基于CMAC的气动人工肌肉变结构位置控制研究

气动人工肌肉是一种具有功率自重比/功率体积比大、响应快的新型气动元件，近年来已引起了人们广泛关注。然而，由于难于实现气动人工肌肉系统的精确控制，从而阻碍了其更加广泛的应用。在气动人工肌肉动态数学模型的基础上，提出采用基于CMAC的气动人工肌肉两层滑模变结构控制方法。CMAC神经网络用于学习气动人工肌肉系统的不确定信息，并作为前馈补偿使跟踪误差快速收敛，通过变结构控制消除网络的学习误差和不可重复随机干扰的影响，确保系统鲁棒性。试验结果表明了该方法的有效性和系统的鲁棒性。

CMAC-BASED VARIABLE STRUCTURE POSITION CONTROL OF A PNEUMATIC MUSCLE ACTUATOR SYSTEM

Pneumatic muscle actuator (PMA) is arrested attention recently. Its main advantage is the high efficiency that it is possible to obtain from a relatively light device. However, the problems with control and compliance of pneumatic system prevent its widespread use, and the nonlinearity in the system limits its controllability. A CAMC-based variable structure position controller is proposed based on the dynamic model. The control scheme comprises a stable controller and a CMAC neural network. The CMAC neural network is employed to approximate and compensate the uncertainties induced by in- accurate modeling of the system as a feed-forward compensator. Fast tracking error convergence is obtained through CMAC. The stable controller is used to obtain the robustness under the bounded disturbances and the approximation error of CMAC. The effectiveness of the proposed control approach is demonstrated by experiment. The results are shown to be accurate, and robust to changes in payload.