Abstract

Metaheuristics are powerful techniques for solving hard real-world problems in many application domains. Their behavior and performance strongly depend on their ability to efficiently explore and exploit the search space. A well-known metaheuristic is Ant Colony Optimization which has been successfully applied to solve many engineering problems. In this paper, we focus on ACO that solves Constraint Satisfaction Problems. In this context ACO has already shown to be able to solve many difficult CSPs, however, some problems are still very hard for this kind of technique. We introduce two strategies that allow improving ACO intensification and diversification process: one for learning in a pre-processing step and a second strategy to focus the search to a feasible space. Our results suggest that both the learning phase as well as the strategy to focus the search allow improving the ACO performance especially to solve hard CSPs. Moreover, these strategies can be applied to ACO for other application domains.