On structurally constrained ℋ ₂ performance bounds for stable MIMO plant models

Abstract

This paper is about optimal control problems in which the controller must satisfy sparsity structure constraints. Conditions are derived under which the optimal controller associated to an unconstrained quadratic performance index is naturally structured and, as a consequence, the sparsity constraint imposed on the controller has no impact on the optimal loop performance. The results are then applied to study the control of triangular plants that have to be controlled by triangular controllers. We derive explicit characterisations of the Youla parameter that defines an optimal triangular controller and of the performance loss associated to the structural restriction imposed on the controller.

Control system design subject to SNR constraints

“…Hence, the equivalent controller K in Figure 2(b) has entries that are identically zero. Thus, the optimal design of K becomes an optimal control problem with sparsity constraints (see, e.g., [25,31]). Such problems are, in general, non-convex and difficult to solve [33].…”

Analysis and design of networked control systems using the additive noise model methodology

“…The reasons for this choice are manifold and include ease of design, simplified tuning, and implementation related aspects such as cabling or geographic plant distribution. It is well known that restricting the controller architecture generally limits the achievable performance (see, e.g., [29][30][31]). Within this context, networks can play significative roles.…”

“…It is well known that restricting the controller architecture generally limits the achievable performance (see, e.g. , 29–31). Within this context, networks can play significative roles.…”

A brief introduction to the analysis and design of Networked Control Systems

“…The reasons for this choice are manifold and include ease of design, simplified tuning, and implementation related issues such as cabling or geographic plant distribution. It is well known that restricting the controller architecture generally limits the achievable performance (see, e.g., [25][26][27]). Within this context, networks can play significative roles.…”