Blog
With no Pokémon TCG expertise, we built a high-throughput NashPG agent and placed 36th out of 6,807 teams.
NashPG turns Independent PPO into a practical equilibrium-finding method by adding a moving KL reference policy.
This blog summarises and explains newly integrated features of the EPyMARL codebase and updates to our environments.
How to improve the zero-shot transfer capabilities of RL agents using unsupervised environment design?
This blog post describes our action policy to ensure reproducibility and transparency in our research.
How can disentangled representations improve generalisation of reinforcement learning algorithms?
We created an environment resembling the Starcraft Multi-Agent Challenge (SMAC), but without using the Starcraft II engine.
Can we improve the performance of MADDPG in grid-worlds by simply changing its discrete gradient estimator, the Gumbel-Softmax? Indeed we can!
This special issue showcases current multi-agent systems research led by university and industry groups based in the United Kingdom.
This blog post explains how to install EPyMARL, run experiments, and prototype new MARL algorithms in EPyMARL.
An overview of the tips and tricks you could use to improve performance, stability, and compute time of your RL algorithms.
How can we make people trust and understand autonomous vehicles through the use of explanations and conversations?
A list of MSc dissertations completed in our group since the group was founded.
This blog post looks at some common inaccuracies in recent multi-agent reinforcement learning research and provides recommendations.
An overview of a range of multi-agent learning environments with some of their properties, main challenges, and practical suggestions.
New online talk series on multi-agent systems research hosted by the Multi-Agent Systems special interest group at the Alan Turing Institute.
We integrate goal recognition and Monte Carlo Tree Search to create an interpretable planning and prediction system for autonomous driving.
In this post, we present two new multi-agent RL environments and a new algorithm which achieves state-of-the-art performance in these environments.
A list of multi-agent reinforcement learning surveys in reverse chronological order.
A list of single-agent reinforcement learning surveys in reverse chronological order.
New technical contributions in autonomous agents modelling other agents, and research perspectives about current challenges and future directions.
The UK Multi-Agent Systems Symposium gathered world-leading UK-based research labs from academia and industry to discuss the current landscape and future of Multi-Agent Systems research.
The symbol is an abstract representation of a process in which two agents converge to an equilibrium point via repeated interaction.
Journal special issue and workshop series presenting new technical contributions in multiagent interaction without prior coordination.
A number of prominent scientists, inventors, and entrepreneurs openly voiced their concerns regarding the potential dangers and risks of artificial intelligence.
Games allow us to study different aspects of interaction systematically and in isolation by providing well-defined environments for interactive decision making.
