Research On Indoor Mobile Robot Point Cloud Mapping And Relocation

Intelligent mobile robots are an important research area of the new generation of artificial intelligence and play an important role in social production and human life.When a mobile robot enters a new environment or loses its position during navigation,it needs to determine its absolute position in the global map,which is the robot’s repositioning problem.In order to improve the adaptability of mobile robots in long-term autonomous environments,relocation has become a hot problem of research.High-precision map is the premise of relocation.This paper studies point cloud mapping and relocation for multi-sensor fusion of indoor mobile robots.The main work is as follows:(1)The pose of the inertial measurement unit is calculated.For sensor calibration,the principles of IMU internal reference calibration and LIDAR external reference calibration are analyzed,and IMU internal reference calibration experiments and LIDAR and IMU external reference calibration experiments are conducted.(2)A typical single-sensor point cloud building algorithm is investigated.Since the corner structure of the point cloud map created by this algorithm is not obvious and degrades in a narrow building,which does not provide enough effective information for repositioning,a multi-sensor fusion-based point cloud building algorithm is proposed.The focus is on a graph optimizationbased approach to achieve fusion of Li DAR and IMU data to build high-precision maps.(3)Point cloud localization algorithms such as Iterative Nearest Neighbor,Normal Distribution Transform,and Scan Context are analyzed.An improved ICP algorithm is proposed for the problem of poor robustness of ICP algorithm in localization.For the problems that the robot needs to return to the initial point of the map for repositioning and may fall into local optimum when positioning,a SC-NII algorithm is proposed,which firstly provides the initial pose by the Scan Context algorithm,then coarse positioning by the NDT algorithm,and finally precise positioning by the improved ICP algorithm.(4)In order to prove the effectiveness of the algorithm,a physical robot platform is built and experiments are conducted in the underground garage and the floor corridor.The experiments verify that the quality of the point cloud maps built by the multi-sensor fusion approach is better than that of the single-sensor-based approach.Under the prior point cloud map,the ICP algorithm and improved ICP algorithm positioning experiments are carried out respectively,and the robustness of the improved ICP algorithm is verified to be higher than that of the ICP algorithm.Finally,relocation experiments are carried out based on the SC-NII algorithm.The feasibility of the relocation algorithm proposed in this paper is verified.