TSDF-based 3D Monte Carlo Localization ROS Package

This package consists of several nodes and tools to perform a 6D Monte Carlo Localization of robots equipped with a 3D LiDAR in 3D TSDF maps. The sensor update is massively accelerated by a GPU-based implementation, but can also be executed on the CPU.

Prerequisites

• ROS Noetic (ros-noetic-desktop-full)
• ROS packages: See package.xml
• OpenMP (for CPU acceleration)
• CUDA (optional, recommended for acceleration)

Installation

1. Clone this repository into your ROS workspace

$ git clone --recursive https://github.com/uos/tsdf_localization.git

2. Make sure have also installed the required external packages or also cloned them into the local ROS workspace

3. Build the ROS workspace

$ catkin build

Demo

A quick startup including how to use tsdf_localization within your package is shown here: https://github.com/uos/tsdf_localization_demo.git

Citation

Please reference the following papers when using tsdf_localization in your scientific work.

@inproceedings{eisoldt2023,
  author={Eisoldt, Marc and Mock, Alexander and Porrmann, Mario and Wiemann, Thomas},
  booktitle={2023 Seventh IEEE International Conference on Robotic Computing (IRC)}, 
  title={{Towards 6D MCL for LiDARs in 3D TSDF Maps on Embedded Systems with GPUs}}, 
  year={2023},
  pages={158-165},
  doi={10.1109/IRC59093.2023.00035}
}

Nodes

mcl_3d

Starts MCL in a given TSDF map.

Subscribed Topics:

initialpose (geometry_msgs/PoseWithCovarianceStamped)

Initial pose guess can be provided using RViz.

/cloud (sensor_msgs/PointCloud2)

PointCloud topic for sensor update.

/odom (nav_msgs/Odometry)

Odometry message for motion update.

(optional) /imu_data (sensor_msgs/Imu)

Services

Start global localization:

/global_localization

Contributions

We are happy about issues and pull requests or other feedback. Please let us know if something did not work out as expected.