The automatic navigation system of agricultural machinery is based on the navigation control system and is equipped with an autonomous operation controller to realize land farming with less human intervention. Aiming to solve the problems that the large agricultural machineries are inconvenient to transport in small fields of hilly areas, inefficient to operate, and restricted to change direction, a set of automatic navigation operation system was constructed with a light and simple crawler-type rape seeder and Beidou RTK. This system was composed of sensing device, walking module, control module and a seeding module. According to the requirements of sowing operation, the automatic operation control strategy was designed by using finite state machine. The crawler-type rape seeder was divided into four states: non-seeding state, seeding state, steering state and completed operation state, and the state transformation was carried out by judging different paths, and the seeding operation was coordinated. The automatic navigation operation not only required vehicle coordination and planter control to achieve accurate sowing operation, but also required navigation controller with high control precision and strong adaptability. Therefore, a fuzzy adaptive pure pursuit controller based on kinematic model and geometric model was designed for light and simple tracked vehicles. A small forward looking distance would make the vehicle approach the target path quickly with large curvature，and it would also cause vehicle oscillation. A large forward looking distance would make the vehicle approach the target path slowly with small curvature. Although the vehicle would not oscillate, the settling time was long. The forward looking distance can be adjusted adaptively by using the fuzzy control method. The lateral deviation, heading deviation and speed were used as the inputs of the fuzzy controller, and the forward looking distance was used as the output of the fuzzy controller. The input and output were fuzzified and fuzzy rules were established through test simulation. The simulation and field comparative test results of fuzzy adaptive pure pursuit controller and pure pursuit controller showed that compared with the pure pursuit controller, the fuzzy adaptive pure pursuit navigation controller had the characteristics of short rise time and small overshoot. The road test results showed that when the speed of the crawler-type rape seeder was 0.8m/s, the maximum tracking deviation of the fuzzy adaptive pure pursuit controller was 0.039m and the average absolute deviation was 0.018m. When the forward speed of the dry field road was 0.5m/s, 0.8m/s and 1.2m/s, respectively, the maximum tracking deviations of linear navigation tracking were not more than 0.082m, 0.086m and 0.092m. The average absolute deviations were not more than 0.031m, 0.032m and 0.034m. The experimental data of the automatic navigation operation system showed that the designed navigation controller had a stable ability of linear tracking. It can meet the requirements of automatic sowing of rape in small fields of hilly areas, and provide technical support for the construction of unmanned farms in hilly and mountainous areas.