基于贝叶斯优化的三维飞行航迹规划  

3D Flight Path Planning Based on Bayesian Optimization Algorithm

作　　者：符小卫[1] 高晓光[1] 

机构地区：[1]西北工业大学电子信息学院,陕西西安710072

出　　处：《兵工学报》2007年第11期1340-1345,共6页Acta Armamentarii

基　　金：中国博士后科学基金资助项目(20060391101);航空科学基金资助项目(05D53022)

摘　　要：提出了一种基于贝叶斯优化算法的飞行器三维航迹规划方法。把飞行航迹编码为离散时间间隔上飞行器速度向量的变化序列。采用固定时间间隔时,这种编码方法把每一步速度向量的变化量都限制在飞行器最大加速性能之内,所以这种编码方法对应的物理轨迹是可飞的。利用每代种群中的可行解集合构造贝叶斯网络,用贝叶斯网络的结构体现染色体基因位之间的联系,用贝叶斯网络参数体现染色体基因位之间的联系程度。设计了一个多变量K2度量评价网络的优劣。用贝叶斯网络产生新的染色体以体现种群的进化,这取代了传统遗传算法的交叉和变异过程。如果种群中最优个体不满足终止条件,则用新一代种群的可行解集合构造贝叶斯网络,直到满足终止条件。仿真结果验证了算法的有效性。A Bayesian optimization algorithm for three-dimensional （3D） flight path planning problem was presented. The flight path was presented by a time sequence of velocity vectors whose elements are speed, heading and climb angle. Using a fixed time interval, this representation allows the planning algorithm to constrain the candidate solution to lie within the acceleration capability of the air vehicle. The Bayesian optimization algorithm is applied to implement explicit learning by building a Bayesian network of the joint distribution of viable candidate path genotype strings. The construct and conditional probabilities of the network indicate the qualitative and quantitative relationship among the path genotypes. A multivariate K2 metric is designed to evaluate the network. A new set of path genotype strings is generated by using the corresponding conditional probabilities. If the stopping conditions are not met, the conditional probabilities for all nodes in the Bayesian network are updated again using the current set of viable path genotype strings. Experimental results demonstrate that this approach is effeetively.

关 键 词：运筹学 航迹规划 遗传算法 贝叶斯网络 贝叶斯优化算法 

分 类 号：V218[航空宇航科学与技术—航空宇航推进理论与工程]