径向基神经网络解决威胁排序问题  

RBF Neural Network for Threat Sequencing

作　　者：王向华[1] 覃征[1] 刘宇[1] 史哲文[1] 

机构地区：[1]西安交通大学计算机系,西安710049

出　　处：《系统仿真学报》2004年第7期1576-1579,共4页Journal of System Simulation

摘　　要：在空战威胁估计中,把一架战机的态势量化为一个向量,每一个决策因素为一个分量。确定各因素之间的相对重要程度,是合成综合威胁指数,完成威胁排序的关键。本文使用径向基神经网络确定各个因素之间的非线性复杂关系。使用层次分析法生成初始的(战机态势量化向量、综合威胁指数)训练样本对。然后使用样本校正器对不合理的样本对自动进行校正。调整后的综合威胁指数,作为最终的学习样本的目标值,以供径向基神经网络训练使用。实验表明,径向基神经网络可以很好的逼近各个因素之间的权重关系。The air combat situation of fighter can be formulated as a vector, an element of which is a decision factor. Properly determining the weights of all factors is crucial for threat assessment. In this paper, radial basis function neural network was employed to approximate the nonlinear complex relationship of all factors. Analytic hierarchy process was used to form initial training patterns. Then an adjustment was given to adjust unreasonable training patterns among them. The resulting training patterns were supplied to RBF neural network. The experimental results show that RBF neural network can successfully approximate the weights of all factors.

关 键 词：径向基神经网络 层次分析法 威胁估计 威胁排序 空战态势 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]