ESPResSo Summer School: Dr. Samuel Tovey

Reinforcement learning for intelligent active matter

Reinforcement learning is seeing a resurgence, largely due to its recent application in the fine-tuning of large language models. The method has, however, been used over the last two decades to train robots, game-playing computers, and even algorithms for generating molecules, to great success. Recently, attention has slowly shifted towards the application of reinforcement learning in micro-scale active matter, particularly in training agents in simulations to perform interesting tasks. In this talk, we will delve into the foundations of reinforcement learning, examining its evolution and the methods that underpin it, seeing that they extend well beyond the field of artificial intelligence. In the second part of the talk, the research being conducted in Stuttgart using reinforcement learning will be presented and discussed, outlining the current state of the art and introducing avenues for future development.