Voronoi Global Path Planning Based On Improved Skeleton Extraction And Local Dynamic Path Planning

Currently the mobile robot can't meet the real-time and accurate requirements at the same time in the process of navigation using path planning algorithm based on raster map and is close to obstacles.And it can't safely avoid dynamic obstacles in complex scenes.Aiming at these problems,a Voronoi global path planner based on improved skeleton extraction and a local path planner combining the motion state information of dynamic obstacle were proposed to improve the real-time performance and safety of mobile robot navigation.The detailed research contents of this paper are as follows:(1)Firstly,an indoor mobile robot experiment platform was built by choosing the right hardware and software.The parameters of stepper motor,bottom control unit,main controller,laser radar and camera were introduced in detail.Then,the related models of robot were introduced,including selection of driving model,solution of differential motion model,robot speed motion model,odometer motion model.Finally,the mainstream SLAM algorithms were analyzed and selected.(2)An improved skeleton extraction algorithm was proposed to solve the problem that robot storage was occupied too much due to the global path generated by the original Voronoi diagram algorithm being too bloated and the route generated by the existing method of skeleton extraction being too redundant.First,the raster map was binarized,corroded and expansion pretreated.Then,a leaner skeleton was regenerated.Next,the improved cubic spline smoothing process was applied to each path planned by navigator.The robustness and feasibility of the proposed algorithm are verified by simulation experiments and experiments in real environment.(3)Aiming at the current local path planning algorithm,the mobile robot treats moving obstacles as transient static ones when avoiding obstacles,lacking initiative and security.A local path planning algorithm combining motion state information of dynamic obstacle was proposed.Firstly,dynamic obstacles were detected through real-time changes of single scan LIDAR point clouds,and their position information was extracted for motion state fitting.Then,the safe distance when the robot and the dynamic obstacles were meeting was increased by expanding the cost map in the direction of the velocity of a dynamic obstacle.Finally,a local path planner combining dynamic obstacle motion state was realized by DWA algorithm.Comparing with the experiments using traditional DWA local path planner,the feasibility and effectiveness of the proposed method are verified.