A framework for enabling energy efficient semantic views in wireless sensor networks for data intensive applications

摘要

Sensor networks have been envisioned to be a promising techniquefor data intensive applications such as disaster management andemergency response and are being designed and deployed for theseapplications. The effectiveness of sensor networksin providing information is determined by human's capacity torecognize and comprehend information from the raw data collected,and act accordingly.Finding relevantinformation from the large amount of data, however, becomes achallenging problem because user interests continues to grow asthe number and variety of sensors increase and users expect toreceive only the data they select to view. Transmitting usersirrelevant data during data processing not only overloads userswith unneeded data but also incurs unnecessary communicationoverhead. Furthermore, the user interests may be correlated when alarge number of users seek information from sensor networks. As aresult, a lot of redundant data transmission can be incurredduring processing in resource-constrained sensor networks. Dataaggregation, though effective in reducing data transmission foraggregated queries, doesn't take the correlation among userinterests into consideration during processing. Therefore,additional techniques need to be proposed to provide efficientinformation delivery for correlated user interests inresource-constrained sensor networks.To bridge the gap between data collected by sensors and the information interests of users, the concept of "semantic view" is proposed in this thesis. The semantic view is a powerful abstraction which allows the fusion of multi-sensor and multi-source data into a virtual data gathering and analysis infrastructure commensurate with the interest of an end user. The main challenge is to enable semantic views in an energy efficient manner in resource constrained sensor networks. To that end, a framework which consists of five protocols and algorithms, "Query Aware Sensing", "Probabilistic Query Dissemination", "Correlated Multi-query Processing", "Location Discovery using Out-of-Range information with multi-lateration" and "End-to-end pairwise key establishment" is presented. In the proposed framework, The ultimate goal is to develop an energy efficient and secure framework towards enabling semantic views in sensor networks for data intensive applications.