li9i / cbgl

Globally localise your 2D LIDAR in a 2D map in no time

CBGL: Fast Monte Carlo Passive Global Localisation of 2D LIDAR Sensor

cbgl is a ROS package written in C++ that allows you to localise your 2D LIDAR sensor in a given 2D occupancy grid map under global uncertainty in position and orientation in minimal time

• You can expect the execution time to roughly have an order of magnitude of $10e \cdot \text{area} \cdot N_s$ microseconds, where $\text{area}$ is the area of the map's free space and $N_s$ is the LIDAR's number of rays. (Strictly speaking the execution time varies according to the geometry of the environment and other factors.) In the video below the environment area is $2000$ m$^2$ and localisation is performed in under four seconds

• CBGL does not require motion for performing global localisation: it's a one-shot approach that only requires a single laser scan measurement and the map of the sensor's environment

cbgl_in_willowgarage.mp4

_{A panoramic 2D LIDAR sensor mounted on a turtlebot 2 is spawned into an environment at a pose whose immediate surroundings are repeated in (almost) the same geometry and proportions at locations other than the sensor's spawning ground. The user calls the global localisation service once before moving the robot at a second challenging pose, at which she calls it for a second time. Both times cbgl is successful in estimating the sensor's pose. Sensor characteristics: $N_s = 360$ rays; noise: $\sim \mathcal{N} (0.0, 0.05^2)$ [m,m$^2$]}

Why use CBGL

Table of Contents

• Pre-installation
• Install
• Run
  • Launch
  • Call
• Input/output at a glance
• Motivation
• More results

Pre-installation

cbgl is installed, launched, and called via Docker:

• if this is your first time running docker then I happen to find this docker installation guide very friendly and easy to follow
• if instead you wish to install and run the package natively in Ubuntu 16.04, see the INSTALLATION_DEPRECATED.md guide.

Install

Build the image with the most recent code of this repository with:

git clone [e-mail]:li9i/cbgl.git
cd cbgl
docker compose build

or pull the docker image and run it with

docker pull li9i/cbgl:latest

docker run -it \
    --name=container_cbgl \
    --net=host \
    --rm \
    li9i/cbgl:latest

Run

Launch

docker compose up

Call

Launching cbgl simply makes it go into stand-by mode and does not actually execute global localisation. To do so simply call the provided service

docker exec -it cbgl_container sh -c "source ~/catkin_ws/devel/setup.bash; rosservice call global_localization"

Input/output at a glance

• [in] A sensor_msgs/LaserScan message published through topic configuration_files/scan_topic
• [in] A nav_msgs/OccupancyGrid message published through topic configuration_files/map_topic
• [out] A geometry_msgs/PoseWithCovarianceStamped message published through topic configuration_files/output_pose_topic
• [out] The transform between the odom frame and the map frame if configuration_files/tf_broadcast is set to true

Motivation

More results

Citation

The theoretical underpinning, key aspects, and experimental performance of CBGL are presented in the preprint cited through

@article{cbgl,
  title={CBGL: Fast Monte Carlo Passive Global Localisation of 2D LIDAR Sensor},
  author={Filotheou, Alexandros},
  journal={arXiv:2307.14247},
  url={https://arxiv.org/abs/2307.14247}
  year={2023}}

Update JUN 2024: CBGL has been accepted for publication in the Proceedings of the 2024 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS 2024)