Planning and Acting with Non-Deterministic Events: Navigating between Safe States

Authors: Lukas Chrpa, Jakub Gemrot, Martin Pilat9802-9809

AAAI 2020 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Experimental Evaluation For our experimental evaluation, we specified 6 problems for each domain. ... The results of the experiments are shown in Table 1.
Researcher Affiliation Academia Luk aˇs Chrpa Faculty of Electrical Engineering Czech Technical University in Prague Jakub Gemrot, Martin Pil at Faculty of Mathematics and Physics Charles University
Pseudocode Yes Algorithm 1 Enhancing the PER approach by Safe State reasoning
Open Source Code Yes Our implementation and benchmark problems are available at https://github.com/martinpilat/j PDDL
Open Datasets No The paper describes problems/domains used for evaluation ('For our experimental evaluation, we specified 6 problems for each domain.'), and states 'Our implementation and benchmark problems are available at https://github.com/martinpilat/j PDDL'. However, it does not refer to a publicly available 'dataset' with a formal citation, DOI, or direct link to the data itself, but rather to custom-defined problems used in their implementation.
Dataset Splits No The paper mentions '100 independent runs' for evaluation but does not specify any training, validation, or test dataset splits in terms of percentages, sample counts, or references to predefined splits typically found in machine learning experiments.
Hardware Specification No The paper mentions using the 'LAMA planner' and 'PRP planner' for plan generation, but it does not provide any specific details about the hardware (e.g., CPU, GPU models, memory) used for running the experiments.
Software Dependencies No The paper mentions 'LAMA planner (Richter and Westphal 2010)' and 'PRP planner (Muise, Mc Ilraith, and Beck 2012)'. While these are specific tools, their exact version numbers are not provided, which is necessary for reproducibility.
Experiment Setup No The paper states 'We made 100 independent runs of each method for each problem' and mentions 'increasing unsafeness limit until we find a plan (or fail if we reach the threshold)'. However, it does not provide specific details such as hyperparameter values, optimization settings, or other concrete system-level training configurations for the planners or algorithms used.