Collective behaviour of active collodial particles via reinforcement learning

The overall aim of this project is to uncover the motivation of collective behaviour using reinforcement learning methods, which is capable to define a cost function being minimized. This approach is realized using an experimental model system of active colloidal particles which undergo a controlled swimming motion (similar to bacteria) when illuminated with a laser beam. Using a feedback-mechanism which is coupled to a neuronal network, this allows to implement reinforcement learning in such a system.

Using such methods, our goal is to understand e.g. the response of groups to predators but also to study the influence of selfish motivations on collective behaviour. Our experimental studies will be corroborated by numerical simulations, which are also performed within this project.