This paper explores the use of deep reinforcement learning to enable a model microswimmer to self-learn effective locomotory gaits for translation, rotation, and combined motions. The AI-powered swimmer can adaptively switch between various gaits to navigate towards target locations, mimicking the gait-switching behaviors of swimming microorganisms. The study demonstrates the robustness of the AI-advised strategy to flow perturbations and its versatility in enabling complex tasks like path tracing without explicit programming, highlighting the potential of AI-powered swimmers for applications in complex fluid environments.

Zonghao Zou, Yuexin Liu, Y.-N. Young, On Shun Pak, Alan C. H. Tsang