem_max
sysctl net.ipv4.ipfrag_time
sysctl net.ipv4.ipfrag_high_thresh

# Should be configured per Installation Guide

Solutions:

  1. Configure CycloneDDS buffers (if not done): 

    cd setup
just cyclonedds-sysctl

  2. Reduce sensor data rate: 

    # Edit sensor parameters
# For cameras:
general.grab_frame_rate: 15  # Reduce from 30

# For LiDAR:
downsample_factor: 2  # Reduce points by half

  3. Disable heavy features: 

    # Disable ZED object detection
ros2 launch autosdv_launch autosdv.launch.yaml enable_zed_object_detection:=false

  4. Use lightweight sensor suite: 

    # VLP-32C + ZED IMU only (no camera processing)
ros2 launch autosdv_launch autosdv.launch.yaml sensor_suite:=vlp32c_zed_imu

Sensor-Specific Issues#

LiDAR#

No Point Cloud#

Symptoms: LiDAR driver launches but no points published

Solutions: 1. Check network connectivity (Ethernet LiDARs): 

ping <lidar_ip>
# Robin-W: 172.168.1.10
# Velodyne: 192.168.1.201
# Blickfeld: 192.168.26.26

  1. Configure Jetson IP on same subnet: 

    # For Robin-W (172.168.1.x)
sudo ip addr add 172.168.1.100/24 dev eth0
sudo ip link set eth0 up

# For Blickfeld (192.168.26.x)
sudo ip addr add 192.168.26.1/24 dev eth0
sudo ip link set eth0 up

  2. Check firewall: 

    # Disable firewall temporarily for testing
sudo ufw disable

# Or allow specific ports
sudo ufw allow from <lidar_ip>

TensorRT Compilation Takes Forever#

Symptoms: First launch takes 10-30 minutes with TensorRT messages

This is NORMAL for first launch. TensorRT compiles models for your specific GPU.

Solutions: - Be patient - only happens once - Models cached in ~/.autoware/models/ - Subsequent launches are fast

Point Cloud in Wrong Orientation#

Solution: Check coordinate frame rotation (especially Robin-W, which requires roll=180°, pitch=-90°)

See LiDAR Sensors - Coordinate Transformation

Camera (ZED)#

Camera Not Detected#

Symptoms: ZED_Explorer shows "No camera detected"

Solutions: 1. Check GMSL connection: 

# Verify ZED Link card
lspci | grep NVIDIA
# Should show ZED Link Duo

  1. Power cycle camera:

    • Unplug camera
    • Wait 10 seconds
    • Reconnect
    • Reboot Jetson

  2. Verify ZED SDK installation: 

    cat /usr/local/zed/settings/version.txt
# Should show 5.1.2

  3. Check permissions: 

    sudo usermod -aG video $USER
# Log out and back in

VNC/Remote Access Issues#

Problem: ZED camera doesn't work over standard VNC

Solution: Use TurboVNC + VirtualGL for hardware acceleration

# Install TurboVNC
sudo apt install turbovnc virtualgl

# Start VNC with VirtualGL
vglrun /opt/TurboVNC/bin/vncserver

# Launch apps with vglrun
vglrun ZED_Explorer
vglrun rviz2

Namespace Problems#

Problem: Topics in wrong namespace, can't remap

Solution: Use composable node loading instead of Python launch file

See Camera Sensors - Namespace Workaround

IMU#

MPU9250 Not Detected#

Symptoms: i2cdetect doesn't show device

Solutions: 1. Check I2C bus: 

# List I2C buses
ls /dev/i2c-*

# Scan bus 1
sudo i2cdetect -y -r 1
# Should show device at 0x68 or 0x69

  1. Verify wiring:

    • VCC → 3.3V (not 5V!)
    • GND → GND
    • SDA → Pin 3 (I2C1 SDA)
    • SCL → Pin 5 (I2C1 SCL)

  2. Check I2C address: 

    # Try alternate address
sudo i2cdetect -y -r 1
# Look for 0x68 or 0x69

Noisy IMU Data#

Symptoms: Large fluctuations, unstable orientation

Solutions: 1. Check IMU configuration: See IMU Sensors - Configuration 2. Secure mounting: Minimize vibration 3. Add software filtering: Low-pass filter in driver config

ZED IMU Not Publishing#

Symptoms: ZED camera works but no IMU topic

Solutions: 1. Enable IMU in ZED config: 

sensors.sensors_image_sync: true
sensors.publish_imu_tf: true
  1. Check IMU relay node: 
    ros2 node list | grep relay
# Should show IMU relay node running

GNSS#

No GNSS Fix#

Symptoms: Fix status remains -1 (no fix)

Solutions: 1. Ensure clear sky view: Move to open area (no buildings/trees) 2. Wait for satellite acquisition: Cold start takes 1-10 minutes 3. Check antenna connection: Verify cable secure 4. Monitor satellite count: Need minimum 4 satellites for 3D fix

No RTK Fix (u-blox)#

Symptoms: GPS fix but no RTK (status != 2)

Solutions: 1. Check NTRIP connection: 

ros2 topic hz /sensing/gnss/ntrip/rtcm
# Should show ~1 Hz
  1. Verify internet connection: NTRIP requires network
  2. Check credentials: Edit ntrip.launch.xml with correct username/password
  3. Wait for convergence: RTK can take 1-10 minutes
  4. Improve antenna placement: Higher, less obstruction

Garmin GPS Not Working#

Symptoms: No data from Garmin GPS 18x

Solutions: 1. Start gpsd: 

sudo /usr/sbin/gpsd -n -G -b /dev/ttyUSB0

  1. Check device: 

    ls -l /dev/ttyUSB*
sudo cat /dev/ttyUSB0
# Should see NMEA sentences

  2. Set baud rate: 

    sudo stty -F /dev/ttyUSB0 9600

  3. Test with cgps: 

    cgps
# Should show GPS data

Build Errors#

Package Not Found#

Symptoms: colcon build fails with "package 'X' not found"

Solutions: 1. Update rosdep: 

rosdep update
rosdep install --from-paths src --ignore-src -r -y

  1. Check package.xml dependencies: Ensure all <depend> tags correct

  2. Install missing ROS packages: 

    sudo apt install ros-humble-<package-name>

CMake Errors#

Symptoms: CMake configuration fails

Solutions: 1. Clean build: 

rm -rf build install log
colcon build

  1. Check CMakeLists.txt: Verify syntax, dependencies

  2. Source ROS 2: 

    source /opt/ros/humble/setup.bash

Getting Help#

If issues persist after trying these solutions:

1. Enable Debug Logging#

ros2 launch autosdv_launch autosdv.launch.yaml --log-level DEBUG 2>&1 | tee debug.log

2. Gather Diagnostic Information#

# System info
lsb_release -a
uname -a

# ROS info
ros2 doctor
ros2 wtf

# Topic info
ros2 topic list
ros2 topic hz /sensing/lidar/*/points_raw

# TF tree
ros2 run tf2_tools view_frames

3. Check GitHub Issues#

Search AutoSDV Issues for similar problems

4. Create Issue#

When reporting issues, include: - Hardware setup (sensor models, connections) - Launch command used - Full error messages/logs - Output of diagnostic commands - Steps to reproduce

Useful Diagnostic Tools#

ROS 2 Tools#

# List nodes
ros2 node list

# Node info
ros2 node info /node_name

# Topic info
ros2 topic info /topic_name
ros2 topic hz /topic_name
ros2 topic bw /topic_name
ros2 topic echo /topic_name

# Parameter inspection
ros2 param list /node_name
ros2 param get /node_name parameter_name

# Service calls
ros2 service list
ros2 service call /service_name service_type "{args}"

# TF debugging
ros2 run tf2_ros tf2_echo frame1 frame2
ros2 run tf2_tools view_frames

System Tools#

# Process monitoring
htop
top

# GPU monitoring
nvidia-smi
watch -n 1 nvidia-smi

# Network
ip addr show
ping <ip>
iftop -i eth0

# USB devices
lsusb
lsusb -t

# Serial devices
ls -l /dev/tty*
sudo dmesg | grep tty

# I2C devices
ls /dev/i2c-*
sudo i2cdetect -y -r 1

# Disk usage
df -h
du -sh *

For sensor-specific troubleshooting, see: - LiDAR Sensors - Robin-W, Velodyne, Blickfeld issues - Camera Sensors - ZED camera troubleshooting - IMU Sensors - MPU9250 and ZED IMU issues - GNSS Sensors - GNSS and RTK troubleshooting

For integration help: - Adding a New Sensor - Step-by-step integration guide - Using Sensors - Getting started with sensors