Timescale Decoupled Routing and Rate Control in Intermittently Connected Networks

We study an intermittently connected network (ICN) composed of multiple clusters of wireless nodes. Within each cluster, nodes can communicate directly using the wireless links. However, these clusters are far away from each other such that direct communication between the clusters is impossible except through “mobile” contact nodes. These mobile contact nodes are data carriers that shuffle between clusters and transport data from the source to the destination clusters. There are several applications of our network model, such as clusters of mobile soldiers connected via unmanned aerial vehicles. Our work here focuses on a queue-based cross-layer technique known as the back-pressure algorithm. The algorithm is known to be throughput-optimal, as well as resilient to disruptions in the network, making it an ideal candidate communication protocol for our intermittently connected network. In this paper, we design a back-pressure routing/rate control algorithm for ICNs. Though it is throughput-optimal, the back-pressure algorithm has several drawbacks when used in ICNs, including long end-to-end delays, large number of potential queues needed, and loss in throughput due to intermittency. We present a modified back-pressure algorithm that addresses these issues. We implement our algorithm on a 16-node experimental testbed and present our experimental results in this paper