具有Preisach类型磁滞输入的不确定非线性系统 自适应神经网络控制

Adaptive neural control for uncertain nonlinear systems with Preisach-type hysteresis input

中文关键词  Preisach算子  自适应控制系统  磁滞  神经网络  非线性系统

英文关键词  Preisach operator  adaptive control systems  hysteresis  neural networks  nonlinear systems

中文摘要

      本文针对一类执行器受Preisach磁滞约束的不确定非线性系统, 提出一种基于神经网络的直接自适应控制 方案, 旨在解决系统的预定精度轨迹跟踪问题. 由于Preisach算子与系统动态发生耦合, 导致算子输出信号不可测 量, 给磁滞的逆补偿造成了困难. 为解决此问题, 本文首先将Preisach模型进行分解, 以提取出控制命令信号用于 Backstepping递归设计, 并在此基础上融合一类降阶光滑函数与直接自适应神经网络控制策略, 形成对磁滞非线性 和被控对象非线性的强鲁棒性能, 且所设计方案仅包含一个需要在线更新的自适应参数, 同时可保证Lyapunov函数 时间导数的半负定性. 通过严格数学分析, 已证明该方案不仅保证闭环系统所有信号均有界, 而且输出跟踪误差随 时间渐近收敛到用户预定区间. 基于压电定位平台的半物理仿真实验进一步验证了所提出控制方案的有效性.

英文摘要

      In this paper, a direct adaptive control scheme based on neural networks is proposed for a class of uncertain nonlinear systems with Preisach hysteresis constraints on the actuators to solve the trajectory tracking problem with prescribed accuracy. Due to the coupling between Preisach operator and system dynamics, the output signal of the operator is immeasurable, which makes it difficult to compensate the hysteresis. To overcome this problem, the Preisach model is firstly decomposed to extract the control command signal for backstepping recursive design, and based on this, a class of reduced-order smooth function and direct adaptive neural network control strategy are fused to establish a strong robustness performance to hysteresis nonlinearity and plant nonlinearities. Moreover, our designed scheme only contains one adaptive parameter to be updated online, and also guarantees the semi-negative definite property of the time derivative of the chosen Lyapunov function. It has been proved, through a rigorous mathematical analysis, that the scheme ensures not only the boundedness of all closed-loop system signals, but also the asymptotic convergence of output tracking error to a user-defined interval with time. The effectiveness of the proposed control scheme is further verified by a semi-physical simulation experiment based on the piezoelectric positioning stage.