In order to let the combine working in the best condition, it is necessary to control its speed according to the crop conditions in the field. The walking office of combine is analyzed in this paper. It is a high order system, some parameters will change during its working, and it is impacted by external disturbance. Then a servocontroller based on RBF neural network is designed. The model is simplified, and there are 2 parameters in it. Then a RBF neural network is established. The speed of combine, its integral value and differential value are inputted to identify the parameters. The inverse controller is established base on the simple model. The result of simulation and real vehicle tests shows that the controller can effectively adapt to changes in the internal parameters, overcome the external disturbance, and achieve the purpose of tracking a given speed.