Efficient Search for Trade-Offs by Adaptive Range Multi-Objective Genetic Algorithms
Efficient Search for Trade-Offs by Adaptive Range Multi-Objective Genetic Algorithms
复制标题
DOI：
10.2514/1.12909
复制
发表时间：
2005
期刊：
影响因子：
--
通讯作者：
D. Sasaki;S. Obayashi
中科院分区：
文献类型：
--
作者：
D. Sasaki;S. Obayashi
权衡是工程设计问题的重要元素之一，该问题具有多个相互冲突的目标，需要同时改进。此外，在许多问题中，如空气动力学设计，由于计算的原因，只有有限数量的评估可以被允许用于工业用途。本文提出了一种新的高效多目标进化算法——自适应范围多目标遗传算法（ARMOGAs），利用少量的函数评估来识别目标之间的权衡。通过四个不同的多目标分析测试问题来检验ARMOGAs的搜索性能。ARMOGAs还与另一种MOEA进行了比较。虽然评价次数有限，但ARMOGAs表现出良好的性能。此外，还应用了顺序二次规划和动态爬坡方法对同一问题进行了权衡。这些基于梯度的方法在确定取舍方面存在一些困难。
Trade-offs is one of important elements for engineering design problems characterized by multiple conflicting objectives that needs to be simultaneously improved. Further, in many problems such as aerodynamic design, due to computational reasons, only a limited number of evaluations can be allowed for industrial use. This paper proposes new efficient Multi-Objective Evolutionary Algorithms (MOEAs), Adaptive Range Multi Objective Genetic Algorithms (ARMOGAs), to identify trade-offs among objectives using a small number of function evaluations. The search performance of ARMOGAs is examined by using four different multi-objective analytical test problems. ARMOGAs are also compared with another MOEA. Although the number of evaluations is limited, ARMOGAs showed good performance. In addition, Sequential Quadratic Programming and Dynamic Hill Climber methods are applied to obtain trade-offs for the same problems. These gradient-based methods had some difficulties in identifying trade-offs.