Distributed Robust Communication-Efficient Multirobot SLAM Combining Real-Time Intersection and Historical Loop Constraints

To perform collaborative tasks in unknown environment, multirobot simultaneous localization and mapping (SLAM) must establish their spatial constraints from the perception data, and locate each robot within a unified and consistent shared environmental map. However, it faces the challenges of insufficient utilization of perception information, low robustness, and poor real-time communication. For example, most current work focuses on historical loop constraints, neglecting the more direct, geometrically stronger, and communication-efficient “real-time intersection” constraints. To overcome this problem, we propose a distributed robust communication-efficient multirobot SLAM framework, integrating real-time intersection constraints and historical loop constraints. The detection and tracking results of anchor robot are used as real-time intersection constraints to update the front-end status estimation, and the historical loop constraints within and between robots are considered to establish a distributed optimization back-end. Additionally, to generate accurate real-time intersection constraints, a two-stage clustering detection and tracking method for teammate robots based on reflective points is proposed. We established an experimental platform consisting of different mobile robots and collected datasets to test the system. These experimental results in various scenes show that combining real-time intersection and historical loop constraints can effectively improve robust positioning, global consistent mapping and efficient communication performance of multirobot SLAM systems.