Trajectory Planning And Compliant Control For Multi-process Human-robot Cooperative Tasks

Human-robot collaboration combines the advantages of human and robots,which can improve production efficiency and quality.The actual production tasks are usually multi-process tasks,and the current research on human-robot collaboration is mostly aimed at single process tasks such as collaborative assembly,handover and transportation,which is difficult to meet the actual needs.In this paper,the corresponding trajectory planning and compliance control are studied for multi-process human-robot collaboration tasks.Combined with finite state machine and imitation learning,the multi-process collaboration tasks are analyzed.The variable stiffness impedance control is used to balance the compliance of interaction and the tracking accuracy of trajectory.Combined with virtual force and impedance control,obstacle avoidance is realized,and efficient and safe human-robot collaboration is realized.Firstly,build multi-process human-robot collaboration platform.Taking the assembly of small tank car as an example,the multi-process collaborative task is analyzed,and the hardware and software of the collaborative system are determined.Then the intrinsic parameters and extrinsic parameters of the camera are calibrated and the kinematics and dynamics model of the manipulator is established,which lays the foundation for the trajectory planning and compliance control of the subsequent collaborative tasks.Secondly,combined with imitation learning and finite state machine analysis of multi-process human-robot collaboration tasks.Firstly,the task is divided into a single process and the finite state machine is used to determine the state and the transfer and action between states.Then,the task-parameterized Gaussian Mixture Model(TP-GMM)is used to learn the human operation skills in a single process to plan the manipulator trajectory,and realize the generalization of different task parameters by single process skills,so as to complete the decomposition and learning of multi-process collaborative tasks.Thirdly,the influence of variable stiffness impedance control on interaction compliance and trajectory tracking accuracy,and the virtual force obstacle avoidance strategy are studied.For the contact human-robot collaboration task,the mean value of the trajectory probability model learned by TP-GMM is used as the reference trajectory of impedance control,and the variable stiffness coefficient is designed according to the covariance matrix in the probability model.The stiffness coefficient increases with the increase of the distance from the reference trajectory,so as to balance the interaction compliance and trajectory tracking accuracy in the cooperation task.For the non-contact human-robot collaboration task,the point cloud processing and collision detection are used to calculate the nearest point pair between the human and the manipulator,which generates virtual force and combines impedance control to realize dynamic obstacle avoidance,Avoiding obstacles with dynamic consistency pseudo-inverse of jacobian matrix in zero space which improves the safety of human-machine cooperation.Finally,the trajectory planning and compliant control strategy for multi-process human-robot collaborative tasks are verified by experiments.Obstacle avoidance experiments of static and dynamic objects,collaborative transportation experiments and small tank car assembly experiments are designed to verify the effectiveness of the proposed method for multi-process collaborative tasks and achieve the flexibility and security of interaction in human-robot collaborative tasks.