Abstract: This paper proposes an improved double elite coevolutionary genetic algorithm. In this algorithm, the population is divided into two elite teams, both of them evolve cooperatively; elite is the best individual of the team, every elite of the two teams has a high differential degree. The elite crossed with the selected individuals of the team respectively, this method enhances the affinity between the population and the global optimal solution;at the same time, when the individual’s difference degree in the same team decreases to the specified threshold, the variation mechanism is introduced to avoid the problem of premature convergence and maintain the diversity of the population effectively. An individual diversity measuring method of δ-phenotype is given, which can compute the diversity accurately for the fitness value belong to the real number. The search ability is improved significantly according to the complex computing environment which has many parameters and large search scale.

Key words: genetic algorithm, elitist strategy, coevolution, population, diversity measure

摘要： 提出一种改进的双精英协同进化遗传算法。在该算法中，种群被划分为两个精英小队，二者协同进化；精英是小队中的最优个体，并且两个小队的精英具有较高的差异度。精英分别与被选的个体进行交叉，增强了种群个体和全局最优解的亲和度；同时，当精英小队中的个体间的差异度下降到规定的预警值时，引入变异操作，有效地保持了种群的多样性，避免了早熟问题。算法中还给出一种δ-表现型多样性测度计算方法，使之可以对个体适应值为实数的群体多样性进行准确计算。针对参数多、大范围的复杂计算环境，算法的搜索能力明显提高。

关键词: 遗传算法, 精英策略, 协同进化, 种群, 多样性测度