基于神经网络的鲁棒自适应逆飞行控制  

Neural network based robust dynamic inversion flight control

作　　者：朱家强[1] 朱纪洪[1] 郭锁凤[2] 孙增圻[1] 

机构地区：[1]清华大学智能技术与系统国家重点实验室,北京100084 [2]南京航空航天大学自动化学院,江苏南京210016

出　　处：《控制理论与应用》2005年第2期182-188,共7页Control Theory & Applications

基　　金：中国博士后科学基金资助项目;国家863计划资助项目(2003AA755024).

摘　　要：提出基于在线神经网络的超机动飞行自适应动态逆鲁棒控制方法.超机动飞行的基本控制律采用非线性动态逆方法设计,对于建模误差或者控制面损伤等因素导致的不确定性逆误差采用神经网络进行自适应补偿.通过动态逆控制律简化计算和飞机控制面故障自适应修复的仿真表明,神经网络通过在线补偿逆误差,能够有效降低非线性动态逆对模型准确性的要求,增强控制系统的鲁棒性.An adaptive dynamic inversion controller design method based on neural network(NN) is proposed for super maneuverable fighter aircrafts.The basic control law is developed in nonlinear dynamic inversion;on_line learning neural networks are implemented to compensate inversion errors due to modeling error or actuator faults.Simulations of compensating simplified calculation and actuator failure are presented.The results demonstrate that through adaptively compensating inversion error by neural networks,the limitation of needing accurate mathematical model in dynamic inversion method is released and the robustness of the whole system is greatly enhanced.

关 键 词：神经网络(NN) 自适应控制 飞行控制 动态逆 超机动 

分 类 号：V249[航空宇航科学与技术—飞行器设计]