Event detection is one of the main components in numerous wireless sensor network (WSN) applications. WSNs for military application are deployed to detect the invasion of enemy forces, health monitoring sensor networks are deployed to detect abnormal patient behavior, fire detection sensor networks are deployed to set an alarm if a fire starts somewhere in the monitored area. Regardless of the specific application, the network should be able to detect if particular events of interest have occurred or are about to. Traditional event services allow for the definition of events including correlated events, registering for events, and upon occurrence of events, detection and notification of events. In WSNs, events are not binary, but are based on sensor fusion from many noisy sensors in complicated environments. Sensor data may be missing, wrong, or out of date. Consequently, the event service must support real-time, data fusion, confidence calculations, and conserve power. The event service must also fundamentally recognize location (i.e., events are a function of where they occur as well). The event service must be highly decentralized in order to work on the limited capacity devices. The service must also minimize false alarms. All these features make building event services for WSNs very challenging.

 

This work focuses on robust real-time event services that would allow the development of WSN applications that can monitor and detect much more sophisticated events. An event service has four main components: event specification, event registration, event detection, and notification delivery.

  1. Event specification: There is a need for a formalized description language that can incorporate the knowledge of sensing with event definitions. Such a language will provide computer scientists with well-defined requirements for designing the sensor network. It would also be used as an interface between people who register events (e.g. domain experts and scientists who recognize the characteristics of events) and the sensor network designers (mostly computer engineers, who are responsible for designing protocols to control sensors according to certain temporal and spatial specifications). We are developing a formal event description language which is an enhanced Petri net. This language, MEDAL, can capture the structural, spatial, and temporal properties of a complex event detection system.
  2. Event registration: Once the events have been described, we need appropriate techniques to translate the descriptions into a format that the sensor nodes can “understand”. 
  3. Event detection: We are studying how using fuzzy values could improve the accuracy, timeliness, and resource requirements of event detection. In addition, we plan to integrate MEDAL with fuzzy logic and investigate what effect this would have on the precision of event detection as well as on the amount of node and network resources required by the event detection services.
  4. Notification delivery: Although notification delivery is very important part, is not the focus of our work. A lot of research has been done in this area, and we believe that there are plenty of algorithms that could be used to securely and timely deliver the nodes’ reports of detected events to their destination (neighbor nodes or base stations).

In addition, as part of the Run Time Assurance project, we have developed methods to automatically generate tests based on the formal Petri net model of a WSN application. These tests are used to improve the robustness of applications by validating at run time that the system functions correctly despite node failures or changes in the environment conditions or the network topology. We have also developed test reduction techniques that allow us to drastically decrease the size of the test suites. We will further develop these reduction techniques as well as test prioritization schemes for the cases when even after test reduction has been applied the number of tests remains too large.