基于Dueling DQN的临近空间飞行器再入轨迹规划

Near Space Vehicle Reentry Trajectory Planning Based on Dueling DQN

针对临近空间飞行器再入段禁飞区规避制导问题,构建了临近空间飞行器再入过程横侧向制导的马尔可夫决策过程(Markov decision process,MDP)模型。基于竞争深度Q网络(dueling deep Q network,Dueling DQN),设计了横侧向制导律及满足射程需求与禁飞区规避需求的再入过程奖励函数。经仿真验证,该横侧向制导律能够通过改变倾侧角符号实现禁飞区规避,并导引飞行器到达目标区域,具备较高精度,验证了方法的有效性。

Aiming at the problem of no-fly zone avoidance guidance in the reentry phase for near space vehicle,the Markov decision process(MDP)model of lateral guidance in the reentry process for near space vehicle was constructed.On the basis of dueling deep Q network(Dueling DQN),the lateral guidance law and the environmental reward feedback function to satisfy the range requirement and the no-fly zone avoidance requirement were designed.The simulation results show that the lateral guidance law can avoid the no-fly zone by changing the sign of roll angle,and guide the aircraft to the target area with high precision,which verifies the effectiveness of the method.