An Admissible HTN Planning Heuristic

Authors: Pascal Bercher, Gregor Behnke, Daniel Höller, Susanne Biundo

IJCAI 2017 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental The evaluation further studies the heuristic both in terms of plan quality and coverage.
Researcher Affiliation Academia Pascal Bercher, Gregor Behnke, Daniel H oller, Susanne Biundo Institute of Artificial Intelligence, Ulm University, Ulm, Germany {pascal.bercher, gregor.behnke, daniel.hoeller, susanne.biundo}@uni-ulm.de
Pseudocode No The paper describes algorithms (e.g., 'We use the following algorithm to compute the cost estimates') but does not provide structured pseudocode blocks or figures labeled as such. It refers to algorithms from other papers.
Open Source Code Yes Its code will be made available online (www.uni-ulm.de/en/in/ki/panda).
Open Datasets Yes The domains include all the ones used by Elkawkagy et al. [2010; 2012] and Bercher et al. [2014b] for their evaluations, which were also used by Alford et al. [2016a]. These are the UM-Translog domain (originally designed specifically for HTN planning), the Smart Phone domain (originally designed for hybrid planning), and the Satellite and Woodworking domains (both were originally designed for the International Planning Competitions (IPCs), which were adapted to hybrid planning).
Dataset Splits No The paper describes the domains and problem instances used for evaluation but does not specify training, validation, or test set splits, nor does it mention cross-validation.
Hardware Specification Yes We used a machine with Xeon E5-2660 v3 CPUs with 2.60 GHz base frequency, a memory limit of 10 GB, and a time limit of 10 minutes (CPU time) per run.
Software Dependencies No The paper mentions using 'PANDA', 'JASPER', and 'Sym BAStar -2' planners, but it does not specify their version numbers for reproducibility.
Experiment Setup Yes We have set the time limit for runs of the planner of each translated problem to two minutes (increasing this bound does not improve the results of the planner).