引用本文:朱盼盼,卜旭辉,梁嘉琪,闫帅明.考虑数据量化的改进无模型自适应迭代学习控制算法[J].控制理论与应用,2020,37(5):1178~1184.[点击复制]
ZHU Pan-pan,BU Xu-hui,LIANG Jia-qi,YAN Shuai-ming.An improved model free adaptive iterative learning control algorithm with data quantization[J].Control Theory and Technology,2020,37(5):1178~1184.[点击复制]
考虑数据量化的改进无模型自适应迭代学习控制算法
An improved model free adaptive iterative learning control algorithm with data quantization
摘要点击 1760  全文点击 723  投稿时间：2019-05-14  修订日期：2019-09-03
查看全文  查看/发表评论  下载PDF阅读器
DOI编号  10.7641/CTA.2019.90347
  2020,37(5):1178-1184
中文关键词  无模型自适应控制  迭代学习  编码解码量化机制  数据量化
英文关键词  model free adaptive control  iterative learning  encoding and decoding quantization mechanism  data quantization
基金项目  国家自然科学基金项目(61573129, 61573130, U1804147, 61833001), 河南省高校科技创新团队项目(20IRTSTHN019), 河南理工大学创新型科技 团队项目(T2019–2, T2017–1), 河南省创新型科技团队项目(CXTD2016054)资助.
作者单位E-mail
朱盼盼 河南理工大学 zhupanpan@home.hpu.edu.cn 
卜旭辉* 河南理工大学 buxuhui@gmail.com 
梁嘉琪 河南理工大学  
闫帅明 河南理工大学  
中文摘要
      针对一类存在数据量化的离散时间单输入单输出非线性系统, 提出一种带有编码解码量化机制的无模型 自适应迭代学习控制(MFAILC)算法. 首先使用伪偏导数将受控非线性系统动态线性化, 进而考虑系统输出数据经 由均匀量化器进行量化处理的过程, 并设计了一种编码解码量化机制, 最后基于这种编码解码量化机制提出了一种 改进的MFAILC算法. 理论上给出了算法的收敛性分析, 结果表明, 当系统存在数据量化时, 所提出的算法仍可保证 系统收敛. 与已有算法相比, 所提算法仅利用较少的输入输出数据, 就可以实现跟踪误差的零收敛. 仿真进一步验 证了算法的有效性.
英文摘要
      For a class of discrete time single input single output nonlinear systems with data quantization, a model free adaptive iterative learning control (MFAILC) algorithm with encoding and decoding quantization mechanism is proposed. First, the nonlinear system is dynamically linearized by using pseudo partial derivative. Then the output data of the system is quantized by a uniform quantizer, and an encoding and decoding quantization mechanism is designed for the system. Finally, an improved MFAILC algorithm is proposed based on the encoding and decoding quantization mechanism. The convergence of the algorithm is analyzed theoretically. The results show that the proposed algorithm can still guarantee the convergence of the system which is subject to data quantization. Compare with the existing results, the proposed algorithm can only use less input and output data to achieve the zero-error convergence. The simulation validates the effectiveness of the algorithm.