Human-robot Compliant Collaboration Based on Feedback of Motion Intention of Human Arm

This paper aims to improve the compliance of human-robot collaboration based on the feedback of human arm motion intention. Firstly, the autoencoder and backpropagation neural network (BPNN) are used to fuse the surface electromyography (sEMG) and the machine vision signals for estimating the human elbow joint torque and motion intention. Then, the human elbow joint torque is feedback to the robot, and the motion intention of the arm is estimated to make the robot act adaptively, and thus human-robot collaboration is realized. Finally, 3 different collaboration patterns are compared by combining objective and subjective evaluation indices through the wood sawing experiments in human-robot collaboration. Comparing with the case of the human-robot collaboration without motion intention feedback, the fluctuation amplitude of the interaction force is reduced by 153.39 N and the task completion time is reduced by 19.25 s in the case with feedback. Experimental results show that the human-robot collaboration with the feedback of human arm motion intention can improve the compliance of human-robot collaboration, and the effect is similar to that of the human-human collaboration.