Energy-efficient And Robust Convergecast Technologies In Wireless Sensor Networks

Recent advances in micro-electro-mechanical systems (MEMS)technology, wireless communications, and digital electronics have enabled thedevelopment of low-cost, low-power sensor nodes that are small in size andcapable of short-distance communications. A sensor network is formed when alarge number of sensor nodes are deployed in the region of interest andconduct cooperative sensing. Sensor networks are the extensions of theInternet. Moreover, the concept of cyber-physical system (or the Internet ofthings) has made sensor network an indispensible part of modern computernetworks. However, sensor nodes are typically energy-constrained, and havelimited computation, storage, and communication capabilities. Therefore, datatransmission, especially the convergecast from sensor nodes to the sink, is avery challenging problem.There are three main challenges facing convergecast in wireless sensornetworks. First, a large number of sensor nodes periodically sense theenvironment, creating a huge amount of data to be transmitted to the sink.This is in sharp contrast to the limited energy supply of sensor nodes. Second,limited by the communication range, most sensor nodes need multi-hop relayto reach the sink. Multiple simultaneous data sources will create the funnelingeffect, i.e. the communication load near the sink is far heavier than other areas.As a result, sensor nodes that are close to the sink will deplete their energysoon, and the lifetime of the entire network is shortened. Third, sensor nodesare prone to damage and various errors, so the transmission reliability androbustness cannot be guaranteed.It can be concluded that energy-efficient and robust convergecasttechnologies are the key to solve the above challenges and to ensure thesuccessful deployment of wireless sensor networks. In this thesis, the authorwill research on the convergecast technologies for several different types of sensor networks, with the emphasis on application and transport layers.The author first considers the convergecast in a small-scale sparselydeployed wireless sensor network. In this type of network, although everysensor node can communicate with almost every other node, the deliveryratios could be very low. Inspired by the emerging compressive sensing theory,we propose a novel joint source-network coding scheme based on randomlinear projection. This scheme is capable of compressing sensor data, as wellas effectively increases network throughput through a joint design withmulti-party routing. More importantly, our proposed scheme can achieve"elastic" data gathering, which means that the data reconstruction precisiongradually increases with the number of packets received. we formulate theenergy constrained multi-source, multi-hop, and multi-path transmission as anetwork utility maximization (NUM) problem. A practical distributedalgorithm is developed to achieve the optimal utility.Following this, the author considers the convergecast in a large-scaledensely deployed wireless sensor network. Such sensor networks typically arecomposed of hundreds to thousands of sensors, generating tremendous amountof sensor data to be delivered to data sink. As sensor data are transmittedtowards the sink through multi-hop relay, the nodes which are close to the sinkwill consume more energy. They will soon run out of energy and lifetime ofthe entire sensor network will be significantly shortened. The author proposesa compressive data gathering scheme based on the compressive sensing theory.This scheme successfully achieves effective global communication costreduction and energy consumption load balancing. It can be found throughtheoretical analysis that compressive data gathering can improve the networkcapacity by several times, and is capable of utilizing various types of datasparsity including cross-domain sparsity.In addition, noticing that sensor networks are evolving towardinformation-intensive networks, the author considers the convergecast inwireless video sensor network, which is a typical information-intensive sensornetwork. The main challenge is the intra-flow and inter-flow interference aswell as the load balancing among simultaneous streams. The author proposesa forepressure transmission control scheme which is composed of transportlayer hop-by-hop control and two flow control protocols. The former can nearly eliminate packet losses in the middle of the transmission link, while thelatter ensures fair rate allocation among several concurrent video streams. Thiscross-layer design achieves the fair and effective video transmission frommultiple source nodes.In summary, the author researches on the challenging convergecastproblem in wireless sensor networks, and proposes several cross-layerenergy-efficient and robust convergecast techniques for different types ofnetworks. At the application layer, the author is the first one to apply theemerging compressive sensing theory to convergecast to achievehigh-efficiency data compression and graceful-degraded data reconstruction.At the transport layer, the author proposes a forepressure transmission controlprotocol which is in sharp contrast to the conventional backpressuretransmission control. As a result, higher energy-efficiency and robustness areachieved. At the network layer, the author proposes a multi-path routingalgorithm to maximize network utility for network coding enabledconvergecast. Both simulation results and experimental results based on realsensor data have proved that the proposed techniques can greatly improve theenergy-efficiency and robustness of convergecast in wireless sensor networks.