On the Throughput and Delay in Ad Hoc Networks With Human Mobility

In this paper, we study the impact of human mobility on throughput and delay for people-centric applications in mobile ad hoc networks (MANETs). We consider a general human mobility model for MANETs, which can capture important features of human mobility, such as time correlation, node correlation, location heterogeneity, and node heterogeneity. Multiple unicasts with general arrival processes are delivered, and nodes are equipped with infinite buffers. Under our system model, we first characterize the network stability region in terms of the probability of each node set visiting each location and the amount of transmission resources at each location. We show that the node correlation and heterogeneity of locations' popularity usually decrease the size of the network stability region, whereas the diversity of locations visited by a node usually increases the size of the network stability region. Then, by solving a stability-related optimization problem, we develop a throughput-optimal policy based on the obtained optimal solution. We obtain the upper and lower bounds of the delay performance under the proposed policy. Finally, using simulations based on a theoretical model and some real traces, we verify the analytical results and compare the performance of the proposed policy with some existing policies.

I. Introduction

Mobile ad hoc networks (MANETs) using the store-carry-forward paradigm are also referred to as delay-tolerant networks (DTNs). Using the store-carry-forward paradigm, a node can carry buffered data while moving in the network until it reaches a suitable next-hop node to relay the carried data. Thus, data can be delivered from a source toward a destination through multiple transmission opportunities, which may not be possible when nodes are static or data transmission cannot tolerate large delay.