Disconnection prediction in mobile P2P networks using publish/subscribe

In many pervasive scenarios (e.g., emergency management or health care), operators need to exchange data and information and collaborate in order to carry on a collaborative job, but communication features can be lacking on the spot. Therefore, mobile ad-hoc networks are valuable solutions to let them coordinate. While executing some activities, nodes can move in the area and, hence, disconnect from the others. In order operators to coordinate with each other, devices need to be continuously connected, although that is not guaranteed, due to the mobile nature of the network. In this paper we present a Bayesian approach to predict device disconnections in mobile ad-hoc networks, and validating experimental results that show the viability of the approach. This prediction feature is useful for an upper coordination layer, which can arrange appropriate actions to prevent disconnections from occurring or, at least, to mitigate the consequences.