无源雷达目标跟踪的线性回归-梯度下降算法

Linear regression-gradient descent algorithm for passive radar target tracking

针对以多颗GNSS卫星作为外辐射源的无源雷达目标跟踪问题,提出一种基于机器学习的无源雷达目标定位跟踪算法。该算法通过线性回归和梯度下降算法训练得到时延和多普勒频偏的模型,根据模型对低信噪比下互模糊函数估计出的多普勒频偏进行修正,之后构建了基于4颗卫星的时延和修正后的多普勒频偏的定位算法模型实现对目标的定位跟踪。仿真结果表明,该算法与直接利用互模糊函数估计的时延和多普勒定位跟踪目标的原始算法相比,位置估计精度提高了11.1倍,速度估计精度提高了3.6倍,能够更加有效快速的定位跟踪空中目标,同时,与粒子滤波的算法相比,该算法不需要长时间的累计,可以在更短的时间内对目标进行定位跟踪,效率更高。

Aiming at the passive radar target tracking problem with multiple GNSS satellites as external radiation sources,a passive radar target tracking algorithm based on machine learning is proposed.The algorithm obtains the model of delay and Doppler by linear regression and gradient descent algorithm,and the value of Doppler shift which estimated by the cross-ambiguity function in the case of low SNR is corrected according to the trained model,and then a location algorithm model based on the value of four satellites’delay and Doppler is constructed to achieve the tracking of target.The simulation results show that the proposed algorithm can improve the position estimation accuracy by 11.1 times and the speed estimation accuracy by 3.6 times,which is more efficient and faster than the original algorithm that directly uses the value of delay and Doppler which estimated by cross-ambiguity function.At the same time,compared with the particle filter algorithm,this algorithm does not need long time accumulation,and can locate the target in a shorter time with higher efficiency.