Learning Multi-Agent Communication through Structured Attentive Reasoning

Authors: Murtaza Rangwala, Ryan Williams

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

Reproducibility Variable Result LLM Response
Research Type Experimental We empirically demonstrate the strength of our model in cooperative and competitive multi-agent tasks, where inter-agent communication and reasoning over prior information substantially improves performance compared to baselines. We conduct benchmarks on a mixture of discrete and continuous actions spaces, with limited or zero agent vision, using both REINFORCE [13] and our improved version of TD3 [12] to compare our approaches to relevant baselines. Our empirical study demonstrates the effectiveness of our novel architecture to solve cooperative and competitive multi-agent tasks with varying team sizes and environments.
Researcher Affiliation Academia Murtaza Rangwala Virginia Tech Blacksburg, VA 24060 murtazar@vt.edu; Ryan Williams Virginia Tech Blacksburg, VA 24060 rywilli1@vt.edu
Pseudocode No The paper does not contain structured pseudocode or algorithm blocks.
Open Source Code Yes Code available at: https://github.com/caslab-vt/SARNet
Open Datasets Yes We evaluate our communication architecture on Traffic Junction and Open AI s multi-agent particle environment, [25], a two-dimensional stochastic environment consisting of agents and landmarks with cooperative tasks.
Dataset Splits No The paper describes training methodologies but does not provide specific dataset split information (percentages, sample counts, or explicit validation sets) needed to reproduce data partitioning.
Hardware Specification No The paper does not provide specific hardware details such as exact GPU/CPU models, processor types, or memory amounts used for running its experiments.
Software Dependencies No The paper mentions the use of specific algorithms and environments (e.g., LSTM, REINFORCE, TD3, Open AI's multi-agent particle environment) but does not provide specific software library names with version numbers needed to replicate the experiment.
Experiment Setup Yes Hyperparameters are noted in Appendix A.2. Refer to Appendix A.1.3 for multi-agent trajectory-TD3. Environment details and additional experiments are described in Appendix B.