pddl-skills

Automated PDDL planning generates plans using classical planners and validates them with a SequentialPlanValidator to ensure correct goal achievement. This skill loads PDDL domains and problems via unified_planning, synthesizes plans with OneshotPlanner, and validates the results.

Yes, unified_planning supports standard PDDL parsing to load domains and problems into a unified planning problem object. This skill uses it to parse PDDL files for classical planning benchmarks and real-world planning tasks.

Saving and serializing PDDL plans to disk allows you to reuse them for audit trails and reproducible experiments. This skill validates plans against goals and provides save functionality to store the sequential plans generated.

Yes, this skill is suitable for classical planning benchmarks where domain and problem loading, plan synthesis, and plan verification are required. It uses OneshotPlanner to generate sequential plans and validates them against the specified goals.

PDDL plan validation checks if the generated sequential plan achieves the defined goals in the problem file. If the plan contains invalid actions or fails to satisfy the goal state, the SequentialPlanValidator will reject it, indicating an issue with the plan synthesis.

Yes, unified_planning is a required dependency to load and parse PDDL domains and problems into a unified planning problem object. It provides the standard PDDL parsing and OneshotPlanner functionality needed for plan generation and validation.