Self-learning how to self-propel at low Reynolds numbers

Synthetic swimmers capable of moving at the microscopic scale offer exciting opportunities for biomedical applications. However, existing swimmers are typically designed with fixed swimming gaits for a particular type of medium or environmental conditions. Their locomotion performance therefore may not be robust to environmental changes. In this talk, we will present a machine learning framework that enables the design of a new class of self-learning, adaptive (or "smart") swimmers at low Reynolds numbers. Unlike the traditional approach, we do not specify the swimming gaits a priori but allow the swimmer to self-learn swimming policies via machine learning algorithms. This framework can serve as a useful tool for designing smart micro-robots with robust locomotive capabilities for biomedical applications.