Abstract: Multi-niche crowding genetic algorithm based on fitness sharing is proposed in this paper，for conquering the limitations of the sharing method and the crowding method，singly used in solving multimodal functions.The algorithm combines the idea of crowding and sharing.During the selection step，it uses the crowding selection policy；during the replacing step，it uses a replacement policy called worst among most similar，after fitness sharing.Theoretical analysis and numerical experiments indicate that the algorithm keeps a good diversity throughout the search，having a good searching ability on various multiple hump functions.

Key words: genetic algorithm, multiple hump function optimization, multi-niche, fitness sharing, crowding

摘要： 针对优化多模函数时单纯使用共享和排挤机制的遗传算法所存在的缺陷，提出了基于适应值共享的多生境排挤遗传算法。基本思想是：按照共享的思想在对个体的适应值进行调整的同时，将排挤选择和相似个体中适应度最差个体被替换的策略分别应用于选择算子和群体的进化中。理论分析和数值实验表明，该算法很好地维持了种群多样性，对于各类多峰函数具有较强的搜索能力。

关键词: 遗传算法, 多峰函数优化, 多生境, 适应值共享, 排挤