Sparse Pointcloud Map Fusion of Multi-Robot System

• Published in: 2018 International Conference on Control, Automation and Information Sciences (ICCAIS) pp. 270 - 274
• Main Authors: Sun, Shiqin, Xu, Benlian, Sun, Yidan, Sun, Zhen
• Format: Conference Proceeding
• Language: English
• Published: IEEE 01.10.2018
• Subjects: Feature extraction; map fusion; matching search strategy; Motion estimation; multi-robot; Optimization; Robot kinematics; Simultaneous localization and mapping; sparse pointcloud map; Task analysis
• ISSN: 2475-7896
• DOI: 10.1109/ICCAIS.2018.8570467

Map building is a crucial function of multi-robot system, and many available autonomous navigations in multi-robot systems assume high precision of the environmental map. Thus, poor performance in map building can heavily affect the navigation results in a large-scale environment. The core issue that needs to be addressed of multi-robot mapping is how to integrate the data of the different robots into a single global map. In this paper, a matching search strategy based on the co-viewing relationship between key frames is proposed to reduce matching time. The key frames to be matched are selected from the maps according to a certain condition instead of being matched individually. Thus, a considerable amount of match time can be saved. After a set of matched map points are obtained, the motion estimation between matched points is solved by nonlinear optimization and an error compensation technique is employed to obtain more accurate camera posture. Finally, the redundant map points after fusion are removed, and both the connection between the key frames and the map points in the two maps are established. The algorithm is tested in an indoor environment and the experiment results show the validness of the proposed method.