Towards efficient management of wireless sensor networks

Abstract

Wireless Sensor Networks (WSNs) have been successfully deployed in various applications. However, managing sensor networks is still a challenging issue. Motivated by the needs of precise carbon emission measurement and real-time surveillance for CO₂ in forests and cities, we develop CitySee, a real-time CO_2-monitoring system using sensor networks in an urban area. In order to conduct environment monitoring in a real-time and long-term manner, CitySee has to address management issues such as sensor deployment and data processing. In this thesis, we aim at studying several fundamental challenges in managing large-scale sensor networks, including sensor deployment, node diagnosis, network management and link monitoring. We first investigate the sensor deployment problem. In CitySee, it can be formulated as a relay node placement problem under hole-constraint. By carefully taking all constraints and real deployment situations into account, we propose an efficient approach which uses additional relay nodes at most twice of the minimum. We then study the node diagnosis problem and propose a novel approach AD which performs diagnosis in an agnostic manner. Specifically, AD does not require network operators to predefine the types and symptoms of possible faults. Instead, it explores the correlation patterns of system metrics and discover potential faults by tracking changes and anomalies of correlation patterns. We further study management center placement schemes to improve the performance of online network management services based on the quality of interactive communications. We define the reachability from a management center to a sensor node using Expected Transmission Ratio (ETR) and then design optimal and heuristic algorithms in which multiple management centers work in a cooperative manner to cover as many sensor nodes as possible. Finally, we exploit the sparse property of link loss rates and advocate a Compressive Sensing based approach to monitor link qualities using mobile sinks. On the identification of lossy links, further management approaches can be applied to enhance network performance.

Date of Award	2013
Original language	English
Awarding Institution	The Hong Kong University of Science and Technology