Abstract: To overcome the problems of searching speed, running time, and earlier premature of traditional multiple ant colonies algorithm, ?an improved algorithm is proposed. This algorithm introduces more than one type of ant colonies with different pheromone updating mechanisms. These different types of ant colonies have different searching traits. They can cooperate smoothly each other. Every ant colony adaptively chooses suitable information exchangeable object from more than one potential selected ant colonies, determines the best pheromone exchanging strategy from a variety of options through similarity coefficient among every ant colonies. By this way, the balance between the diversity and convergence of every ant colony is kept desirable. A series of TSP experiments show that this algorithm can generate solutions with better quality and faster speed.

Key words: multiple ant colonies, heterogeneous colonies, similarity, ant colony algorithm

摘要： 针对原有的多种群蚁群算法收敛速度慢，运行时间长，容易早熟等缺陷，提出了一种新型异类多种群蚁群算法。算法由多类不同特性蚁群构成，不同蚁群具有不同特质，且优势互补，彼此间具有潜在的合作性。不同种类蚁群搜索时，通过子蚁群间的相似度，自适应选择最互补的蚁群进行信息交换，以加强不同种类蚁群间的协作，增强解的多样性，增强跳出局部最优的能力。TSP仿真结果表明，该算法在搜索速度以及搜索质量方面都有明显的提高。

关键词: 多种群, 异类种群, 相似度, 蚁群算法