超机动导弹的神经网络控制系统

 研究了应用于具有超机动能力的导弹的非线性神经网络控制系统.利用神经网络可以任意逼近非线性函数的能力,来学习超机动导弹在大迎角状态下的高度非线性动态特性的逆动态特性,得到对参数变化及未建模动态具有较好鲁棒性的控制系统.为了改善神经网络的学习能力和学习算法的稳定性,对学习增益做了模糊化处理.数字仿真结果说明了该控制方案可以达到预期效果.

Neural-Network Control of Agile Missile

A neural network based controller of agile missile capable of high angle of attack flight is presented. Super agility is needed in close in combat. For the missile to possess super agility (high angle of attack capability) some form of alternate control is needed, vectoring the thrust (or using reaction jets) can provide this capability. With super agility requirements, missiles expected to maneuver in extreme flight conditions where there is a considerable amount of nonlinearity. The objective of this paper is to examine the use of neural network (NN) as an integral part of nonlinear dynamic inverse control. A fuzzy learning gain is adopted to enhance the stability of neural network. The NN are used to generate feedback control to force the plant toward desirable responses. A reference model is used to represent the ordered index of the closed loop system performance. A benefit in the case is that the control system tends to be more robust. A simulation is presented and the results show the perfect tracking performance.