基于气动人工肌肉的混合位置跟踪控制

针对一种气动人工肌肉驱动的弹簧质量位置控制系统，设计了一个带有自适应模糊小脑模型（Cerebellar Model Articulation Controller，CMAC）在线逼近的离散趋近律滑模混合控制器.该混合控制器中离散趋近律滑模策略产生控制器的输出；自适应模糊CMAC用以逼近气动人工肌肉系统中的不确定项.CMAC网络权值的在线学习调整保证了自适应模糊CMAC的逼近性能.对离散抗饱和PID控制器（DASPID）与自适应模糊CMAC离散滑模混合控制器（HybridC）的位置跟踪控制性能进行了对比实验.实验结果表明，HybridC较之DASPID有更好的位置跟踪控制性能.当期望参考输入为正弦信号时，DASPID的最大位置跟踪误差为±15 mm；而HybridC的最大位置跟踪误差仅为±07 mm，平均位置跟踪误差大约仅为±02 mm.并且，离散滑模所固有的抖振现象得到了有效的抑制.

Hybrid Position Tracking Control of a Pneumatic Artificial Muscle

A single degree freedom pneumatic artificial muscle spring mass system was built.The pneumatic artificial muscle is controlled by high speed on/off valve in PWM technique.A hybrid controller is setup to tracking control the pneumatic artificial muscle system which is composed of an adaptive fuzzy CMAC(cerebellar model articulation controller) and discrete sliding mode reaching low strategy.In hybrid controller discrete sliding mode strategy produces the control output;the adaptive fuzzy CMAC is used to approximate the uncertainty of the pneumatic artificial muscle control system.The on-line self adjustment ability guarantees the approximation ability of the adaptive fuzzy CMAC.Finally,the comparative study of control performance between discrete anti saturation PID(DASPID) and hybrid controller(HybridC) was carried out.The experimental results suggest that the hybridC is better than DASPID.While the desired reference input is sinusoidal,the biggest tracking error of PID is ±1.5 mm;however the biggest tracking error of the hybridC is only ±0.7 mm;the average tracking error is about ±0.2 mm.Furthermore as the main drawback of discrete sliding mode strategy,the chatter phenomenon is greatly reduced.