As a kind of novel compliant actuators, pneumatic muscle actuators (PMAs) have been recently used in wearable devices for rehabilitation, industrial manufacturing and other fields due to their excellent actuation characteristics such as high power/weight ratio, safety and inherent compliance. However, the strong nonlinearity and asymmetrical hysteresis cause difficulties in the accurate control of robots actuated by PMAs. In this paper, a method for hysteresis modeling of PMA based on Hammerstein model is proposed, which introduces the BP neural network into the hysteretic system. In order to overcome the limitation of BP neural network’s single-valued mapping, an extended space input method is adapted while the Modified Prandtl–Ishlinskii model is applied to characterize the hysteretic phenomenon. A conventional PID control is implemented to track the trajectory of PMA with and without the feed-forward hysteresis compensation based on Hammerstein model, and experimental results validate the effectiveness of the designed model which has the advantages of high precision and easy identification.

中文翻译：

---

Hammerstein模型，用于气动肌肉执行器的磁滞特性

作为一种新型的顺应性致动器，气动肌肉致动器（PMA）最近因其出色的致动特性（例如高功率/重量比，安全性和固有的顺应性）而用于康复，工业制造和其他领域的可穿戴设备中。但是，强烈的非线性和不对称的磁滞导致在精确控制由PMA驱动的机器人时遇到困难。提出了一种基于Hammerstein模型的PMA滞后建模方法，将BP神经网络引入了滞后系统。为了克服BP神经网络单值映射的局限性，在应用修改后的Prandtl–Ishlinskii模型描述滞后现象的同时，采用了扩展空间输入法。