IEEE Trans Cybern. 2022 Jul;52(7):5654-5667. doi: 10.1109/TCYB.2020.3037096. Epub 2022 Jul 4.
This article concentrates on the adaptive neural control for switched nonlinear systems interconnected with unmodeled dynamics. The investigated model consists of two dynamic processes, namely, the x -system and the unmodeled z -dynamics. In this article, we focus on a scenario that the unmodeled z -dynamics do not contain input-to-state practically stable (ISpS) modes, that is, all modes are not ISpS (non-ISpS). First, we design an adaptive neural controller such that each mode of the closed-loop x -system is ISpS with respect to the state of dynamic uncertainties. Then, fast average dwell time (fast ADT) and slow average dwell time (slow ADT) are simultaneously used to limit the switching law. In this way, both the closed-loop x -system and the unmodeled z -dynamics are ISpS under switching. By assigning the ISpS gains with small-gain theorem, we can guarantee that the whole closed-loop system is semiglobal uniformly ultimately bounded (SGUUB), and meanwhile, the system output is steered to a small region of zero. Finally, simulation examples are used to verify the effectiveness of the proposed control scheme.
本文专注于具有未建模动态的切换非线性系统的自适应神经控制。所研究的模型由两个动态过程组成，即 x 系统和未建模的 z 动力学。在本文中，我们关注的是一个场景，即未建模的 z 动力学不包含输入到状态实际稳定（ISpS）模式，也就是说，所有模式都不是 ISpS（非 ISpS）。首先，我们设计了一个自适应神经控制器，使得闭环 x 系统的每个模式相对于动态不确定性的状态都是 ISpS。然后，快速平均驻留时间（fast ADT）和慢速平均驻留时间（slow ADT）同时用于限制切换律。这样，在切换下，闭环 x 系统和未建模的 z 动力学都是 ISpS。通过使用小增益定理分配 ISpS 增益，我们可以保证整个闭环系统是半全局一致最终有界（SGUUB），同时，系统输出被引导到零的小区域。最后，使用仿真示例验证了所提出的控制方案的有效性。
