CAMPUS : a middleware for automated context-aware adaptation decisions at run-time

Abstract

So far in most context-aware systems, the decisions of when and how to adapt an application are made a priori by developers during the compile time. While such approaches empower developers with sufficient flexibility to specify what they want in terms of adaptation rules, they inevitably place an immense load on developers, especially in an extremely dynamic environment like pervasive computing, to anticipate and formulate all potential run-time situations during development time. In addition, making adaptation decisions in design-time or compile-time makes it difficult for the system to consistently deliver services of an optimal quality. These challenges motivated us to explore an approach to automating context-aware adaptation decisions by a middleware layer at run-time. The resulting middleware CAMPUS, short for Context-Aware Middleware for Pervasive and Ubiquitous Service, achieves the objective with the confluence of three key technologies: compositional adaptation, ontology, and DL/FOL reasoning. More specially, we have proposed and designed a new programming model called ATM (short for Adaptable Task Model) to completely separate context-aware adaptation from the functional concerns of applications. A comprehensive ontological model has been developed to capture important knowledge about context-aware applications built on the basis of the ATM model. Importantly, the middleware layer can perform DL and FOL reasoning on these ontologies to derive the important decisions at run-time. We designed and implemented a middleware prototype that served as a platform for us to evaluate the effectiveness of the system in enabling automated context-aware adaptation decisions and to validate the principles underpinning the design. The CAMPUS implementation has been evaluated with a number of case studies to validate the operation of the system on a realistic environment and to provide us with opportunity to obtain experimental results for further analysis. In particular, we have selected and implemented a context-aware instance messenger application to run over the CAMPUS. We systematically traced the application development cycle and validate the effectiveness of the semantic-based approach to capturing contextual, service and adaptation requirements. In capturing the system's performance, we evaluated the potential overheads introduced by deferring the adaptation decision to run-time in the middleware level. The results are significant in that they show that CAMPUS can be adapted to run on resource-constraint portable devices without significant degradation in its performance.