基于径向基函数神经网络辅助容积卡尔曼滤波的多自主水下航行器协同定位方法

在多自主水下航行器(AUV)协同定位系统中，针对协同定位性能受到系统内部和外部等多种因素制约的问题，提出一种基于径向基函数(RBF)神经网络辅助容积卡尔曼滤波(CKF)的多AUV协同定位方法。当基准参考位置可用时，通过非线性CKF得到滤波新息、预测误差和滤波增益作为RBF神经网络输入层的输入，滤波误差值作为输出对RBF神经网络进行训练；当基准信号中断时，利用训练好的RBF神经网络，对CKF的滤波状态估计值进行补偿，进而得到新的估计状态。利用湖试数据，模拟多AUV协同定位系统输入存在误差情况下的协同定位实验。实验结果表明，所提方法与无RBF辅助的CKF方法相比，平均定位误差减小70%，具有更好的准确性和稳定性。

A Cooperative Localization Method for AUVs Based on RBF Neural Network-assisted CKF

For cooperative localization of autonomous underwater vehicles (AUVs), a multi-AUV cooperative localization method based on radial basis function (RBF) neural network-assisted cubature Kalman filter (CKF) is proposed to solve the problem that the cooperative localization performance is restricted by various factors, such as internal and external factors of cooperative localization system. When a basic reference position is available, the filtering innovation, prediction error and filtering gain, which are extracted from the nonlinear filtering CKF, are used as the inputs of the input layer of RBF neural network, and the filtering error value is used as an output to train the RBF neural network. When the reference signal is interrupted, the trained RBF neural network is used to compensate the estimated value of CKF filter state, and then a new estimated state is obtained. The cooperative localization experiment with the input error of multi-AUV cooperative localization system was simulated based on the lake area test data. The experimental results show that the average positioning error of the proposed method is reduced by 70% compared to average positioning error of RBF without CKF.