Symbolic Search for Oversubscription Planning

Authors: David Speck, Michael Katz11972-11980

AAAI 2021 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental An empirical analysis shows that our symbolic approach favorably competes with explicit state-space heuristic search, the current state of the art for oversubscription planning. Finally, an empirical study on various planning domains is conducted which shows that the presented algorithm results in an optimal planner that exceeds the current state of the art in terms of overall coverage.
Researcher Affiliation Collaboration David Speck1 and Michael Katz2 1 University of Freiburg, Freiburg im Breisgau, Germany 2 IBM T. J. Watson Research Center, Yorktown Heights, NY, USA
Pseudocode Yes Algorithm 1: SYM-OSP for unit cost OSP tasks
Open Source Code Yes 3Available online: https://github.com/speckdavid/symbolic-osp
Open Datasets Yes All experiments are conducted on the oversubscription planning benchmark set, based on the the optimal track from the International Planning Competition (IPC) 1998-2014 (Katz et al. 2019b), where goal facts are replaced with utilities.
Dataset Splits No The paper mentions using the 'oversubscription planning benchmark set' and discusses variations in cost bounds (25%, 50%, 75%, 100%), but it does not specify any training, validation, or test dataset splits or methodologies like k-fold cross-validation.
Hardware Specification Yes All experiments are run on a compute cluster with nodes equipped with two Intel Xeon Gold 6242 32-core CPUs, 20 MB cache and 188 GB shared RAM running Ubuntu 18.04 LTS 64 bit.
Software Dependencies Yes running Ubuntu 18.04 LTS 64 bit. and decision diagram library CUDD (Somenzi 2015)
Experiment Setup Yes The planners are build with 64 bit and run with a time limit of 30 minutes and memory limit of 4 GB.