Event identification and intelligent hybrid control

Abstract

Hybrid dynamical systems consist of two types of systems, a continuous state system called the plant and a discrete event system called the supervisor. Since the plant and supervisor are different types of systems, an interface is required to facilitate communication. An important issue in the design of hybrid control systems is the determination of this interface. Essentially, the interface associates logical symbols used by the supervisor with nonsymbolic events representative of the plant's behaviour. This chapter discusses a method for learning a hybrid system interface where symbols and events are bound in a way which is compatible with the goal of plant stabilization. The method is called event identification and provides an on-line method for adapting hybrid dynamical systems in the face of unforseen plant variations.KeywordsInvariant SubspaceAutonomous Underwater VehicleSwitching SurfacePhase Space TrajectoryHybrid Dynamical SystemThese keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.