Improving Nogood Recording Using 2SAT

Peter J. Stuckey, Lei Zheng

Abstract: Nogood recording is a dynamic learning technique widely applied to solve CSP (constraint satisfaction problems). It is highly effective in reducing the search space for SAT (satisfiability) problems. While SAT is NP-complete, the problem restricted to binary clauses (2SAT) is solvable in linear time. We can improve SAT solving by incorporating 2SAT solving techniques. In this paper we investigate extending nogood recording to make use of binary clause resolution. Our experiments show that nogoods generated from binary resolution can significantly reduce the search space, and size of nogoods generated, as well as the search time.