Event Broker Grids with Filtering, Aggregation, and Correlation for Wireless Sensor Data Eiko Yoneki University of Cambridge Computer Laboratory, Cambridge CB3 0FD, United Kingdom eiko.yoneki@cl.cam.ac.uk Abstract. A significant increase in real world event monitoring capabil- ity with wireless sensor networks brought a new challenge to ubiquitous computing. To manage high volume and faulty sensor data, it requires more sophisticated event filtering, aggregation and correlation over time and space in heterogeneous network environments. Event management will be a multi-step operation from event sources to final subscribers, com- bining information collected by wireless devices into higher-level informa- tion or knowledge. At the same time, the subscriber’s interest has to be efficiently propagated to event sources. We describe an event broker grid approach based on service-oriented architecture to deal with this evolu- tion, focusing on the coordination of event filtering, aggregation and corre- lation function residing in event broker grids. An experimental prototype in the simulation environment with Active BAT system is presented. 1 Introduction Recent progress in ubiquitous computing with a dramatic increase of event mon- itoring capabilities by Wireless Sensor Networks (WSNs) is significant. Sensors can detect atomic pieces of information, and the data gathered from different devices produce information that has never been obtained before. Combining regionally sensed data from different locations may spawn further useful infor- mation. An important issue is to filter, correlate, and manage the sensed data at the right time and place when they flow over heterogeneous network envi- ronments. Thus, an integrated event correlation service over time and space is crucial in such environments. Event correlation services are becoming important for constructing reactive distributed applications. It takes place as part of applications, event notification services or workflow coordinators. In event-based middleware systems such as event broker grids, an event correlation service allows consumers to subscribe to patterns of events. This provides an additional dimension of data management, improvement of scalability and performance in distributed systems. Particularly in wireless networks, it helps to simplify the application logic and to reduce its complexity by middleware services. It is not easy to provide reliable and useful data among the massive information from WSNs. Mining new information from sensed data is one issue, while propagating queries over WSNs is a different issue. Combination of both approaches will enhance data quality, including users’ R. Meersman et al. (Eds.): OTM Workshops 2005, LNCS 3762, pp. 304–313, 2005. c  Springer-Verlag Berlin Heidelberg 2005