Models for GPS Position Estimation

This paper presents the results obtained in our research about application of advanced signal processing to GPS based position estimation. The motivation of this research is to address the GPS positioning problems in vehicle navigation under bad circumstances where the visible satellites are variable frequently or less than four. A new model for position and velocity estimation are developed for nonlinear filtering. The model is nonlinear and has variable measurement number for coping with an arbitrary num- ber of satellites. The model is investigated applying it to unscented Kalman filter for position estimation. We propose an algorithm with expectation maximization (EM) and unscented smoothing to perform the GPS model parameter learning. The first experimental results that comprise the comparison of estimation results obtained with different algorithms are then presented. Comparison shows that our approaches provide better estimation than other solutions.