Using Allan Variance to Analyze the Zero-differenced Stochastic Model Characteristics of GPS

Acta Geodaetica et Cartographica Sinica

XUE Xueming,¹  ZHANG Xiaohong¹  TANG Long¹  ZHU Feng¹ 

[1] School of Geodesy and Geomatics, Wuhan University, Wuhan 430079, China;

关键词: Allan variance;    GPS observation;    Stochastic model;    noise component;    noise parameter;   

DOI  :  10.11947/j.AGCS.2015.20130513

来源: DOAJ

【 摘 要 】

The estimation criteria for solving parameters in zero-differenced GPS positioning is that observations obey Gaussian white noise distribution. But a number of pioneering studies point out that the white noise would be damaged by satellites errors, propagation errors, station environment errors and so on. Meanwhile, un-modeling errors also have adverse effects. These errors not only undermine the assumption estimation criteria, and some non-white noises are likely to be absorbed by state parameters. In result, the accuracy of estimates is influenced. This paper regards white noise, colored noise and un-modeling errors as ZD stochastic model of GPS. Then the Allan variance method is proposed to analyze the posteriori residuals which can represent the Stochastic characteristics of GPS data. Noise component and parameters are mainly investigated. The result shows GPS noise behaves as WN plus GM. The phase and pseudorange WN is 2.392 mm and 0.936 m respectively, GM process noise is 4.450 mm/√s and 0.833 m/√s respectively, correlation time is 52.074 s and 14.737 s respectively. It is found that the phase GM component is associated with satellite, but the rest is associated with station. A number of analysis indicate that the ZD stochastic model characteristics of GPS obeys non-Gaussian white noise distribution and is to be refined.

【 授权许可】

Unknown