Multi-Colony Ant Algorithm using both Repulsive Operator and Pheromone Crossover based on Multi-Optimum for TSP

This paper presents a modified multi-colony ant algorithm, based upon a pheromone arithmetic crossover and a repulsive operator. Iteration of this algorithm can avoid some stagnating states of basic ant colony optimization. An important mechanism of this algorithm is the reinitialization of such stagnating states (worst performing ant colonies), which is accomplished through application of the pheromone arithmetic crossover and the repulsive operator. At the same time, the main algorithm parameters alpha, beta and rho are self-adaptive. The ratio of communication time between processors to the computation time of the processors of this system (master and slaves) is relatively small. Comparing against a parallel asynchronous algorithm, we show the effectiveness of the modified multi-colony ant algorithm.