A study on the self-difference GPS positioning by dynamic and fictitious datum station

Abstract

GPS positioning has a lot of error elements. In order to remove them, this paper advances a new method of GPS positioning-self-difference GPS positioning by a dynamic and fictitious datum station. The distance that a vehicle runs can be divided into many small regions. Every region sets up a fictitious datum station. The foundation of the fictitious datum station demands three values: forecasting value; real-time value; and value of the last datum station). They are in all sent to a neural network that consists of three layer neurons. The output of the neural network is the coordinates of the fictitious datum station. The training of the network uses a BP algorithm. The decision function chooses a nonlinear Sigmoid function. The experiment has proved that the method can significantly improve positioning precision, and that the system also has rapid tracking ability.