Suppressing Uncommanded Roll-Yaw Motion by Jet Flow Control Based on Reinforcement Learning

Abstract

The suppression of uncommanded motion of aircraft at high angles of attack (α) is a research topic of continuous concern in the aviation field. Aiming at the suppression of uncommanded roll-yaw motion of a canard aircraft at high angles of attack, an experimental method of virtual flight test based on reinforcement learning is proposed in this paper. In the virtual flight experiment, the agent was trained to control the jet actuators, so as to suppress the uncommanded roll-yaw motion. Force measurements were conducted to obtain the performance of the jet actuators in a low-speed wind tunnel. The results show that when the spanwise jet actuator and the reverse jet actuator were working on the same side, their control effects were suppressed by each other. Then, the stability augmentation control law was trained through virtual flight experiments based on a reinforcement learning algorithm (TD3), and the uncommanded motion was successfully suppressed. The time histories of the reinforcement learning agent’s action in tests were analyzed, showing that the agent can avoid the coupling relationship between two kinds of jet actuators during tests.