Code, benchmarks and experiment data for the ICAPS 2024 paper "Transition Landmarks from Abstraction Cuts"

This archive contains all components needed to reproduce the experiments from the paper "Transition Landmarks from Abstraction Cuts" presented at the 34th International Conference on Automated Planning and Scheduling (ICAPS 2024). It includes the following elements:

• pommerening-et-al-icaps2024-benchmarks.zip contains the benchmark tasks grouped by domains. They are copied from https://github.com/aibasel/downward-benchmarks.
• pommerening-et-al-icaps2024-code.zip contains the implementation based on Fast Downward 22.06 and provides the implementation of the operator-counting constraint generators presented in the paper. (See below for more details.)
• pommerening-et-al-icaps2024-scripts.zip contains the experiment scripts used to run the experiment. They additionally require Downward Lab 8.0 (https://github.com/aibasel/lab).
• pommerening-et-al-icaps2024-data.zip contains the parsed data as well as HTML reports summarizing it.

Our experiments generally use A* search with the operator counting heuristics, and we compare configurations with different operator-counting constraint generators. The general command to run Fast Downward in this setup is ./fast-downward.py ./path/to/domain.pddl ./path/to/problem.pddl --search "astar(operatorcounting([cdb(patterns=[systematic(2)],cdb_generator=<generators>)]))" where the <generators> correspond to the techniques introduced in the paper as follows:

• tlm-set(disj): [transition_landmark_cdb(use_costs=false)] (tlm-disj in the reports) 
• tlm-set(ovlp): [transition_landmark_cdb(use_costs=true)] (tlm-overlap in the reports)
• dc-set(disj)-all: [landmark_sequence_cdb(all)] (proj-all in the reports)
• dc-set(disj)-connected: [landmark_sequence_cdb(all_connected)] (proj-all_connected in the reports)
• dc-set(disj)-max: [landmark_sequence_cdb(max_cc_only)] (proj-max_cc_only in the reports)
• dc-set(disj)-atomic: [landmark_sequence_cdb(lm_only)] (proj-lm_only in the reports)
• dc-set(disj)-atomic,max: [landmark_sequence_cdb(lm_and_max_cc)] (proj-lm_and_max_cc in the reports)

Note on license: we chose GPL v3.0 or later mainly because we consider our implementation based on Fast Downward the main contribution of this package, and Fast Downward comes with GPL v3.0.