Location management cost estimation for PCS using neural network

Abstract

The fluid-flow model is suitable for vehicle traffic on highways but not for pedestrian movements at variable velocities. Considering the change of a mobile's velocity within a short time is limited due to physical restrictions, therefore a mobile user's future velocity is likely to be correlated with its past and current velocity. Since Gauss-Markov model captures the essence of the correlation of a mobile's velocity in time, we propose a Gauss-Markov process based fluid model that it is suitable for both vehicle traffic on highways and pedestrian in street, and presents the total cost estimation of location management for the Gauss-Markov process based fluid model.Considering the importance of memory level in the Gauss-Markov model, we estimate the parameter using neural network. Considering the measurements of updating and paging cost is not consecutive but contain missing observations, we propose the methods of location management total cost estimation with missing measurement for PCS, and utilize the Kalman filter to deduce the steady covariance of the estimation error of total cost per unit time. A numerical example is given to illustrate the use of the proposed approach. © Springer-Verlag Berlin Heidelberg 2007.