In the world of industrial automation, learning from mistakes is not exclusive to humans. Computers, too, possess the ability to learn and improve from past errors. One prevalent control technique used in industries that rely on batch-based production systems, like manufacturing clothing, computer chips, or baked goods, is iterative learning control (ILC). The predominant method employed in ILC systems is the proportional-type update rule (PTUR), which enhances performance by repeating tasks and adjusting control inputs based on previous errors.
However, as industries increasingly adopt ILC systems for more complex tasks, the need arises for faster and more accurate learning techniques. In a recent breakthrough, a group of scientists has proposed a new approach that leverages the fractional power update rule (FPUR) to unlock the full potential of single-input-single-output linear ILC systems. Their study, published in the IEEE/CAA Journal of Automatica Sinica, outlines this groundbreaking technique.
Convergence rates, which measure the speed at which the difference between desired and actual outputs diminishes over time, are crucial for determining the efficiency of an ILC system. Existing methods for improving convergence rates often fall short in situations requiring high precision. Even with constant or manually selected learning gains, current ILC systems employing linear update methods fail to fully exploit available information. Thus, the researchers investigated nonlinear update methods to achieve enhanced learning and faster convergence rates beyond PTUR.
“Traditional PTUR employs a linear term to update the control input based on tracking errors. In contrast, FPUR employs a fractional term for this update. As any positive number smaller than one possesses a larger fractional power than itself, FPUR exhibits greater update intensity than PTUR for small tracking errors, resulting in accelerated convergence rates,” explains Zihan Li, the lead author of the study and a master’s student at the School of Mathematics, Renmin University of China.
The team devised a novel FPUR method inspired by modern finite-time control (FTC) and terminal sliding mode control (TSMC) strategies. These techniques hold promise in addressing the aforementioned challenges and improving convergence speed. Additionally, the scientists adopted a nonlinear mapping approach to investigate error dynamics over time. This approach enabled them to showcase fast convergence performance and analyze potential limit cycles of tracking errors in ILC systems. Numerical simulations were also conducted to validate the effectiveness of their proposed method.
When asked about the impact of their research on the field of ILC systems, Li states, “This study serves three primary purposes. Firstly, it presents an algorithm employing a nonlinear update method to enhance learning capability. Secondly, it demonstrates that adapting fractional power terms allows convergence rates to be adjusted based on actual performance. Finally, it showcases fast convergence rates comparable to those achieved with FTC and TSMC.”
For the first time, this study illustrates the application of FPUR in ILC systems for single-input-single-output linear systems. The proposed technique holds potential for utilization in other repetitive systems such as autonomous vehicles, unmanned aerial vehicles, and rehabilitation robots.
By revolutionizing the learning potential and convergence rates of ILC systems, this research opens up new avenues for improving efficiency and accuracy in various industries, paving the way for advancements in automation and control technology.
Source: Cactus Communications