Data-driven Optimal Preview Output Tracking of Linear Discrete-time Systems

Abstract

This paper is focused on an approach of designing an optimal preview output tracking controller for discrete-time systems with measurable data. It is assumed that the reference trajectory is previewable, i.e., some finite future as well as present and past values of the reference signal are known in advance. An augmented state-space system including the prior information in a preview horizon is firstly constructed. A linear quadratic regulator (LQR) representation is then developed by taking some manipulation on the original output tracking problem. Next, a Q-function based value Iteration (VI) algorithm is presented to obtain the optimal output tracking control gain using online measurable data. It’s shown that the presented approach does not require a priori information of system dynamics and an initially stabilizing control is also no longer needed. Finally, a numerical simulation is carried, which shows the merits of the proposed scheme.