提出一种新的方法,把分布式Kalman滤波(DKF)方法与后向传播神经网络(BPNN)技术相结合,用于静电陀螺漂移的模型辨识。首先,为了消除测量噪声影响,将同一个静电陀螺带有噪声的多次测量数据集映射到一个虚拟的传感器网络中,然后采用具有嵌入式紧致滤波功能的DKF对映射数据进行滤波预处理。在此基础上,将预处理结果转换为用于训练神经网络的输入数据和输出数据,然后采用BPNN辨识静电陀螺漂移。实验表明,该方法可有效用于陀螺漂移的模型辨识。 A new method is proposed to combine the distributed Kalman filter (DKF) with backpropagation neural network (BPNN) to model the electrostatic gyro drift. First of all, in order to eliminate the influence of measurement noise, multiple measurement datasets with the same electrostatic gyro are mapped to a virtual sensor network, and then the DKF with embedded compact filter is used to filter the preprocessing data. On this basis, the preprocessing results are converted into the input data and the output data for training the neural network, and then BPNN is used to identify the electrostatic gyro drift. Experiments show that this method can be effectively used for model identification of gyro drift.