<ul data-eligibleForWebStory="true"><li>Researchers propose a transformer-based reinforcement learning framework for regulating aerodynamic lift in gust sequences via pitch control, overcoming nonlinear interactions in highly disturbed flows.</li><li>The framework addresses partial observability challenges from limited surface pressure sensors by leveraging techniques like pretraining with an expert policy and task-level transfer learning.</li><li>The learned strategy surpasses proportional control performance, particularly as the number of gusts increases, and can effectively generalize to arbitrary sequences of gusts.</li><li>Quarter-chord pitching control is identified as superior for lift regulation with less control effort than mid-chord pitching control, attributed to the dominant added-mass contribution reachable via quarter-chord pitching.</li><li>The success across multiple configurations demonstrates the generalizability of the transformer-based RL framework for more computationally demanding flow control issues when coupled with acceleration techniques.</li></ul>

Attention on flow control: transformer-based reinforcement learning for lift regulation in highly disturbed flows

Discover more