LiDAR-Based Decentralized Collaborative Localization and Mapping for Multi-AAV With Stable Triangle Descriptor

Considering the limited communication bandwidth and computational resources of drone swarms in global positioning systems (GPS)-denied environments, we propose a decentralized collaborative 3-D LiDAR simultaneous localization and mapping (SLAM) system that achieves accurate and less time-consuming relative state estimation with low communication bandwidth, relying solely on onboard sensors and computing power. The system consists of four key modules: local odometry, initialization of relative state estimation, relative state estimator, and map management. Our system is built on top of a factor graph and uses a global descriptor called the stable triangle descriptor (STD) for 3-D place recognition. Benefiting from STD and keyframe strategy, the proposed system can perform relative state estimation under unknown initial relative poses while requiring only limited information without exchanging raw data. The developed method was tested with different types of LiDARs, demonstrating its adaptability. Furthermore, the proposed system was compared with the state-of-the-art collaborative 3-D LiDAR SLAM system, DiSCo-SLAM, through extensive experiments, validating the superiority of the developed approach.