Situation-Aware Optimizations in Challenged Networks
Challenged networks refer to networks with unconventional difficulties, such as intermittent connectivity, large delay, and others. Their unique communication characteristics create new challenges for the research community and demand novel solutions to achieve efficient routing and maintain existing network services.
In this dissertation, I explore new optimizations to combat these challenges under the unifying theme of achieving situation awareness. In particular, I study four categories of networks that contain a range of disruptions: highly varying mobility, lossy radio links, opportunistic connectivity, and intermittent connectivity.
The first category consists of networks with a varying mobility pattern found in many challenged mobile networks. I propose a model-based approach to capture mobility phase changes in order to maintain efficient routing. When evaluated using a real-world mobility trace, our approach leads to an improvement of up to 120% in packet delivery rate.
The second category consists of networks with lossy links. Since data collection is frequently disrupted by the difficulty of identifying good links, I use supervised learning to maintain accurate link quality information under heavy traffic load when traditional approaches fail. Our approach yields improvements of up to 300% when evaluated on a real-world sensor network testbed.
The third category consists of networks with unpredictable mobility in which data can only be forwarded in a store-and-forward fashion. Existing approaches depend heavily on mobility prediction, which is difficult though, if not impossible. I use erasure coding to forward coded data to more contacts to combat inaccurate predictions. Simulation results
show that our approach has a smaller worst case delay compared to other state-of-the-art algorithms.
The fourth category consists of static sensor networks with intermittent connectivity. In such networks, energy saving opportunities arise during network disconnection. I propose a new transport protocol to leverage such opportunities that yields significant idle energy savings compared to existing approaches.
Overall, this dissertation investigates a range of challenged networks and propose a set of techniques to enable situation-awareness in order to achieve high routing performance and energy efficiency. The outcomes reveal high potential for situation-aware techniques and provide new perspectives on optimizations in challenged networks.