Reversible pushdown automata

Abstract

Reversible pushdown automata are deterministic pushdown automata having the property that any configuration occurring in any computation has exactly one predecessor. In this paper, the computational capacity of reversible computations in pushdown automata is investigated and turns out to lie properly in between the regular and deterministic context-free languages. Furthermore, it can be shown that a deterministic context-free language cannot be accepted reversibly if more than realtime is necessary for acceptance. Closure properties as well as decidability questions for reversible pushdown automata are studied. Finally, the question of whether a given (nondeterministic) pushdown automaton can be algorithmically tested for reversibility is answered in the affirmative, whereas it is shown to be undecidable whether the language accepted by a given nondeterministic pushdown automaton is reversible. © 2010 Springer-Verlag Berlin Heidelberg.