Model Predictive Control of the Degree of Automation Optimizing Robot Health

Abstract

Adjusting operator support in human-machine systems is a promising way of combining operator involvement with high overall system performance. Adaptive automation aims to achieve this goal without burdening the operator with the task of selecting and setting the desired amount of support. In this work, two novel adaptive automations are presented. We use the performance measure of “robot health” to formulate the optimal control problem of maximizing the robot health of a human-robot system through the adaption of operator support to develop two model predictive controllers. The first one considers discrete levels of operator support, or levels of automation, the second one uses the continuous conception of the degree of automation. We report on a proof-of-concept simulation study evaluating the proposed model predictive controllers in a collaborative teleoperation of a mobile robot; the results demonstrate the ability of both model predictive controllers to successfully arbitrate control between the operator and the robot’s controller to maximize robot health.