Code, Benchmarks and Experiment Data for the ICAPS 2024 HSDIP Workshop Paper "Hitting Set Heuristics for Overlapping Landmarks"

This archive contains all benchmarks, code and experimental data used in the
paper "Hitting Set Heuristics for Overlapping Landmarks in Satisficing Planning"
published at the ICAPS 2024 HSDIP workshop (HSDIP 2024).

Structure:

- benchmarks: contains all benchmarks, with each domain grouped in its own
  folder. The benchmarks are copied from
  https://github.com/aibasel/downward-benchmarks.

- code: The implementation is based on Fast Downward 23.06 and provides the
  implementation of two hitting set heuristics (including the admissible 
  approximation for the greedy hitting set heuristic), as well as landmark 
  factories used for allowing overlaps between landmarks.

- experiments: Contains the parsed output of all performed experiments as well
  as scripts for extracting data presented in the paper. The scrips require
  lab (https://github.com/aibasel/lab) to run.
  - experiment-data: contains the parsed data as well as html reports
  summarizing it.
    - disjunction-size: An experiment analyzing the impact of varying maximal
    disjunction size on the number of landmarks.
    - first-iteration: The experiment on which subsection "Simplified
    First Iteration" is based on.
    - lama-like:  The experiment on which subsection "Comparison against LAMA"
    is based on.
    - optimal: The experiment on which subsection "Optimal Planning" is based
    on.
  - paper-data-extraction: helper scripts to extract the data which we report
  in the tables and figures.
    - figure1: extracts the points for the relative scatter plot in Figure 1.
    - table1: extracts coverage results for Table 1. It creates two html
    reports, one which summarizes over all domains (report.html), and one which
    only considers domains without derived variables (no-axioms.html).
    - table2: extracts coverage and IPC quality score results for Table 2. As
    for table1, two reports are created. The file "upper_bounds.json" contains
    data accessed from the planning.domains API (https://api.planning.domains/)
    on April 5, 2024.