Designing future‐context‐aware dynamic applications with structured context prediction. (13th October 2011)

• Record Type: Journal Article
• Authors: Zaplata, Sonja; Meiners, Matthias; Lamersdorf, Winfried
• Other Names: Kapitza Rüdiger guestEditor.
• Is Part Of: Software, practice & experience. Volume 43:Number 10(2013)
• Page Start: 1185
• Page End: 1204
• Publication Date: 2011-10-13
• Subjects: context awareness -- mobile computing -- distributed systems -- context prediction; Computer software -- Periodicals; Computer programming -- Periodicals; Computer programs -- Periodicals; 005.3
• DOI: 10.1002/spe.1126
• Languages: English
• ISSNs: 0038-0644

Abstract

SUMMARY: Middleware support for advanced pervasive applications includes dealing with heterogeneous systems and dynamic changes of execution environments. If applications are context aware, they are able to detect such changes at runtime and react to them accordingly. Furthermore, they can also proactively adapt to upcoming situations by making accurate assumptions about future contexts. However, the design of reusable middleware support for such future‐context‐aware applications is still challenging as a supporting prediction system has to be generic, but at the same time has to provide potential for high accuracy and efficiency. This paper proposes a concept for the development of future‐context‐aware applications on the basis of the novel approach of structured context prediction . As a framework, this approach allows for integrating domain‐specific knowledge and facilitates application, combination, and implementation of suitable prediction methods. In addition, specific runtime mechanisms for distribution of knowledge and adaptation of prediction tasks are proposed, which enable useful context predictions even for dynamic applications. The overall development process as well as practical experiments with the prototype framework is illustrated by two use case scenarios—demonstrating that both high accuracy and efficiency of future context predictions are achievable and even ad hoc context predictions can be supported this way. Copyright © 2011 John Wiley & Sons, Ltd.