2016 Volume 72 Issue 5 Pages I_673-I_681
In this study, we apply the adaptive parameter estimation method to the positioning calculation of pedestrians on network by navigation satellite systems. Like the formulation of DDR which was proposed by Bierlaire and Frejinger, we formulate the positioning accuracy as the stochastic model and introduce into the observation model of general state space model. Then we estimate the positioning accuracy using sequential Bayesian filtering. The advantages of this estimation are follows. Firstly, we don't have to assume the positioning accuracy in advance of the estimation; it is enough to employ a uniform distribution as a prior. Secondly, we can estimate simultaneously the positioning accuracy and the velocity and route choice of the pedestrian. This feature is valuable especially when we deal with the pedestrian behaviour which includes the rambling or sudden change of speed and direction. We applied the proposed method to simulation data and confirmed the estimated results, and showed the improvement in estimation accuracy compared to the conventional method. At last we apply the method to the real data.