摘要：For existing workflow models, it is still difficult to take into account both the hierarchical relations among activities and their execution orders at the same time, as well as the dynamic uncertainties of workflow processes. Aiming at these problems, an approach to modeling flexible workflow using activity decomposition and incremental refinement is presented. First, the activities as well as their decomposition relationship and the decomposition rules are analyzed to establish an activity decomposition model, where flexible activities, with verifying constraints, are used to package uncertainties. Then temporal relations are introduced into the model to ensure the proper execution orders of activities, which are tightly integrated with the hierarchical relationship as well as different granular activities to make the model perform well. The execution mechanism of the model, such as the delivering of events and states constraints of parent-child activities, is discussed in detail, and a common algorithm for activity execution is also presented. Finally, the proposed approach is applied to the PBL learning system, and the results indicate the effectiveness of the proposal.
