A Fully Symbolic Bisimulation Algorithm

We apply the saturation heuristic to the bisimulation problem for deterministic discrete-event models, obtaining the fastest to date symbolic bisimulation algorithm, able to deal with large quotient spaces. We compare performance of our algorithm with that of Wimmer et al., on a collection of models. As the number of equivalence classes increases, our algorithm tends to have improved time and space consumption compared with the algorithm of Wimmer et al., while, for some models with fixed numbers of state variables, our algorithm merely produced a moderate extension of the number of classes that could be processed before succumbing to state-space explosion. We conclude that it may be possible to solve the bisimulation problem for systems having only visible deterministic transitions (e.g., Petri nets where each transition has a distinct label) even if the quotient space is large (e.g., 10 classes), as long as there is strong event locality.