Convex synthesis of localized controllers for spatially invariant systems (Q5960315)

The paper focuses on the problem of decentralized feedback control for multi-dimensional linear dynamical systems, and in particular 2-D (time and space) systems. The paper is very well written. Being accessible to non-specialists, the paper has a tutorial value, even though one can regret that the authors did not seem to be aware of the voluminous existing literature on multi-dimensional linear systems. As a result, many key references are missing. The authors are interested in localized controller design, which means that the control action at any given position is a function of information coming from a small number of neighboring positions. Roughly speaking, localized control can be seen as an intermediate between decentralized control (no information about neighboring system elements) and centralized control (information about all system elements). Because decentralized or localized controller design is generally intractable numerically, the authors focus on sufficient (potentially conservative) but convex linear matrix inequality (LMI) formulations. In order to do this, they follow the classical approach of enforcing structural constraints on the Lyapunov matrix certificate. Parametric dependence is handled with the Kalman-Yakubovich-Popov lemma, a standard tool used in transforming an infinite-dimensional parameter-dependent LMI into a finite-dimensional LMI. Numerical experiments are described, showing clear advantages over standard truncation methods.