Analysis of Boolean Equation Systems through Structure Graphs

We analyse the problem of solving Boolean equation systems through the use of structure graphs. The latter are obtained through an elegant set of Plotkin-style deduction rules. Our main contribution is that we show that equation systems with bisimilar structure graphs have the same solution. We show that our work conservatively extends earlier work, conducted by Keiren and Willemse, in which dependency graphs were used to analyse a subclass of Boolean equation systems, viz., equation systems in standard recursive form. We illustrate our approach by a small example, demonstrating the effect of simplifying an equation system through minimisation of its structure graph.

💡 Research Summary

The paper tackles the longstanding problem of solving Boolean Equation Systems (BES) by introducing a novel graphical representation called a structure graph. A structure graph captures every variable, logical operator (∧, ∨, ¬), and the nesting of sub‑expressions as vertices, while edges encode the dependency and evaluation order among them. The construction of such graphs is driven by a set of Plotkin‑style deduction rules: each equation X = φ is recursively decomposed, creating a vertex for φ, vertices for its immediate sub‑expressions, and labeled edges that reflect the logical connective. Negation is handled by annotating the corresponding vertex with a “¬” label. This rule‑based approach mirrors the operational semantics of process calculi, ensuring that the graph faithfully mirrors the syntactic structure of the original BES.

The central theoretical contribution is the bisimilarity‑preserves‑solution theorem. Two structure graphs are said to be bisimilar if there exists a relation pairing their vertices such that (i) paired vertices carry identical labels, and (ii) their outgoing edges correspond one‑to‑one under the relation. The authors prove, via structural induction on the graph, that any bisimilar pair of graphs yields BES with exactly the same truth‑value assignment for every variable. Consequently, solving a BES can be reduced to solving any bisimilar (and potentially much smaller) graph representation.

This result subsumes earlier work by Keiren and Willemse, who employed dependency graphs to analyse a restricted subclass of BES—those in Standard Recursive Form (SRF). Dependency graphs only capture direct variable‑to‑variable dependencies and cannot represent nested logical operators. By contrast, structure graphs are expressive enough to model arbitrary BES, and when a BES happens to be in SRF, its structure graph collapses to the previously studied dependency graph. Hence the new theorem is a conservative extension of the earlier SRF results.

A practical upshot of the bisimilarity theorem is the possibility of graph minimisation. Using standard partition‑refinement algorithms (akin to those used for automata minimisation), the authors merge bisimilar vertices, thereby shrinking the graph. The resulting minimal structure graph corresponds to a simplified BES with fewer variables and less complex right‑hand sides, yet it is guaranteed to have the same solution as the original system. The paper illustrates this process with a compact example: a BES comprising seven variables and a mixture of conjunctions, disjunctions, and negations is transformed into a structure graph, bisimilarity is computed, and the graph is reduced to four vertices. Solving the reduced system yields exactly the same truth values as solving the original, confirming the correctness of the minimisation.

Beyond the theoretical proof, the authors discuss implementation considerations. The Plotkin‑style rules lend themselves to a straightforward syntax‑directed translation from a BES to a graph, which can be automated in a compiler‑like front‑end. The bisimilarity check can be performed in O(|V| log |V|) time using efficient partitioning, making the approach scalable to moderately large systems. Moreover, because the minimised BES is syntactically smaller, downstream model‑checking or verification tools that consume BES (e.g., parity game solvers, μ‑calculus model checkers) benefit from reduced state‑space and faster convergence.

In the concluding section, the authors outline several avenues for future work. One promising direction is the integration of structure‑graph minimisation into existing verification pipelines, thereby providing an automatic preprocessing step that shrinks BES before they are handed to parity‑game solvers. Another is the quantitative evaluation of state‑space reduction on benchmark suites drawn from software model checking, where BES often arise from encoding of temporal properties. Finally, the authors suggest extending the framework to other fixed‑point logics, such as the modal μ‑calculus, where analogous graph‑based representations could yield similar bisimulation‑based simplifications.

Overall, the paper makes a substantial contribution by bridging the gap between algebraic BES solving and graph‑theoretic techniques. By formalising structure graphs, proving that bisimilarity preserves solutions, and demonstrating practical minimisation, the authors provide both a deeper theoretical understanding and a concrete toolset that can be leveraged to tame the combinatorial explosion typical of Boolean fixed‑point computations.