Research On Motion Planning Algorithm Based On Dual-arm Collaboratio

With the increasingly complex operation tasks,the single manipulator has been more and more unable to meet the production requirements,such as assembling precision instruments,loading and unloading large goods and picking parts.Most of these tasks require the shortest time as possible,and the cooperative work of multiple manipulators has more advantages.Therefore,based on the kinematic model of double manipulator,this paper studies the path planning algorithm of dual arm cooperation,and verifies it through the assembly experiment of shaft hole parts.First,according to the kinematics model of single manipulator,the kinematics model of double manipulator is created.The kinematics model of single arm is established by D-H parameter method,and the inverse kinematics model of single arm is constructed by using the closed solution.According to the kinematics model of the single arm and the constraint relationship between the two arms,the kinematics model of the two arms is obtained.The forward and inverse kinematics are verified by MATLAB.Then,an improved RRT dynamic obstacle avoidance path planning algorithm is proposed.The collision detection model of manipulator is created,and the environment containing dynamic obstacles is analyzed and created.Based on this,an improved RRT algorithm is proposed.The main improvements are: segmentation and merging of search tree,adaptive step size,selection of nearest neighbor with weight,pruning of result path.By comparing the ordinary RRT algorithm with the improved RRT algorithm in MATLAB,the effectiveness of the improved RRT algorithm is verified.Next,for the fully constrained state during the movement of both arms,an improved MADDPG dual arm cooperative control algorithm is proposed.This paper introduces the relevant knowledge of reinforcement learning,including classifying reinforcement learning,especially introducing multi-agent deep reinforcement learning algorithm.In view of the synchronization requirements in the process of parts handling with two arms,the sparsity reward of the algorithm itself and the sample selection in the experience pool,improvements are made in MADDPG algorithm: establishing the priority experience pool,introducing the post experience playback algorithm and establishing a reward function for both arms.The stability and efficiency of the improved algorithm are verified by MATLAB.Finally,for a dual manipulator cooperative task,a dual arm cooperative control algorithm based on motion mode switching is proposed.Taking the assembly of shaft hole parts as an example,the different motion modes of double manipulators in the process of cooperative operation are analyzed;Then,aiming at the dual arm motion of loose coordination,the improved RRT algorithm is adopted,and the dual arms are planned respectively;Next the tight cooperative dual arm motion is divided into partial constraint and full constraint: when in the motion mode of partial constraint,the master arm adopts the improved RRT algorithm to plan the path,the slave arm carries out the path planning according to the motion constraints between the two arms,and when in the motion mode of full constraint,the improved MADDPG algorithm is used;Finally,the corresponding simulation experiments are completed with MATLAB.