To improve the trajectory tracking performance of agricultural vehicles, a new path tracking control method based on the nonlinear model prediction was proposed. The path tracking problem was transformed into solving the trajectory planning and trajectory tracking control problems with satisfying speed and angle constraint. Firstly, the nonlinear kinematic model of agricultural vehicles was discretized, and the recursive model was derived as the predictive equation of the controller. Then the objective function which defined the system control quantities as the state quantities was set up. Finally, the prediction equation was brought into the objective function to transform the problem into a quadratic programming method based on the recursive sequence and the gradient computation was performed to solve the nonlinear constrained optimization. The closed-loop correction of controller was realized by real-time feedback and rolling optimization. Predictive control can be achieved by repeating the process. Matlab simulation results showed that the nonlinear model prediction controller can realize effective tracking and had strong robust stability. The comparison test with the linear model prediction controller indicated that with the control of the controller, the lateral deviation was reduced by 36.8% and the longitudinal deviation was reduced by 32.98% when the velocity was 3m/s and tracking path was circular. At the same time, the corresponding field test was carried out and the results showed that when the test car was tracking the reference path with the velocity of 2m/s, the maximum lateral deviation was -4.28cm and when the velocity was 3m/s, the maximum longitudinal deviation was -6.61cm.