基于RMSprop的高精度GNSS-R镜点快速预测算法  

High-precision GNSS-R Specular Point Prediction Algorithm Based on RMSprop

作　　者：盛志超 周勃 秦瑾 顾斯祺 孟婉婷 SHENG Zhichao;ZHOU Bo;QIN Jin;GU Siqi;MENG Wanting(Shanghai Aerospace Electronic Technology Institute,Shanghai 201109,China)

机构地区：[1]上海航天电子技术研究所,上海201109

出　　处：《无线电工程》2021年第10期1057-1063,共7页Radio Engineering

基　　金：上海市科委高新技术领域基金(19511101200);上海市经信委基金资助项目(200130)。

摘　　要：从全球导航卫星系统反射信号(Global Navigation Satellite System Reflectometry,GNSS-R)测高应用中对镜面反射点的高精度需求出发,利用GNSS-R空间几何位置关系和机器学习中的RMSprop(Root Mean Square Prop)算法对镜面反射点问题进行研究,提出了一种基于RMSprop的高精度镜面反射点预测算法,该算法能够实现对学习率的自适应调整,加速梯度下降,提高算法的收敛性能。仿真结果表明,该算法收敛快,精度高。Based on the requirements for high-precision specular points in Global Navigation Satellite System Reflectometry(GNSS-R)altimetry applications,the issue of specular reflection points is studied by using the GNSS-R spatial geometric position relations and RMSprop algorithm.A high-precision specular point prediction algorithm based on RMSprop is proposed,which can adjust the learning rate adaptively,accelerate the gradient descent,and improve the convergence performance of the algorithm.The simulation results show that the algorithm converges quickly and has high precision.

关 键 词：全球导航卫星反射信号 镜面反射点 RMSprop 梯度下降 

分 类 号：TP721.1[自动化与计算机技术—检测技术与自动化装置]