Fast incremental unit propagation by unifying watched-literals and local repair

Abstract

The propositional satisfiability problem has been studied extensively due to its theoretical significance and applicability to a variety of fields including diagnosis, autonomous control, circuit testing, and software verification. In these applications, satisfiability problem solvers are often used to solve a large number of problems that are essentially the same and only differ from each other by incremental alterations. Furthermore, unit propagation is a common component of satisfiability problem solvers that accounts for a considerable amount of the solvers' computation time. Given this knowledge, it is desirable to develop incremental unit propagation algorithms that can efficiently perform changes between similar theories. This thesis introduces two new incremental unit propagation algorithms, called Logic-based Truth Maintenance System with Watched-literals and Incremental Truth Maintenance System with Watched-literals. These algorithms combine the strengths of the Logic-based and Incremental Truth Maintenance Systems designed for generic problem solvers with a state-of-the-art satisfiability solver data structure called watched literals.

(cont.) Empirical results show that the use of the watched-literals data structure significantly decreases workload of the LTMS and the ITMS without adversely affecting the incremental performance of these truth maintenance systems.