带推力矢量空空导弹的神经网络控制与仿真

Neural Network Control for Air-to-Air Missiles with Thrust Vectoring

对于新型空空导弹为了使导弹获得更高的机动性、敏捷性和更高的导引精度，大多采用推力矢量控制方案。因为神经网络控制对于系统非线性变化具有很强自适应能力，因而在解决带推力矢量空空导弹的控制问题时有较明显的优点。本文在给出推力矢量空空导弹数学模型的基础上，提出了两种适用于带推力矢量空空导弹的神经网络控制方案，并采用其中的双网络逆动态学习控制方法进行了自动驾驶仪设计。为进一步改善该神经网络的学习效果，还引入基于学习经验的模糊规则。数字仿真表明所提出的神经网络控制对于系统内参数非线性变化具有很强的适应性。

The advanced air-to-air missiles possess the characteristics of maneuverability, agility and accurate guidance performance by adopting thruster vector control. Because the neural network control has strong self-learning ability and adaptability to system nonlinear variations, it has significant advantages in the control of air-to-air missiles with thruster vectoring. After modeling of the air-to-air missiles with thruster vectoring, two neural network control methods for the air-to-air missiles are presented. One of them with two networks, dynamic inversion learning structure, is given to design an autopilot for the missile. In order to improve the learning ability of the presented neural network control system, fuzzy rules based on expert learning experience are introduced. Numerical simulation results are given to illustrate that the presented neural network control method possess strong adaptability to system nonlinear variations.