A* Search and Bound-Sensitive Heuristics for Oversubscription Planning

Authors: Michael Katz, Emil Keyder9813-9820

AAAI 2022 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We implemented our approach1 in the Fast Downward planner (Helmert 2006), and evaluated it on a set of publicly available OSP benchmarks (Katz et al. 2019b). ... The per-domain and overall coverage, as well as per-task node expansions comparing the two blind search approaches are shown in Table 1 and Figure 2, respectively. ... Domain-level and overall coverage comparison, as well as per-task node expansions for the various configurations and problem suites, are shown in Table 2 and Figure 3, respectively.
Researcher Affiliation Industry Michael Katz1 Emil Keyder2 1IBM Research, Yorktown Heights, NY, USA 2Invitae Corporation, San Francisco, CA, USA
Pseudocode No The paper describes algorithms mathematically and conceptually but does not include any explicit pseudocode blocks or sections labeled 'Algorithm'.
Open Source Code Yes Code available at https://github.com/emilkeyder/fd-2018-osp.
Open Datasets Yes evaluated it on a set of publicly available OSP benchmarks (Katz et al. 2019b). (Katz, M.; Keyder, E.; Pommerening, F.; and Winterer, D. 2019b. PDDL benchmarks for oversubscription planning. https://doi.org/10.5281/zenodo.2576024.)
Dataset Splits No The paper does not specify exact percentages, absolute sample counts, or refer to predefined splits for training, validation, and test datasets. It mentions using 'publicly available OSP benchmarks' but does not detail how these were split for the experiments.
Hardware Specification Yes The experiments were performed on Intel(R) Xeon(R) CPU E7-8837 @2.67GHz machines, with time and memory limits of 30min and 3.5GB.
Software Dependencies No The paper mentions 'Fast Downward planner (Helmert 2006)' as the implementation base but does not provide specific version numbers for other key software dependencies or libraries.
Experiment Setup Yes For h MS, we use exact bisimulation with an abstract state space threshold of 50k states and exact generalized label reduction (Sievers, Wehrle, and Helmert 2014), limiting the time for abstraction creation to 600 seconds.