Reinforcement learning of a multi-link swimmer

The use of machine learning techniques in the development of microscopic swimmers has drawn considerable attention in recent years. In particular, reinforcement learning has been shown useful in enabling a swimmer to learn effective propulsion strategies through its interactions with the surroundings. In this talk, we will report results on integrating reinforcement learning into the design of a multi-link swimmer. With minimal degrees of freedom, the learning algorithm identifies the locomotory gaits of the classical Purcell's swimmer. We will discuss other effective strategies identified by reinforcement learning with increased degrees of freedom.