CDSAT and QSMA for satisfiability modulo theories and assignments

The success of Conflict-Driven Clause Learning (CDCL) for propositional satisfiability motivates the quest for methods that realize the principle of conflict-driven reasoning for richer logics. After generalizing CDCL to first-order logic with SGGS, I worked on the CDSAT (conflict-driven satisfiability) method, that generalizes CDCL, MCSAT, and the Nelson-Oppen scheme, to realize conflict-driven satisfiability in unions of theories modulo assignments. The current version of CDSAT is for quantifier-free problems. Thus, I also worked on the QSMA (quantified satisfiability modulo theories and assignments) algorithm, that decides the satisfiability of formulas with arbitrary quantification modulo a complete theory and an initial assignment. QSMA will be integrated in CDSAT in future work.

CDCL, SGGS, and CDSAT are instances of a paradigm of conflict-driven reasoning:

An implementation of CDSAT was explored in a prototype SMT/SMA solver named Eos written by Giulio Mazzi:



Maria Paola Bonacina