一种解决多目标优化问题改进的进化规划算法

目前,大多数多目标进化算法采用非优超排序的方法逼近Pareto前沿,此方法存在的一个致命弱点是需要花费大量的时间检验非劣解,效率很低。论文提出了一种新的多目标进化规划算法,将初始群体划分为可替换部分与不可替换部分,并用外部文件存储进化过程中得到的非劣解,大大减少了检验非劣解所需的工作,加快了算法的收敛速度。仿真试验表明,与传统的基于非优超排序的多目标进化规划算法相比,该算法在效率上有很大的改善,并能更好地逼近Pareto前沿。

An Improved Evolutionary Programming for Multiobjective Optimization

Evolutionary algorithm has the ability to find multiple Pareto-optimal solutions in one single run and it is adaptive to deal with the mulitiple objectives.The usual evolutionary multiobjective optimization uses a ranking procedure to classify a population of individuals based on their Pareto dominance,but the ranking procedure normaly consumes most of the running time.In this paper,we propose an impoved evolutionary programming for the multiobjective optimization,using an exteral file to store the nondominated vectors found during the evolutionary process.The population memory is divided into two parts:a replaceable and non-replaceable portion.In this way we can reduce the checkings for non-dominace.The validity of the algorithm is shown by computer simulation results.