Research On Terminal Guidance Method Based On Neural Network

In view of the development trend of high-altitude,high-speed and high mobility of intercepted targets in the battlefield,the final guidance method based on neural network is studied.On the basis of the traditional guidance law,the nonlinear relationship is established through neural network,and the parameters of the guidance law are adaptively adjusted to improve the performance of the traditional guidance law under the conditions of overload control,line of sight angle dependence,and angle of drop constraints.The research on BP neural network guidance method under the condition of low dependence of line of sight rotation rate,RBF neural network guidance method for overload control,and RBF neural network guidance method under the condition of angle of drop constraint are carried out in turn.Based on the traditional guidance theory,combined with the nonlinear fitting theory of neural network and adaptive control theory,the appropriate guidance law is optimized and simulated.The simulation results show that the neural network is feasible and effective in optimizing the traditional guidance law.The research results have reference significance for enhancing the defense performance of the system,improving the strike accuracy and studying the ideal trajectory.The main research contents of this paper are as follows:(1)Research on the guidance method of BP neural network under the condition of low dependence of missile visual linear angular rate.By establishing the missile target pursuit model,based on the nonlinear relationship approximation idea in the BP neural network theory,combined with the traditional proportional guidance law,the BP neural network guidance method with low dependence on the rotation rate of line of sight angle is studied.The simulation results are compared with the pure proportional guidance law,and the advantages of the BP neural network guidance law are analyzed.(2)RBF neural network guidance law for overload control is studied.Through the analysis of the relationship between overload and proportional coefficient,line of sight angle rotation rate in the guidance process of comparative example guidance law,a three-dimensional missile target relative motion model based on line of sight coordinate system is established,and the expression form of proportional guidance law acceleration in line of sight coordinate system is studied.Based on this,a RBF neural network model is designed to realize K value adaptive adjustment.By comparing the trajectory,normal overload and attack time with pure proportional guidance law,the effectiveness of RBF neural network guidance law in suppressing line-of-sight speed chatter and overload control is verified without reducing the guidance performanceResearch on guidance method of RBF neural network under angle of fall constraint.By analyzing the relative motion model of the missile and target,the relative motion equation of the missile and target is obtained,and the 3D offset proportional guidance law under the falling angle constraint is derived.According to the structural characteristics of the RBF neural network,the acceleration expression is reasonably constructed,and the algorithm formula based on the RBF neural network guidance law is derived.By comparing with the constructed 3D offset proportional guidance law,it is verified that both of them can meet the requirements of the falling angle constraint when hitting.