Novel feedforward controller for straight-fiber-type artificial muscle based on an experimental identification model
Novel feedforward controller for straight-fiber-type artificial muscle based on an experimental identification model
复制标题
基于实验辨识模型的新型直纤维型人工肌肉前馈控制器
DOI：
10.1109/robosoft.2018.8404893
复制
发表时间：
2018
期刊：
影响因子：
--
通讯作者：
Taro Nakamura
中科院分区：
文献类型：
--
作者：
Ryuji Suzuki;M. Okui;S. Iikawa;Yasuyuki Yamada;Taro Nakamura
本文利用实验识别模型，对直纤维型人工肌肉前馈控制器进行了改进，实现了收缩量、刚度和收缩力的控制。与McKibben人造肌肉相比，这种直纤维型人造肌肉具有更高的收缩力和更高的收缩率。在之前的研究中，我们开发了一种基于力学模型的直纤维型人工肌肉前馈控制器。然而，该控制器不能准确地控制刚度和收缩力。为了弥补前馈控制精度的不足，需要一个反馈控制器，这增加了系统的复杂性。此外，以前的控制器的计算非常复杂，微控制器无法跟上顺序计算。当控制器用于辅助服等设备时，这是不实际的。为了解决这些问题，本文提出了一种基于实验辨识模型的前馈控制器，其计算比以往的前馈控制器简单。实验识别模型使前馈控制器通过识别模型中使用的参数来提高精度。此外，我们还比较了所提出控制器与先前控制器的精度。
This paper reports on an improvement to a feedforward controller for a straight-fiber-type artificial muscle that can control the amount of contraction, stiffness, and contraction force by use of an experimental identification model. This straight-fiber-type artificial muscle has a higher contraction force and a higher contraction rate than a McKibben artificial muscle. In a previous study, we developed a feedforward controller for a straight-fiber-type artificial muscle based on a mechanical model. However, this controller could not accurately control the stiffness or the contraction force. A feedback controller was necessary to compensate for the lack of feedforward control accuracy, which increased the system complexity. In addition, the calculations of the previous controller were so complex that the microcontroller could not keep up with the sequential calculations. This is not practical when the controller is used in devices such as an assist suit. In this paper, to solve these problems, we propose a novel feedforward controller based on an experimental identification model whose calculations are simpler than the previous ones. An experimental identification model enables the feedforward controller to improve the accuracy by identifying the parameters used in the model. Also, we compare the accuracy of the proposed controller with the previous one.