Simulation study of CMAC control for the robot joint actuated by McKibben muscles

The paper simply introduces the CMAC model. It explains how to train the CMAC off-line by the data from fuzzy control theory. Then two control systems based on CMAC for controlling a humanoid robot joint actuated by a pair of McKibben muscles are designed. One is based on pure CMAC and another one based on both the CMAC and PID. The simulations with different frequencies of the input function show that the control systems are stable. As to the two controllers, the paper discusses their on-line learning algorithms and compares them with each other.