Collaborative Visual Inertial SLAM with KNN Map Matching

Abstract

Simultaneous localization and mapping (SLAM) is widely used in the field of intelligent robots. With the expansion of SLAM applications, especially the requirements of multi-user interaction proposed in the AR field, the realization of multi-user collaborative SLAM technology has become a necessary demand. Multi-agent collaboration in indoor scenes can effectively improve the efficiency of mapping and the actual user experience through map fusion. We design a system for real-time multi-user collaborative mapping and map fusion. The system includes communication module, mapping module, match module, and fusion module. Most previous solutions are based on overlapping region detection between clients to realize map fusion and collaborative mapping, and employ DBoW2 to find overlapping regions. However, errors are introduced from the mismatching problem because the fixed similarity threshold of DBoW2 is difficult to set. In this paper, we propose the KNN map match method to solve mismatch, and apply it to the collaborative visual-inertial SLAM system. The function of the system is verified on the public dataset. We evaluated the accuracy of the map fusion on public dataset, and the results show that the proposed method can improve the matching accuracy.