Application Performance-Driven Design and Management of Sensor Networks.

摘要

With increasing high-fidelity and real-time decision-making applications for Wireless Sensor Networks (WSNs), it becomes necessary to investigate WSN deployment designs with particular focuses on satisfying application-specific performance requirements, such as reconstruction fidelity, detection performance, and response latency.;The primary concern of this work is the development of a framework that structures the fundamental steps in performance-driven designs for event detection WSNs, and facilitates analysis of their expected performance prior to actual deployments. At the heart of the proposed framework is a set of procedures to alleviate the selection of detection schemes that provide guarantees on accuracy and latency requirements.;Techniques based on hypothesis testing and Maximum Likelihood estimation have been used, respectively, for detection and tracking purposes in WSNs. In this study, the expected performance of these techniques is analytically studied for WSN deployments. The proposed extensions allow the designer to select the appropriate system parameters such that required performance metrics are ensured.;A major factor affecting the decision-making performance is the presence of faulty behaviors in sensor data. An important component of the present work is the development of a general two-tiered fault detection that is particularly integrated with the proposed event detection scheme. Such a coupled procedure progressively detects faulty behaviors, isolates faulty sensors, and improves detection accuracy.;Several aspects of the proposed framework are verified and validated through its application to real and synthetic data.