Bridging Intermittently Connected Mobile Ad hoc Networks (ICMAN) with Sociological Orbits

An Intermittently Connected Network (ICN) is often modeled as a weighted graph, where the capacities and durations of edges (radio links) between nodes (users) are time varying due to user mobility. As such, there is no guarantee of a contemporaneous end-to-end path from the source to the destination through intermediary peers. This sets an ICN 1 strictly apart from a conventional mobile ad hoc network (MANET). Consequently, traditional MANET routing protocols are rendered useless within ICN. Earlier work [15] on temporally disconnected networks proposed intelligent means of data dissemination, which led to various other propositions based on similar concepts. More recent research within ICN [9] has considered algorithms based on deterministic user mobility, which is neither applicable to MANET, nor to ICMAN. Literature suggests similar study on how mobility (controlled or not) affects routing protocols and network performance (e.g., network capacity) in various types of ad hoc networks including sensor networks with mobile sinks (or base stations), and delay tolerant networks [1], [6], [13], [18]. However, they did not deal with specific user mobility patterns. In [8], [10], [11], [14], the main focus was on the so-called “contact probability” of two users, which is oblivious to the specific locations (or “hubs”) they visit. In this work, we focus on Intermittently Connected Mobile Ad hoc Networks (ICMAN), which lacks infrastructure like a MANET, in addition to having non-deterministic user mobility like in an ICN. In [3] we studied wireless users’ mobility traces in ETH Zurich campus to validate our sociological orbital claims in [4], [5]. We found that users regularly visit a small set of socially significant places (called “hubs”) forming a “hub list” and showed that such hub lists may be clustered via a technique involving a Mixture of Bernoulli distributions. We shall refer to such clusters that is represented by a hub list along with its associated visit probabilities as a “mobility profile” and shall use these profiles to propose efficient routing solutions for ICMAN users. For a conventional MANET (as opposed to intermittently connected networks), we proposed efficient routing solutions [4], [5] that leveraged on our concept of “sociological orbits” in the mobility patterns of wireless users. In this work, we illustrate the usefulness of our mobility profiles in also performing “hub-level” routing within an