IEEE Trans Neural Netw Learn Syst. 2015 Dec;26(12):3097-108. doi: 10.1109/TNNLS.2015.2403712. Epub 2015 Mar 16.
High-gain observers have been extensively applied to construct output-feedback adaptive neural control (ANC) for a class of feedback linearizable uncertain nonlinear systems under a nonlinear separation principle. Yet due to static-gain and linear properties, high-gain observers are usually subject to peaking responses and noise sensitivity. Existing adaptive neural network (NN) observers cannot effectively relax the limitations of high-gain observers. This paper presents an output-feedback indirect ANC strategy under a nonseparation principle, where a hybrid estimation scheme that integrates an adaptive NN observer with state variable filters is proposed to estimate plant states. By applying a single Lyapunov function candidate to the entire system, it is proved that the closed-loop system achieves practical asymptotic stability under a relatively low observer gain dominated by controller parameters. Our approach can completely avoid peaking responses without control saturation while keeping favourable noise rejection ability. Simulation results have shown effectiveness and superiority of this approach.
高增益观测器已广泛应用于在非线性分离原理下构造一类反馈线性化不确定非线性系统的输出反馈自适应神经网络控制（ANC）。然而，由于静态增益和线性特性，高增益观测器通常容易受到峰值响应和噪声敏感性的影响。现有的自适应神经网络（NN）观测器不能有效地放宽高增益观测器的限制。本文在非分离原理下提出了一种输出反馈间接 ANC 策略，其中提出了一种混合估计方案，将自适应 NN 观测器与状态变量滤波器相结合，以估计 plant 状态。通过将单个 Lyapunov 函数候选应用于整个系统，证明了在由控制器参数主导的相对较低的观测器增益下，闭环系统实现了实用渐近稳定性。我们的方法可以在不使用控制饱和的情况下完全避免峰值响应，同时保持良好的噪声抑制能力。仿真结果表明了该方法的有效性和优越性。
