Active Goal Recognition Design

Authors: Kevin C. Gall, Wheeler Ruml, Sarah Keren

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

Reproducibility Variable Result LLM Response
Research Type Experimental To characterize the behavior of this optimal AGRD algorithm, we implemented it in C++ and ran experiments on Intel E8500 3.16 GHz CPUs. and Figure 3 plots the geometric mean of the runtime with 95% confidence intervals as a function of this upper bound.
Researcher Affiliation Academia Kevin C. Gall1 , Wheeler Ruml1 and Sarah Keren2,3 1Department of Computer Science, University of New Hampshire, USA 2School of Engineering and Applied Sciences, Harvard University, USA 3School of Computer Science and Engineering, Hebrew University of Jerusalem, Israel
Pseudocode Yes Algorithm 1: Optimal AGRD
Open Source Code No The paper states, 'we implemented it in C++', but does not provide any concrete access information (e.g., specific repository link, explicit code release statement, or code in supplementary materials) for the source code.
Open Datasets No The paper uses generated grid environments ('uniform grids' and 'Room grids') for its experiments, but it does not specify a publicly available or open dataset with concrete access information (e.g., specific link, DOI, repository name, formal citation).
Dataset Splits No The paper focuses on evaluating algorithm runtime on generated grid environments and does not describe training, validation, or test dataset splits typically used for model training and evaluation.
Hardware Specification Yes To characterize the behavior of this optimal AGRD algorithm, we implemented it in C++ and ran experiments on Intel E8500 3.16 GHz CPUs.
Software Dependencies No The paper states 'we implemented it in C++', but does not provide specific software dependency details like library or solver names with version numbers needed to replicate the experiment.
Experiment Setup Yes We use grid pathfinding with 4-way movement as a testbed for our experiments. We used two patterns of obstacles. In uniform grids, each cell had a 0.2 probability of being blocked. Room grids contain walls with randomly placed openings, scaled down from those in the Moving AI repository [Sturtevant, 2012]. From among unblocked cells, the subject and observer s start locations, possible goals, and cells the observer could block were chosen randomly. The observer follows the same movement rules as the subject, and they can occupy the same cell. To block a cell, the observer must be adjacent to it. Instances were generated with 2, 3, or 4 goals...