On Using Action Inheritance and Modularity in PDDL Domain Modelling
Keywords:Knowledge representation for planning and scheduling, Applications and case studies of plannin and scheduling techniques
AbstractThe PDDL modelling problem is known to be challenging, time consuming and error prone. This has led researchers to investigate methods of supporting the modelling process. One particular avenue is to adapt tools and techniques that have proven useful in software engineering to support the modelling process. We observe that concepts, such as inheritance and modularity have not been fully explored in the context of modelling PDDL planning models. Within software engineering these concepts help to organise and provide structure to code, which can make it easier to read, debug, and reuse code. In this work we consider inheritance and modularity and their use in PDDL action descriptions, and how these can have a similar impact on the PDDL modelling process. We define an extension to PDDL and develop appropriate tools to compile models using these extensions, both directly from the command line and through the Visual Studio Code PDDL extension. We report on our use of inheritance and modularity when modelling a planning model for a companion robot scenario. We also discuss the benefits of exploiting the inheritance hierarchy in other modules within our robot system.
How to Cite
Lindsay, A. (2023). On Using Action Inheritance and Modularity in PDDL Domain Modelling. Proceedings of the International Conference on Automated Planning and Scheduling, 33(1), 259-267. https://doi.org/10.1609/icaps.v33i1.27203