up-pddl-stream

This repository contains an integration of PDDLStream as a planning engine for the Unified Planning framework.

This integration has been developed to support the experimental evaluation of the following paper:

E. Tosello, A. Valentini, A. Micheli. A Meta-Engine Framework for Interleaved Task and Motion Planning using Topological Refinements. ECAI 2024

---

Requirements

up-pddl-stream relies on PDDLStream (follow the instruction inside the pddlstream submodule) and TAMPEST.

Usage

up-pddl-stream is integrated with the Unified Planning framework. You must register the planner engines with the Unified Planning environment:

from unified_planning.shortcuts import *

# Register the engine
env = get_environment()
env.factory.add_engine("pddlstream", "up_pddl_stream.engine", "PDDLStreamEngine")

problem = ...  # your task and motion planning problem

# Solve the problem
with OneshotPlanner(name="pddlstream") as planner:
    result = planner.solve(problem)
    print(result.plan)

Reproducing Results from ECAI 2024

To reproduce the experimental results from the ECAI 2024 paper, use the following command-line interface:

python3 test.py --domain <domain-name> --dim <dim> --d <n> --c <m> --tp <task-planner-name> --mp <motion-planner>

The possible domains are: doors, maze, rover.

The possible dimensions are: 2D, 3D. 3D setups are available only for maze and rover.

The possible set of benchmark options are:

• doors: --d [1 2 4 6 8 10] --c [0 1 2 3]

• maze: --d [1 2 3 4 5 6 7 8 9 10] --c [0 1 2 3 4 5 6 7 8 9 10]

• rover: --d [2 4 6 8 10] --c [0 1 2 3 4]

The possible task planners are: incremental, binding, focused, adaptive.

The possible motion planners are: LazyRRT, RRT.

Refer to the paper for the detailed meaning of the domain parameters.

License

up-pddl-stream is released under the Apache-2.0 License. See the LICENSE file for full details.

Contact

For questions, bug reports, or contributions, please open an issue on GitHub or contact the maintainers.