This repository contains algorithms designed for map-based robot localization, specifically when dealing with maps composed of triangle meshes or complete scene graphs. These maps may be provided by architects who have designed the building in which the robot operates, or they can be autonomously generated by the robot through Simultaneous Localization and Mapping (SLAM) methods. It’s crucial to note that map-based localization differs from SLAM; it focuses on estimating the robot’s pose within a potentially large map, whether the initial pose is roughly known (tracking) or entirely unknown from the start aka kidnapped robot problem. Map-based localization is essential for precisely planning the robot’s missions on a given map.

MICP-L

MICP-L: Mesh-based ICP for Robot Localization Using Hardware-Accelerated Ray Casting. An approach to directly register range sensor data to a mesh in order to localize a mobile robot using hardware-accelerated ray casting correspondences (See publications).

Hilti: 6DoF Localization	MulRan: Large-scale scenes

Requirements:

• At least one range sensor is equipped and running
• Triangle mesh as map
• Prior odometry estimation of the robot given as TF

IMU prior is also possible as long as it is integrated as TF-Transform, e.g. with Madgwick Filter.

Publication

Please reference the following paper when using the MICP-L method in your scientific work.

@inproceedings{mock2024micpl,
  title={{MICP-L}: Mesh-based ICP for Robot Localization Using Hardware-Accelerated Ray Casting}, 
  author={Mock, Alexander and Wiemann, Thomas and Pütz, Sebastian and Hertzberg, Joachim},
  booktitle={2024 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS)}, 
  year={2024},
  pages={10664-10671},
  doi={10.1109/IROS58592.2024.10802360}
}

The paper is available on IEEE Xplore and as preprint on arXiv. The experiments are available at https://github.com/amock/micp_experiments (mostly for ROS 1).

Usage

The micp_localization Node starts the process of localizing the robot in a mesh using MICP and a given pose estimate. It is usually started through a launch file since it requires a large set of parameters.

Launch

The following Launch-File

<launch>

<node pkg="rmcl_ros" exec="micp_localization" name="micp_localization" output="screen">
    <param name="map_file" value="/path/to/mesh/map.dae" />
    <param from="/path/to/config/file.yaml" />
    <remap from="pose_wc" to="/initialpose" />
</node>

</launch>

runs the MICP localization node. After that a pose has to be given, e.g. by the RViz “2D Pose Estimate” Tool that publishes the results on the /initialpose topic. Doing that, make sure to set the fixed frame to the map coordinate system. RMCL itself doesn’t provide any tools to visualize the maps (triangle meshes). If you want to see the map in RViz, use for example the rviz_mesh_tools_plugins of the mesh_tools.

At startup, MICP does a few sanity checks for the input parameters. Every possible mistake in configuration can then be inferred by this output. For example, once there is no data available on the given PointCloud2-Topic it will print data: no instead.

Parameters

The following sections describe example configuration files. More example files for configuration are placed in the rmcl_ros/config/examples.

Parameters - 3D LiDAR only - CPU

MICP Localization using a 3D LiDAR and doing the MICP steps completely on the CPU. Here the 3D LiDAR is a Velodyne VLP-16 with 16 scan lines. The horizontal number of points is reduced to 440 and might be adjusted for your own Velodyne.

Parameters - 2D LiDAR + Wheels - GPU

MICP also supports to localize a robot only equipped with a 2D LiDAR in a 3D map. To correct the third dimension the wheels can be used to pull the robot towards the map’s ground plane. Thus, you should only run it on a robot that always drives on the ground and e.g. cannot fly. In this example, all MICP steps are computed on GPU. The robot has four wheels of which the highest points are located relative to base_footprint as listed in origs. By setting a virtual scanner to the wheel positions scanning downwards with a constant scanning range equal to the wheel diameter it is possible to pull the robot to the mesh.

RMCL - Project

Installation

Dependencies:

• Download and install Rmagine (v > 2.2.9): Compile from source (not debian packages).
  • Recommended: Install OptiX backend if NVIDIA GPU is available
  • For rmagine version >= 2.2.9 it is possible to put rmagine into your ROS workspace for easier compilation
• ROS2 (check compatible branches)

Clone this repository into your ROS workspace and build it.

$ colcon build

Optional, but recommended: RMCL itself doesn’t provide any tools to visualize the triangle mesh maps. If you want to see the map in RViz, use for example the rviz_mesh_tools_plugins of the mesh_tools.

Examples

To learn how to use RMCL ROS nodes in your project, visit https://github.com/amock/rmcl_examples.

To learn how to use RMCL library in your Node: src/nodes/examples.

Mesh Navigation

To navigate a robot automatically and safely through uneven terrain, the combination RMCL + Mesh Navigation Stack is very suitable: https://github.com/naturerobots/mesh_navigation. As we presented on ROSCon 2023:

Roadmap

This package will be expanded by more functionalities to localize a robot in mesh maps. The planned Roadmap is as follows:

• MICP-L (Tracking)
• RMCL (Global Localization)

News

2024-11-25: Restructuring - ROS 1 + ROS 2

We had to do minor structural changes to the repository in order to better integrate new features into RMCL. This repository is now devided into

• “rmcl” which is a ROS-agnostic library that can be compiled and installed as regular CMake project,
• “rmcl_ros” which contains all the nodes,
• “rmcl_msgs” which are the message moved from to this repository. The original msgs repository is not required anymore.

Using the latest rmcl version might break your launch files as the nodes are now located in “rmcl_ros” package. However, it’s rather simple to fix that. The new versions of RMCL are v2.1.0 for ROS 2 and v1.3.0 for ROS 1.

2024-02-11: ROS2 release - v2.0.0

The main branch is humble now! Since it is not backwards compatible we decided to increase the version of RMCL to 2.0.0. The noetic version will still exist with the “noetic” branch. The “noetic” branch will be maintained until the end of 2024.

2024-01-05: ROS2 - humble

The first ROS2 port has been released! If you are using ROS2, check out the humble branch of this and all linked repositories. After the new branch has been tested well enough, I will make it the main branch. The current version will persist in the noetic branch.