A stabilizing iteration scheme for model predictive control based on relaxed barrier functions

TL;DR

This paper introduces a new iterative scheme for model predictive control that ensures stability and constraint satisfaction using relaxed barrier functions, even with limited optimization iterations per sampling step.

Contribution

It develops a stabilizing iteration scheme for MPC with relaxed barrier functions, explicitly accounting for suboptimality and algorithm dynamics in stability analysis.

Findings

Asymptotic stability of the closed-loop system is proven.

Limited optimization iterations can achieve satisfactory stability.

Numerical example demonstrates effective control with a single iteration.

Abstract

We propose and analyze a stabilizing iteration scheme for the algorithmic implementation of model predictive control for linear discrete-time systems. Polytopic input and state constraints are considered and handled by means of so-called relaxed logarithmic barrier functions. The required on-line optimization is based on warm starting and performs only a limited, possibly small, number of optimization algorithm iterations between two consecutive sampling instants. The optimization algorithm dynamics as well as the resulting suboptimality of the applied control input are taken into account explicitly in the stability analysis, and the origin of the resulting overall closed-loop system, consisting of state and optimization algorithm dynamics, is proven to be asymptotically stable. The corresponding constraint satisfaction properties are also analyzed. The theoretical results and a presented numerical example illustrate the fact that asymptotic stability as well as a satisfactory closed-loop performance can be achieved by performing only a single optimization algorithm iteration at each sampling step.