非仿射纯反馈系统的间接自适应神经网络控制

Indirect adaptive neural networks controller for non-affine pure-feedback systems

中文关键词  非仿射非线性  自抗扰  全调节径向基函数神经网络  微分器

英文关键词  non-affine nonlinear  active disturbance rejection (ADR)  fully tuned radial basis function neural network (RBFNN)  differentiator

中文摘要

      针对非仿射纯反馈系统, 提出了一种新的设计方案. 与现有文献中方法不同, 该方案不是直接利用逼近技巧构建理想的反馈控制器. 首先通过自抗扰思想将非仿射纯反馈系统转化成含有未知控制系数以及未知非线性的仿射系统, 并且证明了可行性. 然后结合微分器和全调节径向基函数神经网络, 利用自适应反演技巧设计了自抗扰控制器, 微分器的引入避免了传统反演的计算复杂性. 最后, 从理论上证明了所设计的控制器能够保证闭环系统所有信号半全局一致有界, 并且证明了系统状态渐进收敛到零点的残集内. 仿真例子验证了算法的有效性.

英文摘要

      A novel design scheme is investigated for a class of non-affine pure-feedback systems. Being different from approaches in references, it does not directly employ approximation-based techniques to construct the ideal desired feedback control. Firstly, it is proved that the non-affine pure-feedback systems can be transformed into affine systems with unknown virtual control coefficients and unknown uncertainties based on active disturbance rejection ideal (ADR). Combining a differentiator with fully-tuned radial-basis-function neural network (RBFNN), we design an active disturbance rejection controller (ADRC) by employing the adaptive backstepping method. The explosion of complexity in traditional backstepping design is avoided by utilizing differentiator. Based on Lyapunov stability analysis, it is proved that the controller guarantees all signals of the closed-loop system to be semi-globally uniformly bounded. It is also proved that the states can asymptotically converge to an arbitrary small region around zero. The effectiveness of the proposed algorithm is validated by a simulation example.