CCMD-SLAM: Communication-Efficient Centralized Multirobot Dense SLAM With Real-Time Point Cloud Maintenance

Abstract

This paper presents a communication-efficient centralized multi-robot dense SLAM system with real-time point cloud maintenance (CCMD-SLAM), to address the limitations of data transmission and real-time creation and updating of dense maps in multi-robot SLAM. This method solves the problem of high bandwidth consumption for information transmission in multi-robot SLAM by pre-processing and compressing the transmitted data and filtering the RGB-D information using the co-viewing degree of keyframes. The proposed method utilizes loop closure detection to integrate information from multiple robots and establish a global point cloud. Additionally, a keyframe and point cloud storage mechanism is designed to facilitate real-time maintenance of global point cloud data. Through comprehensive evaluations using standard datasets and real-world experiments, CCMD-SLAM significantly alleviates data transmission pressures, enables flexible global point cloud management across multiple robots, and effectively achieves dense mapping for multi-robot systems.