Kinematic-ICP

Paper   •   Contact Us

[Kinematic-ICP](https://www.ipb.uni-bonn.de/wp-content/papercite-data/pdf/kissteam2025icra.pdf) is a LiDAR odometry approach that explicitly incorporates the kinematic constraints of mobile robots into the classic point-to-point ICP algorithm.

How to Build

Our system operates on ROS2, supporting ROS Humble, Iron, and Jazzy. To build and run Kinematic-ICP, follow these steps:

1. Clone the Repository:

   cd <your_ros_workspace>/src
   git clone https://github.com/PRBonn/kinematic-icp
   cd ..

2. Ensure all Dependencies are Installed:

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

3. Build the Workspace:

   colcon build

4. Source the Setup Script:

   source ./install/setup.bash

TF Requirements

Kinematic ICP can enhance existing odometry using a 3D LiDAR. However, there are specific requirements regarding motion and transformations since we use a kinematic motion model for the pose correction. Below are the key requirements:

1. Planar Movement: The robot is expected to move primarily on a planar surface.

2. Existing Odometry: An existing odometry source must be provided, such as the platform's wheel odometry. In the ROS ecosystem, this means that another node must publish the tf transformation between base_link and odom. (Note: The names may vary and can be adjusted in the pipeline parameters.)

3. Static Transform for LiDAR: To utilize the platform's motion model effectively, the system needs to compute the pose in base_link. Therefore, a static tf transform between base_link and the LiDAR frame (extrinsic calibration) is required. If this calibration is significantly inaccurate, it may compromise system performance.

Finally, Kinematic ICP will publish a new tf transformation between base_link and odom_lidar.

Running the System

This system offers two entry points for deployment, depending on your use case: one for real-time operation and one for offline processing.

1. Real-Time Deployment: online_node

Use the online_node to run the system on a robotics platform. The only required parameter is the lidar_topic. You can start the system using the following command:

ros2 launch kinematic_icp online_node.launch.py lidar_topic:=<TOPIC>

To enable simultaneous visualization through RViz, use the visualize flag set to true:

ros2 launch kinematic_icp online_node.launch.py lidar_topic:=<TOPIC> visualize:=true

2. Offline Processing: offline_node

For post-processing and analysis, the offline_node processes a ROS bag file at CPU speed, ensuring no frames are dropped. This mode is ideal for reviewing trajectory results, debugging, and speeding up bag file processing. You can launch the offline node with the following command:

ros2 launch kinematic_icp offline_node.launch.py lidar_topic:=<TOPIC> bag_filename:=<ROSBAG>

RViz can also be used in this mode by setting the visualize flag to true. Additionally, the system will output a file in TUM format containing the estimated poses, named \_kinematic_poses_tum.txt. This file is saved in the same directory as the ROS bag file by default.

To specify a custom directory for the output file, use the output_dir parameter:

ros2 launch kinematic_icp offline_node.launch.py lidar_topic:=<TOPIC> bag_filename:=<ROSBAG> output_dir:=<OUTPUT_DIRECTORY>

2D LiDAR support

You can run both the online_node and the offline_node on a 2D Laser topic (with message type LaserScan) by setting the use_2d_lidar flag to true. For example:

ros2 launch kinematic_icp online_node.launch.py lidar_topic:=<LASER_2D_TOPIC> use_2d_lidar:=true

Citation

If you use this library for any academic work, please cite our original paper.

@article{kissteam2024arxiv,
  author      = {Guadagnino ,Tiziano and Mersch, Benedikt and Vizzo, Ignacio and Gupta, Saurabh and Malladi, Meher V.R. and Lobefaro, Luca and Doisy, Guillaume and Stachniss, Cyrill},
  title       = {{Kinematic-ICP: Enhancing LiDAR Odometry with Kinematic Constraints for Wheeled Mobile Robots Moving on Planar Surfaces}},
  journal     = arXiv Preprint,
  year        = {2024},
  volume      = {arXiv:2410.10277},
  url         = {https://arxiv.org/pdf/2410.10277},
}