A trellis notion for distributed system diagnosis with sequential semantics

Abstract

We consider modular automata, obtained as a product of elementary components, and adopt the usual sequential semantics: runs of these systems are sequences of events (in contrast to partial orders). The set of all runs of an automaton can be described by a trellis, which forms the support of many recursive estimation or decision algorithms. In this paper, we revise this notion to adapt it to the case of modular automata, and obtain a factorization property on this revised trellis. This factorization opens the way to distributed estimation algorithms that were described up to now in partial order semantics. © 2006 IEEE.