Efficient Constraint Generation for Stochastic Shortest Path Problems

Authors: Johannes Schmalz, Felipe Trevizan

AAAI 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Our experiments show that CG-i LAO ignores up to 57% of i LAO s actions and it solves problems up to 8 and 3 faster than LRTDP and i LAO .
Researcher Affiliation Academia Johannes Schmalz, Felipe Trevizan School of Computing, Australian National University johannes.schmalz@anu.edu.au, felipe.trevizan@anu.edu.au
Pseudocode Yes Algorithm 1: i LAO ... Algorithm 2: CG-i LAO
Open Source Code Yes Our code and benchmarks are available at Schmalz and Trevizan (2023).
Open Datasets Yes Triangle Tire World with Head-start (TW) In the original Triangle Tire World domain (Little, Thiebaux et al. 2007; Buffet 2008)... Probabilistic PARC Printer (PARC) (Trevizan, Thi ebaux, and Haslum 2017)
Dataset Splits No The paper mentions running "50 runs with different random seeds for each combination of planner and heuristic" and describes different problem sizes and difficulties within domains. However, it does not specify explicit training, validation, and test dataset splits with percentages or counts for reproducing data partitioning.
Hardware Specification Yes The experiments were conducted in a cluster of Intel Xeon 3.2 GHz CPUs and each run used a single CPU core.
Software Dependencies Yes The LP solver used for computing hroc was CPLEX version 20.1.
Experiment Setup Yes We use ϵ = 0.0001 and convert SSPs into dead-end free SSPs (Trevizan, Teichteil-K onigsbuch, and Thi ebaux 2017) with a penalty of D = 500 for all domains except Parc Printer variants where D = 107 due to the large cost of single actions.