by Andreas Serov, Joachim Clemens, Kerstin Schill
Abstract:
Simultaneous localization and mapping (SLAM) is a crucial task for autonomous systems in the field of robotics. In the context of multi-robot systems, decentralized approaches have gained popularity due to eliminating the need for a central base station and promoting scalability. This paper presents a decentralized multi-robot graph-based SLAM termed mrg_slam, which leverages the well-established single-robot SLAM frame-work hdl_graph_slam. In our approach, each robot independently runs the LIDAR-based graph SLAM algorithm, enabling an efficient exploration of unknown environments. The robots can exchange unique graph nodes and edges, incorporating this shared information into their individual SLAM solutions. Specifically, each robot integrates the exchanged graph nodes with associated point clouds and edges from other robots into its own graph, contributing to the development of a globally consistent map for each individual robot. We demonstrate the effectiveness of our algorithm through an evaluation of the multi-robot SLAM in a simulated environment and on real-world experiments. We publish the source code that uses the robot operating system 2 (ROS2), which is designed for multi-system communication: https://github.com/aserbremen/Multi-robot-graph-SLAM.
Reference:
Multi-Robot Graph SLAM Using LIDAR (Andreas Serov, Joachim Clemens, Kerstin Schill), In 10th International Conference on Automation, Robotics and Applications (ICARA), 2024.
Bibtex Entry:
@inproceedings{serov2024multi,
author={Serov, Andreas and Clemens, Joachim and Schill, Kerstin},
booktitle={10th International Conference on Automation, Robotics and Applications (ICARA)},
title={Multi-Robot Graph {SLAM} Using {LIDAR}},
year={2024},
pages={339-346},
keywords={vamex3, Point cloud compression, Base stations, Simultaneous localization and mapping, Laser radar, Source coding, Scalability, Operating systems, multi-robot decentralized SLAM, autonomous systems, cooperative exploration},
doi={10.1109/ICARA60736.2024.10553070},
url={10.1109/ICARA60736.2024.10553070">https://doi.org/10.1109/ICARA60736.2024.10553070},
abstract={Simultaneous localization and mapping (SLAM) is a crucial task for autonomous systems in the field of robotics. In the context of multi-robot systems, decentralized approaches have gained popularity due to eliminating the need for a central base station and promoting scalability. This paper presents a decentralized multi-robot graph-based SLAM termed mrg_slam, which leverages the well-established single-robot SLAM frame-work hdl_graph_slam. In our approach, each robot independently runs the LIDAR-based graph SLAM algorithm, enabling an efficient exploration of unknown environments. The robots can exchange unique graph nodes and edges, incorporating this shared information into their individual SLAM solutions. Specifically, each robot integrates the exchanged graph nodes with associated point clouds and edges from other robots into its own graph, contributing to the development of a globally consistent map for each individual robot. We demonstrate the effectiveness of our algorithm through an evaluation of the multi-robot SLAM in a simulated environment and on real-world experiments. We publish the source code that uses the robot operating system 2 (ROS2), which is designed for multi-system communication: https://github.com/aserbremen/Multi-robot-graph-SLAM.}
}