基于遗传算法优化BP神经网络的GNSS干扰源定位技术

GNSS Interference Source Localization Technology Based on Genetic Algorithm Optimized BP Neural Network

全球导航卫星系统(GNSS)应用已全面深入到国家安全和国民经济当中,但由于GNSS信号到达地面后信号强度很弱,极易受到无意或有意的人为干扰。当出现压制干扰时会影响接收机正常工作,从而导致某一区域导航定位效果受到影响,因此对干扰源的排查和消除十分重要。针对上述压制干扰,通过在监测区域分布一定数量低成本接收机,利用其接收的载噪比数据特征实现干扰源的位置估计。考虑到信号传播过程中的衰减模型是非线性的,提出了基于遗传算法(Genetic Algorithm,GA)优化反向传播(Back Propagation,BP)神经网络的干扰源定位方法,通过神经网络学习得到监测区域载噪比特征的复杂非线性关系,GA对神经网络的初始权值和阈值进行优化,最终在监测区域通过梯度下降法搜索出干扰源位置。结果表明,GA优化后的网络预测误差更小,能够初步定位干扰源位置且平均定位误差率(Average Localization Error Rate,ALER)约为0.23%,验证了模型的合理性和有效性。

The application of GNSS has been fully penetrated into the national security and national economy,but due to the weak signal strength of GNSS signal after reaching the ground,it is extremely vulnerable to unintentional or intentional human interference.When there is suppression interference and it will affect the normal operation of the receiver,resulting in a certain area of navigation and positioning being affected,so the identification and elimination of interference sources is very important.For the suppressed interference in the above,the location estimation of the interference source is achieved by distributing a certain number of low-cost receivers in the monitoring area and using the characteristics of their received carrier-to-noise data.Considering that the attenuation model during signal propagation is nonlinear,an interference source location method based on Genetic Algorithm(GA)optimized Back Propagation(BP)neural network is proposed.The complex nonlinear relationship of the carrier-to-noise ratio characteristics in the monitoring area is obtained through neural network learning and the genetic algorithm optimizes the initial weights and thresholds of the neural network,finally the interference source location is searched in the monitoring area by the gradient descent method.The results show that the prediction error of the optimized network by genetic algorithm is smaller,and it can initially locate the interference source location with an Average Localization Error Rate(ALER)of about 0.23%,which verifies the reasonableness and effectiveness of the model.