基于神经网络的反步自适应大机动飞行控制  

Backstepping adaptive high maneuvers flight control based on neural network

作　　者：孙勇[1] 章卫国[1] 章萌[1] 

机构地区：[1]西北工业大学自动化学院,陕西西安710072

出　　处：《系统工程与电子技术》2011年第5期1113-1117,共5页Systems Engineering and Electronics

基　　金：航空科学基金(20090753008)资助课题

摘　　要：针对飞机大机动飞行时模型非线性和参数不确定性的特点,提出了一种基于全调节神经网络的反步自适应控制方法。飞机模型不确定部分由全调节径向基函数(radical basis function,RBF)神经网络在线补偿,控制律及神经网络参数自适应律由反步法回馈递推得到,并利用一种自适应参数策略的混沌粒子群算法优化控制器固定参数,改善动态性能,最后通过加权伪逆控制分配方法得到最终控制信号。仿真结果表明:在较大的模型气动参数不确定及控制增益矩阵未知时,所设计的控制律仍能理想地跟踪飞机大机动指令飞行,神经网络参数估计误差指数收敛到有界紧集,系统具有快速的收敛性和良好的鲁棒性。A backstepping adaptive control method based on fully tuned neural network is proposed in the presence of model nonlinearity and parameters uncertainty for high maneuvers flight.Parameter uncertainties are compensated for online by the fully tuned radical basis function（RBF）neural network.The control law and the adaptive law of neural network are recursively achieved through a backstepping method.The fixed parameters optimization is done using a chaotic particle swarm optimization algorithm with adaptive parameter strategy for achieving a good transient performance.The final control surface deflections are derived by a weighted pseudoinverse control allocation method.Simulation results show that precise high maneuvers can be performed with fast convergence and good robustness properties in spite of large aerodynamic parameters uncertainty and unknown control gain matrix.Moreover,the estimation errors of neural networks＇ parameters are remained in compact sets.

关 键 词：神经网络 反步自适应 大机动飞行 

分 类 号：TP273[自动化与计算机技术—检测技术与自动化装置] V249.1[自动化与计算机技术—控制科学与工程]