Learning-based Adaptive Optimal Impedance Control to Enhance Physical Human-robot Interaction Performance

摘要

This paper presents a framework of adaptive optimal impedance control to enhance physical human-robot interaction (pHRI) performance. The overall structure of the proposed control scheme consists of an outer control loop and an inner control loop. In the outer control loop, a cost function that considers human motion and interaction force is minimized to optimize the overall human-robot interaction performance. An adaptive impedance controller is designed based on a Q-learning algorithm to realize impedance adaptation and guarantee the impedance parameters converge to the optimal value with completely unknown dynamics of the human limb. Then, a torque controller is developed in the inner control loop to enable the robot respond follow the obtained impedance model. In this controller, a novel barrier Lyapunov function (BLF) is employed to guarantee the error constraint and radial basis function neural networks (RBFNNs) are utilized to approximate the unknown robot dynamics. Stability and uniform boundedness of the closed-loop system are validated. Numerical simulation studies are performed to verify the effectiveness of the proposed controller.