模型驱动的动态多目标演化算法和它在求解耗时的星载天线优化设计中的应用

Antenna design, which is usually a kind of nonlinear multiobjective optimization problem with constraints, is normally an expensive problem as well due to the time-consuming computation of electromagnetic. This project will construct a theoretical system in that a dynamic multiobjective optimization problem converges the expensive problem of the antenna design. A predict model will be designed for predicting performance of a solution in dynamic multiobjective evolutionary algorithm. Then the algorithm will largely reduce calculations. At the same time, the model based dynamic multiobjective evolutionary algorithm will be implemented in parallel for reducing period in solving the problem. By way of solving the dynamic multiobjective optimization problem, the model based algorithm will solve the expensive problem of antenna design. The performance of the model based algorithm will be tested and improved in designing some antennae of micro satellites, such as CubeSat antenna, NASA ST5 antenna, and so on, considering minimization of satellite. By following the program of this proposal, great achievement would be obtained in reduction of antenna design period, automation of satellite antenna design and invention of new antenna, while theory for solving non-linear dynamic multiobjective expensive problem would be enriched. The research achievement could also be used in the regions of scientific programming, military decision, engineering design, and so on, where there exist a lot of non-linear multiobjective expensive problems.

天线设计问题是一类非线性、多目标、多约束的优化问题；同时因其耗时的电磁仿真计算，它还是昂贵优化问题。本课题从此类天线设计难题出发，建立并论证动态多目标优化问题逼近天线设计昂贵问题理论体系；设计适合于该类问题的预测模型，研究基于预测模型的动态多目标演化算法的计算形式，使算法最大限度地避免耗时计算；并行实现基于模型预测的动态多目标演化算法，缩短解题周期；通过基于模型的动态多目标演化算法求解动态多目标优化问题的途径，达到求解天线设计昂贵优化问题的目的；针对卫星小型化趋势和本演化天线技术特点,重点设计微小卫星天线，测试并改进昂贵演化天线技术。沿着这个技术路线展开研究，将能在缩短天线设计周期、星载天线自动设计及设计新型天线上获得突破性进展，同时丰富非线性动态多目标昂贵优化理论成果。研究成果还可以应用到其它大型工程和科学规划、军事对抗与指挥决策、工业和工程设计等众多普遍存在昂贵优化领域。

研究背景：.天线设计问题是麦克斯韦非线性反问题，也是非线性、多目标、多约束的优化问题，同时因其耗时的电磁仿真计算，它还是昂贵优化问题， 它是优化领域和计算智能领域里的热点主题。.主要研究内容：.建立了约束优化问题转化为等价的动态多目标优化问题的一般框架，设计和实现了高斯代理模型协助的动态多目标演化算法，利用多目标优化丰富的理论和方法求解昂贵约束优化问题，从而求解昂贵天线优化问题；开发了演化天线并行优化软件，基于演化天线软件设计和优化了一些天线并研制了一些样机。..重要结果：.首次提出约束优化问题转化为等价的动态多目标优化问题的一般框架，有效地利用了多目标演化算法求解约束优化问题，优化效果总体上优于同类算法效果；设计和实现了高斯代理模型协助的动态多目标演化算法，现实有效地求解了昂贵天线设计问题，性能优于常规演化算法，开发了演化天线并行优化软件，提高了计算效率。..关键数据：. 约束优化问题转化为等价的动态多目标优化问题的一般框架下的动态多目标演化算法性能在与同类算法比较的Friedman排序第一，研究成果将在《IEEE Transactions on Cybernetics》上发表； 高斯代理模型协助的动态多目标演化算法在一周内能完成一个昂贵（每次电磁评估大约半小时）天线优化。..科学意义或应用前景：.在理论上首次创立了约束优化问题转化为等价的动态多目标优化问题的一般框架，设计和实现了动态多目标演化算法求解常规约束优化问题，设计和实现了高斯代理模型协助的动态多目标演化算法求解昂贵优化问题；在应用上这些成果不仅可以用来求解（昂贵）天线优化问题，还可以用来求解任何领域里的（昂贵）约束优化问题；演化天线软件可以商用化，成为天线设计者天线智能设计工具。