Composing Nested Web Processes Using Hierarchical Semi-Markov Decision Processes

Contents

Many methods proposed for automated composition of Web processes use classical AI planning approaches such as rulebased planning, PDDL, and HTN planning. Web processes generated by classical planning methods suffer from the assumption of deterministic behavior of Web services and do not take into account fundamental QoS issues like service reliability and response time. In this paper, we propose a new model and method based on hierarchical semi-Markov decision processes (H-SMDPs) to address these concerns and handle the composition problem in a more natural and realistic way. We also demonstrate that HSMDP composition outperforms the HTN planning approach in terms of both the optimality of the plan and the robustness to dynamic nature of Web processes.