Scholarlink

<p>This paper focuses on investigating novel identification methods for multivariable systems interfered by colored noises. Through decomposing a multivariable system into several multiple-input single-output systems and filtering the input–output data by utilizing the estimated noise transfer function, the filtered identification model is derived. Based on the obtained model, a two-stage filtering-based maximum likelihood stochastic gradient (TS-F-ML-SG) algorithm is proposed for parameter estimation by using the interactive estimation idea and the maximum likelihood principle. In addition, an existing maximum likelihood stochastic gradient (AM-ML-SG) algorithm is presented for comparison. By analyzing the total operations (flops), the proposed TS-F-ML-SG algorithm has higher computational efficiency than the existing AM-ML-SG algorithm. Besides, simulation results test that the proposed algorithm has higher estimation accuracy and captures dynamics of the system well.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Highly Efficient Two-Stage Filtering-Based Maximum Likelihood Stochastic Gradient Algorithm for Multiple-Input Multiple-Output Systems

  • 期刊论文

  • 发表日期：

    2025年3月22日

查看全文
  • Huihui Wang,
  • Ximei Liu

摘要

This paper focuses on investigating novel identification methods for multivariable systems interfered by colored noises. Through decomposing a multivariable system into several multiple-input single-output systems and filtering the input–output data by utilizing the estimated noise transfer function, the filtered identification model is derived. Based on the obtained model, a two-stage filtering-based maximum likelihood stochastic gradient (TS-F-ML-SG) algorithm is proposed for parameter estimation by using the interactive estimation idea and the maximum likelihood principle. In addition, an existing maximum likelihood stochastic gradient (AM-ML-SG) algorithm is presented for comparison. By analyzing the total operations (flops), the proposed TS-F-ML-SG algorithm has higher computational efficiency than the existing AM-ML-SG algorithm. Besides, simulation results test that the proposed algorithm has higher estimation accuracy and captures dynamics of the system well.