A swarm of robots offers significant advantages over a single robot, enabling faster exploration of larger areas and enhanced robustness against single-point failures. Accurate relative positioning is critical for executing collaborative missions without collisions. When Visual Simultaneous Localization and Mapping (VSLAM) is employed to estimate each robot’s poses, the inter-agent loop closing method is commonly used to improve relative positioning accuracy by refining pose estimates and merging local maps based on shared feature points. However, this approach demands considerable computational and communication resources. In this paper, we introduce Collaborative SLAM using Visual Odometry and Ranges (CoVOR-SLAM) to address these challenges. CoVOR-SLAM significantly reduces both the amount of data transmitted between robots and their computational loads, as it requires only pose estimates, covariances, and range measurements instead of feature or map points for inter-agent loop detection. The necessary range measurements can be derived from pilot signals within the communication system, eliminating the need for complex additional infrastructure. We evaluated CoVOR-SLAM using real images and ultra-wideband range data collected from two rovers, as well as in a larger multi-agent setup utilizing public image datasets and realistic simulations. The results demonstrate that CoVOR-SLAM accurately estimates robot poses while requiring considerably less computational power and communication capacity than inter-agent loop closing techniques.
CoVOR-SLAM: Cooperative SLAM Using Visual Odometry and Ranges for Multi-Robot Systems
IEEE Transactions on Intelligent Transportation Systems ; 26 , 7 ; 9358-9369
01.07.2025
4431794 byte
Aufsatz (Zeitschrift)
Elektronische Ressource
Englisch
Study on visual odometry and optimization of SLAM system
IEEE | 2022
|Gamma-SLAM: Visual SLAM in unstructured environments using variance grid maps
British Library Online Contents | 2009
|