In order to improve the accuracy of driving straight when agricultural machinery works, an agricultural machinery path tracking algorithm based on an improved pure tracking model was designed. The linear tracking method of agricultural machinery was studied, which was based on the establishment of agricultural machinery kinematics model and pure tracking model. Aiming at the problem that GPS navigation accuracy was susceptible to noise interference, the Kalman filter was used to smooth the heading error and the lateral error so that higher accuracy heading and lateral errors can be obtained. In order to improve the adaptive ability of the pure tracking model, a fitness function was constructed based on the root mean square error of the lateral error and the heading error, and a weight function was designed, and the lateral error was used as the main decision parameter to determine the forward view in the pure tracking model in real time distance. In order to reduce the calculation time of the particle swarm, which the local search of the particle swarm can be performed as soon as possible, the inertia weight coefficient in the particle swarm optimization (PSO) algorithm was improved. When conducting field planting trials, an automatic navigation control system for agricultural machinery was designed where Dongfanghong 1104-C was used as the experimental platform. Seeding experiment proved that: when the path tracking algorithm based on improved pure tracking model was adopted, the agricultural machinery travel speed was 0.7m/s, the maximum lateral error of the linear tracking was 0.09m;when the driving distance exceeded 5m, the maximum lateral error was 0.02m. The proposed improved pure tracking model had good applicability to the automatic navigation control of agricultural machinery, which can effectively improve the straight-line driving accuracy during agricultural machinery operation.