A PDDL Based Tool for Automatic Web Service Composition

One of the motivations for research in semantic web services is to automatically compose web service operations to solve given problems. The idea of using AI planning software to this end has been suggested by several papers. The present paper follows this approach but argues that the diversity of the web service domains is best addressed by a flexible combination of complementary reasoning techniques and planning systems. We present a tool that transforms web service composition problems into AI planning problems and delegates them to the planners most suitable for the particular planning task. The tool uses PDDL, a language supported by a wide range of planning engines, as a transfer format. The present paper describes the tool and its strategies to cope with the problems of incomplete information, various types of web service indeterminism, stateful services and structurally rich goal specifications.