Bisimulation equivalence of first-order grammars is Ackermann-hard

TL;DR

This paper proves that the problem of determining bisimulation equivalence for first-order grammars is Ackermann-hard, indicating extremely high computational complexity, by reducing from the reachability problem of reset counter machines.

Contribution

It establishes Ackermann-hardness for bisimulation of first-order grammars, extending the understanding of complexity beyond previously known decidability results.

Findings

Bisimulation equivalence is decidable but Ackermann-hard.

The proof involves a reduction from reset counter machine reachability.

The result highlights the high complexity of bisimilarity even without explicit epsilon-transitions.

Abstract

Bisimulation equivalence (or bisimilarity) of first-order grammars is decidable, as follows from the decidability result by Senizergues (1998, 2005) that has been given in an equivalent framework of equational graphs with finite out-degree, or of pushdown automata (PDA) with only deterministic and popping epsilon-transitions. Benedikt, Goeller, Kiefer, and Murawski (2013) have shown that the bisimilarity problem for PDA (even) without epsilon-transitions is nonelementary. Here we show Ackermann-hardness for bisimilarity of first-order grammars. The grammars do not use explicit epsilon-transitions, but they correspond to the above mentioned PDA with (deterministic and popping) epsilon-transitions, and this feature is substantial in the presented lower-bound proof. The proof is based on a (polynomial) reduction from the reachability problem of reset (or lossy) counter machines, for which the Ackermann-hardness has been shown by Schnoebelen (2010); in fact, this reachability problem is known to be Ackermann-complete in the hierarchy of fast-growing complexity classes defined by Schmitz (2013).