Enhancing Planning Based Automated Service Composition Models and Techniques

Service-oriented architecture (SOA) has been widely adopted by government and industry to enable rapid systems development and deployment via composing existing services. To further reduce manual efforts in service composition, planning techniques are used to automate the service composition process. However, some gaps still exist in automated service composition research. First, real world systems are complex and need to consider multiple functionalities. Existing service composition models do not support the specification of multiple functionalities and existing planning techniques cannot be used directly to generate a composite service with multiple functionalities. Secondly, a lot of work exists for improving the performance of planners for automated service composition, but none of them consider the scalability problem due to the number of services. With the growing adoption of SOA and open source development, more and more concrete services are becoming available, which makes the scalability issue highly pressing. Thirdly, existing service models are based on software services, while physical services have quite different characteristics. Though some works consider modeling physical services, they are still confined to the same issues of the software services. When considering automated service composition, these models are insufficient.

In this dissertation, the three issues in automated service composition are thoroughly investigated and methods for coping with them are developed. For the first issue, we extend existing service models to support multi-functionality specification and develop planning techniques to facilitate service composition for multi-functionality systems. To cope with the second issue, we develop an approach that integrates service clustering and planning techniques to improve the performance of the automated service composition process and make it scalable with the number of services. We also develop a specification model for physical services and their compositions to ensure that automated service composition can be correctly applied to cyber physical systems and Internet of things applications. Our work significantly enhances the state-of-the-art technologies in automated service composition, making it more efficient and more applicable to a wider variety of application domains.