Fuzzy CMAC (FCMAC) was developed by Ker in 1997  and it has advantages in terms of simple learning structure, fast convergence speed, easy implementation by hardware and the preservation of derivative information. It has been successfully applied to control field. To enhance on-line learning ability of FCMAC , this study presents a learning efficiency improvement approach based on Aitken acceleration method for fuzzy CMAC (FCMAC). This approach employs the forecast ability of Aitken acceleration method to shorten the training time as well as promote the learning accuracy of FCMAC. The experimental results show that the proposed Aitken acceleration method not only help FCMAC to promote the learning convergence speed, but also has more accurate learning ability than the original FCMAC.