Aiming at the waist of hominoid robot,this paper raised a 3 DOF parallel platform based on pneumatic artificial muscles.And an AFCMAC was introduced for posture tracking control of the parallel platform.

By arranging the input space,AFCMAC could learn such uncertainties as hysteresis and

pressure fluctuation,as well as nonlinear coupling characteristics.In the beginning of control

process,the bigger tracking error and pressure fluctuation phenomenon can be avoided

by the design of discrete anti saturation PID parallel supervisory control and offline

identification,which make it possible for AFCMAC to online adjust its structure parameters in real time.Finally,the fix point rotation experiments and anti disturbance experiments

 were carried out.And the results suggest the good control performance and online self learning and adjustment ability of AFCMAC.

着眼类人机器人的腰部应用,设计了一种气动人工肌肉三自由度并联平台。引入一种自适应模糊CMAC(AFCMAC),实现了并联平台的姿态控制。通过规划输入空间,实现了AFCMAC对迟滞力、气压波动等不确定因素和非线性耦合因素的感知。离散抗饱和PID并行监督和离线辨识避免了在控制运行初期出现较大的跟踪误差和气压波动,从而使AFCMAC的在线实时自学习调整成为可能。进行了定点转动姿态跟踪实验和抗干扰实验,实验结果表明,AFCMAC具有较好的姿态控制性能和在线学习调整能力。

 

气动人工肌肉, 三自由度并联平台, 姿态控制, 自适应模糊CMAC, 迟滞特性