Abstract: In this paper, the stability of nominal adaptive plus iterative learning control is discussed. Using the excellent characteristics of cascaded CMAC, a new kind of adaptive control method based on cascaded CMAC is also proposed. This method divides the robot model into deterministic part and uncertain part. So the deterministic information can be employed to enrich advantages in control strategies and the uncertain part can be adaptively learned by using cascaded CMAC. Simulation had verified that using this control system for trajectory control of 6 DOF parallel manipulator, we can got more effective results than ordinary CMAC+PD control.