Surrogate-Based Deep Reinforcement Learning for Active Flow Control on Slender Bodies

Slender bodies are a fundamental element in the design of aerospace vehicles for a wide range of applications. This study investigates the application of active flow control to manage asymmetric vortex-induced side forces on slender bodies at high angles of attack. Through a combination of computational fluid dynamics (CFD) simulations and deep reinforcement learning (DRL), a robust framework is developed to optimize flow control strategies. First, a detailed analysis of the flowfield around a slender body with a protuberance is presented, demonstrating how active blowing can significantly alter the flow dynamics, reducing asymmetry and improving aerodynamic performance. In further investigations, DRL is utilized to determine the optimal blowing rates required to minimize side force, with Gaussian-process-based surrogate models employed to reduce the computational demands of direct CFD simulations. These models offer a practical alternative for real-time optimization. The results validate the effectiveness of the proposed control strategies and highlight their potential impact on the design and operation of aerospace vehicles. In a simulated trajectory, side force reduction exceeded 90%, demonstrating the potential of the approach. This research contributes to the broader field of aerodynamic control, providing insights and methodologies that can be applied to enhance stability and performance in complex flight environments.