Design Of Indoor Autonomous Navigation System For Mobile Robot Based On ORB-SLAM2

Simultaneous Location and Mapping(SLAM)technology is an important subject in the field of mobile robot navigation and positioning.Compared with laser SLAM,visual SLAM-based navigation and positioning system has the advantages of low cost and rich data information,thus many works have been introduced in universities and enterprises.Although some progress has been made,there still exist lots of problems when landing the pure visual SLAM navigation and positioning system.Therefore,in this paper,we propose an ORB-SLAM2-based navigation system design by applying the depth camera,and carries out simulation experiments.The main research contents are as follows:(1)We evaluate that the ORB feature point outperforms other feature extraction methods(such as SURF and SIFT algorithm)in the visual SLAM domain,and it also achieve higher matching accuracy after being filtered by the mismatched culling algorithm.(2)In order to solve the problem that the sparse point cloud map is insufficient of space occupation information in the navigation task,we introduce a dense map construction and optimization algorithm based on ORB-SLAM2.First,based on the dense mapping architecture of the RGB-D SLAMV2 system,the dense mapping module is added to the ORB-SLAM2 system.Then we compare the improved ORBSLAM2 system with the RGB-D SLAMV2 system(with ORB feature point)in terms of dense map quality and positioning accuracy.Finally,for the problem of excessive outliers and contour blur in dense point cloud map,a three-layer point cloud filtering algorithm is proposed to optimize dense maps,which can effectively reduce the storage capacity,remove the outliers,and enhance the quality of dense point cloud maps.(3)To address the problems of information redundancy and "drag shadow" of moving objects when utilizing the dense point cloud map in a navigation task,the point cloud map is transformed into the octree map which is easy to maintain,expand and store,and it provides a precondition for the subsequent navigation tasks of the mobile robot.In order to reduce the invalid map construction work,the sparse feature map multiplexing module is added on the basis of the ORB-SLAM2 system.Meanwhile,we discard text format and save the map in binary format,which can effectively speed up map reading and reduce storage.Finally,based on the improved ORB-SLAM2 system combined with the robot operating system,we complete the navigation system design of the simulated robot in the indoor environment and verified that the positioning accuracy of the system is high.