基于OFRBF-Elman网络的UUV动力学模型辨识  

Identification of Dynamic Model of UUV Based on OFRBF-Elman Neural Network

作　　者：边信黔[1] 牟春晖[1] 张勋[1] 严浙平[1] 

机构地区：[1]哈尔滨工程大学自动化学院,黑龙江哈尔滨150001

出　　处：《计算机测量与控制》2011年第9期2248-2251,共4页Computer Measurement &Control

基　　金：黑龙江省博士后基金(LBH-Z09242)

摘　　要：水下无人航行器(UUV)是具有较强非线性的复杂动态系统,而神经网络具有理论上逼近任意非线性的能力;为了提高UUV的动力学模型精度,运用了基于输出反馈的RBF-Elman(OFRBF-Elman)神经网络的系统辨识方法,即对Elman神经网络进行改进,将网络输出进行延时反馈,作为输入与隐层进行联接;将径向基函数作为隐层节点的激活函数,并以线性最小二乘法调整隐层到输出层的连接权值;然后,将该方法应用于UUV空间六自由度的动力学模型辨识中;最后,通过仿真证明了该网络结构的辨识算法具有很好的逼近能力和快速的训练速度。Unmanned underwater vehicle was a highly complex nonlinear dynamic system, and neural network had the ability to arbitrary approximate nonlinear system in theoretically. In order to improve the accuracy of dynamic model of UUV, used the method of System iden- tification based on output feedback RBF--Elman neural network. That improved the E^man neural network, made the output of the network delay feedback, as the input association with the hidden layer. The radial basis function as the activation function of hidden nodes, takes line- ar least squares to adjust these connection weights of the hidden layer to output layer. Then, the method was applied to identification of dy- namic model of six degrees of freedom of UUV. Finally, the simulation proved that the network structure identification algorithm has a good approximation ability and fast training speed.

关 键 词：系统辨识 水下无人航行器 输出反馈RBF—Elman网络 动力学模型 非线性系统 

分 类 号：TP242[自动化与计算机技术—检测技术与自动化装置]