Reentry Guidance for Hypersonic Vehicle based on Reinforcement Learning

For the problem of reentry guidance with online no-fly zone constraints, this paper combines the improved A algorithm with DDPG and proposes a trajectory planning algorithm with high real-time property. Firstly, this paper constructs 2D environment grid map with online no-fly zone information. Secondly, this paper proposes a waypoint decision algorithm by conducting improved A algorithm on the above map. The reentry guidance of the aircraft is thus simplified to the trajectory planning subtasks between waypoints. Finally, this paper proposes an aircraft trajectory generation algorithm. The algorithm first constructs the MDP model based on the requirements of the subtasks; the builds the DDPG agent training process. Through offline training simulation, the optimal parameters of the agent net is finally obtained. Finally, the aircraft uses the above agent to output the online bank angle guidance law and the expected trajectory of the aircraft.