Development, Evaluation and Application of Multi-objective Optimization Methods by Genetic Algorithms

1.提出了一种多编码遗传算法(MCGA)，它是一种多目标遗传算法，每个个体专注于多个目标中的一个，并根据每个目标有不同的译码。通过计算机实验将其与传统的并行选择方法进行了比较，证明了两者在搜索性能上的互补性，从而提出了一种两者相结合的方法。将该方法应用于2目标背包问题和2目标流水作业调度问题，并与顺序并行选择方法进行了性能比较。(1)在背包问题的应用中，系统地生成不同相关系数的数据集，以控制Pareto集的分布。(2)在流水作业调度的应用中，针对每个目标提出了启发式译码方法。通过相对精确度、覆盖率、多样性和获得解的个数对所获得的解进行定量评价。仿真结果表明，该方法能有效地获得不同的解，特别是对于具有较大Pareto集的问题，无论是由问题的统计特性、问题规模的大小还是解空间的高维引起的。

1. Multiple Coding Genetic Algorithm (MCGA) is proposed as a multi-objective GA method, in which each individual specializes on one of many objectives and has a different decoding according each objective. The comparison with a conventional parallel selection method by computer experiment shows the complementary performance in searching, therefore a hybrid method of the both is also proposed.2. The proposed method is applied 2-objective knapsack problems and 2-objective flowshop scheduling problem, and its performance is compared with the ordinal parallel selection.(1) In the application to knapsack problems, the data sets with different correlation coefficients are generated in a systematic way to control the distribution of a Pareto set.(2) In the application to flowshop scheduling, heuristic decoding methods are proposed for each objective.3. Quantitative evaluations of the obtained solutions are given by relative accuracy, cover ratio, diversity and the number of acquired solutions.4. Simulation result shows that the proposed method is effective to obtain diverse solutions especially in the problem with a large Pareto set, which results either from a statistical character of a given problem, from a largeness of a problem scale, or from a high dimensionality of a solution space.