A Graph-based Reduction in Planics Abstract Planning, Based on Partial Orders of Services (extended abstract)

The paper deals with the abstract planning problem – a stage of the Web Service Composition in the Planics framework. The planning task is viewed from the graph perspective, searching a graph built of vertices representing service and object types, and edges connecting service types to object types processed by each corresponding service. The preprocessing search identifies all the service and object types, which can potentially participate in a plan bounded by the given length. Then, the planning problem is split into subproblems, at the basis of the relation of independence between the sets of object types listed in user query as the desired result of the composition. The ontology is divided into disjoint sub-ontologies, each of which contains only the types relevant for the respective set. If there is a sub-plan for every sub-ontology, the resulting plan is composed out of these sub-plans. The presented approach uses similarily defined graphs, and makes use of planners as external tools, as in [Szr15] describing graph-based pruning of ontologies, however the aim is different. In [Szr15], there are removed all the service and object types which, for the given user query, cannot occur in any plan bound by a fixed length. The current work optimizes planning in what is left by this reduction.