The network-based GPS technique provides a broad Spectrum of corrections for real-time kinematic. And the atmospheric refraction error is the main factors to be sufficient to support the ambiguity resolution (AR) and accuracy of the long-distance RTK.However, due to the strong spatial correlated of the tropospheric delay, the elevation difference between the reference plane and the rove station will cause the deviation of tropospheric error in the system so that the accuracy of troposphere correction will be lowered. In this paper, a new tropospheric error model based on neural network is presented. The neural network model takes into account not only the level factors, but also the elevation factor. It establishes the model in the spatial space. And the experimental results show that the accuracy of tropospheric delay model is within 5cm regardless of interpolation points in the network or network.

No abstract