Trajectory Planning Based on Continuous Decision Deep Reinforcement Learning for Stratospheric Airship

Aiming at the characteristics of stratospheric airships which are greatly influenced by continuous wind fields, this paper proposes a trajectory planning method based on continuous deep reinforcement learning. Firstly, the state space, action space and reward function are designed. After that, the twin delay deep determined policy gradient algorithm based on time series is used the trajectory planning. The algorithm can be used to output actions under continuous space. The experimental results show that the algorithm is stable and efficient, and the feasibility and generalizability of the algorithm are demonstrated.