pddl

pddl aims to be an unquestionable and complete parser for PDDL 3.1.

Install

from PyPI:

pip install pddl

from source (main branch):

pip install git+https://github.com/AI-Planning/pddl.git

or, clone the repository and install:

git clone https://github.com/AI-Planning/pddl.git
cd pddl
pip install .

Quickstart

You can use the pddl package in two ways: as a library, and as a CLI tool.

As a library

This is an example of how you can build a PDDL domain or problem programmatically:

from pddl.logic import Predicate, constants, variables
from pddl.core import Domain, Problem, Action, Requirements
from pddl.formatter import domain_to_string, problem_to_string

# set up variables and constants
x, y, z = variables("x y z", types=["type_1"])
a, b, c = constants("a b c", types=["type_1"])

# define predicates
p1 = Predicate("p1", x, y, z)
p2 = Predicate("p2", x, y)

# define actions
a1 = Action(
    "action-1",
    parameters=[x, y, z],
    precondition=p1(x, y, z) & ~p2(y, z),
    effect=p2(y, z)
)

# define the domain object.
requirements = [Requirements.STRIPS, Requirements.TYPING]
domain = Domain("my_domain",
       requirements=requirements,
       types=["type_1"],
       constants=[a, b, c],
       predicates=[p1, p2],
       actions=[a1])

print(domain_to_string(domain))

that gives:

(define (domain my_domain)
    (:requirements :strips :typing)
    (:types type_1)
    (:constants a b c)
    (:predicates (p1 ?x - type_1 ?y - type_1 ?z - type_1)  (p2 ?x - type_1 ?y - type_1))
    (:action action-1
        :parameters (?x - type_1 ?y - type_1 ?z - type_1 )
        :precondition (and (p1 ?x ?y ?z) (not (p2 ?y ?z)))
        :effect (p2 ?y ?z)
    )
)

As well as a PDDL problem:

problem = Problem(
    "problem-1",
    domain=domain,
    requirements=requirements,
    objects=[a, b, c],
    init=[p1(a, b, c), ~p2(b, c)],
    goal=p2(b, c)
)
print(problem_to_string(problem))

Output:

(define (problem problem-1)
    (:domain my_domain)
    (:requirements :strips :typing)
    (:objects a - type_1 b - type_1 c - type_1)
    (:init (not (p2 b c)) (p1 a b c))
    (:goal (p2 b c))
)

As CLI tool

The package can also be used as a CLI tool. Supported commands are: - pddl domain FILE: validate a PDDL domain file, and print it formatted. - pddl problem FILE: validate a PDDL problem file, and print it formatted.

Features

Supported PDDL 3.1 requirements:

[x] :strips
[x] :typing
[x] :negative-preconditions
[x] :disjunctive-preconditions
[x] :equality
[ ] :existential-preconditions
[ ] :universal-preconditions
[ ] :quantified-preconditions
[x] :conditional-effects
[ ] :fluents
[ ] :numeric-fluents
[x] :non-deterministic (see 6th IPC: Uncertainty Part)
[ ] :adl
[ ] :durative-actions
[ ] :duration-inequalities
[x] :derived-predicates
[ ] :timed-initial-literals
[ ] :preferences
[ ] :constraints
[ ] :action-costs

Development

If you want to contribute, here's how to set up your development environment.

Install Pipenv
Clone the repository: git clone https://github.com/AI-Planning/pddl.git && cd pddl
Install development dependencies: pipenv shell --python 3.7 && pipenv install --dev

Tests

To run tests: tox

To run only the code tests: tox -e py37

To run only the code style checks: tox -e flake8

Docs

To build the docs: mkdocs build

To view documentation in a browser: mkdocs serve and then go to http://localhost:8000

Authors

License

pddl is released under the MIT License.

Acknowledgements

The pddl project is partially supported by the ERC Advanced Grant WhiteMech (No. 834228), the EU ICT-48 2020 project TAILOR (No. 952215), the PRIN project RIPER (No. 20203FFYLK), and the JPMorgan AI Faculty Research Award "Resilience-based Generalized Planning and Strategic Reasoning".