An optimal sliding mode controller applied to human motion synthesis with robotic implementation

Abstract

The operational space formulation is applied to a practical robot system in order to generate realistic human reaching motion based on the minimisation of `effort', a function of gravity and weighting gains. We present a novel optimal sliding mode controller that uses techniques of steepest descent to achieve this minimisation without affecting the task controller. The sliding mode optimal controller is verified both theoretically and by practical evaluation on simulated and physical two degree of freedom (dof) robotic arms. These arms produce redundant reaching motion that is similar to that observed from human subjects. We also present our modifications to the effort function, for implementation of smooth joint limits. A separate sliding mode controller for task control is also presented. Both sliding mode controllers guarantee robustness to model uncertainty and actuator disturbances e.g. friction.