[Paper Review] Model-checking an Epistemic μ-calculus with Synchronous and Perfect Recall Semantics

Rodica Bozianu, Cătălin Dima+1 more|arXiv (Cornell University)|Apr 10, 2012

Logic, programming, and type systems25 references4 citations

TL;DR

This paper presents a decidable fragment of the epistemic µ-calculus with synchronous and perfect recall semantics, where epistemic modalities can apply to non-closed formulas under a syntactic restriction preventing common knowledge constructions. The key contribution is a generalized Finite Model Theorem that reduces model checking on infinite tree unfoldings to finite-state interpretations via determinization of agent projections, enabling decidability through predicate transformer semantics and in-splitting constructions.

ABSTRACT

We show that the model-checking problem is decidable for a fragment of the epistemic μ-calculus. The fragment allows free variables within the scope of epistemic modalities in a restricted form that avoids constructing formulas embodying any form of common knowledge. Our calculus subsumes known decidable fragments of epistemic CTL/LTL. Its modal variant can express winning strategies in two-player games with one player having imperfect information and non-observable objectives, and, with a suitable encoding, decidable instances of the model-checking problem for ATL with imperfect information and perfect recall can be encoded as instances of the model-checking problem for this epistemic μ-calculus.

Motivation & Objective

• To identify a larger fragment of the epistemic µ-calculus with decidable model checking beyond closed formulas.
• To address the expressivity gap in existing decidable fragments that cannot express winning strategies in imperfect-information games.
• To develop a semantic framework based on tree unfoldings and predicate transformers that supports non-closed formulas under restricted epistemic nesting.
• To establish a generalized Finite Model Theorem linking tree and finitary interpretations for the proposed fragment.
• To prove non-elementary hardness of the model-checking problem, confirming its complexity ceiling.

Proposed method

• Introduce a syntactic restriction: epistemic modalities for agents a and b may only be applied to non-closed subformulas if their observability relations are compatible, i.e., Πa ⊆ Πb or vice versa.
• Adapt predicate transformer semantics to the epistemic µ-calculus, defining interpretations over both tree unfoldings and finitary models.
• Construct a tree of in-splitting mappings TInsφ that tracks state transformations across formula nodes, ensuring compatibility with the semantics.
• Define a state-transformer tree Tstrφ that computes the truth of subformulas at each node using bottom-up evaluation over the formula tree.
• Use a generalized Finite Model Theorem to show that the tree interpretation of a formula is equivalent to a finitary interpretation over a determinized projection of the original model.
• Prove decidability by reducing model checking to checking membership of the initial state in the final state-transformer output.

Experimental results

Research questions

• RQ1 Can the model-checking problem for the epistemic µ-calculus be decided when epistemic modalities apply to non-closed formulas, provided common knowledge is avoided?
• RQ2 Is there a semantic framework that generalizes the Finite Model Theorem to epistemic µ-calculus with perfect recall and synchronous semantics?
• RQ3 Can winning strategies in two-player games with imperfect information and non-observable objectives be encoded in this fragment?
• RQ4 Does the proposed restriction on epistemic nesting (via observability relation inclusion) suffice to ensure decidability without sacrificing key expressivity?
• RQ5 Is the model-checking problem for this fragment non-elementary hard, as in related logics like LTLK?

Key findings

• The model-checking problem for the proposed fragment of the epistemic µ-calculus is decidable, even when epistemic modalities apply to non-closed formulas.
• The decidability result is achieved via a generalized Finite Model Theorem that links infinite tree interpretations to finite-state models through determinization of agent projections.
• The construction relies on in-splitting mappings and predicate transformers that preserve truth across formula nodes, ensuring correctness of the bottom-up evaluation.
• The fragment avoids expressing common knowledge by syntactically restricting epistemic nesting to compatible observability relations, i.e., Πa ⊆ Πb or vice versa.
• The model-checking problem is non-elementary hard, as shown by a reduction from the emptiness problem for star-free regular expressions.
• The approach avoids reliance on tree automata or two-player games, which are insufficient for capturing the full expressivity of the fragment.