On the capacity of ad hoc wireless networks under general node mobility

We revisit the problem of characterizing the capacity of an ad hoc wireless network with n mobile nodes. Grossglauser and Tse (2001) showed that, by exploiting user mobility, it is possible to maintain a constant per-node throughput as the number of nodes grows. Their scheme allows to overcome the throughput decay (at least as 1/root n) that affects networks with static nodes, which was first pointed out by Gupta and Kumar (2000). Subsequent works have analyzed the delay-capacity trade-off that arises in mobile networks under various mobility models. Almost invariably, however, available asymptotic results strongly rely on the assumption that nodes are identical, and move according to some ergodic process that is equally likely to visit any portion of the network area. In this paper, we relax such 'homogeneous mixing' assumption on the node mobility process, and analyze the network capacity in the more realistic case in which nodes are heterogeneous, and the motion of a node does not necessarily cover uniformly the entire space. We propose a general framework to characterize the capacity of networks with arbitrary mobility patterns, considering both the case of finite number of nodes (also with the support of experimental traces), as well as asymptotic results when the number of nodes grows to infinity.