An Improved Robust Adaptive Kalman Filter for GNSS Precise Point Positioning

For precise point positioning (PPP), Kalman filter is usually used to estimate the position and velocity parameters. But, the positioning accuracy and convergence time are highly susceptible to gross errors in observations and dynamic model errors in motion states. In addition, the weights of different types of observation are generally determined by the prior variance are usually inaccurate, which affects the performance of PPP. In order to achieve a more desirable positioning solutions, an improved robust adaptive filtering method is proposed. First in this method, classification robust equivalent weight model based on t-test statistic is developed to construct the equivalent weight matrix for each types of observation separately. Second, variance component estimation weighting method based on innovations is proposed to determine the scale factor of each types of observation. Finally, adaptive factor function model based on innovations is adopted to acquire the optimal adaptive factor. The availability of the proposed classification robust equivalent weight model and variance component estimation weighting method are verified with the static test data, and the performance of the improved robust adaptive filtering method is further validated with the dynamic test data. The results indicate that compared with the conventional robust adaptive filter, the improved one has a better performance in positioning accuracy, convergence time, and the stability of PPP.