Decision-Making Policies for Heterogeneous Autonomous Multi-Agent Systems with Safety Constraints

Authors: Ruohan Zhang, Yue Yu, Mahmoud El Chamie, Behçet Açıkmese, Dana H. Ballard

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

Reproducibility Variable Result LLM Response
Research Type Experimental We evaluate the effectiveness of proposed algorithms first using a simple MDP, and then using a dynamic traffic assignment problem. The numerical results demonstrate that safe decision-making algorithms in this paper significantly outperform other baselines.
Researcher Affiliation Academia 1Department of Computer Science, University of Texas at Austin, Austin, United States 2Department of Aeronautics and Astronautics, University of Washington, Seattle, United States
Pseudocode Yes Algorithm 1 SC-Forward Induction (SCF) ... Algorithm 2 SC-Worst Case (SCWC) ... Algorithm 3 SC-Projection (SCPRO) ... Algorithm 4 SC-Backward-Forward Induction (SCBF)
Open Source Code No The paper does not provide any specific links to source code repositories, nor does it explicitly state that the code for the methodology described is released or available.
Open Datasets No The paper describes two experimental setups: a "simple 11-state MDP" and a "Dynamic Traffic Assignment (DTA) problem" which uses a "multi-intersection environment [Hausknecht et al., 2011]" and a "discretized version of the Automatic Intersection Manager simulator [Dresner and Stone, 2008]". While the simulation environments and parameters are described or referenced, the paper does not explicitly provide a link, DOI, or repository for a fixed, publicly available dataset used for training/simulation, nor does it refer to a well-known, pre-collected benchmark dataset with direct access information.
Dataset Splits No The paper states conducting experiments with a certain number of "trials" (e.g., "100 trials", "50 trials") but does not specify any training, validation, or test dataset splits, nor does it mention cross-validation techniques.
Hardware Specification No The paper does not provide any specific details about the hardware (e.g., CPU, GPU models, memory, or cloud instances) used to run the experiments.
Software Dependencies No The paper mentions that algorithms are "based on LP and can be implemented using any standard LP solvers" but does not specify any particular software, libraries, or their version numbers (e.g., Python 3.x, specific solver names with versions) that are required to replicate the experiment.
Experiment Setup Yes We first evaluate the algorithms using a simple 11-state MDP... 100 agents are placed at home state initially... We set γ = 0.99. ... We discretized the state space as in Fig. 3... The problem MDP with γ = 0.99 is defined as: S: a state is a tuple hposition, destination, typei. A: {North, South, East, West, Stay}. ... The roads have different carrying capacities. ... Agents have two sizes: a large agent occupies 1 = 3 units of capacity, and a small one occupies 2 = 1 unit. ... The first one has 6 roads and 9 intersections... The capacity upper bound for each block is 500. ... Half of the agents are large agents and the other half are small. The planning horizon for SC algorithms is 50 time steps.