The Active-disturbance rejection control (ADRC) has the advantage of strong robustness, anti-interference capability, and it does not rely on the accurate math model of controlled plant. But the parameter self-turning of ADRC isn’t as easy as PID controller because there are more parameters to turn in ADRC. In this paper the parameters are self-turning by the Radial Basis Function (RBF) Neural Network. The results of the simulation indicate that the controller has good anti-interference capability and fast response. The robustness of the system is improved.