Research on indoor robot SLAM of RBPF improved with geometrical characteristic localization

Abstract

LIDAR and odometer are used as the main sensors in this paper, with the two-wheeled self-balancing robot as the research and experiment platform, this topic has researched the way to SLAM when the robot is in unfamiliar environment with uncertain position and orientation. On the issue of setting the dynamic threshold value to zone the LIDAR scanning points in the process of constructing the geometry map, with the consideration of the character of RPLIDAR which is used in our research, this paper has analyzed and proposed the specific dynamic threshold values, which make the zoning of LIDAR scanning points more reasonable. In order to improve the robustness of SLAM, we have used regular particle filter (RPF) as the location algorithm. In order to solve the problem of unable to add auxiliary information under the traditional MCL framework, we have taken full use of the high accuracy character of the geometry matching and locating, and have used the result of it to improve the importance density function of RPF. Based on the idea of Rao-Blackwellization, the improved RBPF-SLAM has been proposed. The validity and feasibility of the improved RBPF-SLAM have been proved with simulations and experiments.