An Analytical Random Direction-Based Method in User Mobility Modeling for Wireless Networks

Abstract

In mobile communications, mobility modeling is involved in several aspects such as signaling and traffic load analysis. In particular, the accuracy of mobility models become essential for the evaluation of system design alternatives and network implementation cost issues. In this paper, we propose a mobility model which is appropriate for the practical analysis of the full range of mobile communications design issues. The model provides different levels of details for the user mobility behavior and can be adapted to any environment topology. This random direction based model can be applied in analytical purpose. This model is described by motion direction and user velocity. The motion direction of users is related to real street patterns and is modeled by discrete time and discrete state Gauss Markov. The velocity of users is modeled as Gaussian with variable mean and variance with respect to traffic factor. Simulation results confirm the preferences of the proposed method in both simulation and analytical estimations.

Mobility Tracking Based on Autoregressive Models

“…The MMSE estimators for Q 2 , denoted byQ 2 , can be obtained from (13) when residual error is defined as e e i ¼ 4 s s 2;i À A 2 s s 2;iÀ1 . Similarly, the recursive estimation for noise covariance matrix designed for the AR-1 model can be used with the Position-AR covariance matrix when the residual errors are calculated with A 2 andŝ s 2;n and the initial estimateQ…”

“…The maximization in (24) yields an estimate for A of the form (10) and an estimate for Q of the form (13). As long as Ás s n is kept small, convergence properties for the EM algorithm carry over to the proposed mobility tracking scheme.…”

“…The additional information may consists of a terrain map or street layout [11], [12], traffic conditions [13], hotspots or regions of interest [14], [15], obstacle's shapes and locations [16], behavioral rules to represent typical human responses [17], etc. However, the applicability of such models may be limited since the parameters for one scenario may not be usable in other environments.…”