Abstract: Aiming at the nonlinearity and the surplus torque in rudder electric loading systems of unmanned aerial vehicle(UAV), a self-adaptive cerebellar model articulation controller (CMAC) was proposed, which was parallel to an adaptive neuron controller. This hybrid control strategy adopted the desired value and the actual output as the incentive signals of CMAC, and put the current system error as the training signal of CMAC. The method was proposed by using the error to adjust the learning rate on line, which eliminated the excess self-learning phenomena. The mathematical model of rudder electric loading systems for UAV was established and the detailed control structure was put forward. Simulation results show that the proposed hybrid controller can effectively eliminate the surplus torque, enhance the control stability of the systems and fairly improve the dynamic loading performances of the systems, which is highly suitable for real-time control of nonlinear systems.