A model predictive control based on input-output characterization of finite-horizon linear systems

Authors:	Kojima Akira, Tokyo Metropolitan Institute of Technology, Japan Kageyama Yohei, Hitachi Ltd., Japan
Topic:	1.2 Adaptive and Learning Systems
Session:	Optimal and Adaptive Control
Keywords:	control theory, constraints, model predictive control, discrete-time system, singular value decomposition.

Abstract

An LQ control problem for constrained discrete-time systems is discussed based on model predictive control with reduced order quadratic programming (QP) problem. By introducing a singular value decomposition for linear systems, where zero-order holder is embedded in the input signals, an approximation method of model predictive control is derived. The feature of resulting control system is illustrated with numerical examples.