Code and Evaluation Results of the Work "Star-Topology Decoupled State-Space Search for AI Planning and Model Checking"

This archive contains the source code (decoupled-fast-downward.zip) of a decoupled-search version of the Fast Downward planner (http://www.fast-downward.org/). The code is also available in the following public repository (tag "diss-experiments"):

https://gitlab.com/dgnad/decoupled-fast-downward

The experiments were conducted using the Python lab package in version 6.2 (https://github.com/aibasel/lab). Lab produces JSON-style "properties" files, of which one is included in this archive for every type of experiment in the work (optimal/satisficing planning, proving unsolvability, factoring statistics, exhausting the reachable state space). These files contain the parsed results of all evaluation runs.

This archive finally contains the scripts used to generate the tables and plots in the work (scripts.zip). These take as input the unzipped properties files.

The benchmarks were taken from the following sources:

https://github.com/aibasel/downward-benchmarks

http://fai.cs.uni-saarland.de/downloads/unsat-benchmarks.tar.bz2

The following planners/tools were used that are not part of the code distributed in this archive:

DUAL_BFWS: https://bitbucket.org/ipc2018-classical/team20/src/ipc-2018-seq-sat/

Complementary2: https://bitbucket.org/ipc2018-classical/team32/src/ipc-2018-seq-opt/

Symbolic Fast Downward (revision "758885dc"): https://gitlab.com/atorralba/fast-downward-symbolic

SymPA: https://github.com/AI-Planning/unsolve-ipc-2016

LAMA (revision "117c17ee"): https://github.com/aibasel/downward

Cunf: https://github.com/cesaro/cunf

Mole: http://www.lsv.fr/~schwoon/tools/mole/mole-140428.tar.gz

Partition-Pruning: not publicly available; code obtained by private communication with the authors.