2010 GPS Solutions: A windowing-recursive approach for GPS real-time kinematic...
Zebo Zhou, Yunzhong Shen, Bofeng Li
All conventional Kalman filtering methods depend to a great extent on dynamic models for describing the motion state of vehicle. However, low-cost GPS navigation systems do not provide velocity and acceleration measurements to construct dynamic models. Therefore, it is rather difficult to establish reasonable dynamic models. A windowing-recursive approach (WRA) which employs previous positions to predict the current position is proposed, and the transition matrix is modeled for transforming the previous positions to the current one. Two typical transition matrices are constructed by numerical polynomial fitting and extrapolation. A real vehicular GPS experiment is carried out to demonstrate the WRA performances in two relative positioning scenarios. The data are processed by the least squares approach and by WRA using the two developed transition matrices. The results show that the WRA performed excellently in a high sampling rate data. In case of a lower sampling rate, higherorder polynomial fitting and extrapolation models work better than lower-order models for a given window. In addition, the extrapolation models can alleviate the computation burdens significantly relative to the polynomial fitting models.
