AAS2PDDL

Generates PDDL planning problems from Asset Administration Shell (AAS) models. The tool reads a set of AASX files, extracts planning-relevant information (types, predicates, actions, objects, initial states, goals), and produces standard PDDL domain and problem files.

The expected AAS structure is defined in AAS-Planning-Metamodel.

Installation

pip install -r requirements.txt

Requires Python 3.9+, BaSyx Python SDK for AAS parsing, and the Unified Planning Framework for PDDL generation and solving.

Usage

Point the tool at a directory containing AASX files:

python src/generate_pddl.py --input examples/mps500/aasx/

This writes domain.pddl and problem.pddl to a pddl/ directory next to the input. To also solve the problem:

python src/generate_pddl.py --input examples/mps500/aasx/ --solve
python src/generate_pddl.py --input examples/mps500/aasx/ --solve --optimal

--optimal uses A* with LM-cut instead of the default LAMA-first heuristic. Other UPF-compatible planners can be selected with --planner <name>.

Design Space Exploration

If the input directory contains subdirectories (each with its own AASX files), the tool runs in DSE mode — it generates and optionally solves all variants and outputs a comparison table:

python src/generate_pddl.py --input examples/mps500/ --solve --optimal

Project Structure

src/
├── generate_pddl.py     # CLI entry point
├── aas_loader.py         # AASX loading and reference resolution
├── aas_extractor.py      # Five-phase data extraction from AAS
├── pddl_builder.py       # UPF problem construction and PDDL export
└── utils.py

examples/
├── mini_example/aasx/    # Minimal example (5 types, 7 predicates, 2 actions)
└── mps500/aasx/          # MPS500 production system (11 types, 11 predicates, 9 actions)

License

MIT