设计了一种求解模糊多目标资源受限项目调度问题的遗传局域搜索(GLS)算法,目标是生成近似有效解集以便决策者在决策过程中有更多的选择.算法利用线性加权效用函数将多目标组合优化问题转换为单目标组合优化问题,通过系统的方法生成目标权系数向量,对于每次生成的权系数向量,调用GLS算法求解以极小化效用函数为单一目标的子问题,由此生成的近似有效解集更加具有多样性.实验结果表明:本文算法可以针对多目标资源受限项目调度问题生成较好质量的近似有效解集,在多数指标上优于其它两种对照算法.
A genetic local search (GLS) algorithm is designed to solve fuzzy multi-objective resource-constrained project scheduling problems. The goal of the algorithm is to generate a set of ap- proximate efficient solutions so that decision makers (DM) can choose a good compromise solution for the problem. The algorithm specifies weight values in a systematical way, and with each specified weight vector, the algorithm uses the GLS procedure to optimize the sub problem which takes the weighted linear utility function as its single objective, so that the resultant set of approximate efficient solutions has more devisity. Computational results show that the presented algorithm can effectively generate sets of higher quality approximate efficient solutions for this kind of problems, and outperforms the two other algorithms on most of criteria.