A recursive algorithm for diagnosis in hierarchical finite-state machines

Abstract

The authors have recently developed a framework for failure diagnosis in hierarchical finite-state machines (HFSMs). This diagnosis method reduces random access memory requirements and changes the complexity of the diagnosis computations from exponential to polynomial if the subsystems of the HFSM do not have common unobservable events. In this paper, a recursive algorithm is provided for diagnosis computations (and more generally, observer design) in multilevel HFSMs. Some of the procedures provided in this work for diagnosis computations are equivalent to those proposed in the authors' previous work. However these computations have been reformulated in terms of an updator function so that they lend themselves better for extension to a recursive algorithm for the more general case of multi-level HFSMs.