Reentry Trajectory Optimization And Guidance Method Based On Convex Optimization And Reinforcement Learning

Because of its fast flying speed,long range,and powerful maneuver penetration capability,the lift gliding aircraft will be of great strategic significance in future wars.The re-entry trajectory optimization and guidance of a lift-glide vehicle is a research focus in the aerospace field,and it is also a difficult point.This is mainly due to the complex constraints faced by the same vehicle when re-entering the atmosphere and its own nonlinear kinematics model.This paper mainly focuses on the re-entry trajectory optimization and guidance of the lift-glide vehicle.It is divided into the following aspects to carry out research work.In this paper,we first establish an appropriate coordinate system,and use Newton’s second law to model and analyze the re-entry motion of lift glider,and get the particle kinematics equation of the aircraft motion.At the same time,for the sound velocity and density of the atmospheric environment where the vehicle is located Perform mathematical analysis on other characteristics and obtain mathematical models to prepare for the study of aircraft reentry trajectory optimization and guidanceThe traditional reentry trajectory optimization problem often needs a lot of computation time,and it is difficult to optimize the attack angle and the roll angle simultaneously under complex constraints.In this paper,a reentry trajectory optimization method based on convex optimization is designed.The kinematics equations,process constraints,no fly zone and waypoint constraints and performance indicators of the reentry process are convex.The original reentry trajectory optimization problem is transformed into a second-order cone programming problem(SOCP),and then solved by convex optimization solver.This method can optimize the angle of attack and the angle of pitch quickly and simultaneously under multiple constraints to obtain the optimal trajectory,and can better meet all the constraints in the flight process.The simulation results show that the method is fast and accurate,and can be applied to engineering practice.In the process of the actual reentry flight,due to the self modeling and the deviation of the external environment,the aircraft can not fly accurately according to the optimized nominal trajectory.In this paper,a reentry trajectory tracking guidance method based on convex optimization is designed to ensure that the flight task can still be completed under the deviation condition.In this paper,the kinematics equations,constraints and performance indexes of the reentry guidance are convex by a series of convex methods,and the quadratic constrained quadratic programming problem(QCQP)is obtained.For the quadratic performance index of QCQP,the rotation conic constraint is introduced,which makes the QCQP problem be solved quickly.In this paper,an integrated system of trajectory optimization and guidance based on convex optimization is designed to adjust flight strategy on-line actively to deal with large deviation and temporary flight task switching.Simulation results show that the method has high precision and good robustness,and can be used in reentry guidance with complex deviation.In the process of reentry guidance,the aircraft is expected to get the guidance trajectory which can bypass the no fly zone and track the nominal trajectory well.In this paper,the strategy gradient descent depth neural network decision(DDPG)algorithm in reinforcement learning is applied to the problem of reentry guidance to deal with the problem of lateral avoidance of no fly zone.In this paper,firstly,the guidance trajectory without considering no fly zone is obtained by the guidance method based on convex optimization,and then the ddpg algorithm is used to fine tune the guidance command,so that the aircraft can bypass the no fly zone.With the development of computer technology,more artificial intelligence methods will be applied in the field of aerospace engineering.Finally,this paper summarizes the research work of the whole paper,points out the main innovation of this paper,and provides some ideas for the future research in this direction.