Code and data for the KR 2023 paper: "Eliminating Redundant Actions from Plans using Classical Planning"

Brief explanation of the contents of each folder:

• action-elimination: all the code needed to run the planning approach for the minimal reduction problem.
• domains: the domains and problems used in the experiments.
• plans: the input plans used in the experiments.
• results: HTML report and properties of the experiments results generated with Downward Lab.

    
 About the benchmarks:

• The benchmarks and plans are taken from: https://gitlab.com/ctu-fee-fras/public/speeding-up-redundant-action-detection-icaps-2022.
• We only keep domains without conditional effects. 

 For the code of the MaxSAT experiments we refer you to the original repository:  https://github.com/biotomas/freelunch
 
 Key notes of our code:

• The module responsible for creating the reformulated planning tasks is src/translate/action_elim.py
• To see the options for creating a MR task from a problem and a plan, run `./action_elim.py -h`.
• Example to create a MR task with FPALs
  • ./action_elim.py -t output.sas -p sas_plan --subsequence --enhanced --enhanced-fix-point --add-pos-to-goal --reduction MR
  • This generates file action-elimination.sas with the reformulated task
• You can also solve an instance with Scorpion and run MR on the resulting plan:
  • ./fast-downward.py --search --translate --eliminate-actions --alias lama-first domain.pddl problem.pddl -- --action-elimination-options --subsequence --reduction MR --action-elimination-planner-config --search "astar(hmax())"
  • All --action-elimination-options are processed by the action_elim.py module.