Real-time single-step deep reinforcement learning framework for control of Tollmien-Schlichting waves

This work presents a real-time, model-free control framework that utilizes single-step deep reinforcement learning to suppress Tollmien-Schlichting waves in a two-dimensional boundary layer. By learning an opposition-control strategy directly from sensor feedback, the controller outperforms a classical adaptive Filtered-x Least Mean Square (FXLMS) approach in both effectiveness and robustness. The results demonstrate a compact and experimentally feasible AI-based solution for real-time control of convective instabilities.