Adaptive echo state network control for a class of pure-feedback systems with input and output constraints
Adaptive echo state network control for a class of pure-feedback systems with input and output constraints
复制标题
DOI：
10.1016/j.neucom.2017.09.083
复制
发表时间：
2018-01
期刊：
影响因子：
6
通讯作者：
Qiang Chen;Linlin Shi;J. Na;X. Ren;Yurong Nan
中科院分区：
文献类型：
--
作者：
Qiang Chen;Linlin Shi;J. Na;X. Ren;Yurong Nan
针对一类同时考虑输入和输出约束的约束纯反馈系统，提出了一种自适应回声状态网络控制方案。采用一个描述收敛速度、最大超调量和稳态误差的性能函数来提高瞬态跟踪性能。此外，一种改进的动态表面滑模方法，通过将高阶滑模（HOSM）微分器的控制器设计的每一步，这样，在传统的动态表面控制（DSC）中的滤波器性能的滞后影响，可以消除。最后，未知的非线性，包括输入饱和动态估计使用回声状态网络，这是很容易训练，而无需调整输入层和隐层之间的权重。比较仿真结果表明了该方法的有效性和上级性能。
In this paper, an adaptive echo state network control scheme is proposed for a class of constrained pure-feedback systems, in which both input and output constraints are considered simultaneously. A prescribed performance function characterizing convergence rate, maximum overshoot and steady-state error is employed to enhance the transient tracking performance. Moreover, an improved dynamic surface sliding mode approach is developed by incorporating high-order sliding mode (HOSM) differentiators into each step of controllers design, such that the sluggish effect of the filter performance in conventional dynamic surface control (DSC) can be eliminated. Finally, the unknown nonlinearities including the input saturation dynamics are estimated by using an echo state network, which is easily trained without adjusting the weights between the input layer and the hidden layer. Comparative simulations are provided to illustrate the effectiveness and superior performance of the proposed method.