$mu$-calculus on data words

We study the decidability and expressiveness issues of $\mu$-calculus on data words and data $\omega$-words. It is shown that the full logic as well as the fragment which uses only the least fixpoints are undecidable, while the fragment containing only greatest fixpoints is decidable. Two subclasses, namely BMA and BR, obtained by limiting the compositions of formulas and their automata characterizations are exhibited. Furthermore, Data-LTL and two-variable first-order logic are expressed as unary alternation-free fragment of BMA. Finally basic inclusions of the fragments are discussed.

💡 Research Summary

The paper investigates the μ‑calculus over data words and data ω‑words, focusing on decidability, expressive power, and the relationship with existing automata‑theoretic formalisms. A data word is a finite sequence over Σ×D, where Σ is a finite alphabet and D is an infinite set of data values; a data ω‑word is the infinite counterpart. The crucial feature is that only the equality relation on data values (the “class” relation) is observable, making the model invariant under arbitrary permutations of D.

The authors first define a modal μ‑calculus tailored to this setting. The language contains propositional variables, atomic propositions, and four navigation modalities: Xg (global successor), Xc (class successor), Yg (global predecessor), and Yc (class predecessor). Negation is allowed only on atomic propositions, and fixpoint operators µ (least) and ν (greatest) are introduced in the usual way. The semantics interprets each formula as a set of positions in a data word, using the four navigation relations.

The main decidability results are as follows. The full μ‑calculus, as well as the fragment that uses only least fixpoints (the µ‑fragment), are shown to be undecidable (Theorem 3.6). The proof reduces the halting problem of a two‑counter machine to the satisfiability problem, exploiting the ability of least fixpoints to encode unbounded counting. In contrast, the ν‑fragment (only greatest fixpoints) is decidable. The authors prove that every language definable in the ν‑fragment can be recognized by a data automaton (DA) (Theorem 3.8). Consequently, the ν‑fragment is closed under union, intersection, and mirroring, but not under complement. The same results hold for data ω‑words, where ν‑definable languages coincide with those recognized by data ω‑automata.

To obtain fragments that are also closed under complement while retaining decidability, the paper introduces two syntactic restrictions:

1. Bounded Reversal (BR) fragment – In any fixpoint formula, the number of switches between future modalities (Xg, Xc) and past modalities (Yg, Yc) is bounded by a fixed constant. This limits the number of “reversals” of the navigation direction. BR is strictly less expressive than the ν‑fragment (Theorem 4.8) but is closed under complement, and its decidability follows from inclusion in the ν‑fragment.

2. Bounded Mode Alternation (BMA) fragment – Here the number of switches between global modes (Xg/Yg) and class modes (Xc/Yc) is bounded. BMA ⊆ BR ⊆ ν‑fragment (Theorem 4.5). The authors show that BMA contains Data‑LTL (a linear‑time temporal logic with both word‑wise and class‑wise operators) and, consequently, FO² (two‑variable first‑order logic) (Theorem 6.4). Moreover, Data‑LTL with only unary modalities and FO² are shown to be equivalent to the unary alternation‑free part of BMA. For data ω‑words, BMA is also contained in data ω‑automata, though it is not contained in the ν‑fragment.

All decidable fragments are shown to have satisfiability problems equivalent (under elementary reductions) to reachability in vector addition systems (VAS), i.e., to the emptiness problem for data automata. Hence, their complexity is elementary (though non‑primitive‑recursive in the worst case).

The paper also provides a detailed comparison with other models of data languages: deterministic finite‑memory automata, data monoids, non‑deterministic finite‑memory automata, and alternating one‑register automata. It highlights the trade‑offs between expressive power, closure properties, and decidability, arguing that the ν‑fragment and its subclasses (BR, BMA) are strong candidates for a notion of “regular data languages”.

Finally, the authors discuss open problems, such as extending the BR fragment to data ω‑words, exploring richer fixpoint alternation hierarchies, and developing practical verification tools based on the presented fragments. Overall, the work offers a comprehensive logical framework for reasoning about data words, unifies several existing formalisms, and delineates the precise boundaries between decidable and undecidable extensions of the μ‑calculus in this setting.