Deep Reinforcement Learning for Multi-Agent Interaction - Stefano Albrecht

Speaker: Dr Stefano V. Albrecht School of Informatics, University of Edinburgh Date: 20th October 2021 Title: Deep Reinforcement Learning for Multi-Agent Interaction Abstract: Our group specialises in developing machine learning algorithms, specifically deep reinforcement learning, to enable autonomous agents to interact with other agents (e.g. cooperate, compete) to solve a given task. Questions we tackle include: How can multiple agents learn together to solve a given task in a scalable and robust way? How can a single agent learn to collaborate effectively in a team in which other agents may have diverse types and may enter/leave at any time? I will also give a glimpse into our research with UK-based company Five AI on developing safe, interpretable planning and prediction algorithms for autonomous driving in urban environments.