四旋翼飞行器的RBF神经网络鲁棒自适应控制

RBF neural network robust adaptive control of quadrotor aircraft

针对具有模型不确定性和有界外部扰动的四旋翼飞行器,提出了一种基于径向基函数神经网络的鲁棒自适应全局控制方法(RRAC)。所提方法结合了神经网络控制对未知非线性的强拟合能力和鲁棒控制的全局稳定性,解决了神经网络控制仅能实现半全局一致最终有界的问题,实现了控制精度和鲁棒性的双重提升。所设计的控制器由在近似域内工作的神经网络控制器和在近似域外工作的鲁棒控制器组成。引入一种新型切换函数来实现两者之间的平滑切换,以保证闭环系统的所有信号是全局一致最终有界的。利用Lyapunov函数和Barbalat引理严格证明了非线性四旋翼飞行器系统的稳定性。仿真表明,所设计的控制器在模型不确定性和有界外部扰动下对参考轨迹依旧保持良好的跟踪性能,且跟踪误差趋近于零。

The paper presents a robust adaptive global control method based on radial basis function(RBF)neural network for quadrotors with model uncertainties and bounded external disturbances.The method combines the strong fitting ability of neural network control to unknown nonlinearities and the global stability of robust control,which solves the problem that neural network control is only semi-globally uniformity ultimately bounded,and achieves the double improvement of control accuracy and robustness.A robust controller that operates outside of the approximation domain and a neural network controller that operates within it make up the planned controller.A smooth switching function is introduced to achieve smooth switching between the two to ensure that all signals of the closed-loop system are globally uniform and ultimately bounded.Using the Lyapunov function and Barbalat's lemma,the stability of the nonlinear quadrotor aircraft system is strictly proved.Under model uncertainty and constrained external disturbance,simulations demonstrate that the suggested controller still maintains a good tracking performance for the reference trajectory,and the tracking error approaches zero.