摘要
针对冗余惯性导航信息识别问题,提出一种基于Siamese网络的导航信息一致性判断方法。采用一段时间内的惯导数据变化趋势,模拟人脑对数据信息一致性的判断。首先建立惯组测量模型,获得不同状态的惯组测量数据,导航解算等处理后得到一段时间内7类21维的数据信息构造样本对。然后通过设计Siamese网络结构,采用卷积神经网络提取样本对特征向量进行网络训练。最后进行实验仿真验证。结果表明,本文采用552对小样本数据进行训练的结果,在与训练样本存在差异的497对测试样本上达到了98.19%的准确度,说明网络具有一定的泛化能力。神经网络训练结果稳定,在测试集上的准确度高。为冗余惯性导航系统信息一致性判断提供了新思路。
Aiming at the problem of judging the consistency of redundant inertial navigation information,a method for judging the consistency of multi-sensor data based on Siamese network is proposed. The neural network is used to simulate the judgment of the human brain on the consistency of data information,and the simulation verification is implemented. Firstly,an inertial measurement model is established. After processing the inertial measurement data,7 types of 21-dimensional data information are obtained to establish a sample pair. Then,by designing the Siamese network architecture,the convolutional neural network is used to extract the feature vector of the sample pair,and the network training is performed. Finally,the experimental simulation verification is implemented. The results show that this 552 pairs of data are used to train the network and 98. 19% accuracy is achieved on 497 pairs of test samples that are different from the training samples,which indicate that the network has a certain generalization ability. The neural network training results are stable and have high accuracy on the test set,which confirms that a new idea is supplied for judging the consistency of redundant inertial navigation information.
作者
赵欣艺
高晓颖
踪华
李冰
李宇明
Zhao Xinyi;Gao Xiaoying;Zong Hua;Li Bing;Li Yuming(Beijing Aerospace Automatic Control Institute,Beijing 100854,China;National Key Laboratory of Science and Technology on Aerospace Intelligence Control,Beijing 100854,China)
出处
《航天控制》
CSCD
北大核心
2021年第3期27-32,42,共7页
Aerospace Control
