This is a repository to store PDDL domain files that are extended with a domain-wide goal and constraints describing which instances are legal for the domain.

With those extensions the domains comply with the formalism presented in the paper C. Grundke, G. Röger, M. Helmert. Formal Representations of Classical Planning Domains. In Proceedings of the 34th International Conference on Automated Planning and Scheduling (ICAPS 2024), pp. 239-248. 2024.

The extension to PDDL types has been introduced in the paper C. Grundke, M. Helmert, G. Röger. Domain-Independent Instance Generation for Classical Planning. In Proceedings of the 22nd International Conference on Principles of Knowledge Representation and Reasoning (KR 2025). 2025.

Domain Syntax

This section only describes the syntax of the extension, i. e. domain-wide goals and legality constraints. For the general syntax of PDDL domain files please refer to the definition of PDDL2.2 (S. Edelkamp, J. Hoffmann. PDDL2.2: The Language for the Classical Part of the 4th International Planning Competition. Technical Report 195, University of Freiburg, Department of Computer Science. 2004.).

The legality constraints use the same syntax as PDDL axioms (which use the keyword :derived) but with the keyword :axiom instead. Like the existing PDDL axioms, legality constraints can define derived predicates. They must define at least one nullary derived predicate to be used as legality predicate. The truth value of this legality predicate determines if a given task is legal for the domain or not. Like any other predicate it must be mentioned in the :predicates section. Additionally, there must be a section :legality-predicate that contains only this predicate to identify it as the legality predicate.

To specify a domain-wide goal add a section :domain-goal that contains a single first-order formula. Same as PDDL's task-specific goals it is satisfied if a state satisfies the formula of the goal.

Domains in the Repository

The folders ending with ipc23-learning contain domain files that were copied from the IPC 2023 learning track benchmark repository and extended. The added legality-constraints and domain-wide goals replicate the constraints and goal information mentioned implicitly in the instance generators of the IPC 2023 learning track (which are also in the benchmark repository).

Note on childsnack-ipc23-learning

The instance generator for childsnack of the IPC 2023 learning track is not consistent with the base-case instances of the learning track. The generator produces instances where the number of bread portions, content portions, and children is the same. In the base cases however the number of bread portions, content portions, and sandwiches is the same. The constraints in childsnack-ipc23-learning/domain.pddl model the latter version, i. e. the number of sandwiches matches the number of bread portions and the number of content portions.