In the actual environment, due to the influence of uncertain factors such as the amount of agricultural machinery operation and the change of soil conditions, the driving condition of following agricultural machinery is unstable, and the response of following cooperative operation is slow, which lead to the difficulty of control. Therefore, in consideration of uncertainty and response performance, a kind of hierarchical control architecture for agriculture machinery was put forward, by building longitudinal dynamic model of agricultural machinery under the field to follow. Furthermore, taking the performance of following distance, following speed, fuel economy, following acceleration as the optimization target, the upper controller based on model predictive control (MPC) algorithm was derived to restrain uncertainty and interference noise. At the same time, to track torque (current) to improve the follow response ability of agricultural machinery, the lower layer controller based on the feedforward and feedback PI controller was designed which took the output of the upper controlleracceleration as the target. The simulation and experiment showed that this control method can effectively solve the following control problem of farm machinery operation. Compared with the sliding mode variable structure controller (SMC), this method can realize stable following driving with smaller speed error and smaller acceleration error. Speed error was controlled in the range of -0.29132~0.18001m/s and acceleration error was controlled in the range of -0.05678~0.05628m/s2. Moreover, the error of stable follow distance was within ±0.45m. The control effect was proved well according to the above data.