Identification of hybrid systems

Project members: V. Verdult, B. De Schutter, T.J.J. van den Boom

Hybrid models consist of a combination of continuous dynamics (differential or difference equations) and discrete dynamics (discrete-events, logical switches, automata). They form an important class of dynamical systems for which analysis, verification and control techniques have been widely studied. However, identification techniques for hybrid systems have not received a lot of attention.

The goal of this research is to develop identification techniques for several classes of hybrid systems. The identification techniques will be based on our experience with state-space identification methods for linear and nonlinear systems, especially the subspace type of methods. Identification methods will be developed that determine a discrete-time multivariable state-space model using input and output measurements We have performed some preliminary work on the identification of hybrid systems that can be described by max-plus-linear discrete event systems.