利用RBF神经网络对GPS广播星历插值

摘要

Using the RBF(Radial Basis Function) neural network to interpolate the GPS satellite coordinate computed by broadcast ephemeris was proposed in order to solve the problem of Runge Phenomenon ( RP) in poly-nomial interpolation. The GPS satellite coordinate calculated by GPS broadcast ephemeris in 25th, July 2011 were selected to test the RBF neural network interpolation method. The interpolation results compared with the GPS sat-ellite coordinate calculated by direct computation method show that RBF neural network interpolation method can a-void the RP, improve the interpolation stability and accuracy as well meet the demands for GPS point positioning.%针对多项式插值计算任意时刻GPS卫星坐标过程中出现的Runge现象,提出了采用径向基函数(Radial Basis Function,RBF)神经网络对GPS广播星历计算所得卫星坐标进行插值.选用2011-07-25日广播星历,利用径向基函数神经网络插值计算GPS卫星坐标,并将其与直接法计算的GPS卫星坐标比较,结果表明该方法能有效避免Runge现象,提高了插值稳定性,插值精度高,可满足GPS单点定位的需求.