A humanoid driving model was established by collecting the harvester position information and driver operation information from skilled drivers. A navigation controller based on the humanoid driving model was designed by using neural networks to solve the problems of tedious operation，low operation quality and low efficiency of wheeled combine harvester in rice harvesting operation. According to the requirements of harvesting operation，a control method of field set row operation was designed，which can better complete the harvesting operation while ensuring the turning accuracy. Compared with the conventional PID and conventional pure tracking model，the humanoid driving model control converges 0.42 s faster and the overshoot is reduced 4.0 cm，which has the characteristics of fast convergence and small overshoot. The results of road right-angle route steering test showed that when the driving speed was 0.62 m/s，0.82 m/s and 1.02 m/s，the overshoot after the turn was no more than 3.93 cm，and it still had high robustness under different driving speeds. The results of field test showed that the right-angle steering navigation tracking post-turn overshoot was no more than 8.1 cm，8.9 cm，9.6 cm. The average absolute deviation of the linear tracking part was no more than 3.1 cm，3.0 cm，3.3 cm in the rice field at the forward speed of 0.6，0.8，1.0 m/s. It is indicated that the designed navigation controller based on the humanoid driving model established can better complete the automatic navigation of rice harvesting operation，and provide technical support for unmanned harvesting of rice production.