非线性飞行混沌运动的修正RBF神经网络控制  

Modified RBF Neural Networks Control for the Nonlinear Flight of Chaotic Motion

作　　者：黄烈德[1] 

机构地区：[1]同济大学,上海200092

出　　处：《航天控制》2001年第4期35-40,共6页Aerospace Control

摘　　要：如何求神经网络控制使非线性空间飞行混沌运动不会发生。该法是在输出层用回归权代替常数权再用EM算法来估计回归权的参数 ,这样修正的RBF的神经网络控制就可使非线性空间飞行不会出现混沌现象。这种算法R Langari,L Wang&J Yen (1997) [1] 在研究径向基函数网络时提出过。其突出的优点是把复杂的多参数的最优化问题分离为N个小型最优化问题 ,这里N是隐藏层单元数。In this paper we present how to find the neural control, such that the chaotic motion of nonlinear space flight will not develop. This approach is a modified radial basis function (RBF) network in which the regression weights are used to replace the constant weights in the output layer. Apply expectation-maximization (EM) algorithm to estimate the parameters of the regression weights. This method have been proposed by R.Langari, L. Wang and J.Yen. A salient feature of this algorithm is that it decomposes a complicated multiparameter optimization problem into N separate small-scale optimization problems, where N is the number of hidden units.

关 键 词：非线性航天飞行 修正RBF神经网络 混沌运动 回归权 网络控制 

分 类 号：V448.22[航空宇航科学与技术—飞行器设计]