Standardized documentation and accessibility of AI planning domain models : value cards and a centralized registry

Abstract

AI planners use planning domain models to generate a plan. This makes the domain model an important component for solving problems correctly and for the development and testing of AI planners. There are two major challenges for understanding the domain model. One is the unavailability of documentation and standardization for communication. Another challenge is the lack of a centralized repository where domain models are stored together. The purpose of the thesis is to address these challenges. The thesis designs a standardized framework termed value card to provide supporting information about the domain model. Value cards for an initial set of domain models are developed using scientific reports accompanying the documents. They also highlight the realistic aspects of the domain model. To address the latter challenge, we develop a registry system to store the domain models and corresponding value cards. This application gives a common platform to researchers and practitioners publishing domain models and users of the domain model. By integrating value cards in the registry, domain models can be categorised based on certain characteristics like application category or model language. Results of the user survey performed on the system highlight the ease of use. Through these outputs, we improve domain model understanding and availability together with promoting fairness, transparency and standardisation in AI planning domain models.