This research investigates the design of reentry trajectory for non-thrusting gliding hypersonic reentry vehicle through reinforcement learning (DR) algorithm. Firstly, we established the three degree of freedom dynamic model to calculate the dynamic motion of the gliding hypersonic vehicle in reentry phase. The process constraints and boundary conditions when the vehicle should meet during the re-entry process were discussed. Secondly, the trajectory design problem of re-entry vehicles was studied based on the reinforcement learning framework. The trajectory design framework for reinforcement learning of state space and action space were selected, and a total reward function considering terminal position reward and process direction reward was designed. The process constraints and boundary conditions were combined to design an Actor-Critic reinforcement learning architecture for trajectory planning of gliding hypersonic re-entry vehicle through reward function. Then, the designed reinforcement learning framework model was trained using deep deterministic policy gradient (DDPG), proximal policy optimization (PPO) and twin delayed deep deterministic policy gradient (TD3) algorithms respectively. After the model was trained, three reinforcement learning models were used to design the reentry trajectory of the vehicle in a typical scenario. The performance of the reentry trajectory designed by the two algorithms was compared, and the advantages and disadvantages of the design results of the two algorithms were analysed.
Hypersonic vehicle reentry trajectory design based on reinforcement learning
01.01.2023
1470912 byte
Aufsatz (Konferenz)
Elektronische Ressource
Englisch
Steady glide trajectory planning method for hypersonic reentry vehicle
British Library Online Contents | 2015
|Hypersonic Vehicle Reentry Trajectory Optimization Based on Particle Swarm Algorithm
Springer Verlag | 2025
|Reentry Trajectory Planning of Hypersonic Gliding Vehicle Based on IFDS Algorithm
Springer Verlag | 2025
|