Study On Grid-Clustering Routing Protocol And Data Aggregation For Wireless Sensor Networks

Advances in wireless networking and embedded microprocessor design have enabled the creation of wireless sensor networks (WSN). Due to its special challenges and wide range of applications, WSN is attracting more and more researcher's interest and becoming one of hottest research areas.A typical wireless sensor network is composed of a large number of low-cost sensor nodes, which are densely deployed either inside the phenomenon or very close to it. Sensor nodes usually transmit collected data to the sink by multihop communication using wireless medium. Every node must possess the self-organizing ability and work as a router in traditional computer network.The nodes are typically resource constrained and are operated by limited and irreplaceable battery power, which makes energy consumption a concern. And the limited power, limited computation ability, the limited bandwidth, and the limited memory in the sensor node bring more technology challenges than other kinds of computer networks. Most existent and mature technologies are not well suited for the unique features and application requirements of sensor networks.This dissertation mainly focuses on two basic and key issues on WSN: routing and data aggregation.In every kind of computer network, routing always is one of key technologies and determines the performace of whole network. The unique features and application requirements of sensor networks need a routing protocol which can fulfill goals including high energy efficiency, high scalability, robustness, and supporting data aggregation.In this dissertation, the features and goals of routing protocol in WSN are well studied. And existent typical routing protocols for WSN are classified, analyzed and compared. Especially, the routing protocols with virtual hierarchy are discussed detailedly. Then, a grid-clustering routing protocol (GROUP) for a large-scale sensor network, one of main innovations of this dissertation, is proposed.In GROUP, primary sink (PS) selects grid seed (GS) periodically or on demand within near area based on the residual energy of sensor nodes and random factor. Then cluster heads around the whole sensor network, which form an approximate grid with a certain width, are chosen based on the location of grid seed. GROUP combines multiple factors (i.e. residual energy, location, and random factor) to select the cluster heads rather than single factor in other clustering routing protocols. GROUP has high scalability and can work well in some complex scenarios, e.g. the scenario with multiple sinks and the scenario with mobile sink. In order to deal with the possible failure of cluster head and improve the reliability, recovery mechanism is applied in GROUP.The performance of GROUP is evaluated by theoretical analysis and simulations using NS-2 in this dissertation. The impact of important parameters on GROUP is studied, and the comparison between GROUP and two similar routing protocols (i.e., TTDD and LEACH) is made. The results of simulations show that GROUP has the better performance than LEACH and TTDD in energy efficiency, balance of energy consumption, packet delivery fraction, scalability, etc.Due to the high redundancy in the raw collected data in sensor node, data aggregation is one of effective approaches to reduce the communication between sensor nodes, to save the energy of sensor nodes and to prolong the lifetime of the sensor network. With data aggregation, only the few processed data will be transmitted to the sink rather than the numerous raw collected data.In this dissertation, the principle and value of data aggregation are studied, and the existent typical approaches of data aggregation are compared. Then a novel data aggregation approach based on neural network, another innovation in this dissertation, is proposed. We name this approach as Neural-Network Based Aggregation (NNBA).NNBA, which can work on the virtual-hierarchy based routing protocol like as GROUP, is an effective approach for data aggregation in WSN. Every cluster in WSN is regard as one three-layer peptron in NNBA. Neural-network based functions are applied in both cluster head and cluster member to process the large number of raw data. And only the processed result, which can represent the features of the raw data, will be transmitted to sinks. Various functions of peptron in NNBA are designed under the application scenario of the wireless sensor network for mornitoring forest fire.The performance of NNBA is well evaluated by analysis and simulations. The results of evaluations show that NNBA is one effective approach of data aggregation in WSN, especially in the application of periodical reporting.