Workflows are increasingly becoming a universal means for driving and coordinating complex processes, not only in the business world but also in areas like pervasive computing. Pervasive flows run in parallel with the user's real-world actions and are synchronized using automatically collected context information about her current activities (context events). Respective workflows cannot be rigidly defined since the user needs to retain her flexibility and must not be obstructed by the workflow. However, if the order of activities is not defined until the activities are actually executed, correctly assigning the uncertain context events becomes a major challenge. We propose FlexCon - a novel event assignment approach for such human-oriented workflows that is based on hybrid workflow models and Dynamic Bayesian Networks. FlexCon exploits the dependency between context events to provide more accurate information as to which events need to be consumed by which workflow activities. Our evaluations show that FlexCon improves the event accuracy on average by 54% and the number of successful completed flows on average by 88%. Thus, FlexCon represents a major step towards unobtrusive pervasive applications.
展开▼
