Exploiting data fusion to improve the sensing performance of wireless sensor networks

Abstract

Wireless sensor networks (WSNs) have been increasingly deployed for mission-critical surveillance applications such as environmental monitoring and security surveillance. These applications often impose stringent sensing performance requirements including low false alarm rate, high detection probability and short detection delay. However, as the low-cost sensors are often deeply integrated with physical environments, the sensing performance of a WSN is inevitably affected by various physical uncertainties, which include stochastic sensor noises, unpredictable environment changes and dynamics of the monitored phenomenon. In practice, collaborative data fusion algorithms that can deal with sensing uncertainties and enable sensor collaboration have been widely employed in WSNs to improve system sensing performance. However, these algorithms often have complex complications to the system-level sensing performance. As a result, the systematic performance analysis of these fusion-based WSNs as well as the design of performance-directed sensing algorithms have received little research effort. In this thesis, we first analytically study the impact of data fusion on the system sensing performance of WSNs. We then develop various performance-directed algorithms for fusion-based WSNs such as sensor calibration and motion planing algorithms. As two fundamental performance measures of WSNs, sensing coverage and detection delay jointly characterize how well a sensing field is monitored by a network. Most previous analytical studies on sensing coverage and detection delay are conducted based on overly simplistic sensing models, e.g., the disc model, which do not capture the stochastic nature of sensing as well as the possible collaboration among sensors. This thesis explores the fundamental limits of sensing coverage and detection delay of fusion-based WSNs by deriving the scaling laws between sensing coverage, detection delay, network density and signal-to-noise ratio (SNR). The analysis shows that data fusion can significantly improve sensing coverage and reduce detection delay by exploiting the collaboration among sensors. In particular, for signal path loss exponent of k (typically between 2:0 and 5:0), we have ρf = O(ρd1-1/k), where ρf and ρd are the densities of uniformly deployed sensors that achieve full coverage under the fusion and disc models, respectively. Moreover, the ratio of network densities to achieve full coverage or minimum detection delay under the two models satisfies ρf/ρd = O(SNR2/k). The results imply that data fusion is effective in the scenarios with low SNRs. In contrast, the disc model is only suitable when the SNR is sufficiently high. The results help understand the limitations of the previous analytical results based on the disc model and provide key insights into the design of fusion-based WSNs. This thesis then develops various performance-directed sensing algorithms for fusionbased WSNs. First, this thesis proposes an adaptive system-level calibration algorithm that enables a fusion-based WSN to achieve optimal sensing performance in the presence of various unpredictable system and environmental dynamics. Different from traditional device calibration approaches that work in open-loop fashion, this new algorithm features a feedback control loop that exploits sensor heterogeneity to deal with the unpredictable dynamics in calibrating system performance. However, the optimal sensing performance of a calibrated all-static network may be still unsatisfactory due to inefficient deployment or severe changes of network conditions, e.g., coverage holes caused by death of nodes. This thesis then proposes to exploit reactive sensor mobility to enable the reconfigurability of a fusion-based WSN in target detection. Specifically, mobile sensors are directed to move toward a possible target when a preliminary detection consensus is reached by a group of sensors. The accuracy of final detection result is improved as the mobile sensors have higher SNRs after the movement. This thesis presents a near-optimal movement scheduling algorithm that minimizes the movement of mobile sensors subject to system detection performance requirements.

Date of Award	4 Oct 2010
Original language	English
Awarding Institution	City University of Hong Kong
Supervisor	Jianping WANG (Supervisor), Guoliang XING (Supervisor) & Xiaohua JIA (Co-supervisor)