We introduce an approach which combines ACO (Ant Colony Optimization) and IBM ILOG CP Optimizer for solving COPs (Combinatorial Optimization Problems). The problem is modeled using the CP Optimizer modeling API. Then, it is solved in a generic way by a two-phase algorithm. The first phase aims at creating a hot start for the second: it samples the solution space and applies reinforcement learning techniques as implemented in ACO to create pheromone trails. During the second phase, CP Optimizer performs a complete tree search guided by the pheromone trails previously accumulated. The first experimental results on knapsack, quadratic assignment and maximum independent set problems show that this new algorithm enhances the performance of CP Optimizer alone.
展开▼
