Skip to content

LIVE_TOPICS_USAGE.md

Latest commit

 

History

History
211 lines (163 loc) · 6.14 KB

LIVE_TOPICS_USAGE.md

File metadata and controls

211 lines (163 loc) · 6.14 KB

AirSLAM Live Topics Usage Guide

This guide explains how to use AirSLAM with live ROS topics from Intel RealSense D455 camera instead of pre-recorded datasets.

Overview

We've added live topic support to AirSLAM while preserving the original file-based functionality:

  • Original: visual_odometry - reads from EuRoC dataset files
  • NEW: visual_odometry_live - subscribes to live ROS topics

Files Added/Modified

New Files Created:

  • include/ros_dataset.h - ROS topic-based dataset header
  • src/ros_dataset.cc - ROS topic-based dataset implementation
  • demo/visual_odometry_live.cpp - Live topic visual odometry node
  • launch/visual_odometry/vo_d455_live.launch - Live D455 launch file
  • launch/visual_odometry/vo_d455_complete.launch - Complete setup launch file

Modified Files:

  • CMakeLists.txt - Added message_filters dependency and new executable

Original Files (Preserved):

  • All original files remain unchanged and functional
  • Original visual_odometry node still works with datasets

Setup Instructions

Step 1: Build the Updated AirSLAM

cd ~/catkin_ws
source /home/robot/ros_catkin_ws/install_isolated/setup.bash
catkin_make
source devel/setup.bash

Step 2: Start D455 Camera (ROS2)

In Terminal 1:

source /opt/ros/humble/setup.bash

# Start D455 with clean IR images (no projector)
ros2 run realsense2_camera realsense2_camera_node --ros-args \
  -p enable_infra1:=true \
  -p enable_infra2:=true \
  -p enable_depth:=false \
  -p enable_color:=false \
  -p enable_gyro:=true \
  -p enable_accel:=true \
  -p unite_imu_method:=2 \
  -p infra_width:=848 -p infra_height:=480 -p infra_fps:=30 \
  -p publish_tf:=true \
  -p enable_auto_exposure:=true \
  -p depth_module.emitter_enabled:=0 \
  -p depth_module.emitter_always_on:=false

Step 3: Start ROS1/ROS2 Bridge

In Terminal 2:

source /home/robot/ros_catkin_ws/install_isolated/setup.bash  # ROS1
source /opt/ros/humble/setup.bash                            # ROS2
source ~/bridge_ws/install/setup.bash                        # Bridge workspace

# Start the bridge
ros2 run ros1_bridge dynamic_bridge --bridge-all-topics

Step 4: Run AirSLAM Live

In Terminal 3:

cd ~/catkin_ws
source /home/robot/ros_catkin_ws/install_isolated/setup.bash
source devel/setup.bash

# Run live visual odometry
roslaunch air_slam vo_d455_live.launch

Configuration

Topic Names (Configurable)

The live node subscribes to these topics by default:

  • Left camera: /camera/camera/infra1/image_rect_raw
  • Right camera: /camera/camera/infra2/image_rect_raw
  • IMU data: /camera/camera/imu

To use different topic names:

roslaunch air_slam vo_d455_live.launch \
  left_image_topic:="/your/left/topic" \
  right_image_topic:="/your/right/topic" \
  imu_topic:="/your/imu/topic"

Parameters

Parameter Description Default
use_live_topics Enable topic mode (vs file mode) true
left_image_topic Left camera topic /camera/camera/infra1/image_rect_raw
right_image_topic Right camera topic /camera/camera/infra2/image_rect_raw
imu_topic IMU topic /camera/camera/imu

Technical Details

Stereo Synchronization

  • Uses message_filters for stereo image synchronization
  • Approximate time policy with 10ms tolerance
  • Handles slight timestamp differences between left/right cameras

IMU Integration

  • IMU data is buffered independently
  • Associated with image frames based on timestamps
  • IMU data between consecutive frames is grouped together

Thread Safety

  • All data buffers are mutex-protected
  • Safe for concurrent ROS callbacks and processing

Performance Considerations

  • Buffer sizes limited to prevent memory issues (100 stereo pairs, 1000 IMU samples)
  • Processing rate limited to 30 Hz to prevent overwhelming
  • Queue-based processing ensures real-time performance

Usage Examples

Basic Live SLAM

roslaunch air_slam vo_d455_live.launch

Live SLAM without RViz

roslaunch air_slam vo_d455_live.launch visualization:=false

Custom Save Directory

roslaunch air_slam vo_d455_live.launch saving_dir:="/path/to/save"

File-based Mode (Original)

# Still works with datasets
roslaunch air_slam vo_euroc.launch

Troubleshooting

No Data Received

Problem: AirSLAM starts but receives no camera data Check:

  • Bridge is running and topics are available: rostopic list
  • Camera is publishing: rostopic echo /camera/camera/infra1/image_rect_raw
  • Topic names match in launch file

Synchronization Issues

Problem: Only receiving left or right images Check:

  • Both cameras publishing at same rate
  • Timestamps are reasonable (not too different)
  • No network delays in topics

High CPU Usage

Problem: System performance issues Solution:

  • Reduce camera framerate: infra_fps:=15
  • Increase processing rate limit in code
  • Close RViz if not needed

Bridge Connection Issues

Problem: Topics not bridging from ROS2 to ROS1 Solution:

  • Ensure both ROS1 and ROS2 environments are sourced
  • Check bridge workspace is built correctly
  • Restart bridge if topics change

Comparison: File vs Live Mode

Aspect File Mode (Original) Live Mode (New)
Input EuRoC dataset files ROS topics
Timing Fixed dataset order Real-time streams
IMU Pre-synchronized Real-time sync
Performance Deterministic Variable (real-time)
Debugging Repeatable Live data only
Setup Simple Requires camera+bridge

Next Steps

After live visual odometry works:

  1. Map Refinement: Create live version of map refinement
  2. Relocalization: Add live relocalization capability
  3. ROS2 Direct: Bypass bridge with native ROS2 support
  4. Multi-camera: Extend to multiple camera systems

Backup/Restore

All original functionality is preserved. To revert:

  1. Use original launch files (vo_euroc.launch)
  2. Original visual_odometry executable unchanged
  3. All modifications are additive (commented, not deleted)

The live topic system is designed to coexist with the original file-based system.