The analysis of optimization based controllers

Abstract

Many control techniques employ online optimization in the determination of a control policy. We develop a framework which provides sufficient convex conditions, in the form of linear matrix inequalities, for the analysis of constrained quadratic based optimization schemes. These results encompass standard robustness analysis problems for a wide variety of receding horizon control schemes, including polytopic, structured, and measurement uncertainty for schemes with or without end constraints, observers, or input and output constraints. A simple example illustrates the methodology.