IEEE Trans Cybern. 2017 Oct;47(10):3160-3171. doi: 10.1109/TCYB.2016.2647626. Epub 2017 Jan 12.
In this paper, a new adaptive approximation-based tracking controller design approach is developed for a class of uncertain nonlinear switched lower-triangular systems with an output constraint using neural networks (NNs). By introducing a novel barrier Lyapunov function (BLF), the constrained switched system is first transformed into a new system without any constraint, which means the control objectives of the both systems are equivalent. Then command filter technique is applied to solve the so-called "explosion of complexity" problem in traditional backstepping procedure, and radial basis function NNs are directly employed to model the unknown nonlinear functions. The designed controller ensures that all the closed-loop variables are ultimately boundedness, while the output limit is not transgressed and the output tracking error can be reduced arbitrarily small. Furthermore, the use of an asymmetric BLF is also explored to handle the case of asymmetric output constraint as a generalization result. Finally, the control performance of the presented control schemes is illustrated via two examples.
本文针对一类具有输出约束的不确定非线性切换下三角系统，提出了一种基于神经网络的自适应逼近跟踪控制器设计方法。通过引入一种新的障碍李雅普诺夫函数（BLF），首先将受约束的切换系统转换为一个新的无约束系统，这意味着两个系统的控制目标是等效的。然后应用命令滤波技术解决传统回溯法中所谓的“复杂性爆炸”问题，并直接使用径向基函数神经网络来建模未知的非线性函数。所设计的控制器确保所有闭环变量最终都是有界的，同时不会超过输出限制，并能将输出跟踪误差任意减小。此外，还探索了使用非对称 BLF 来处理非对称输出约束的情况，作为推广结果。最后，通过两个实例说明了所提出的控制方案的控制性能。
