Moose is a generalised planner that employs goal regression for synthesising programs represented as first-order decision lists. The programs can be executed as policies or for pruning the state space during search. The corresponding reference is
@inproceedings{chen-etal-aaai2026,
author = {Dillon Z. Chen and Till Hofmann and Toryn Q. Klassen and Sheila A. McIlraith},
booktitle = {Proceedings of the 40th AAAI Conference on Artificial Intelligence},
title = {Satisficing and Optimal Generalised Planning via Goal Regression},
year = {2026},
}
apptainer build moose.sif Apptainer
[Optional] If Moose is used for numeric planning, Python2 must be installed first. This is because Moose depends on Numeric Downward for generating training plans, which in turn depends on Python2.
wget https://www.python.org/ftp/python/2.7.9/Python-2.7.9.tgz
tar xzf Python-2.7.9.tgz
cd Python-2.7.9
./configure --enable-optimizations
make altinstall
ln -s /usr/local/bin/python2.7 /usr/local/bin/python2
cd -
Create a virtual environment and then install requirements.
python3 -m venv .venv
source .venv/bin/activate
sh install.sh
Policies are trained and used for planning as follows.
# Training
./moose.sif train benchmarks/<DOMAIN>/domain.pddl
# Planning via policy execution
./moose.sif plan <DOMAIN>.model benchmarks/<DOMAIN>/domain.pddl benchmarks/<DOMAIN>/testing/<PROBLEM>.pddl
# Planning via optimal search
./moose.sif plan <DOMAIN>.model benchmarks/<DOMAIN>/domain.pddl benchmarks/<DOMAIN>/testing/<PROBLEM>.pddl --search symk
Call ./moose.sif <MODE> -h for more instructions. If installation was done manually, replace ./moose.sif <MODE> with python3 <MODE>.py.
# Training
./moose.sif train benchmarks/ferry/domain.pddl
# Planning via policy execution
./moose.sif plan ferry.model benchmarks/ferry/domain.pddl benchmarks/ferry/testing/p2_30.pddl
# Planning via optimal search
./moose.sif plan ferry.model benchmarks/ferry/domain.pddl benchmarks/ferry/testing/p0_30.pddl --search symk