On the precision in polyhedral partition representation and the fragility of PWA control - CentraleSupélec Access content directly
Conference Papers Year : 2017

On the precision in polyhedral partition representation and the fragility of PWA control

Rajesh Koduri
Sorin Olaru

Abstract

Explicit model predictive control (EMPC) solves a multi-parametric Quadratic Programming (mp-QP) problem for a class of discrete-time linear system with linear inequality constraints. The solution of the EMPC problem in general is a piecewise affine control function defined over non-overlapping convex polyhedral regions composing a polyhedral partition of the feasible region. In this work, we consider the problem of perturbations on the representation of the vertices of the polyhedral partition.Such perturbations may affect some of the structural characteristics of the PWA controller such as " non-overlapping within the regions " or " the closed-loop invariance ". We first show how a perturbation affects the polyhedral regions and evoke the overlapping within the modified polyhedral regions. The major contribution of this work is to analyze to what extend the non-overlapping and the invariance characteristics of the PWA controller can be preserved when the perturbation takes place on the vertex representation. We determine a set called sensitivity margin to characterize for admissible perturbation preserving the non-overlapping and the invariance property of the controller. Finally, we show how to perturb multiple vertices sequentially and reconfigure the entire polyhedral partition.
Fichier principal
Vignette du fichier
root.pdf (1.75 Mo) Télécharger le fichier
Origin : Files produced by the author(s)
Loading...

Dates and versions

hal-01648012 , version 1 (24-11-2017)

Identifiers

Cite

Rajesh Koduri, Sorin Olaru, Pedro Rodriguez-Ayerbe. On the precision in polyhedral partition representation and the fragility of PWA control. 56th IEEE Conference on Decision and Control (CDC 2017), Dec 2017, Melbourne, Australia. ⟨10.1109/cdc.2017.8264026⟩. ⟨hal-01648012⟩
104 View
291 Download

Altmetric

Share

Gmail Facebook X LinkedIn More