Fluid dynamic control and optimization using deep reinforcement learning

This paper presents a review of recent research on applying deep reinforcement learning in fluid dynamics. Reinforcement learning is a technique in which the agent autonomously learns optimal action strategies while interacting with the environment, mimicking human learning mechanisms. Combined with artificial intelligence technology, it is providing a new direction in fluid dynamic control and optimization, which were challenging due to the nonlinear and high-dimensional characteristics of the fluid. In the section on fluid dynamic control, control strategies for drag reduction and research on controlling biological motion are reviewed. The optimization section focuses on shape optimization and automation of computational fluid dynamics. Current challenges and possible future developments are also described.