Chinese Journal of Computers   Full Text
Title	A Fault-Tolerant Algorithm for Event Region Detection in Wireless Sensor Networks
Authors	CAO Dong-Lei CAO Jian-Nong JIN Bei-Hong
Address	(Graduate University of Chinese Academy of Sciences, Beijing 100080) (Institute of Software, Chinese Academy of Sciences, Beijing 100080) (Department of Computing, Hong Kong Polytechnic University, Hong Kong)
Year	2007
Issue	No.10(1770—1776)
Abstract &   Background	Abstract Detecting the region of emergent events is an important application of wireless sensor networks. In recent years, research on fault-tolerant event region detection algorithms becomes a hot topic. By assuming that the occurrence of an event is spatially correlated, previous work distinguish fault and event by exchanging readings among neighboring sensors. Considering that in many cases, an event is both spatially and temporally correlated, this paper proposes a distributed and localized algorithm for fault-tolerant event region detection. Aiming at reducing the network traffic, this algorithm determines a faulty sensor by using statistical hypothesis test for matching the reading sequence of sensors and statistical characters of the event. The analysis shows that the proposed algorithm is more energy-efficient than existing ones. The simulation results show that the algorithm can detect as much as 93% of the event region and 88% of faults, when 10% of sensors are faulty. keywords wireless sensor network; fault-tolerance; event region detection; energy-efficient background Event region detection is an important application of wireless sensor networks (WSNs), especially when the monitored events are harmful to human beings, such as harmful chemical pollution and forest fire. In these scenarios, wireless sensor networks can be tasked to answer queries about events in the monitored environment. Especially one particular class of queries is important: Determining the event regions in the environment with a distinguishable characteristic. As sensor nodes in a WSN are small devices with limited energy, they are prone to failures and energy-exhausting. As a result, the two fundamental challenges in event region detection problem are sensor faults and limited energy supply. To distinguish faults and event, previous work assumed that the event is spatially correlated. Neighboring sensors can exchange their readings to determine the event region by using the Bayesian algorithm or majority-voting. But these message exchanges will consume much energy of the sensor nodes and decrease the lifetime of the WSN. This paper assumes that the event is both spatially and temporally correlated as it is the case in many applications. As a result, the event can be modeled as a stochastic process called event process. In addition, each sensor periodically measures the environment and hence its readings are recognized as a reading sequence. Given the statistical characteristic of the event process, each sensor can locally test its reading sequence against this statistical characteristic to detect faults.