Learning to Ground Multi-Agent Communication with Autoencoders
Abstract
Communication requires having a common language, a lingua franca, between agents. This language could emerge via a consensus process, but it may require many generations of trial and error. Alternatively, the lingua franca can be given by the environment, where agents ground their language in representations of the observed world. We demonstrate a simple way to ground language in learned representations, which facilitates decentralized multi-agent communication and coordination. We find that a standard representation learning algorithm -- autoencoding -- is sufficient for arriving at a grounded common language. When agents broadcast these representations, they learn to understand and respond to each other's utterances and achieve surprisingly strong task performance across a variety of multi-agent communication environments.
Cite
Text
Lin et al. "Learning to Ground Multi-Agent Communication with Autoencoders." Neural Information Processing Systems, 2021.Markdown
[Lin et al. "Learning to Ground Multi-Agent Communication with Autoencoders." Neural Information Processing Systems, 2021.](https://mlanthology.org/neurips/2021/lin2021neurips-learning/)BibTeX
@inproceedings{lin2021neurips-learning,
title = {{Learning to Ground Multi-Agent Communication with Autoencoders}},
author = {Lin, Toru and Huh, Jacob and Stauffer, Christopher and Lim, Ser Nam and Isola, Phillip},
booktitle = {Neural Information Processing Systems},
year = {2021},
url = {https://mlanthology.org/neurips/2021/lin2021neurips-learning/}
}