Research On Composite Event Detection In Wireless Sensor Networks

Wireless Sensor Network (WSN) is a self-organizing network composed of tiny sensornodes, which deployed in a monitored area and can communicate with each other throughwireless devices. WSNs have features like large-scale deployment, self-organization, lowpower consumption, etc. In real life, WSNs have a wide range of applications, such as thedetection of forest fires, agricultural automation, intrusion detection, volcano monitoring.The special features make WSNs become a research focus of many universities andresearch institutions. Event detection is one of WSNs’ research content. Composite eventdetection is more realistic due to its consideration of many attributes of the event.Therefore, the study of composite events has more practical significance.This paper analyses the characteristics of composite events. In the case of nodesrandomly deployed, a criteria is proposed to determine the area of composite events.Combined with concepts and methods of dominating set, we propose a composite eventdetection algorithm based on dominating set (the DS algorithm). Through the alternativeroute and the path discovery mechanism, the results of event detection can be transmittedto the base station reliably. Compared with the CollECT algorithm and the CDS algorithmin the literature, the DS algorithm improves the precision of detection, reducescommunication cost and energy consumption of the network.For the unreliability of some sensed data and the vagueness of definitions of events innature, we propose the two-dimensional-GASalgorithm based on the hierarchicalclustering scheme and a decision-making mechanism for composite events based on fuzzydecision making. When most sensed data is correct, the proposed two-dimensional-GASalgorithm can effectively filter out the data of failed nodes, thus reduce the influence of thefalse data for event detection. The proposed decision-making mechanism based on fuzzydecision making not only retains the superiority of the mechanisms based on fuzzy logic in the literature, but also does not require a large rule base. And it has a low computationalcomplexity.