Composing business processes with partially observable problem spaces in a Web services environment.

摘要

Web services have received much interest to support business-to-business or enterprise application integration, but how to combine these services optimally in a continually growing search space continues to be a challenge. This thesis research investigates composing business processes from individual services as a planning problem where a planner determines the execution order and other constraints among services in a process. When there is a large number of Web services available, it is not easy to find an execution path of Web service composition that can satisfy the given request, since the search space for such a composition problem is in general increasing exponentially. In practice, the planner has to work with a problem space that is not fully enumerable. This thesis presents a method, GA-Planner, that combines Genetic Algorithms (GAs) with symbolic AI planning to optimize composition results within an incompletely observed problem space. GA helps to navigate the search through the whole space. At each loop of GA, Web service data are queried and a new subspace is built. The planner works with the subspace and calculates a feasible solution. We test our method on a travel domain. The result is an optimized solution, though global optimization is not guaranteed.