Robust, Near Time-Optimal Control of Nonlinear Second-Order Systems: Theory and Experiments

Abstract

A technique is presented for controlling second-order, nonlinear systems using a combination of bang-bang time-optimal control, sliding-mode control, and feedback linearization. Within the control loop, a state space evaluation of the system classifies the instantaneous dynamics into one of three regions, and one of three corresponding control algorithms is invoked. Using a prescribed generation of desirable sliding surfaces, the resulting combined controller produces nearly time-optimal performance. The combination controller is provably stable in the presence of model uncertainty. Experimental data are presented for the control of a General Electric GP132 industrial robot. The method is shown to achieve nearly time-optimal motion that is robust to modeling uncertainties. Representative transients compare favorably to bang-bang control and PD control.