Code and Data for the IJCAI 2025 Paper "Abstraction Heuristics for Classical Planning Tasks with Conditional Effects"

Code

The file scorpion.tar.gz contains a copy of the Scorpion planning system. Please see README.md and BUILD.md for instructions on how to compile the planner. Here is the short version for building the planner and running the strongest configuration from the paper:

./build.py
./fast-downward.py SAS_TASK --transform-task preprocess-h2 --search "astar(scp_online([\\
projections(sys_scp(max_time=100, max_time_per_restart=10, max_pdb_size=2M, max_collection_size=20M, \
pattern_type=interesting_non_negative, create_complete_transition_system=true), create_complete_transition_system=true), \
cartesian(subtasks=[landmarks(combine_facts = false, order=random), goals(order=random)], max_states=infinity, max_transitions=50M, max_time=100, cache_splits=true, pick_flawed_abstract_state=FIRST_ON_SHORTEST_PATH_BACKWARD, bw_progression_flaw_fallback=true, pick_split=max_refined, tiebreak_split=min_cg), \
cartesian(subtasks=[original()], max_states=infinity, max_transitions=50M, max_time=100, cache_splits=true, pick_flawed_abstract_state=FIRST_ON_SHORTEST_PATH_BACKWARD, bw_progression_flaw_fallback=true, pick_split=max_refined, tiebreak_split=min_cg), \
cartesian(subtasks=[original()], max_states=infinity, max_transitions=50M, max_time=100, cache_splits=true, pick_flawed_abstract_state=SEQUENCE, pick_sequence_flaw=LAST_FLAW, pick_split=max_refined, tiebreak_split=min_cg) \
], saturator=perimstar, max_time=100, max_size=1M, interval=10K, orders=greedy_orders())"

The latest version of the code is available at https://github.com/jendrikseipp/scorpion. The latest version supporting conditional effects is available at https://github.com/Martin1887/scorpion/tree/conditional_effects_merged_upstream.

Scripts

The file scripts.tar.gz contains the Python code that was used for running the experiments.

Benchmarks

The file benchmarks.tar.gz contains the compressed SAS benchmarks used for the paper and a script to decompress all files.

Experiment data

The file data.tar.gz contains the raw experiment data and the basic report for the experiments in the paper.