基于神经网络的飞行器再入制导研究  

Study of reentry guidance based on neural network

作　　者：曾志峰[1] 汤一华[1] 徐敏[1] 陈士橹[1] 

机构地区：[1]西北工业大学航天学院,陕西西安710072

出　　处：《飞行力学》2011年第3期64-67,共4页Flight Dynamics

摘　　要：论述了神经网络理论在飞行器再入制导方面的应用。在分析各方法优缺点的基础上,提出了一种基于广义回归神经网络(GRNN)模型的再入制导方法。神经网络通过一组选定的轨迹样本进行有导师训练,训练好之后作为控制指令生成器,输入为再入过程中飞行器的状态参数,输出为倾侧角控制量,迎角控制量则由迎角剖面给定。仿真结果验证了该方法在再入飞行器存在再入初态误差和损伤情况下的可行性及鲁棒性,具有良好的工程应用前景。The applications of neural network to the reentry guidance of reentry vehicle are discussed. Based on this, a Generalized Regression Neural Network （GRNN） reentry guidance method is advanced. In this method, the GRNN is trained by using a group of reentry trajectory samples under the learning with a teacher method. After that, it is used as a controller to get bank angle in real time by inputting the time-varying states of the reentry vehi- cle, and meanwhile the angle of attack is scheduled with the vehicle velocity. Numerical simulation results show the feasibility and robustness of the GRNN reentry guidance method with initial error and structural damage, respectively, and therefore this method has a good prospect for engineer applications.

关 键 词：神经网络 广义回归神经网络 有导师训练 再入制导 

分 类 号：V448.235[航空宇航科学与技术—飞行器设计]