Georgios Papoudakis

In a multi-agent system, each agent should be able to reason about the goals, beliefs and policies of the other agents to make informed decisions. This project investigates how we can use unsupervised and self-supervised learning methods for agent modelling in multi-agent systems. The primary focus of the project is on partially-observable tasks where each agent has only access to a partial view of the other agents and the environment's state. Modelling other agents in such tasks is a crucial milestone for scaling agent modelling in real-world applications.