On Multi-Behavior Based Multi-Colony Ant Algorithm for TSP

To avoid premature convergence and stagnation problems in classical ant colony system, a novel multi-behavior based multi-colony ant algorithm (MBMCAA) is proposed. The ant colony is divided into several sub-colonies; the sub-colonies have their own population evolved independently and in parallel according to four different behavior options, and update their local pheromone and global pheromone level respectively according to immigrant operator. This parallel and cooperating optimization scheme by using different behavioral characteristics and inter-colonies migration strategies can help the algorithm skip from local optimum effectively. The experimental results for TSP show the validity of this algorithm.