Autonomous Avoidance Decision Method for Aircraft Using Reinforcement Learning
There are many unpredictable threats or obstacles in the course of the mission of the aircraft.In order to solve the problem of autonomous avoidance decision of aircraft facing threat targets,firstly,a trajectory prediction algorithm based on deep Long Short-Term Memory(LSTM)neural network is proposed to predict the future trajectory of threat targets by considering the interaction between aircraft and threat targets.Secondly,the Markov decision process of evasive maneuver in the interception scenario was constructed combined with the prediction information.Then,the avoidance decision method based on progressed double delay depth deterministic strategy gradient(P-TD3)was proposed to maximize the benefits of the circumvention process to achieve intelligent autonomous avoidance decisions for the aircraft.Finally,the simulation experiments verify that the decision-making method improves the convergence speed of the network and has an 84%success rate of avoidance,which improves the probability of successful avoidance of potential threats and enhances the autonomy and safety of the aircraft.
国家科技期刊平台
NSTL
万方数据
