Predictor-based control of time-delay systems: a survey

Yang Deng, Vincent Léchappé, Emmanuel Moulay, Zhang Chen, Bin Liang, Franck Plestan, Qing‐Long HanInternational Journal of Systems Science 2022 · DOI

TL;DR

This survey provides a comprehensive overview of predictor-based control techniques for time-delay systems, a critical challenge in modern control applications like networked systems and robotics. It categorizes and analyzes various methods, from traditional ODE and PDE approaches to integration-free and adaptive strategies, highlighting their strengths, weaknesses, and practical implementations.

What Problem Does This Paper Address?

Many modern control systems, such as those in wireless communication, robotics, and multi-agent systems, are inherently affected by time delays. These delays, whether in sensing, actuation, or communication, can significantly degrade system performance and even lead to instability. Traditional control methods often struggle with long or time-varying delays. Predictor-based control offers a powerful solution by actively compensating for these delays through state prediction, aiming to create a delay-free closed-loop system. This paper addresses the need for a structured and up-to-date review of the diverse and rapidly evolving field of predictor-based control for time-delay systems.

Methods

The paper systematically categorizes predictor-based control methods based on their underlying mathematical framework and approach to handling delays:

  • Ordinary Differential Equation (ODE)-based approaches: This includes foundational methods like the Smith predictor and more advanced techniques such as Finite Spectrum Assignment (FSA) and Model Reduction. These methods often involve predicting future states to construct control laws.
  • Partial Differential Equation (PDE)-based approaches: This framework transforms the time-delay system into a coupled ODE-PDE system, allowing for a more rigorous stability analysis and enabling the handling of nonlinearities and state-dependent delays. Backstepping transformations are a key tool here.
  • Integration-Free Predictor-Based Controllers: Recognizing the practical challenges of implementing infinite-dimensional integral terms in real-time, this category focuses on methods that avoid or approximate these integrals. This includes:
  • Truncated Predictor-Based Control: Discards or approximates integral terms under certain conditions.
  • Observation-Predictor-Based Control: Uses observers to estimate state predictions, avoiding direct computation of integrals.
  • Discrete Predictor-Based Control: Discretizes the system and controller, transforming continuous-time integrals into finite sums.
  • Adaptive and Robust Control: Many methods incorporate adaptive laws to estimate unknown delays or robust design principles to handle uncertainties and disturbances.

Datasets

Not discussed in the paper.

Key Insights and Ideas

  • The core idea of predictor-based control is to "look ahead": By predicting future states, controllers can proactively counteract the destabilizing effects of time delays, effectively creating a delay-free system.
  • ODE vs. PDE frameworks offer different strengths: ODE methods are often more direct, while PDE methods provide a more rigorous framework for analyzing complex systems, especially those with nonlinearities and state-dependent delays.
  • Practical implementation is a major driver of research: The difficulty of numerically implementing infinite-dimensional integrals has spurred significant research into "integration-free" methods, making predictor-based control more accessible in real-world applications.
  • Unknown delays are a significant challenge: Handling unknown or time-varying delays requires sophisticated estimation and adaptation techniques, often leading to trade-offs between conservatism and performance.
  • The field is highly interdisciplinary: Predictor-based control finds applications in diverse areas like networked control systems, multi-agent systems, and robotics, driving the development of specialized techniques for each domain.

Key Results

The paper doesn't present specific quantitative results from experiments but rather synthesizes the state-of-the-art in predictor-based control. The key "results" are the diverse methodologies developed and their demonstrated applicability to various classes of time-delay systems, including:

  • Stabilization of linear and nonlinear systems with constant, time-varying, and uncertain time delays.
  • Effective control of networked control systems (NCSs) and multi-agent systems (MASs) despite communication delays.
  • Application to robotic systems, including underactuated and Euler-Lagrange systems.
  • Development of robust and adaptive control strategies to handle uncertainties and external disturbances.
  • Advancements in practical implementation through integration-free and discrete-time approaches.

Limitations and Future Work

The authors identify several limitations and suggest future research directions:

  • Conservatism in delay estimation: Current delay estimation algorithms can sometimes lead to conservative control designs.
  • Computational complexity: Some advanced methods, particularly adaptive backstepping, can be computationally intensive.
  • Numerical implementation challenges: While integration-free methods exist, ensuring robust and efficient numerical implementation remains an area of active research.
  • Mixed network imperfections and cyber attacks: Future work should focus on developing predictor-based controllers that are resilient to combined network constraints (packet loss, quantization) and cyber attacks, especially for critical cyber-physical systems (CPSs).
  • Dynamic event-triggered control: Exploring dynamic event-triggered mechanisms that can adaptively tune control execution based on system states and network conditions is a promising avenue.
  • Finite-time and fixed-time control: Further research is needed to develop robust finite-time and fixed-time predictor-based controllers that maintain their performance under model uncertainties and external disturbances.
  • Robust prediction schemes for time-varying delays: Extending existing robust prediction schemes to handle time-varying delays and investigating their delay-robustness is an important area for future study.
  • Predictor-based control for underactuated robotic systems: This remains a relatively underexplored area, with a need for systematic techniques to address the complexities of these systems.

Summary updated

Abstract

With the developments of wireless data communication and network technology, time-delays are widely found in nowadays' control systems, e.g. networked control systems, mobile robot systems, and multi-agent systems. Predictor-based control is an effective method dealing with long time-delays because it can generally lead to a delay-free closed-loop system by introducing a prediction for future states. Recently, various predictor-based control methods have been developed for numerous control systems subject to different time-delays, which motivates this survey. This paper presents a comprehensive review of the up-to-date results on the predictor-based control of time-delay systems. Firstly, the ordinary differential equation-based approaches for designing and analysing predictor-based controllers are summarised. Secondly, one reports an alternative method of predictor-based control, in which the systems/controllers are understood in the sense of partial differential equations. Next, several integration-free predictor-based controllers are introduced: by abandoning the infinite-dimensional integral terms, the control laws become easier to realise in practice. Hereafter, the paper discusses the real-time implementations and the practical applications of predictor-based control methods to several particular control systems. Finally, this paper suggests some new trends of predictor-based control for future research.
BibTeX
@article{deng2022predictor,
  author = {Yang Deng and Vincent Léchappé and Emmanuel Moulay and Zhang Chen and Bin Liang and Franck Plestan and Qing‐Long Han},
  title = {{Predictor-based control of time-delay systems: a survey}},
  journal = {International Journal of Systems Science},
  year = {2022},
  doi = {10.1080/00207721.2022.2056654},
}
Figure