A Deadzone Modification for Learning Control of Time-Varying Uncertain Systems

This paper presents a learning control method for time-varying uncertain systems, whose uncertainties are bounded by a parameterized function and the unknowns are assumed to be time-varying. A deadzone-modified iterative update mechanism to handle initial condition errors is incorporated in the control design, without assuming that the bound on initial condition errors is small enough in consideration of tracking performance. The resultant tracking error is shown to converge into a pre-specified deadzone, while boundedness of all the signals in the closed-loop learning system is guaranteed. The performance improvement is made possible by our design, due to the introduction of the width-deducing deadzone which boundary layer is governed by an initial rectified attractor defined based on constructive controllability analysis. The learning control method is also presented to deal with time-varying parametric uncertainties