基于遗传算法的空战编队优化研究

【作者】 张科施；

【导师】 王正平；

【作者基本信息】 西北工业大学 ， 飞行器设计， 2003， 硕士

【摘要】 本论文主要进行两方面的理论研究：一方面是发展一种用遗传算法优化大规模空战编队战术的方法；另一方面是对遗传模拟退火算法的寻优性能进行研究。 首先，本论文发展了一种用遗传算法优化大规模空战编队战术的方法，并用模拟退火算法对其进行了改进，使得收敛性能大大提高。该方法是借鉴层级编队思想，对多机编队实行层级编码，在一定的空战模型的基础上，用遗传算法优化交战结果，最终得到一个对给定编队作战的最优编队。关键点在于：①超视距作战用战绩来反映交战结果和编队优势；②用遗传算法来进行基于战绩的优化。该方法可为分析现有空战编队的优势和挖掘新的空战编队提供参考。最后，用两个16机编队作战的算例验证了该方法的有效性。 其次，本论文对遗传模拟退火算法的寻优性能进行了研究。遗传算法是一种模拟生物在自然环境中的遗传和进化过程而形成的一种自适应全局优化概率搜索算法。该方法与常规优化算法相比最大的优点是能处理离散变量的优化问题。本文从理论上阐述了遗传算法、模拟退火算法和遗传模拟退火算法的生物基础和构成要素，分析了它们的寻优速度、收敛能力等特性。最后，用计算试验对比了遗传算法和遗传模拟退火算法的收敛性能，对比了采用不同状态函数和不同Markov链的遗传模拟退火算法的收敛性能，并对结果进行了分析。更多还原

【Abstract】 On one hand, the optimization method for large-scale air combat formation tactics based on genetic algorithm is developed in this paper. A hierarchical formation tactics is employed in this method and the hierarchical code is introduced to encode formations. Then genetic algorithm improved by simulated annealing algorithm is used to optimize the engagement results of BVR (Beyond View Range) intercepts so that a better formation can be gained. This paper integrates two key ideas. The first is that performance reflected engagement outcome and formation advantage, and the second is using genetic algorithm for performance-based optimization of blue team formation tactics. This method may act as a reference for analyzing the advantage of formations and exploring new formations. At last two calculating examples are used to testify the effectiveness of this method.On the other hand, the property of genetic algorithm improved by simulated annealing algorithm for finding the best value is studied in this paper. Genetic algorithm is an adaptive global optimal algorithm that simulates the inheritance and evolution of the biology. The whole course of finding the best value depends on probability. It differs from other traditional optimal algorithms mainly in dealing with discrete points. Genetic algorithm, simulated algorithm and genetic algorithm improved by simulated annealing algorithm are described and their abilities of finding the best value are analyzed in this paper. At last, several examples of genetic algorithm improved by simulated annealing algorithm, which adopted different functions deciding whether to accept a state and different Markov chains, are used to analyze the property of this algorithm.更多还原