中文摘要：

当观测信息冗余度小或出现异常时，仅依靠GPS单历元数据难以获得可靠的定位结果，需借助历史信息改善定位精度和可靠性。导出了Bayesian递推算法的一般形式，建立了基于移动窗口的牛顿向前插值模型与常速度、二次曲线最小二乘拟合模型的先验信息获取方式，引入了当前历元相关观测方程的抗差估计，并采用实测GPS动态数据计算分析，结果表明：抗差估计能有效抑制粗差对GPS定位结果的影响；拟合模型对误差具有平滑作用，其中二次曲线模型明显优于常速度模型，但两者均需实时计算拟合系数；而牛顿向前插值模型系数恒定，计算简单，显著提高了定位精度及可靠性。

英文摘要：

It is difficult to obtain the reliable positioning restflts when observations are influenced by abnormity or low redun- dancy. Historical information should be applied to improve the positioning precision and reliability. Basing on Bayesian theorem, a general recursive algorithm was developed. According to various prior information accessing, the Newton forward interpolation model, the constant velocity and quadratic curve fitting model were constructed in a chosen time window. Robust estimation for correlated observations was also introduced in the present epoch. In the end, a real-life example is tested, and the results show that robust estimation is effective to reduce the influence of outliers ; Fitting model can smooth errors, and quadratic curve is superior to constant velocity ; Nevertheless both of them need to compute the fitting coefficients ; however, the Newton forward interpolation model holds a constant coefficient and significantly enhances the precision and reliability.