Context Evidence and Location Authority: the disciplined management of sensor data into context models

Abstract

Ubiquitous computing requires that data be handled across a very wide range of data rates and weights of associated meaning. A suitable system architecture is layered from event data through simple sensors, smart sensors, and smart environment agents. Upward through these layers the issues for representation and management of the data shift from the distribution and fast, bulk storage of very frequent simple data, to relatively infrequent logical deductions made against relatively complex models. The lower layers of systems for event data will be reused in different applications. In addition, abstract models of location may have a lot in common between different applications, which a common Location Authority can represent as a common model of the physical environment that changes only slowly (through manual administrative maintenance). On the other hand models of the devices and sensors within locations, and the moving population of people, are far more dynamic, require automatic updating, and different applications choose quite different attributes and relationships to model. We describe how the Merino/Personis architecture for an Intelligent Environment context supports the changes of representation of knowledge across this range and the different programming styles suited to the different levels.