•   When: Wednesday, August 10, 2016 from 01:00 PM to 03:00 PM
•   Speakers: Mansour Abdulaziz
•   Export to iCal

Abstract

The underlying motivation for my dissertation research is to investigate the combined use of wireless network coding with multi-channel communication to perform data collection in low-power and lossy networks (LLNs). Network coding allows eavesdropping on non-source, non-destination wireless nodes in order to recombine overheard data with its own data, which increases the ability of the ultimate destination to recover a high amount of information, even in the case of high bit error rates. Coordinated multi-channel communication has the potential to dramatically increase the throughput for many types of LLN applications. My work seeks to unify network coding and coordinated multi-channel communication. I consider several important classes of LLN applications, including scenarios with static data collection sinks and mobile data collection sinks.

This dissertation has three primary contributions to improve communication performance in data collection LLN systems. The first is a protocol called MuCode, which is designed to support single or multiple sinks that support a Convergecast (many-to-one) communication pattern. I describe a synchronized channel switching policy that takes advantage of wireless overhearing to perform network coding operations. I show that optimally solving certain aspects of this problem is NP-Hard, and present a set of heuristics for building the delivery mechanism. I have evaluated MuCode against several other schemes and my results show significant performance improvements under a variety of scenarios.

The next two contributions use the results of the MuCode approach and extend them into mobile environments. The second contribution is the protocol MuTrans, which is employed in environments that require mobile data collection in a way that is predictable but uncontrollable. I formally analyze the complexity of this problem in terms of how to minimize data collection latency, which shows it is an NP-Hard problem. I present a synchronized dynamic round-robin scheduling policy for uploading data to a mobile collector that is based on assigning a method for load balancing. My evaluation of data aggregation in the presence of packet errors shows that MuTrans can significantly reduce the latency for data collection, thus providing strong support for mobile data collection.

Finally, I present the MuCC protocol, which is designed to support data collection when mobile data collectors motions are both controllable and predictable. After designing and evaluating a two-layer architecture, I provide algorithms on cluster head selection, cluster membership assignment, and trajectory planning. My experimental tests indicate that MuCC substantially outperforms similar approaches in hierarchical data collection.