Iterative learning using simplex gradient for human-robot cooperation

Abstract

Variable impedance control has been widely used in Human-Robot Collaboration to improve safety and adaptability. In this paper, a new method for adjusting impedance parameters based on iterative learning is proposed to reduce human effort. We consider a human-robot collaborative task where human operator moves the end-effector from a start position to a goal position. First, impedance parameters are made as a function of time by using several gaussian functions. Parameters of these gaussian functions are then iteratively updated with simplex gradient which is calculated from past data. Finally, experiments with a 2-DOF Planar Robot arm are conducted to verify performance of proposed method.