Research On Top-k Query Processing Technics For Energy Efficiency In Wireless Sensor Networks

With the development of microelectronics, computers, networks, and wireless communications,wireless sensor networks have been deeply investigated in wide-scale applications. However, thesensor nodes are often battery-powered which are energy limited. Meanwhile, a mass of data isgenerated by wireless sensor netwoks, if all sensor readings are delivered to base stations, a vastamount of communications will emerge. In particular, wireless communication plays a most importantrole in enery consumption of sensor networks. Hence, when performing data queries, in order toprolong the lifetime of a wireless sensor network, it is significant to reduce communications. Thisdissertation describes energy-efficient and balanced top-k query techniques, and focuses on top-kquery techniques based on adaptive filters. The main innovations of the dissertation are as follows:(1) Utilizing the spatial correlations of pairwise sensor nodes, the prediction models and thecriteria of high spatial correlation are constructed. According to the predicting models and criteriaabove, an energy balanced algorithm named EBSTopk(ε, δ) is proposed, which is based on iterativerandom sampling technique. The algorithm not only reduces the global energy consumption inwireless sensor networks, but also achieves balanced energy consumption among all sensor nodesafter continuous processing top-k queries, and efficiently prolongs the lifetime of networks.(2) Methods of improving the accuracy of queries and lowering the sampling rate are proposedfor energy-efficient and balanced top-k query processing algorithms. In order to improve the accuracyof queries, a spatiotemporal model based on Kalman filter approach is proposed, according to thespatial correlations of sensor readings and the temporal correlations of individual sensor readings. Thesensor network partition is optimized by combining sensor data relevancy with sensor spatialdistances, so that sensors in the same region have high correlation and are close to each othergeographically. The proposed method can lower the sampling rate and reduce radio communicationsgreatly.(3) An energy-efficient top-k query technique based on adaptive filters is proposed. Due toupdating filters consuming a vast amont of energy, an algorithm named FUGPR based on Gaussianprocess regression is proposed. When the filters change, the sensor readings are predicted to calculatethe updating costs of filters, and then to decide whether the filters need to be updated or not. Thus, theenergy consumption for updating filters is decreased.