Hybridising events and knowledge as a basis for building autonomic systems

Event-based systems are a popular substrate for distributing information derived from sensors to be used in driving adaptive behaviour. This paper argues that using events directly provides a poor model of context, and that a hybrid approach that uses events to populate and maintain a distributed knowledge base offers a more stable solution. The inherent uncertainties in both sensor data and reasoning imply that traditional knowledge-based system techniques applied to contextual systems be extended to deal with more uncertain reasoning.