基于自适应动态规划的非线性鲁棒近似最优跟踪控制
Nonlinear robust approximate optimal tracking control based on adaptive dynamic programming
摘要点击 43  全文点击 52  投稿时间:2018-01-23  修订日期:2018-05-16
查看全文  查看/发表评论  下载PDF阅读器
DOI编号  10.7641/CTA.2018.80075
  2018,35(9):1285-1292
中文关键词  自适应动态规划  最优跟踪控制  未知非线性系统
英文关键词  Adaptive dynamic programming  optimal track control  unknown nonlinear system
基金项目  国家自然科学基金项目(61573090), 高校基本科研业务费项目(N160801001)资助
学科分类代码  
作者单位E-mail
洪成文 东北大学自动研究中心 54544311@qq.com 
富月 东北大学自动研究中心 fuyue@mail.neu.edu.cn 
中文摘要
      为克服现有近似最优跟踪控制方法只能跟踪连续可微参考输入的局限,本文针对一类具有未知动态的连续时间非线性时不变仿射系统,提出了一种新的基于自适应动态规划的鲁棒近似最优跟踪控制方法。首先采用递归神经网络建立系统模型,然后建立评价神经网络对最优性能指标进行估计,从而得到最优性能指标偏导数的估计值,进而得到近似最优跟踪控制器,最后利用系统输出与参考输入之间的跟踪误差设计鲁棒补偿器对神经网络的建模误差和评价网络的估计误差进行补偿。分别针对两个非线性系统进行仿真实验,仿真结果表明了所提方法的有效性和优越性。
英文摘要
      In order to overcome the limitation of the existing approximate optimal tracking control method that can only track continuously differentiable reference inputs, for a class of continuous-time nonlinear time-invariant affine systems with unknown dynamics, a new robust approximate optimal tracking control method based on adaptive dynamic programming is proposed. Firstly, the system model is established by using the recurrent neural network. Then, the optimal performance index is estimated by the established critic neural network. The estimated value of the partial derivative of the optimal performance index, and the approximate optimal tracking controller can consequently be obtained. Finally, The error between the output and the reference input is used to designed the robust compensator to compensate the modeling error and the estimation error of the neural networks. Simulation experiments are conducted for two nonlinear systems respectively. The simulation results show the effectiveness and superiority of the proposed method.