Model-reference based wave-variable force control

Abstract

Internal friction, backlash and structural compliance within a robot makes force control a tricky non-colocated control problem. Traditional force controllers are typically tuned for a specific environment and invariably have instability problems when contacting stiffer environments. This paper presents a new force controller designed to use control effort to compensate for energy losses due to friction without modifying the robots underlying passive dynamics. The controller uses a lossless model of the robot's inertia to predict the robot's motion in real time based on the wave-variables flowing between robot and environment. The model motion is used as a desired input for a wave-variable based controller. To calculate the wave variables the controller uses measurements of both the contact force and the acceleration at the end-effector. The resulting controller is compared analytically and experimentally to more traditional controllers.