Localization for Intermittently Connected Ad Hoc Networks

We consider localization for intermittently connected networks. Traditional localization approaches rely on network infrastructure such as access points (AP) as reference nodes or anchors, and do not work well when infrastructure is sparse. We propose a new class of localization scheme, where mobile nodes act as anchors for each other. Each node maintains an anchor table, storing its encounter history with APs and mobile nodes. When two nodes meet, they share their encounter history with each other, and jointly estimate the current location to form a new anchor. This dramatically increases the number of anchors for a mobile node, and reduces the localization error when position estimation is triggered by a user application. Both anchor formation and position estimation are formulated as a maximum likelihood (ML) estimation problem, by exploiting constraints of node location based on encounter history and mobility profile of nodes. Numerical examples are provided to illustrate properties of the ML estimator. We performed simulations and showed that localization error is independent of the anchor table size under a random walk mobility model. Thus, it suffices to store the most recent anchor, which simplifies implementation and is conducive to privacy. More importantly, localization error decreases as node density increases, and approaches the transmission range of mobile nodes when node density is sufficiently high