A progressive reliability framework for wireless sensor networks

This dissertation investigates a new framework for achieving high data rates and negligible error probabilities by distributing the processing over multi-hop networks. In particular, we consider the case of reliable data transmission in energy constrained Wireless Sensor Networks (WSNs). Low rate channel coding can increase reliability and eliminate the need of costly retransmissions of sensor data. However, low rate channel coding on end-to-end basis puts a considerable burden in terms of transmit energy on resource constrained sensor nodes. We propose a setup that progressively provides reliability as information traverses the multi-hop wireless sensor network. Precisely, we propose an Optimal Progressive Error Recovery Algorithm (OPERA) under which, individual intermediate sensors that are relaying data toward the base station, partially and optimally channel-decode the incoming packets as data reaches the final destination. We use iteratively decodable Low Density Parity Check (LDPC) codes in order to illustrate the efficiency of the proposed architecture. The proposed OPERA setup optimally distributes the decoding iteration budget over the entire network with minimal energy expenditure. We provide a comparison between our iteration assignment algorithm with both random iteration assignment and end-to-end channel coding, and show that OPERA performs considerably better. In addition, further motivated by resource limitation of sensor nodes and the well-known sensor reachback problem, we propose a version of OPERA that is proportionally fair to individual sensor nodes using rate adaptivity in channel coding. We use systematic puncturing of LDPC codes to develop a rate compatible framework that is fair to individual nodes by both, progressive decrease in parity as information reaches the destination node, and restricting the per node processing based on their location in the multi-hop WSN. We present various scenarios for a WSN to achieve rate-compatibility and discuss associated complexity/energy usage and distortion/reliability tradeoffs. Further motivated by distributed architectures, we propose a distributed version of OPERA in which decoding iterations are assigned in a pair-wise fashion to individual nodes. Under the proposed Distributed Progressive Error Recovery Paradigm (D-OPERA), nodes collaborate, in a distributed pair-wise manner to allocate the processing budget to individual nodes and obtain near optimal performance. We further investigate the performance of proposed framework when multiple paths are available for data originating nodes to the destination. We apply the OPERA framework to both still images as well as video streams and present an architecture for reliably transmitting video in WSNs without fast depleting their energy resources.