A Hybrid Optimization Method for Combinatorial Optimization Problems with Disjunctive Constraints

Abstract

Global search and local search are two of major approaches to combinatorial optimization problems. We propose here a new hybrid optimization method of these two approaches. The method (1)divides the search space into small subspaces using disjunctive constraints contained in the problem, (2)wanders from subspace to subspace, (3)optimizes each subspace visited using a global search solver. With the method, one can apply (pseudo) local search to any combinatorial optimization problems which contain disjunctive constraints using only a global search solver. More over, by using constraint programming systems as those global search solvers, one can apply the method just describing the problem without programming. Experimental results showed that the hybrid solver can find better solutions more quickly than existing global search solvers such as clp(FD).