Giovanni Briglia

Hi :)

I am Giovanni Briglia, from November 2024 I am a PhD student at the National PhD in AI; I will engage in a partnership between the University of Pisa and the University of Modena and Reggio Emilia under the supervision of Franco Zambonelli and Stefano Mariani.

For more details, check my CV.

My focus comprehends Causality-Driven Reinforcement Learning in multi-agent system scenarios. In particular, the primary goal of my research is to create a theoretical and practical framework for incorporating causal knowledge into both model-free and model-based (deep) Reinforcement Learning (RL) algorithms, in single and multi-agent systems.

Specifically, this research seeks to answer the following critical questions:

• Can we enhance the efficiency, safety, and generalization capabilities of arbitrary model-free RL algorithms by incorporating in action selection strategies causal models of core environmental dynamics?
• Can this approach be applied to any combinations of target environment, RL algorithm, and task?
• What is the minimal causal model necessary to achieve a measurable improvement?

Overall, the idea is that (deep) RL algorithms should focus on solving the specific problem optimally, while causal models take care of efficiency, safety and generalisability.

For more details, check my PhD research proposal.