Research On Dual-arm Coordination Control Method Of Hyper-redundant Manipulator

Compared with the traditional working mode of single manipulator,the coordinated operation of manipulator can expand the working space,improve the load capacity and increase the flexibility of operation.Therefore,the coordinated operation of dualarm or even multi-arm robot has become one of the most popular research hotspots in the academic and industrial fields.In addition,because of the higher maneuverability and obstacle avoidance ability of the hyper-redundant manipulator,the dual-arm coordination of the hyper-redundant manipulator can accomplish extremely dexterous cooperative tasks in a complex environment.However,there are complex kinematic and dynamic constraints among the hyper-redundant manipulators,and the trajectory planning,control and load distribution of the dual-arm coordination are also great challenges.To this end,the kinematics and dynamics modeling,obstacle avoidance planning method,interaction algorithm with environment and load distribution method of the dual-arm coordination of hyper-redundant manipulator are studied in this thesis.The main research contents of this thesis are as follows:The workflow of dual-arm coordination operation of hyper-redundant manipulator is analyzed,and the kinematics and dynamics models of dual-arm coordination are established.According to the characteristics of dual-arm cooperation process,the dual-arm coordination of hyper-redundant manipulator can be divided into free motion phase,transition phase and constrained motion phase.Using Denavit-Hartenberg(D-H)parameter method,the kinematic models of two hyper-redundant manipulators are established respectively,meantime,the kinematic constraint relationship is clarified.Using the projected inverse dynamics theory,the dynamic model of each hyper-redundant manipulator is decomposed into constrained space and unconstrained space for control.Then,using the grasping matrix of the dual-arm cooperation,the constraint force acting on the manipulation target is decomposed into internal force and external force,and the dynamic constraint relationship of the dual-arm coordination is analyzed at the same time.An obstacle avoidance trajectory planning algorithm is designed that enables the hyper-redundant manipulator to achieve coordinated operation in complex space.The algorithm combines attractive and repulsive fields,in which the virtual force generated by the attractive field is applied to the end-effector of the manipulator,and the Jacobian matrix can be used to convert the attractive force into the joint velocity,so that the pointto-point and trajectory tracking motion planning of the manipulator can be realized;the repulsive force generated by the repulsive field acts on the joints of the manipulator and is also converted into the corresponding joint velocity to avoid collisions with obstacles in the environment,its own joints and other manipulators.Simulation and experiment results show that the proposed obstacle avoidance trajectory planning algorithm can avoid collisions with obstacles while ensuring the trajectory planning accuracy of the end-effector,which proves the effectiveness of the proposed algorithm.A virtual semi-active damping impedance learning algorithm for robot interaction with unknown environment is proposed.The virtual semi-active damping impedance learning algorithm includes a virtual semi-active damping controller for position and force hybrid tracking and an impedance learning algorithm for estimating environmental impedance and position parameters.The additional damping of the virtual semi-active damping controller is actively modulated based on the environmental impedance parameters and the position information of the manipulator,and the environmental impedance parameters are updated using an impedance parameter learning algorithm based on the BFGS method.Simulation and experiment results show that the proposed algorithm has good contact performance and position/force tracking effect when the manipulator interacts with environments with different stiffness,which proves the effectiveness of the control algorithm and the accuracy of the environmental parameter learning algorithm.An optimization method for load distribution of dual-arm coordination operation based on reinforcement learning is developed.The method is based on hierarchical reinforcement learning framework.The upper strategy determines the motion form of the operation target in the dual-arm coordination operation task,and the lower strategy is a parameterized feedback controller,which determines the contact force and load distribution form according to the current motion information.Load distribution optimization is to find a set of low-level policy parameters with better performance corresponding to the motion form of the operation target.Therefore,a contextual reinforcement learning method based on relative entropy policy search is used to learn the distribution of low-level policy parameters with respect to contextual features.The simulation results show that the proposed reinforcement learning method can perform load distribution well when realizing various motion forms of the operation target,which reflects the correctness of the proposed load distribution method.The joint experiment of dual-arm coordination operation of hyper-redundant manipulator is carried out.Firstly,the whole structure of the dual-arm coordination system is introduced;then,the experimental principles of free motion phase,transition phase and constrained motion phase are explained;finally,three phases of experimental verification are carried out,and the experiment results are analyzed.The experiment results show that the coordinated operation model,the obstacle avoidance trajectory planning algorithm,the virtual semi-active damping impedance learning algorithm and the load allocation optimization method developed in this thesis are accurate and effective,and they can be applied to engineering practice.