Algorithmic Planning of Process Models - A Semantic-Based Approach

Diplomarbeit, Februar 2007

Currently the design and modelling of process models is done manually. Process changes that are necessary due to dynamic market environments are therefore expensive in terms of cost and time. SEMPA, an algorithm that is capable of automatically creating process activities from given sets of semantically annotated actions, is developed and described in this thesis.

In this context process models can be regarded as conditional plans including sequences of actions as well as more complex control structures such as parallel split, synchronization, exclusive choice, simple merge, multi-choice and synchronizing merge. In contrast to other approaches in the domain of AI planning and web service composition, the algorithm does not only build conditional plans, but also recognizes all of the afore mentioned control structures. In a subsequent step, those resutling process activities may serve as a basis for concrete implementations in the form of composite web services.