Predicting encounters in opportunistic networks
An opportunistic network is composed of human-carried mobile devices that interact in a store-carry-and-forward fashion. A mobile node stores data and carries it around; when it encounters another node, it may decide to forward the data if the encountered node is the destination or has a better chance of bringing the data closer to the destination. In order to obtain efficient routing in such a network, we should be able to predict the future behavior of a node. This would help the algorithm decide if the data contained by the node should be further carried or forwarded, and to which node it is to be forwarded. In this paper, we present a mobile interaction trace collected at the University Politehnica of Bucharest in the spring of 2012, and analyze it in terms of the predictability of encounters and contact durations. We show that there is a regular pattern in the contact history of a node and then we prove that, by modelling the time series as a Poisson distribution, we can efficiently predict the number of contacts per time unit in the future. These assumptions are demonstrated both on the trace presented in this paper, as well as on a different trace recorded in another type of environment, showing that predictability doesn't happen only in strict and controlled situations.