PDL on Finite Linear Orders with Lookahead

• PDL on Finite Linear Orders is an extension of standard PDL that incorporates lookahead operators to reason about regular expression equivalences over finite sequences.
• The framework introduces explicit identity and anti-identity restrictions, enabling precise control over lookahead modalities in regular programs.
• Automata-theoretic methods demonstrate that the logic’s satisfiability is EXPTIME-complete, supporting robust formal verification and complexity analysis.

Propositional Dynamic Logic (PDL) on finite linear orders is an extension of standard PDL specifically designed to reason about the equivalences of regular expressions with lookahead (REwLA) over finite sequences. The framework provides a Hilbert-style axiomatization that is both sound and complete for specifying logical properties of actions and tests represented as regular programs enhanced with lookahead mechanisms, capturing both match-language and largest substitution-closed equivalences relevant to REwLA semantics. The logic introduces explicit operators for restriction to the identity relation and its complement, enabling fine-grained control over lookahead modalities, and its computational properties align with established complexity bounds for regular languages with lookahead (Nakamura, 21 Jan 2026).

1. Syntax and Program Constructs

The language of PDL on finite linear orders uses two disjoint sets: $A$ for program (action) variables $a,b,\ldots$, and $P$ for propositional variables $p,q,\ldots$. Programs, denoted $\alpha, \beta$, are defined recursively as: $$\alpha, \beta \Coloneqq a \mid \alpha;\beta \mid \alpha+\beta \mid \alpha^{+} \mid \alpha^{\circ} \mid \alpha^{\smallsmile}$$ where:

• $\alpha;\beta$ is sequential composition,
• $\alpha+\beta$ is nondeterministic choice,
• $\alpha^{+}$ is Kleene plus (one-or-more iteration),
• $\alpha^{\circ}$ denotes restriction to the identity (semantically, $\alpha \cap \mathit{Id}$),
• $\alpha^{\smallsmile}$ denotes restriction to the complement of the identity ($\alpha \setminus \mathit{Id}$).

Formulas $\varphi, \psi$ are constructed as: $$\varphi, \psi \Coloneqq p \mid \bot \mid \varphi \to \psi \mid [\alpha]\varphi$$ Negation, disjunction, and diamond modalities are encoded as usual via: $\lnot\varphi := \varphi \to \bot$, $\varphi\lor\psi := \lnot\varphi\to\psi$, $$\langle\alpha\rangle\varphi := \lnot[\alpha]\lnot\varphi$$.

2. Relational Semantics on Finite Linear Orders

The semantics unfolds over finite strict linear orders: $W$ is a finite set equipped with a transitive, irreflexive, connex relation $<$. The models are tuples $M=(W, \{a^M\}_{a\in A}, \{p^M\}_{p\in P})$, with constraints that $<^M$ is a strict linear order and $a^M \subseteq <^M$ for all $a$.

Program terms are interpreted as subsets of $<^M$:

• $$R;S = \{(x,z) \mid \exists y.\; (x,y)\in R \land (y,z)\in S\}$$
• $R+S = R \cup S$
• $R^{+} = \bigcup_{n\ge1} R^n$
• $R^{\circ} = R \cap \{(x,x)\mid x\in W\}$
• $R^{\smallsmile} = R \setminus \{(x,x)\mid x\in W\}$

Formulas are interpreted recursively:

• $\llbracket p \rrbracket^M = p^M$
• $\llbracket \bot \rrbracket^M = \emptyset$
• $$\llbracket \varphi\to\psi \rrbracket^M = W \setminus \llbracket\varphi\rrbracket^M \cup \llbracket\psi\rrbracket^M$$
• $$\llbracket[\alpha]\varphi\rrbracket^M = \{x \mid \forall y.\; (x,y)\in \llbracket\alpha\rrbracket^M \implies y\in\llbracket\varphi\rrbracket^M\}$$

The restriction operators are only applied to subrelations $a^M\subseteq <^M$, ensuring well-defined interpretations in the strict order setting (Nakamura, 21 Jan 2026).

3. Hilbert-Style Axiomatization

Derivability in this logic, denoted $\vdash_{\sf PDL^*} \varphi$, is specified by a finite Hilbert-style system $\mathcal{H}$, comprising axioms and inference rules:

• Inference Rules
  • (MP) Modus Ponens: from $\varphi$ and $\varphi \to \psi$, infer $\psi$
  • (Nec) Necessitation: from $\varphi$, infer $[\alpha]\varphi$
• Axiom Schemata
  • Propositional axioms: all propositional tautologies and closure under uniform substitution
  • K-axioms:
  • $$[\alpha](\varphi\to\psi)\to([\alpha]\varphi\to[\alpha]\psi)$$
  • $$[\alpha+\beta]\varphi\leftrightarrow([\alpha]\varphi\land[\beta]\varphi)$$
  • $$[\alpha;\beta]\varphi\leftrightarrow[\alpha][\beta]\varphi$$
  • Test: $[\varphi?]\psi\leftrightarrow(\varphi\to\psi)$
  • Kleene plus axioms (Segerberg “+” style):
  • $$[\alpha^+]\varphi\leftrightarrow([\alpha]\varphi\land[\alpha][\alpha^+]\varphi)$$
  • $$( [\alpha]\varphi\land[\alpha^+](\varphi\to[\alpha]\varphi) ) \to[\alpha^+]\varphi$$
  • Identity and anti-identity (lookahead) axioms:
  • $$[\alpha^\smallsmile]\varphi\leftrightarrow[\alpha^\smallsmile]\bot?\;\varphi$$ (enforcing $\alpha^\smallsmile=\alpha\setminus Id$)
  • $$[\alpha^\circ]\varphi\leftrightarrow[\alpha^\circ]\bot?\;\varphi$$ (enforcing $\alpha^\circ=\alpha\cap Id$)
  • Löb’s axiom (well-foundedness, for the $\alpha^\smallsmile$-free part):

  $$[(\alpha^\smallsmile)^+]\bigl([\,(\alpha^\smallsmile)^+]\varphi \to \varphi\bigr) \to [(\alpha^\smallsmile)^+]\varphi$$

This system is both sound and complete for the class of finite (strict) linear orders with the extended operators (Nakamura, 21 Jan 2026). The restriction of lookahead operators to atomic actions is critical for avoiding ill-definedness due to lack of reflexivity on strict linear orders.

4. Completeness and Reduction to Identity-Free PDL

Completeness is established via a reduction to an identity-free variant of PDL suitable for strict linear orders and employing a canonical model with Löb’s lemma to guarantee well-foundedness. The steps are as follows:

• Translation to Normal Form: For every program $\alpha$, compute a pair $(\alpha_1, \alpha_2)$ such that $\alpha_1$ captures the identity-part (lookahead along $\mathit{Id}$), and $\alpha_2$ the strict-part. This yields $$[\alpha]\psi \leftrightarrow [\alpha_1]?\;\psi \land [\alpha_2]\psi$$ derivable in the Hilbert system.
• Identity-Free Reduction: By encoding $[\alpha_1]?$ as fresh propositional variables, the completeness argument for the extended system is reduced to a completeness proof for PDL without $\alpha^\circ$ or $\alpha^{\smallsmile}$ over finite strict linear orders.
• Canonical Model and Löb’s Lemma: The identity-free system is axiomatized by K-axioms, Segerberg’s $\alpha^+$ axioms, and Löb’s axiom. The completeness proof constructs a finite canonical model using atoms and ensures well-foundedness of the order via Löb's axiom.

The composition of these reductions shows every formula valid on all finite linear or strict linear orders is derivable in the extended Hilbert system (Nakamura, 21 Jan 2026).

5. Decidability and Complexity

The model theory enforced by finite-order axioms ensures all models are bounded finite linear frames. Decidability and computational complexity are established using automata-theoretic arguments:

• EXPTIME-Completeness: Satisfiability (and hence validity) of the full extended PDL on finite linear orders is EXPTIME-complete. Any formula can be translated in polynomial time to an alternating tree automaton, and the automaton emptiness problem is EXPTIME-complete.
• PSPACE-Completeness (Deterministic Case): Imposing determinism (at most one outgoing edge per action label) aligns the logic exactly with match-language equivalence of REwLA, reducing to alternating string automata, yielding PSPACE-completeness.
• Equivalence with REwLA: When restricting to pure regular expressions with lookahead, standard/substitution-closed equivalence checking inherits these EXPTIME or PSPACE bounds, matching those known for REwLA itself.

The complexity landscape is summarized as:

Framework	Complexity
Extended PDL on finite linear orders	EXPTIME-complete
Deterministic (match-language REwLA equivalence)	PSPACE-complete
REwLA (equivalence checking)	EXPTIME/PSPACE

Both the logic and its complexity coincide with those of regular expressions with lookahead under suitable restrictions (Nakamura, 21 Jan 2026).

6. Significance and Applications

This axiomatization provides rigorous logical tools for regular expression equivalence reasoning in finite contexts with lookahead, a feature critical to advanced formal verification, programming language theory, and automata-based specification frameworks. The restriction operators enable expressing both positive and negative lookahead at the logical level, directly corresponding to common constructs in extended regular expressions. The tractability results align with established automata-theoretic complexity, supporting practical assessment of expressiveness and analysis in automated reasoning settings (Nakamura, 21 Jan 2026).