Double-loop Human-Robot Shared Control for Trajectory Tracking of an Underwater Robot

This paper addresses the challenge of achieving a compromise between tracking a desired trajectory and responding to a human operator's command for a remotely operated underwater robot. The proposed human-robot shared control scheme is structured as a double-loop system, comprising a kinematic controller and a dynamic controller. The outer-loop kinematic controller incorporates the control commands provided by the human operator. To enable a rapid and smooth transition between trajectory tracking and human command following, we employ a convex blending law that uses the human intent measure to adjust the weight of human input and the automatic control signal. The inner-loop dynamic controller tracks velocity reference in the presence of significant system uncertainties, external disturbances, and limitations in actuator saturation. The sparse online Gaussian process technique is leveraged to learn and compensate for unknown dynamics. By integrating the posterior means and variances into an integral sliding mode control framework, the designed dynamic controller ensures asymptotic convergence of velocity tracking errors with a speci-fied probability. Simulation results demonstrate the effectiveness of the proposed shared-control method.