Lifted Fact-Alternating Mutex Groups and Pruned Grounding of Classical Planning Problems

Authors: Daniel Fišer9835-9842

AAAI 2020 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental The grounding of PDDL tasks and the inference of lifted fam-groups was implemented1 in C and experimentally evaluated on a cluster of computing nodes with Intel Xeon Scalable Gold 6146 processors. We implemented both Helmert s original algorithm, referred to as H, and our improved algorithm, referred to as H+.
Researcher Affiliation Academia Daniel Fiˇser Czech Technical University in Prague, Faculty of Electrical Engineering, Prague, Czech Republic danfis@danfis.cz
Pseudocode Yes Algorithm 1: Inference algorithm.
Open Source Code Yes 1https://gitlab.com/danfis/cpddl.git, branch aaai20
Open Datasets Yes We used domains from all IPCs from 2006 to 2018.
Dataset Splits No The paper discusses using “domains from all IPCs from 2006 to 2018” but does not specify explicit training, validation, or testing dataset splits in terms of percentages or counts for model reproduction.
Hardware Specification Yes experimentally evaluated on a cluster of computing nodes with Intel Xeon Scalable Gold 6146 processors.
Software Dependencies No The paper mentions implementation “in C” and provides a link to its git repository (cpddl), but it does not specify version numbers for ancillary software components, libraries, or solvers.
Experiment Setup Yes Both H and H+ ran with the limit on the number of considered invariant candidates set to 10 000. We used domains from all IPCs from 2006 to 2018. We compared the inferred (ground) fam-groups in terms of the mutex group cover number... Lastly, we measured a coverage with the Fast Downward planner (FD) (Helmert 2006)... and set the time limit to 30 minutes and the memory limit to 8 GB.