Controlling Growing Tasks with Heterogeneous Agents

Authors: James Parker, Maria Gini

IJCAI 2016 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Empirical results are given in Robo Cup Rescue for agents with different capabilities.
Researcher Affiliation Academia University of Minnesota jparker@cs.umn.edu and gini@cs.umn.edu
Pseudocode Yes Algorithm 1: Heterogeneous LFF (HLFF), Algorithm 2: Heterogeneous RT-LFF (HRT-LFF)
Open Source Code No The paper does not provide an explicit statement about releasing source code or a link to a code repository for the described methodology.
Open Datasets No The paper describes using the Robo Cup Rescue simulator and modeling fires in clusters, along with empirical derivation of parameters. It does not refer to a publicly available dataset with specific access information like a link or formal citation.
Dataset Splits No The paper describes simulation runs and tests (e.g., "100 simulation steps in 20 different tests", "70 tests for each type of agent") but does not specify formal training/validation/test dataset splits in the typical machine learning sense.
Hardware Specification No The paper does not specify any particular hardware (e.g., GPU, CPU models) used for running the experiments.
Software Dependencies No The paper mentions using the "Robo Cup Rescue Agent Simulator" and "RMASBench simulator extension" but does not provide specific version numbers for these or any other software dependencies.
Experiment Setup Yes To estimate p, we allowed buildings to burn unhindered for 100 simulation steps in 20 different tests. and we used exponential regression to find the best fit with the data. The work rate, wa, was empirically derived for three classes of agent: agents who could use their full hose capacity, agents who could only use only 50% of it, and agents who could only use 10% of it. A fixed small number of fires were repeatedly extinguished in 70 tests for each type of agent. Each configuration ranged between 2-4 100% capability agents, 5-8 50% capability agents and 10-20 10% capability agents. Each map was seeded with 2-4 initial fires and we let 30 simulation time steps pass before agents could move.