关键词: 自适应, 蚁群系统, 模糊隶属度, 信息熵, 旅行商问题

Abstract: Because the Ant Colony System（ACS） is easy to fall into the local optimization and slow convergence, an Adaptive Fuzzy Ant Colony System（AF-ACS） is proposed for Traveling Salesman Problem（TSP）. The core of the algorithm is to introduce the concept of fuzzy membership and information entropy. AF-ACS will adaptively introduce fuzzy membership to ACS with information entropy as the probability. This balances the relationship between the population diversity of the algorithm and the rate of convergence. The probability of introducing fuzzy membership degree early in the algorithm is small, which ensures the diversity of the algorithm. The probability of introducing fuzzy membership degree in the later stage of the algorithm is large, which improves the convergence speed of the algorithm. By comparing 14 different scale TSP test sets with ACS and ECACS（Entropy-based Adaptive Chaotic Ant Colony Algorithm）, AF-ACS obtains the optimal solution or better solution with fewer iterations. This proves the feasibility and efficiency of AF-ACS.

Key words: adaptive, ant colony system, fuzzy membership degree, information entropy, travelling salesman problem