Basic_ROS2_LiDAR

The "lidar_processing" package provides a basic ROS 2 node for processing LiDAR point cloud data.

Features

• Subscribes to a given topic to receive LiDAR data.
• Converts ROS 2 PointCloud2 messages to Open3D point clouds (2 possibilities: convert only xyz coords, or all the fields. Check the available functions in src/point_cloud_utils.py).
• Filters points based on a configurable distance threshold.
• Publishes the filtered point cloud to /filtered_pointcloud topic.
• Provides configurable parameters for real-time constraints, debugging, distance filtering, and topic/frame_id selection.

Code Structure

lidar_processing/
├── __init__.py               # Exposes module components
├── lidar_node.py             # Main node for processing point clouds
├── main.py                   # Entry point to start the ROS2 node
├── point_cloud_utils.py      # Utility functions for point cloud conversions

Requirements

• ROS 2 Humble

• Open3D

• NumPy

• sensor_msgs_py

    pip install numpy open3d sensor_msgs_py 
  

Compilation

After cloning the repo in a desired location:

cd Basic_ROS2_LiDAR/ros2_ws
colcon build
source install/setup.bash

Configurable Parameters

Parameter	Default Value	Description
real_time_constraint	0.1 s	Maximum allowed processing time per message.
debug_mode	False	Enables debug logs with processing times.
input_topic	/velodyne_points	Topic name for incoming LiDAR point clouds.
frame_id	velodyne	Frame ID for incoming LiDAR point clouds.
distance_threshold	10.0 meters	Maximum distance from the origin to keep points.

A launch file (lidar_processing.launch.py) is provided to start the node with desired parameters.

To check the topic name or the frame_id of your rosbag, use the following commands:

# Topic name
ros2 bag info [bag_file]

# Frame ID (while running the rosbag):
ros2 topic echo /[topic_name]

Execution

# Terminal 1 (program)
ros2 launch lidar_processing lidar_processor_launch.py

# Terminal 2 (dataset)
ros2 bag play <bag_file_path> [--loop]

# Terminal 3 (visualization)
rviz2

Checking Real-Time Performance

• Use debug_mode ('debug_mode': True in launch/lidar_processor_launch.py)
• With debugging enabled, the node will print timing information. If the total processing time exceeds real_time_constraint, a warning is issued.
  • Real-time performance evaluation:
    • If processing time ≤ frame interval (1 / Message Frequency), the algorithm meets real-time.
    • If processing time > frame interval (1 / Message Frequency), it may cause delays or frame drops.

Checking the sensor/rosbag message frequency

ros2 topic hz /[topic_name]

Output example (~10MHz):

    average rate: 9.705
    min: 0.091s max: 0.109s std dev: 0.00513s window: 11
    [...]

Demo

Results of the LiDAR filtering process: original point cloud (left) vs filtered by distance point cloud (right).

Tested with:

• OS: Ubuntu 22.04 LTS.
• Python: v3.10.12
• Dataset: KITTI.

Intensity view

XYZ View

*NOTE: Depending on the application, converting only the 3D coordinates can be very convenient. Native Python conversions are very costly and will impact the ability to meet real-time processing constraints. Using C++ bindings for this task can significantly improve performance. Other techniques that can be used when working with Python include avoiding loops, using optimized libraries, or downsampling the point cloud if maximum precision is not a strict requirement.

Maintainer

Silvia R. Alcaraz (@silviaralcaraz)