A method for controller design is proposed based on the approach of functional combination, which has interpreted the proportional, integral and differential expressions in PID controller as the pre-reverse regulating, overcoming deviation and turnaround accelerating functions. According to the control requirements, appropriate functions should be chosen and then combine them in linear or nonlinear ways in some optimal sense. A controller is designed by this means called as the reverse regulation and error elimination controller (RREEC). Two methods are given for the unknown parameters of the RREEC. One is adjusted by fuzzy inference rules online, the other is optimized by fuzzy neural networks. The simulation results compared with traditional adaptive controller show the stronger adaptability of this controller and prove the suitability for nonlinear systems.