Because of the perturbations and disturbances of electro-hydraulic servo systems, the normal controller can?t obtain good output responses under various conditions. A hybrid control strategy base on improved Radial Basis Function (RBF) network was presented for the tracking problem. An offline optimization algorithm was employed to choose the number of hidden units of RBF network. Based on the theory of optimization in groups, different online learning algorithms were used to adjust centers and estimate connection weights. The hybrid controller combined this improved RBF network with traditional PID control, and it was applied to electro-hydraulic servo system of ship test bed. Simulation results show that the servo system has good dynamic performance and nice robustness with this hybrid controller.