Research On Motion Planning Of Multi-DOF Manipulator In Limited Task Space

Motion planning is the vital content for the multi-DOF manipulators to achieve the grasping action or manipulation operation,and it is also a hot research area in the field of robotics.The sampling-based motion planning algorithm has the issues of slowly planning and long planning time due to the influence of the speed of nearest neighbor search,the strategy of sampling expansion and the performance of collision detection algorithm.Based on these facts,this paper proposes a MGMM-RRT*algorithm that combines RRT*algorithm with Long-term and Short-term Memory mechanisms,GMM models,improved strategy of sampling expansion and improved collision detection algorithm respectively,which effectively improves the efficiency of motion planning.The nearest neighbor search and strategy of sampling expansion are essential steps in RRT*algorithm.In this paper,the K-d tree is constructed for the existing nodes of RRT*tree,when new sampling points are generated,the nearest neighbor search method based on K-d tree is used to improve the speed of searching for the nearest nodes.In the process of RRT*tree expansion,obstacle information is incorporated into the nodes.By synthetically judging obstacle information and extension step,some invalid nodes can be avoided,and the density of RRT*tree would be reduced,which reduces the time of the nearest neighbor search and improves the expansion speed.Collision detection is a consequential link in the RRT*algorithm,the faster the collision detection,the shorter the running time of the algorithm.In this paper,the Gaussian mixture model combining with long and short time memory is introduced into RRT*algorithm,in which the traditional collision detection method is replaced by the GMM based method.GMM model is used to guide sampling and fast collision detection in high-dimensional joint space,which can effectively improve the success rate of motion planning,on the other hand,reduce the number of collision detection and the time of collision detection.Afterwards,different GMM of multiple scenes is stored according to instant memory,short-term memory and long-term memory,and the scene matching algorithm is used to realize fast adaptive extraction of GMM in different scenarios to improve the adaptability for changing environments.On the basis of theoretical research,this paper completes the experiment about motion planning of the manipulator in MATLAB and ROS environment,and converts the information about point cloud of the scene into Octomap for motion planning.The storage and matching of scene model are realized by combining the Word-bag model with the iterative nearest neighbor algorithm,which improves the adaptability of the algorithm in multi-scene.The experimental results illustrate that the proposed algorithm can effectively improve the success rate of motion planning and reduce the number and intensity of iterations of the RRT*tree in the process of expansion and improve the efficiency of motion planning.Finally,this paper summarizes the research work carried out,and looks forward to the future research.