Dominic Rigby

Graph Based Deep Reinforcement Learning Aided by Transformers for Multi-Agent Cooperation

Date read: 8th August 2025

arXiv link

Key Points

• Position of the agents and the set of goals are encoded into a graph
• A graph transformer processes these and attends agents to goals.
• A DQN operates of the graph embedding to choose an action for each agent (up, left, right or down in this case)