06.1 - Flight Dynamics and Control (Control & Modelling)
DEEP REINFORCEMENT LEARNING FOR EVTOL HOVERING CONTROL
D.S. Alarcon¹, J.H. Bidinotto¹; ¹University of São Paulo, Brazil
This work uses Deep Reinforcement Learning to control the hovering phase of an eVTOL. A 6DoF model of the Airbus Vahana is implemented in OpenAI Gym and the PPO method tested, with different hyperparamters. Then, compared with DDPG, TD3, SAC and a PID controller baseline. The result showed that the PPO was the best method for this task. This work is the first step of full eVTOL control with DRL.