Modeling steady-state and transient behaviors of user mobility
Authors:
Open Access
- 22 May 2006
- conference paper
- Published by Association for Computing Machinery (ACM)
- p. 85-96
Abstract
Recent studies on mobility modeling have focused on characterizing user mobility from real traces of wireless LANs (WLANs)and creating mobility models based on such characterization. However, most of the work does not study how user mobility is correlated in time at different time scales. For example, the future APs with which a user will be associated are predicted without the knowledge of when the association will take place and for how long. In this paper, we build a mathematical model for characterizing both steady state and transient behaviors of user mobility in WLANs. Specifically, we mode user mobility by a semi-Markov process, and obtain the transition probability matrix and the sojourn time distribution from the association history of WLAN users available at Dartmouth college [21]. With the steady-state characterization of user mobility in WLANs, we can estimate the long-term wireless network usage among different access points. By comparing the steady-state distributions of semi-Markov models built based on trace data collected at different time scales, we are able to characterize the degree of correlation in time and location.We also perform a transient behavior analysis of the semi-Markov process (that characterizes user mobility), and devise a timed location prediction algorithm that accurately predicts the future locations of users both the future access points they will associate themselves with and the association duration. We demonstrate the utility of timed location prediction, by showing how it can be utilized to predict the distribution of future user ocations with the time information figured in, and redistributing loads among neighboring APs. An improvement of 80% (in terms of load balance) is observed in a wide spectrum of traffic loads in the simulation.Keywords
This publication has 15 references indexed in Scilit:
- Model T2005
- A mobility model based on WLAN traces and its validation2005
- Empirical modeling of campus-wide pedestrian mobility: observations on the USC campus2005
- Classifying the Mobility of Users and the Popularity of Access PointsLecture Notes in Computer Science, 2005
- Towards a model of user mobility and registration patternsACM SIGMOBILE Mobile Computing and Communications Review, 2004
- The changing usage of a mature campus-wide wireless network2004
- Numerical Treatment of Homogeneous Semi-Markov Processes in Transient Case–a Straightforward ApproachMethodology and Computing in Applied Probability, 2004
- Towards realistic mobility models for mobile ad hoc networks2003
- A survey of mobility models for ad hoc network researchWireless Communications and Mobile Computing, 2002
- Characterizing user behavior and network performance in a public wireless LAN2002