Plausible mobility inference from wireless contacts using optimization

Abstract

Studies of mobile wireless protocols benefit from real-world traces that measure and record node locations over time. Such traces form an essential part of evaluation studies for many types of mobile and wireless networks, including but not limited to opportunistic networks.Mobility information provides the possibility of richer simulations in such studies, while contact information, irretrievably, loses considerable details like the exact locations of the nod es and their velocities. Despite this, contact information can be measured more easily compared to detailed locations of the nodes. We bridge the gap between relatively more available contact traces and potentially more useful mobility traces by presenting a solution for inferring plausible mobility from contact information. Our technique uses a simple non-linear optimization approach to formulate this as a feasibility problem with fundamental constraints on the mobility of the nodes. We introduce our technique and its underlying assumptions, and evaluate its performance in recovering mobility from synthetic and real traces. We also show that our approach brings orders of magnitude improvements in inference performance compared to state of the art.