Online Safe Flight Control Method Based on Constraint Reinforcement Learning

• DOI: 10.3390/drones8090429
• Publication Date: 2024-08-26T11:53:37Z
• Authors: Jiawei Zhao; Haotian Xu; Zhaolei Wang; Tao Zhang
• Keywords: online safe flight control; meta-learning; 0208 environmental biotechnology; TL1-4050; 02 engineering and technology; constrained reinforcement learning; Motor vehicles. Aeronautics. Astronautics

UAVs are increasingly prominent in the competition for space due to their multiple characteristics, such as strong maneuverability, long flight distance, and high survivability. A new online safe control method based on constrained reinforcement learning is proposed intelligent safety of UAVs. This adopts policy optimization main framework develops a algorithm with extra budget, which introduces Lyapunov stability requirements limits rudder deflection loss ensure improves robustness controller. By efficiently interacting constructed simulation environment, law model trained. Subsequently, condition-triggered meta-learning used adjust raw ensuring successful attitude angle tracking. Simulation experimental results show that using laws perform aircraft tasks has an overall score 100 points. After introducing learning, adaptability comprehensive errors aerodynamic parameters wind improved by 21% compared offline learning. The can be learned UAVs, during flight.