Abstract Pneumatic artificial muscle (PAM) actuators are a kind of biomimetic actuators, which are being widely used in the applications of biomimetic robots and medical auxiliary devices. However, PAM systems usually have high nonlinearities, uncertainties, and time-varying characteristics, which bring challenges for accurate dynamic modeling and controller design. To deal with the above issues, in this paper, a neuroadaptive control method is proposed to handle the system uncertainties and achieve satisfactory tracking performance. First, in order to compensate the unknown nonlinear term involved in the dynamic model of the PAM system online, a three-layer neural network is utilized. Next, by means of the filtered signal, the algebraic loop problem can be solved effectively. Then, based on a sliding mode surface, a nonlinear robust controller is designed. By using the proposed method, the asymptotic convergence of tracking errors of the PAM system is guaranteed, and the tracking errors are always restricted within preset bounds during the control process. Moreover, the stability of the closed-loop system is proven theoretically by utilizing Lyapunov techniques. Finally, a series of hardware experiments are implemented on a self-built PAM testbed to validate the effectiveness and robustness of the proposed neuroadaptive control method.

中文翻译：

---

基于硬件实验的气动人工肌肉系统神经自适应控制方法

摘要 气动人工肌肉（PAM）执行器是一种仿生执行器，广泛应用于仿生机器人和医疗辅助设备。然而，PAM系统通常具有较高的非线性、不确定性和时变特性，这给精确的动态建模和控制器设计带来了挑战。针对上述问题，本文提出了一种神经自适应控制方法来处理系统的不确定性并获得令人满意的跟踪性能。首先，为了在线补偿PAM系统动态模型中涉及的未知非线性项，利用了三层神经网络。接下来，通过滤波后的信号，可以有效地解决代数环问题。然后，基于滑模面，设计了一种非线性鲁棒控制器。该方法保证了PAM系统跟踪误差的渐近收敛，控制过程中跟踪误差始终限制在预设的范围内。此外，利用李雅普诺夫技术从理论上证明了闭环系统的稳定性。最后，在自建的 PAM 测试平台上进行了一系列硬件实验，以验证所提出的神经自适应控制方法的有效性和鲁棒性。