Process planning and management

In times of dynamically shifting markets, companies, especially those integrated in electronic supply chains, have to adapt or even restructure their business processes frequently. First, regarding the sales market of a company, changing customer needs and new offers of emerging competitors need to be considered and demand for quick reactions in the form of enhanced services and innovative products. Second, a national and international network of business partners gets more and more important in order to be able to offer best-of-breed-products and customized solutions instead of commodities. Such approaches, along with an efficient design of the supply chain, constitute distinguishing factors between competitors. Third, the market of suppliers for a company, especially of IT suppliers, is expanding. Only ten years ago, the share of proprietary and individual software has been considerably higher than today. In the future, it should be possible to design and modify company specific applications by composing web services provided by external software suppliers according to predefined business processes. In any of the three "markets" business processes constitute the starting point for permanent modifications along the supply chain.

In order to counter the above mentioned requirements, traditional techniques and tools for process modeling and optimization seem to be insufficient or inadequate to some extent. Reasons are that these requirements demand for an increased adaptability of existing process models in terms of dynamics and flexibility. Traditional methods for process (re)design imply a significant degree of manual work, caused for instance by modeling techniques or by high demand for communication and clarification because of different terminologies. According to the ongoing research in the area of Semantic Business Process Management (SBPM), a higher degree of automation concerning the use of process models can contribute to the solution. More precisely, we envision the (partly) automated design and adaptation of process models. As this task can be regarded as a kind of planning problem, we speak of an automatic planning of process models that are composed of individual actions (PA). The basis for SBPM and in particular the planning of process models constitute Semantic Web standards like the Web Ontology Language (OWL) that enable a semantically enriched description of process models and their elements. These standards have already been adopted for Web Service Composition. Yet, in contrast to Web Service Composition approaches, the planning of process models is conducted on a conceptual level independently from the underlying technology. The composition of web services is accomplished for a specific problem, that means a number of web services are arranged together to deliver one distinct and previously defined output. For the planning of process models, however, we abstract from one specific process execution and its implementation. Thus, process models are initially technology independent and may partially be realized by different combinations of available web services (from different providers) that may be chosen afterwards by means of economic aspects like cost and risk. This two-step approach is advantageous as it increases flexibility and bears optimization potential. It is moreover reasonable to assume that the step from descriptions of process models and process actions to concrete implementations using web services is relatively small. Doubtlessly the conceptual and technological basis for an automatic planning of process models is already present in the areas of AI planning and (Semantic) Web Service Composition. Several approaches in both domains indeed exist, but the planning problem and particularly the planning of process models is far from being solved. For instance, Web Service Composition approaches are often restricted to a manageable number of sequentially executed web services and do not focus complex compositions, including essential control structures like parallel split. Therefore, we are currently working on an approach called SEMPA (SEMantic-based Planning Approach) that supports and enables the (partly) automated planning of process models (set of feasible solutions) using semantically described process actions. We speak of a partly automated approach, as the planned process models are considered as proposals that afterwards need to be assessed from an expert regarding business aspects. The following figure clarifies the problem setting and specifies some terms and concepts used. Based on a semantic description of the domain (in terms of the OWL ontology) and the description of actions which are available in a process library, the planning algorithm is supposed to find process models for a given problem definition. The problem definition includes, besides the process library and the ontology, an initial state that characterizes the overall inputs of the process and one to many goals. Goals are representing process outputs.