一种改进的多目标合作型协同进化遗传算法

Improved cooperative coevolutionary genetic algorithm for multi-objective

针对传统多目标算法早熟收敛及多样性不足的问题,提出了一种改进的非支配排序合作型协同进化遗传算法(Improved Non-dominated Sorting Cooperative Coevolutionary Genetic Algorithm,INSCCGA)。该算法利用外部档案存储每一代进化过程中产生的精英个体,并对其不断进行更新,以加快算法的收敛速度。同时提出了一种新型子种群之间协同进化的方式,增强候选解的多样性。利用ZDT系列标准测试函数,与经典的多目标进化算法NSGA-II以及多目标协同进化算法NSCCGA进行了对比,结果表明改进算法具有更好的收敛性以及均匀的解分布。

Aiming at the problem of premature convergence and insufficient diversity in traditional multi-objective optimization algorithm, it proposes an improved non-dominated sorting cooperative coevolutionary genetic algorithm. The algorithm uses an external archive storage elite individuals which generate each evolutionary process, and the elitism individuals are updated constantly in the external archive, thus speeding up the convergence rate. Meanwhile, this algorithm improves the diversity of candidate solutions by proposing a new kind of co-evolution between sub-populations. Compared with well-known multi-objective evolutionary algorithm NSGA-II and multi-objective coevolutionary algorithm NSCCGA on a suite of standard ZDT test function, the proposed algorithm has the better convergence and better uniform distribution.