In practice, wireless networks operate under multiple, mostly severe, constraints (bandwidth, energy resources, etc). Consequently, efficient techniques have to be employed in communicating data with sufficiently high data rates (depending on the application) while complying with the imposed constraints. Distributed Video Coding (DVC) and Network Coding (NC) are amongst the most dominant techniques employed in constrained data networks. In this dissertation, we address these two techniques with the objective of realizing practical and efficient data networking solutions that fit in resource constrained wireless networks. In one part of the dissertation we address DVC over Visual Sensor Networks (VSNs) from a practical point of view, i.e., unlike a large body of research work related to this topic where the focus is on theoretical analysis and simulation, we study the practical aspects that arise when deploying DVC over real visual sensors. To that end, we develop a Resource-constrained DVC (RDVC) codec, deploy it over some of the widely used visual sensors, and conduct precise energy measurements that are used throughout our study.One RDVC-related challenge that we address in this dissertation is source rate estimation, i.e., trying to efficiently identify the source rate to be used in encoding a frame. This question is crucial since sending more bits than necessary will lead to inefficiency in compression while sending fewer bits will lead to failure in decoding. We propose a practical solution that completely eliminates the need for the costly feedback messages.Another challenge we address is the global choice between a DVC encoding option that involves intensive computations and leads to less transmission versus another choice with minimal computations that implies higher transmission-energy overhead. We carry out an operational energy-distortion analysis for a variety of options available to RDVC on visual sensors.Polar codes are the first codes proven to achieve capacity while having low encoding and decoding complexity. This motivates us to employ polar codes in our DVC platform. We compare the performance of polar codes with the more established and more investigated Low Density Parity Check Accumulate (LDPCA) codes in the context of DVC. Our results show that polar codes offer a clear advantage when coding smaller size image blocks. Consequently, polar codes could represent a viable solution for distributed sensor networks that capture low-resolution video signals. In the final part of the dissertation, we survey state-of-the-art NC solutions designed to achieve high throughput transmission in multicasting over wireless mesh networks while maintaining 100% packet delivery ratio. We propose HopCaster; a scheme that employs the cache-and-forward transport strategy. We show that HopCaster achieves significant performance gains compared to schemes that employ the end-to-end transport strategy.
Read
