Scholarlink

<p>Controlling complex and unknown chaotic systems remains challenging both theoretically and practically, especially when the intrinsic mechanisms of the system are elusive and/or the physical validation of the control strategies is infeasible. Developing effective and efficient control paradigms for these systems thus becomes an essential issue. In this work, we propose an approach using reservoir computing-based digital twins to replicate such unknown systems, providing a validation platform for the design of appropriate controls with only observables. Numerical experiments demonstrate the effectiveness of the proposed digital twins in controlling chaotic systems ranging from low- to high-dimensional cases, applying traditional control strategies including linear feedback methods and advanced nonlinear and time-delay techniques. We further investigate how to select and implement appropriate control strategies as well as their intensities according to specific requirements such as driving the system towards desired dynamics. This approach brings new insights into the field of data-driven control and offers new tools for designing practical control strategies.</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

Reservoir computing as digital twins for controlling nonlinear dynamical systems

  • 期刊论文

  • 发表日期：

    2026年2月18日

查看全文
  • Haoran Sun,
  • Xunzhe Zhou,
  • Ruizhi Cao,
  • Jiawen Hou,
  • Chun Guan,
  • Siyang Leng

摘要

Controlling complex and unknown chaotic systems remains challenging both theoretically and practically, especially when the intrinsic mechanisms of the system are elusive and/or the physical validation of the control strategies is infeasible. Developing effective and efficient control paradigms for these systems thus becomes an essential issue. In this work, we propose an approach using reservoir computing-based digital twins to replicate such unknown systems, providing a validation platform for the design of appropriate controls with only observables. Numerical experiments demonstrate the effectiveness of the proposed digital twins in controlling chaotic systems ranging from low- to high-dimensional cases, applying traditional control strategies including linear feedback methods and advanced nonlinear and time-delay techniques. We further investigate how to select and implement appropriate control strategies as well as their intensities according to specific requirements such as driving the system towards desired dynamics. This approach brings new insights into the field of data-driven control and offers new tools for designing practical control strategies.