Operator-Potential Heuristics for Symbolic Search

Authors: Daniel Fišer, Álvaro Torralba, Jörg Hoffmann9750-9757

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

Reproducibility Variable Result LLM Response
Research Type Experimental 8 Experimental Evaluation We implemented our search algorithm in C.3 Operators and facts are pruned with the h2 heuristic in forward and back-ward direction (Alc azar and Torralba 2015), and the translation from PDDL to FDR uses the inference of mutex groups proposed by Fiˇser (2020). We used all planning domains from the optimal track of International Planning Competitions (IPCs) from 1998 to 2018 excluding the ones containing conditional effects after translation. We merged, for each domain, all benchmark suites across different IPCs. This leaves 48 domains overall. We used a cluster of computing nodes with Intel Xeon Scalable Gold 6146 processors and CPLEX (I)LP solver v12.10. The time and memory limits were set to 30 minutes and 8 GB, respectively.
Researcher Affiliation Academia 1 Saarland University, Saarland Informatics Campus, Saarbr ucken, Germany 2 Czech Technical University in Prague, Faculty of Electrical Engineering, Czech Republic 3 Aalborg University, Denmark
Pseudocode Yes Algorithm 1: Symbolic forward A with a consistent operator-potential heuristic.
Open Source Code Yes 3https://gitlab.com/danfis/cpddl, branch aaai22-symba-op-pot
Open Datasets Yes We used all planning domains from the optimal track of International Planning Competitions (IPCs) from 1998 to 2018 excluding the ones containing conditional effects after translation.
Dataset Splits No The paper mentions using standard IPC domains but does not provide specific details on how these were split into training, validation, or test sets.
Hardware Specification Yes We used a cluster of computing nodes with Intel Xeon Scalable Gold 6146 processors and CPLEX (I)LP solver v12.10.
Software Dependencies Yes We used a cluster of computing nodes with Intel Xeon Scalable Gold 6146 processors and CPLEX (I)LP solver v12.10.
Experiment Setup Yes The time and memory limits were set to 30 minutes and 8 GB, respectively. We used a time limit of 30 seconds for applying mutexes on the goal BDD and 10 seconds for merging transition relation BDDs (Torralba et al. 2017).