Multi S-Graphs: an Efficient Real-time Distributed Semantic-Relational Collaborative SLAM

• DOI: 10.48550/arxiv.2401.05152
• Publication Date: 2024-01-01
• Authors: Miguel Fernández-...; Hriday Bavle; David Pérez-Saura; José Luis Sánchez...; Pascual Campoy; Holger Voos
• Keywords: Robustness; Initialization; Situation Awareness

Collaborative Simultaneous Localization and Mapping (CSLAM) is critical to enable multiple robots operate in complex environments. Most CSLAM techniques rely on raw sensor measurement or low-level features such as keyframe descriptors, which can lead wrong loop closures due the lack of deep understanding environment. Moreover, exchange these measurements among requires transmission a significant amount data, limits scalability system. To overcome limitations, we present Multi S-Graphs, decentralized system that utilizes high-level semantic-relational information embedded four-layered hierarchical optimizable situational graphs for cooperative map generation localization structured environments while minimizing exchanged between robots. support this, novel room-based descriptor which, along with its connected walls, used perform inter-robot closures, addressing challenges multi-robot kidnapped problem initialization. Multiple experiments simulated real validate improvement accuracy robustness proposed approach reducing data compared other state-of-the-art approaches. Software available within docker image: https://github.com/snt-arg/multi_s_graphs_docker