Research On Intelligent Anti-interception Of Hypersonic Vehicle Based On Deep Reinforcement Learning

The hypersonic aircraft is a vehicle that can be used for hypersonic flight and flight Mach number,which can reach more than 5 aircraft,and in recent years,the defense of the hypersonic aircraft has been developed,and the vehicle’s anti-intercept technology is needed to ensure the threat of the vehicle.In this paper,the characteristics of the single and real-time computational difference of the anti-intercept technology of hypersonic aircraft are common,and the method of online anti-intercept method is required to study the rapid time variation characteristics of the hypersonic aircraft.Based on the strategy maneuvering method of deep strengthening learning,the threat avoidance process of the vehicle is established,and the mathematical model of the supersonic vehicle is established for the anti-interceptor process,and the threat of the no-fly zone facing the mid-flight of the supersonic vehicle is established,and the upso algorithm is used to make the flight area of the advanced aircraft,and set the range of the no-fly zone of the hypersonic vehicle,and set up the no-fly zone.The location of the flying area is separat ed by standardises,forming multiple path points,and using the upso algorithm to carry out the navigation path of the pre-able matrix point,and generate a path that can reach the target point.The navigation and environmental information are entered into the neural network and obtained the proposed output.In order to improve the output of the upso algorithm,the dijkstra algorithm is used to optimize the problem of upso algorithm.Using the dijkstra algorithm,the path generation process of the dijkstra algorithm is used by using the dijkstra algorithm,and the corresponding path planning point is generated.This paper studies the threat avoidance strategy based on deep strengthening learning.The deep reinforcement learning method is a new way to analyze and resolve the problem of anti-intercept.Using the DDPG algorithm to design the threat of the hyacoustic vehicle,the intelligent body threat model is established,and the reward function is selected.According to the performance and environmental factors of the vehicle and threat,by analyzing the constraints,the scope of the action value of the hyacoustic aircraft is determined,and the corresponding capacity space is generated,and the action space of the deep reinforcement learning is constructed,the action space will be the environment that can interact with the intelligent body.In the action space,the intelligent body is stored in the action space,and the intelligence is explored,and the threat is committed to the intelligence body.After training,the corresponding deep neural network is developed.The actual threat avoidance effect of the test network in the platform i s improved.In order to improve the speed of the algorithm,the td3 algorithm is optimized.The TD3 algorithm is optimized by cutting the double q,and the problem of the network in the acto algorithm is optimized,and the intelligent body achieves better performance after simulation.