Visual SLAM And Path Planning Based On Depth Camera

In order to realize the autonomous movement of robot,the localization and path planning of indoor robot were studied.Since GPS signals will fail indoors,other sensors will be selected to carry out robot positioning under indoor conditions.Both laser radar and camera sensors contain rich information.Meanwhile,laser radar and camera have limitations.However,single-line laser rada can only measure the obstacle information in one horizontal plane,which is not comprehensive enough,while the depth camera has a small range of left and right,which leads to slower map construction and lower accuracy.In this paper,the map obtained by the fusion of depth camera and lidar is used to carry out robot positioning and path planning.In the part of building dense map for visual SLAM,ORB-SLAM2 is improved in this paper,and the function of building dense point cloud is added on the basis of the original algorithm,and the map fusion is realized based on Bayesian inference principle.After the accurate map of the environment is obtained,the improved adaptive Monte Carlo algorithm is used to locate the robot.Compared with the traditional Monte Carlo algorithm,the adaptive Monte Carlo algorithm effectively solves the "kidnapping problem" in robot localization,and accelerates the localization by controlling the number of particles in the algorithm.In the part of path planning,A* algorithm and DWA algorithm are studied.A* algorithm can only be applied to static environment,which cannot deal with suddenly appearing obstacles and is prone to collision in practical application.However,DWA algorithm is only applicable to local path planning,and it is easy to fall into local optimal value without global guidance,and even cannot reach the destination.Firstly,A* algorithm is used to generate the global optimal path,and some key points are extracted from the optimal path as the endpoint of DWA algorithm to realize the dynamic obstacle avoidance under the global guidance of A* algorithm.The fusion algorithm gives full play to the advantages of the two,makes up for the shortcomings,and gets better results.The effectiveness of the proposed algorithm is verified by laboratory experiments on ROS system.