A methodology for provably stable behaviour-based intelligent control

This paper presents a design methodology for a class of behaviour-based control systems, arguing its potential for application to safety critical systems. We propose a formal basis for subsumption architecture design based on two extensions to Lyapunov stability theory, the Second Order Stability Theorems, and interpretations of system safety and liveness in Lyapunov stability terms. The subsumption of the new theorems by the classical stability theorems serves as a model of dynamical subsumption, forming the basis of the design methodology. Behaviour-based control also offers the potential for using simple computational mechanisms, which will simplify the safety assurance process.