A Model for Operator Endpoint Stiffness Prediction during Physical Human-Robot Interaction

Abstract

Physical contact established during interaction between a human operator and a haptic device creates a coupled system with stability and performance characteristics different than its individual subsystems taken in isolation. Proper incorporation of operator dynamics in physical human-robot interaction (pHRI) conditions requires knowledge of system variables and parameters, some of which are not directly measurable. Operator endpoint impedance, for instance, cannot be directly measured in typical haptic control conditions. Several endpoint impedance estimation techniques have been explored in previous literature, based on measured kinematics and/or other correlated metrics. Arm muscle activity, measured through surface electromyography (sEMG), has been used in previous literature to estimate endpoint stiffness, which is the static component of impedance. Co-activation (co-contraction) of antagonistic arm muscles forming a pair around a joint is known to be the driving factor in modulation of endpoint stiffness. However, previous work employing muscle co-contraction to predict endpoint stiffness has mainly been absent, due in part to the inefficacy of operator models to incorporate muscle co-activation into the prediction scheme.