Qualitative theory of hybrid dynamical systems. (Q1425545)

This study investigates the qualitative behaviour of hybrid systems, i.e. systems evolving on \(\mathbb{R}^n\times Q\) where \(Q\) is a discrete set. This evolution is described by the coupling of the dynamics of an ODE and that of an automaton. Once the continuous state reaches a prescribed set \(S\), the state of the automaton switches to another state and another ODE is activated. The main theme is the characterization of periodic behaviour of such systems and, when it is obtained, the study of its stability. The first chapter considers the case when the continuous dynamics is constant, the constants being associated to \(Q\) supposed finite, and when \(S\) is given by a union of hyperplanes (as many as the number of elements of \(Q)\). Some convenient assumptions are postulated to guarantee well-posedness, and next existence and stability of limit cycles are studied. A Lyapunov function is used for stability analysis. The practicality of some of the above mentioned assumptions should have been explored. Let us observe that the term ``linear'' chosen by the authors for this kind of hybrid system is prone to confusion: the continuous dynamics is not linear. An example of aperiodic (chaotic) behaviour is described. In the next chapter, the switching law for the automaton is multivalued and the continuous dynamics nonlinear. Assumptions are provided so that the so-called reducible ``differential automata'' can be decomposed into finitely many differential automata with discrete cyclic dynamics. Periodicity is also obtained for the continuous component. In the third chapter, \(n=2\) and the results just obtained are used to get a Poincaré-Bendixson theorem. The continuous and discrete dynamics are taken care by these two assumptions. In chapter four, the results of chapter two are applied to the configuration of chapter one with some assumptions (convexity, polyhedra) on the sets associated to the switching (like \(S)\). One gets criteria for the existence and number of limit cycles as well as their stability. The three next chapters apply the results to manufacturing systems, queuing networks, and extend them using more realistic models like time-varying arrival rates. For cases with multiple servers, a decentralized control policy is proposed. The last chapter addresses open issues: uniqueness of the limit cycle, nonlinearity of the continuous dynamics, assessment of the conservativeness of the assumptions for guaranteeing nonchaotic behaviour. This latter aspect is important because it would help assessing the impact of this study. There is a danger of proving what one wants to prove by adapting the hypotheses to the desired result. This is \textit{per se} a despicable scientific attitude. Sometimes the lengthy assumptions in this book seem to be justified by the needs of the cause. On the other hand, this is perhaps needed in a first approach to difficult problems. This mongraph is an exploration, and its merit is to convince the reader of the difficulty and richness of the topic under investigation.