In order to improve the navigation accuracy and control precision, some related study on the autonomous navigation of agricultural robot control was carried out. Firstly, the Sage_Husa self-adaptive filtering and strong tracking Kalman filtering were introduced. The former has the characteristics of high precision but huge computation, the latter has strong adaptive capacity, but the filtering accuracy is low. So combined the advantage of these two filtering algorithms and used the strict convergence criterion, an improved self-adaptive Kalman filtering algorithm was proposed. This filtering algorithm can effectively reduce the state error of system, restrain the signal divergence, so it can guarantee the real-time and stability of the system, and has better filtering convergence and precision. By the simulation experiment, the results show that the proposed filtering algorithm was flexible and reliable. Secondly, in order to prevent the PID controller integral saturation phenomenon, the method of variable structure switch was introduced, and the variable structure PID (VSPID) controller was proposed. By the simulation experiment, the results show that the VSPID can quickly withdrew from the saturation state, and it can quickly reach the expectations and maintain this state. The problem of controller supersaturated was solved effectively. It can greatly improve the control efficiency, and maintain the good tracking performance. Thirdly, combining the improved self-adaptive filtering with anti-windup VSPID controller, the stability and precision of navigation system was improved. The results of simulation experiment show that proposed method makes the system quickly withdrew from the saturated zone, and with a better tracking performance. Finally, the proposed method was applied to the mobile robot navigation system and the field experiments on the school playground. The experimental results show that the proposed method can improve GPS positioning precision and enhance the antiinterference capability, after further adjusting parameters and tracking the given path reaches the desired effect. It demonstrates that the proposed method greatly improves the capability of restraining filtering divergence, anti-interference and control precision.