Rerailing Automata

Abstract

In this paper, we introduce rerailing automata for $ω$-regular languages. They generalize both deterministic parity (DPW) and minimized history-deterministic co-Büchi automata (with transition based acceptance, HdTbcBW) while combining their favorable properties. In particular, rerailing automata can represent arbitrary $ω$-regular languages while allowing for polynomial-time minimization, just as HdTbcBW do. Since DPW are a special case of rerailing automata, a minimized rerailing automaton is never larger than the smallest deterministic parity automaton for the same language. We also show that rerailing automata can be used as a replacement for deterministic parity automata for the realizability check of open systems. The price to be paid to obtain the useful properties of rerailing automata is that the acceptance condition in such automata refers to the dominating colors along all runs for a given word, where just as in parity automata, the dominating color along a run is the lowest one occurring infinitely often along it. A rerailing automaton accepts those words for which the greatest of the dominating colors along the runs is even. Additionally, rerailing automata guarantee that every prefix of a run for a word can be extended to eventually reach a point from which all runs for the word extending the prefix have the same dominating color, and it is even if and only if the word is in the language of the automaton. We show that these properties together allow characterizing the role of each state in such an automaton in a way that relates it to state combinations in a sequence of co-Büchi automata for the represented language. This characterization forms the basis of the polynomial-time minimization approach in this paper.