Reachability in Higher-Order-Counters

TL;DR

This paper analyzes the computational complexity of reachability problems in higher-order counter automata, revealing how zero-tests and automaton levels influence complexity and language hierarchies.

Contribution

It establishes complexity bounds for reachability in higher-order counter automata with and without zero-tests, and connects these results to language hierarchy separations.

Findings

Control-state reachability with zero-test is (k-2)-fold exponential space-complete.

Without zero-test, reachability is (k-2)-fold exponential time-complete.

Level 2 HOCS reachability is PTIME-complete.

Abstract

Higher-order counter automata (\HOCS) can be either seen as a restriction of higher-order pushdown automata (\HOPS) to a unary stack alphabet, or as an extension of counter automata to higher levels. We distinguish two principal kinds of \HOCS: those that can test whether the topmost counter value is zero and those which cannot. We show that control-state reachability for level $k$ \HOCS with $0$-test is complete for \mbox{$(k-2)$}-fold exponential space; leaving out the $0$-test leads to completeness for \mbox{$(k-2)$}-fold exponential time. Restricting \HOCS (without $0$-test) to level $2$, we prove that global (forward or backward) reachability analysis is $\PTIME$-complete. This enhances the known result for pushdown systems which are subsumed by level $2$ \HOCS without $0$-test. We transfer our results to the formal language setting. Assuming that $\PTIME \subsetneq \PSPACE \subsetneq \mathbf{EXPTIME}$, we apply proof ideas of Engelfriet and conclude that the hierarchies of languages of \HOPS and of \HOCS form strictly interleaving hierarchies. Interestingly, Engelfriet's constructions also allow to conclude immediately that the hierarchy of collapsible pushdown languages is strict level-by-level due to the existing complexity results for reachability on collapsible pushdown graphs. This answers an open question independently asked by Parys and by Kobayashi.