Research on CMAC control algorithm for a soft joint

As to the robot joint actuated by a pair of McKibben muscles, a CMAC neural network and PID based compound control algorithm is designed. The CMAC as a feed forward compensator realizes the reverse dynamic model of the robot joint. The PID controller realizes the feed back control in order to guarantee the stability of the system. The output of the compound controller is exerted to the robot joint. CMAC regulates its network weights by on line learning and forces the PID output to tend to zero, so that the total control action is generated by CMAC. Some simulations are done and they have proved that the compound control algorithm has good tracking performance and anti-disturbance ability. And it behaves fast responding.