AFLOW: An Automated Web Service Composition System Based On Hierarchical Task Networking Planning

With the growing maturity of web services'technology, more and more services have been deployed in the network. Meanwhile as the users'requirements keep changing and become more complex, a single simple web service is far beyond to adapt and meet the needs of practical applications. How to effectively compose web services automatically to accomplish complicated functionalities attracts more and more researchers'attentions. In this paper, the Hierarchical Task Network (HTN) planning is utilized for the research of web service composition based on the Artificial Intelligence (AI) Planning. The goal is to find a way to enable the planner to automatically compose existing services according to the users'requirements and integrate the workflow technology which makes the planning results automatically deployed and implemented to complete the users'tasks.To achieve these research objectives, our research and related works are as follows:1. This paper analyzes the existing Web Service composition methods comprehensively and describes in detail about the two main composition methods—AI Planning and workflow. Finally, brief introductions for AI Planning and relevant technologies specifications are given, elaborating the background about HTN Planning more specifically.2. According to the characteristic of web services—the number of services is large, but the scale of service type is small, this paper provides a separating model based on the multi-types based on Ontology Web Language for Services (OWL-S).3. Based on the separating model based on the multi-types, this paper provides a two-step composition method which combines the AI Planning and workflow. It uses HTN planning to construct an abstract sequence of service types and then binds service instances to it by using the matchmaking algorithm based on greedy strategy.4. Based above research, a service composition system based on the HTN Planning—is designed and a demo is implemented. In the final, we verify the feasibility of AFLOW through an example—Online Flower Order.