Indoor Mobile Robot Localization And Map-building

With the rapid development of robotics,modern mobile robots have been used in more and more applications in various fields.Research on mobile robots is in a multidisciplinary field.Therefore,research on mobile robots has always been a hot topic for scholars at home and abroad.The positioning and map building of the indoor environment have always been the focus of research and are the basis for mobile robots to serve humans.Therefore,this paper focuses on the status and map building of indoor robots and has achieved some results.The first part of the study is the representation and building of the line map.This paper use line segments to describe the structured indoor environment and proposes a geometric feature extraction method based on delay prediction.Compared with the traditional method,the proposed method can effectively eliminate the interference of abnormal points or effective points with large noise.The method reduces the number of geometric features in the local map while making the fitted partial map closer to the actual environment map.The second part of the research content is the global localization of mobile robots.This paper proposes a global localization method based on the relative position relationship of line segments.The method uses a new scanning matching technique for global localization,that is,based on the relative position relationship between the complete line segments and the visible complete line segments to achieve the matching between the local map and the global map,avoiding the frequent coordinate transformation between the local map and the global map,while limiting the number of line segment features participating in the match.The third part of the research content is incremental simultaneous localization and map building.The proposed method decomposes SLAM into three loop steps: local map building,robot pose estimation,and global map updating.In the local map building phase,a local geometric map is constructed using a method of geometric feature extraction based on delay prediction and a least-squares method.In the pose estimation stage,a complete line segment in the local map is used as a whole to match the entire line segment in the global map,and an efficient and robust data association method is proposed.It is first based on the generation and exploration of an explanatory tree.Using a constraint based on a full line segment relation table to minimize its search space while greatly reducing feature matching processing time.Finally,the best pose of the robot is obtained by using the matching error minimization and GaussNewton method.In the global map update phase,the line segment features in the local map are classified and the weight vector is used to update the global map.This not only improves the efficiency of the map update but also improves the accuracy of the map update.The updated global map merges redundant line segments,which will not only improve the accuracy of the map but also reduce the matching complexity in the next positioning and map updating.Experiments have proved the effectiveness and robustness of the method.