Rerouting planning of suborbital debris hazard zone based on reinforcement learning DDPG

Abstract

Abstract Aiming at the debris hazard area caused by the explosion and disintegration of the suborbital vehicle, the plane convex hull divide and conquer method is used to process the physical boundary which is in accordance with the air traffic control rules. On this basis, a new method of civil aircraft rerouting planning based on reinforcement learning DDPG algorithm is proposed. According to the state-action space DDPG algorithm, the reward function is optimized to achieve the effect of reasonable routing. With the increase in training times, the reward curve gradually becomes stable and converges in a certain time step, which verifies the effectiveness of the route change planning method.