Run a single planner

./plan.py images/ipc2018-opt-complementary2.img benchmarks/driverlog-strips/{domain-,}0-p01.pddl sas_plan
./plan.py images/ipc2018-lapkt-bfws.img --config poly-bfws benchmarks/driverlog-strips/{domain-,}0-p04.pddl sas_plan

Run a sequential portfolio

./run-portfolio.py --portfolio portfolio_driver/portfolios/hapori-stonesoup-opt.py --overall-time-limit 30m benchmarks/gripper-strips/{domain-,}0-p01.pddl

Please follow the naming and path convention for new portfolios. This also allows calling portfolios via an alias:

./run-portfolio.py --alias hapori-stonesoup-opt --overall-time-limit 30m benchmarks/gripper-strips/{domain-,}0-p01.pddl

run-portfolio.py detects from the portfolio filename whether the portfolio should stop after finding the first plan ("*-agl.py" and "*-opt.py") or continue with subsequent planners ("*-sat.py").

Planner data

• optimal (30m time limit): experiments/data/01-opt-planners-eval/properties-hardest.json.xz
• agile/satisficing (5m time limit): experiments/data/02-sat-planners-eval/properties-hardest.json.xz

Learn Stone Soup portfolios

cd learners/stonesoup
./learn-portfolios.sh

Adding a new learner

• Add code to learners/mylearner.
• If mylearner produces a sequential portfolio, add it to portfolio_driver/portfolios/hapori-mylearner-{agl,opt,sat} using the same format as hapori-stonesoup-opt.py.
• Otherwise, write your own script for executing your planner (under learners/mylearner). This script should use plan.py to execute planners.
• Add Apptainer.hapori-mylearner-{agl,opt,sat} files for your planner.
• Install Apptainer following the instructions in the existing Apptainer.* files.
• Build and test your Apptainer image with ./build-hapori-image.sh Apptainer.hapori-myleaner-{agl,opt,sat} hapori-mylearner-{agl,opt,sat}.img.

Build and test all Hapori Apptainer images

./build-all-hapori-images.sh

This can take a while since the Scorpion planner always uses 200s of preprocessing time.