SENSOR NOTIFICATION SYSTEM

Sensor Notification System

Sensor Notification System

Advantages

Sensor Networks can be considered as a tool for observing states of the real world. For example, the occupancy of nesting burrows is observed in, or the temperature distribution across a vineyard is observed in. In particular, we are often interested in the detection of predicates over the state of the real world, such as “Notify me if more than four birds are in their burrows” or “Notify me if the temperature in a given area is below 0?C” (Allen 2003 832-843). We refer to such problems as detecting a certain real-world state. The detection of such real-world states is a non-trivial problem due to the distributed nature of sensor networks.  

Wireless sensor networks can be considered as a tool for detecting certain states in the real world. We examine the use of event-based approaches for this task. In the literature, a number of event notification systems have been presented that facilitate the specification and automatic detection of event patterns so called composite events. While events are a valuable abstraction in sensor networks, we show that composite events are less suited to detect real-world states with sensor networks. We illustrate an alternate solution that retains the advantages of an event-based approach, but which provides better support for the specification and detection of real-world states.  

Platform of Sensor Nodes

Data fusion involves communication of the local state samples in order to aggregate two or more samples from distinct nodes. However, limited resources in sensor networks make it a necessity to reduce the communication overhead to a minimum. Hence, rather than transmitting local state information in regular intervals, only indications of changes in the observed local state are usually transmitted. Such state changes are commonly referred to as events. Hence, state detection is closely related to the notion of events. Events proved to be a useful abstraction in many traditional application domains such as database systems and distributed systems. In the context of distributed systems, various event notification systems (Madden Franklin Hellerstein and Hong 2000 74-102) have been devised that assign network nodes the roles of event producers and event consumers.

Event producers generate typed event notifications upon the occurrence of certain local events. Event consumers specify their interest in certain types of event notifications using so-called subscriptions. An event notification service mediates between producers and consumers by providing efficient delivery of event notifications from producers to consumers based on subscriptions. Important classes of event systems supports content-based subscriptions, which do not only specify certain types of events, but can also, specify certain instances of an event type by means of expressions over event parameters. For example, an event may contain the location of the sensor node. A content-based subscription may then declare interest in events from sensor nodes that are located within a certain area. Some of these distributed event systems support the notion of composite events, where the consumer specifies a pattern of multiple events (Ahrentzen 2000 723-52). The event notification system matches event notifications against these patterns and informs the consumer when a match is found. 

Such content-based event systems with support for composite-event detection are a potential tool for supporting the detection of real-world states in sensor networks, since they support the communication of local state changes (by means of event notifications) and the detection of global real-world states (by means of composite-event ...