Integration of programming and decision theoretic planning for Web services composition

The high level logic language DT-Golog is proposed to deal with non-deterministic nature of Web services and dynamic nature of environments. The natural logical representation for stochastic nature of Web services composition and various composition patterns by DT-Golog are discussed. The state features of Markov decision processes (MDP) model for Web services composition are compactly expressed by DT-Golog fluents. Such a state space abstraction method avoids enumerating the whole state space. The optimal policy of Web service composition is generated by MDP planning within the constrained state space imposed by DT-Golog program. Finally, the advantage of integration of programming and planning of Web services composition by DT-Golog is illustrated with a motivating example.