Statistical Modeling of Aggregated Inter-contact Time in Different Opportunistic Networks and Its Applications

Abstract

A pioneering body of work in the area of mobile opportunistic networks has shown that characterizing aggregated inter-contact time between mobile nodes is crucial. The most common approaches to study aggregated inter-contact time (ICT) in literatures are focused on a certain type of mobile wireless networks. Up to now, it still lacks a general approach to model the aggregated ICT effectively in different type of networks. This paper proposes a new approach to model aggregated ICT based on extreme value theory (EVT), which uses the generalized Pareto distribution as the unique asymptotic model of the tail distribution. Mean excess function is proposed to choose the threshold value and also other parameter estimation methods to the statistical model are discussed and applied to real mobility traces collected from different opportunistic networks, such as social pocket switched networks and vehicular ad hoc network. A decision rule based on confidence level of parameter is proposed to determine the statistical characteristics of aggregated ICT. By performing extensive experiments using EVT-based model, the aggregated ICT distributions for different type of networks are obtained. To show how well the EVT-based model actually fits the data in the tails, the complementary cumulative distribution function in a logarithmic scale is used to validate the model. Also according to the Kolmogorov---Smirnov test, it illustrates that EVT-based distributions with different parameter values fit the tails well at the desired significance level 0.05 for the taxi mobility in VANET and the human node in the other type of opportunistic networks. Finally, some prospective applications based on the statistical model are discussed in designing new network protocols.