Hypersonic Vehicle Control Based on Deep Reinforcement Learning

The hypersonic vehicle is an important combat force in the future battlefield. At present, traditional guidance and control need to be designed prior to achieving the operational mission, lacking the ability of independent decision-making and rapid response, and not able to adapt to the development needs of complex battlefield situations in the future. With the rapid development of artificial intelligence, the decision-making ability of deep reinforcement learning has been applied in many aspects. In this paper, for the decision-making problem of autonomous flight maneuvering control of hypersonic aircraft, the deep deterministic policy gradient algorithm is used to design an autonomous flight maneuvering control decision-making algorithm to achieve the trajectory planning for the vertical climb and cruise tasks of the aircraft. Through the simulation test, the autonomous flight from the random initial position to the target position is realized, which proves that the training results have certain generalization. In the end stage, the longitudinal climbing section is extended to three-dimensional space, and the training simulation is carried out, showing the feasibility of the algorithm in the actual situation.