Cluster-based Energy-efficient Composite Event Detection for Wireless Sensor Networks Irfana Memon, Traian Muntean ERISCS Research Group Aix-Marseille Universit´ e France Emails: {irfana.memon, traian.muntean}@univ-amu.fr Abstract—Wireless sensor networks (WSNs), well known communicating architectures today, are often used to detect the occurrence of some environmental events, such as pollution monitoring, forest fires detection, location and tracking, etc. In order to reduce irrelevant alarms, multiple attributes are used in the event detection process. In WSNs, communication is often by far more expensive and difficult to control than local computation within nodes. Therefore, it becomes critical to reduce the amount of data exchange within a WSN, in order to optimize the use of power and energy resources within nodes. Energy optimization is thus one of the most important aspects of the WSN design. There are already literature and projects dealing with the detection of composite events using data ag- gregation at intermediate nodes. In this paper, a cluster-based energy-efficient composite event detection (CEC) for wireless sensor networks scheme is proposed, which performs local computation at sensor nodes and local data aggregation at level of each cluster heads in order to reduce the communication overhead. Simulation results show that jointly, considering both local computation at sensor nodes level and local data aggregation at intermediate nodes will further reduce the total energy consumption and thus prolong the network lifetime. Keywords-Wireless sensor networks (WSNs); data aggregation; local computation; composite event detection. I. I NTRODUCTION A wireless sensor network (WSN) consists of a large number of sensor nodes which are distributed in a given space for measuring environmental parameters, such as temperature, light, sound, humidity, and so on [1]-[3]. Many applications have already been envisioned and described for WSNs in a wide range of areas, such as environment moni- toring [4], health care applications [5], military surveillance applications [6], positioning and tracking [7], etc. Depending on the application domain, it may be necessary for sensor nodes within the WSN to react quickly or with critical timing constrains to detected events [8]. Moreover, the data collected by the WSN must be fresh when the corrective action is taken. One of the critical tasks in designing a WSN is to monitor, detect and report various useful occurrence of events in a timely and reliable fashion. An event can be defined as an exceptional change in the environmental parameters. Events can be simple (atomic) or composite [9]. An atomic event can be detected merely based on the observation of one attribute, for example high temperature, if the temperature is higher than a specified threshold, an atomic event is detected. A composite event is the combination of different atomic events. A detailed description of composite events is given in Section III. An event alarming application needs an answer to a question which can be derived by a set of predicates. For example, in fire alarming applications, users are not interested in knowing the exact reading of attributes (temperature, smoke) of monitored area, but they want an exact and valid answer to the question: is there fire in the monitored area? In this case, we assume that an event has some significant characteristics that can be used as threshold to distinguish between normal and abnormal environment parameter. Event detection sensor networks require periodic data update (fresh data) from the net- work. Sending data periodically to a remote base station may incurs high communication overhead, and high energy consumption for event-driven applications. One of the key problems in event-driven applications is energy efficient data extraction, (i.e. how can a base station obtain the event report with a low energy consumption). Hill et al. [10] have shown that a sensor node spends approximately the same amount of energy for sending a single bit of data as it does to execute 800 instructions. Thus, in order to decrease energy consumption and thus increase network lifetime, the amount of data exchanged should be minimized. Data aggregation techniques are very effective in reducing communication overhead (i.e., the data sensed by the sensors are combined at intermediate nodes before sending to a remote base station, BS). A number of data aggregation algorithms have been proposed in the literature [11]-[16]. Cluster-based topologies help to deal with in-network data aggregation (i.e., sensors are grouped in clusters and data aggregation is done locally within clusters). Some of the advantages to be achieved from clustering in WSNs are the reduction in energy for message transmission and construct- ing a virtual backbone for data routing purpose [17]. Many clustering algorithms have been proposed for WSNs, such 241 Copyright (c) IARIA, 2012. ISBN: 978-1-61208-207-3 SENSORCOMM 2012 : The Sixth International Conference on Sensor Technologies and Applications