Extending GENET with lazy arc consistency

Many important applications, such as graph coloring, scheduling and production planning, can be solved by GENET, a local search method which is used to solve binary constraint satisfaction problems (CSP's). Where complete search methods are typically augmented with consistency methods to reduce the search, Local search methods are not. We propose a consistency technique, lazy are consistency, which is suitable for use within GENET. We show it can improve the efficiency of the GENET search on some instances of binary CSP's, and does not suffer the overhead of full me consistency.