基于气动人工肌肉的自适应模糊小脑模型神经网络位置跟踪控制

Adaptive Fuzzy CMAC Position Tracking Control Based on Pneumatic Artificial Muscle

针对一种气动人工肌肉驱动的弹簧质量位置控制系统,设计了一种自适应模糊小脑模型神经网络(AFCMAC)控制器.离散抗饱和PID(DASPID)并行监督控制设计保证了控制运行初期不会出现较大的跟踪误差和气压波动,使AFCMAC的在线实时学习调整成为可能.在线实时的自适应算法逐步提高了AFCMAC的控制性能,从而最终完全过渡到AFCMAC控制.通过规划AF-CMAC的输入空间,保证了AFCMAC对迟滞力和气压波动等不确定因素的感知能力,为实现AF-CMAC控制奠定了基础.对DASPID与AFCMAC控制器的位置跟踪控制性能进行了对比实验.结果表明,在非线性系统条件下,AFCMAC较之DASPID有着更好的跟踪控制性能和较低的实现难度.

A single degree freedom pneumatic artificial muscle spring mass system was built. An adaptive fuzzy CMAC （AFCMAC） was setup to track and control the pneumatic artificial muscle system. With the parallel supervision of discrete anti saturation PID （DA_SPID） there is few too big tracking error and pres- sure fluctuation in the beginning of control process, which makes the online real time self adjustment of AFCMAC possible. The online real time self adjustment ability improves the control performance of the adaptive fuzzy CMAC, and finally the spring mass system is entirely controlled by AFCMAC. In the end, the comparative study of control performance between DASPID and CMAC was carried out. The experi- mental results suggest that in nonlinear condition, AFCMAC could get better tracking performance than DASPID.