정확도와 비용간의 tradeoff를 고려한 이벤트 감지 분산 알고리즘

Abstract

In this paper, we present a distributed algorithm for detecting events in wireless sensor network that provides trade-off between detection accuracy and costs. Our model is that the region of interest is monitored by a large number of randomly distributed 0/1 sensors with low-power battery, limited functionality and memory, just enough for sensing, communicating and performing simple arithmetic operations. For any event detection problem, one needs some sensors awake to detect the event at the time that it happens, and to wake up its neighbors to detect further events. Scheduling for the network to save the total power-cost or to maximize the monitoring time has been studied extensively. Our scheme is that some predetermined uniformly distributed sensors, called critical sensors, are awake all the time and when a critical sensor detect any event, it broadcasts to the neighbors to check their sensing area. Then the neighbors check their area and decide with certain criteria whether they wake up or remain in sleeping mode . This algorithm running in each sensor is very simple and uses at most 2 bit of broadcasting. Sensors do not need any information such as GPS data or the relative position from the neighbors, but the number of its communication neighbors which is initially determined when the network is set. Another advantage of our algorithm is that we adapt two kinds of measure for the wake-up decision. By adjusting the threshold values, our algorithm can be applied for many applications because these thresholds provide trade-off between the accuracy of event detection and the cost of energy and communication.