Boolean Abstraction for Temporal Logic Satisfiability

Increasing interest towards property based design calls for effective satisfiability procedures for expressive temporal logics, e.g. the IEEE standard Property Specification Language (PSL). In this paper, we propose a new approach to the satisfiability of PSL formulae; we follow recent approaches to decision procedures for Satisfiability Modulo Theory, typically applied to fragments of First Order Logic. The underlying intuition is to combine two interacting search mechanisms: on one side, we search for assignments that satisfy the Boolean abstraction of the problem; on the other, we invoke a solver for temporal satisfiability on the conjunction of temporal formulae corresponding to the assignment. Within this framework, we explore two directions. First, given the fixed polarity of each constraint in the theory solver, aggressive simplifications can be applied. Second, we analyze the idea of conflict reconstruction: whenever a satisfying assignment at the level of the Boolean abstraction results in a temporally unsatisfiable problem, we identify inconsistent subsets that can be used to rule out possibly many other assignments. We propose two methods to extract conflict sets on conjunctions of temporal formulae (one based on BDD-based Model Checking, and one based on SAT-based Simple Bounded Model Checking). We analyze the limits and the merits of the approach with a thorough experimental evaluation.