一种引入奖励与惩罚机制的蚁群算法

Ant Colony Algorithm with Strategy of Award and Penalty

蚁群算法是一种新型的仿生类算法,大量实验表明该算法具有较强的搜索最优解的能力,但同时与其它进化算法一样存在搜索速度慢,易于陷于局部最优的缺陷。为了克服蚁群算法在这方面的不足,该文通过引入奖励与惩罚机制,在蚂蚁搜索最优解的过程中,根据每次循环后的搜索结果,对蚁群算法中信息素更新的方法进行自适应调整,以达到从可行解中寻求尽可能好的解(满意解)的目的。通过与ACS算法的对比实验表明本算法在搜索速度和性能方面都有更好的效果。

Ant colony algorithm is a novel simulation algorithm. Lots of experiments have shown that the algorithm has great ability of searching better solution, but at the same time it is slow in searching speed and prone to fall into local optima as other evolutionary algorithms. In order to overcome the shortcoming of ant colony algorithm, a strategy with award and penalty for updating pheromone according to the searching result after every circle is presented in this paper to get a better solution in the feasible solution space. Simulation experiments show that the improved algorithm has a better solution than ACS.