Ontology-Driven Complex Event Processing in Heterogeneous Sensor Networks

Modern scientific applications of sensor networks are driving the development of technologies to make heterogeneous sensor networks easier to deploy, program and use in multiple application contexts. One key requirement, addressed by this work, is the need for methods to detect events in real time that arise from complex correlations of measurements made by independent sensing devices. Because the mapping of such complex events to direct sensor measurements may be poorly understood, such methods must support experimental and frequent specification of the events of interest. This means that the event specification method must be embedded in the problem domain of the end-user, must support the user to discover observable properties of interest, and must provide automatic and efficient enaction of the specification. This paper proposes the use of ontologies to specify and recognise complex events that arise as selections and correlations (including temporal correlations) of structured digital messages, typically streamed from multiple sensor networks. Ontologies are used as a basis for the definition of contextualised complex events of interest which are translated to selections and temporal combinations of streamed messages. Supported by description logic reasoning, the event descriptions are translated to the native language of a commercial Complex Event Processor (CEP), and executed under the control of the CEP. The software is currently deployed for micro-climate monitoring of experimental food crop plants, where precise knowledge and control of growing conditions is needed to map phenotypical traits to the plant genome.