模型预测前向神经网络算法及其在组合导航中的应用

发布时间：2018-02-17 05:22

本文关键词： 前向神经网络模型预测滤波权值修正 SINS/CNS/BDS组合导航　出处：《中国惯性技术学报》2014年02期 　论文类型：期刊论文

【摘要】：针对系统误差的不确定性可能会引起滤波精度降低或发散的问题,提出一种新的基于模型预测滤波的前向神经网络算法。首先,采用模型预测滤波估计网络在正向传递过程中的模型误差,并对其进行修正,以弥补模型误差对隐含层权值更新的影响;然后,利用模型预测滤波为神经网络提供精确的训练样本,学习待估计系统的非线性关系。将提出的算法应用于SINS/CNS/BDS组合导航系统,并与扩展卡尔曼滤波进行比较,仿真结果表明,提出的算法得到的姿态误差、速度误差和位置误差分别在[-0.25′,+0.25′]、[-0.05 m/s,+0.05 m/s]和[-5 m,+5 m]以内,滤波性能明显优于扩展卡尔曼滤波算法,表明该算法能提高组合导航定位的解算精度。
[Abstract]:Aiming at the problem that the uncertainty of system error may lead to the decrease or divergence of filtering accuracy, a new forward neural network algorithm based on model predictive filtering is proposed. The model predictive filter is used to estimate the model error in the forward transmission process and to correct it to compensate for the influence of the model error on the updating of the hidden layer weight. The proposed algorithm is applied to SINS/CNS/BDS integrated navigation system and compared with extended Kalman filter. The simulation results show that the proposed algorithm can provide accurate training samples for neural network and study the nonlinear relationship of the system to be estimated. The attitude error, velocity error and position error obtained by the proposed algorithm are within [-0.25, 0.25'], [-0.05 m / s, 0.05 m / s] and [-5 m, 5 m], respectively. The filtering performance of the proposed algorithm is obviously better than that of the extended Kalman filter algorithm, which shows that the algorithm can improve the accuracy of integrated navigation positioning.
【作者单位】：西北工业大学自动化学院;航天恒星科技有限公司;
【基金】：国家自然科学基金(61174193) 中国航天科技集团公司卫星应用研究院开放基金资助
【分类号】：TN967.2

【参考文献】