Variable Admittance Control Improving Stability in Physical Human-Robot Interaction

Abstract

This paper presents a variable admittance controller for eliminating instability in physical Human-Robot Interaction. In order to predict the instability, human hand stiffness is estimated in real time by modeling the level of muscle co-contraction of the human’s arms. A modified time-domain vibration index is then proposed based on the forces applied by the operator on the robot. The virtual damping coefficient in admittance controller is adjusted according to the human hand stiffness and the vibration index to ensure both the performance and stability of the interaction system. The proposed system is evaluated on an experimental set-up using an UR10 robot and the results demonstrate significant improvement in stable and low-effort cooperation.