摘要
将一种直接基于非线性模型的带神经网络补偿信号的逆系统方法用于具有强非线性和受不确定扰动干扰的飞机自动着陆控制,并对神经网络补偿的方式进行了改进。采用神经网络补偿动态逆反馈线性化后伪系统的逆误差,使得非线性系统在参数受到摄动或外部扰动的情况下仍能保持良好的控制效果。可以证明该方法在理论上是收敛的,对于任意的状态初值和给定的期望输出信号,能控制到指定的精度。神经网络的权值是局部收敛的,同时大量仿真表明,经过较少的神经网络离线训练,即能够获得很好的控制效果。最后通过飞机着陆下滑段的仿真验证表明,该方法具有强的鲁棒性和良好的跟踪精度。
A neural network compensated inverse system method was proposed to control aircraft＇s automatic landing with strong nonlinearity and unknown perturbation, which was directly based on nonlinear model. The compensation mode of neural network was improved. The inversion error of pseudo linear systems was compensated by a neural network after dynamic inversion feedback linearization, thus good control performance was held by the nonlinear system with unprecise parameters and outside perturbation. The convergence of the method was proved theoretically. The given control precision can be satisfied with arbitrary initial state and given preference command. The neural networks＇ weights are locally converged, and it is confirmed by simulation that the method has a good control performance with little offline training of neural network. At last, simulation of aircraft＇s auto - landing in gliding mode shows that the method has strong robustness and good tracking precision.
出处
《电光与控制》
北大核心
2005年第6期20-24,共5页
Electronics Optics & Control
关键词
逆系统
神经网络
鲁棒性
飞行控制
非线性系统
inverse system
neural networks
robustness
flight control
nonlinear system
