RGB-D Based Autonomous Exploration and Mapping of a Mobile Robot in Unknown Indoor Environment

Abstract

An autonomous exploration and mapping method based on RGB-D (RGB-depth) information for mobile robots is proposed. Firstly, the 3D point cloud map is constructed in real time based on the information from the RGB-D sensor through location point generation, map building and loop closure detection. Then, the autonomous exploration is modeled as a partially observable Markov decision process (POMDP). An autonomous exploration strategy is established based on partial map simulation and global frontier search. Further, real-time action constraints are specified and imposed to a dynamic window motion controller, which drives the motion of the mobile robot and avoids local optimums, and ensures stable map construction with RGB-D data. Finally, the proposed exploration method is realized on a mobile robot system in laboratory experiments. The experimental results demonstrate the effectiveness of the proposed autonomous exploration and mapping method in unknown environments.