Towards Engineering AI Planning Functionalities as Services

Abstract

As Artificial Intelligence (AI) planning is utilised in industry to address complex real-world problems, the need for the construction of advanced and deployable AI planning systems emerges. This task, however, proves to be difficult due to a general lack of established mechanisms in AI planning for system design, interoperability, and deployment. In this context, we provide an overview of key engineering challenges that one is faced with when developing and using systems that incorporate AI planning functionalities. To help address the challenges, we propose to leverage service-orientation together with architectural patterns. In particular, we identify a set of planning functionalities, and we present an initial concept for engineering the planning functionalities as services. Having planning services would enable not only to quickly compose and deploy service-oriented planning systems but also expand them with more features as required by application domains.