Abstract: In order to improve GPS dynamic position precision, the model of the GPS dynamic position error is presented by using time series analysis. The stationarity of error series is verified through inverse serial verification firstly. The error series can be described as AR(10) based on the fact that the auto correlation function is tail-dragged and that the partial correlation is tail-truncated. The parameters are calculated using moment estimation. Adaptive verification proves that the dynamic position error can be presented by AR(10). Applying the model to filter the output data of the GPS receiver with 60km／h, experimental results show the mean square error can be deduced about 27％.

Key words: GPS position, error model, AR model, mean square error