Dienstekomposition in intelligenten Umgebungen basierend auf KI-Planung

The aim of research in the field of smart environments is to explore techniques which allow to build dynamic ensembles of devices in a fashion that users can interact with them in an unobtrusive and natural way. These devices are likely to be heterogeneous and offer different services. Based on user intentions and the capabilities of the devices, the challenge is to provide more advanced services. Therefore often cooperation is required which is based on the composition of services. AI planning is one possibility to realize service composition. For unobtrusive service composition in smart environments fast and reliable implementations of planning algorithms are required. Experimentally the behavior of several planners were investigated. It concludes that no silver bullet exists. Different planners show different strengths in terms of efficiency as well as functionality. The analysis of the experiments led to different strategies. Among them the exchange strategy was best suited to shorten runtime of the planners and to increase the number of found solution as well. In addition the possibilities of modelling service composition problems was evaluated based on several szenarios from literature, what led to a guideline for distributed modelling of service descriptions. Based on the experiences from the runtime experiments and the modeling a composer is introduced which uses different AI planners to compose services in smart environments. The composer can be seen as an enhanced meta planner that leverages SOA technologies.