Research On Algorithms Of Indoor Mobile Robot Positioning And Path Planning

With the popularization and application of mobile robots in many fields,the related industries of mobile robots have been vigorously developed.Simultaneous Localization and Mapping(Simultaneous Localization and Mapping(SLAM),as the core technology of mobile robots,has become a research hotspot for scholars at home and abroad.It enables mobile robots to rely on sensors carried by themselves and through the movement of the robot in an unfamiliar environment.It can realize autonomous positioning and establish environmental map in the process of movement.Aiming at the low positioning accuracy of feature point-based visual SLAM system in dynamic environment,this paper combines optical flow method and frame difference method to remove the influence of dynamic objects.In order to solve the problem that A* algorithm has too many redundant points to explore in the exploration process of path planning and that the planned inflexion points are not smooth,A* algorithm is improved and smoothing the inflexion points.The main research contents are as follows:1)The principle of the depth camera Kinect is analyzed,the conversion process of the robot coordinate system and the world coordinate system,the establishment of the odometer model are introduced,and the motion model of the robot is deduced and studied.The basic theoretical framework of traditional SLAM system is studied and sorted out,and there is a large error in the estimation of robot pose for ORB-SLAM2 system in the indoor dynamic environment.A method combining optical flow method and frame difference method is proposed to eliminate dynamic objects and remove the influence of dynamic objects on robot localization.Through the ORB-SLAM2 feature point matching,using optical flow method for preliminary weed out part of the dynamic characteristic points,and then pick up points to calculate affine parameters,for the current frame image motion compensation,finally using interframe difference method to extract the dynamic body,eliminate the feature points on the dynamic object,let these feature points are not involved in the robot pose estimation.2)The principle of A* algorithm is introduced.Aiming at the problem that A*algorithm searches for redundant nodes in the process of path planning,the evaluation function of A* algorithm is modified.The weight factor that changes with the current node position is added into the evaluation function,which can effectively reduce the generation of redundant points and reduce the calculation resources.In addition,for many inflexion points of the planned path,the robot's driving is not utilized,and the Bezier curve is combined with A* algorithm to smooth the inflexion points.3)The validity of the improved ORB-SLAM2 system and the improved A*algorithm were verified by public standard data sets and real environment experiments.The improved ORB-SLAM2 system was tested by selecting data set environments with different sequences in the data set,and the motion trajectory errors in different sequence environments were compared with the original algorithm,so as to judge the improvement of positioning accuracy of the improved system.Map environments of different scales were established in MATLAB respectively,and the Operating results of the improved A* algorithm and A* algorithm were compared.The effectiveness of the improved A* algorithm was verified on the double-differential wheel Robot experimental platform based on ROS(Robot Operating System).