Due to the complex starting conditions of the power shift tractor, there are many problems such as modeling uncertainty, parameter perturbation and time-varying disturbance in the system. Therefore, a novel starting control method employing the time-varying disturbance rejection was proposed to improve the starting performance of the power shift tractor and the drivers’ operating comfort. Firstly, the power shift transmission (PST) starting dynamic model was established by analyzing the dynamic characteristics of the starting process. Using the starting jerk and slipping friction work as the evaluation indicators to construct the performance function of the controller. Then a high-order disturbance observer (HDO) was introduced to estimate the time-varying disturbance and its derivatives in the starting process. The linear quadratic regulator (LQR) was used to obtain the optimal starting control law. The results of numerical calculation showed that HDO could effectively estimate the time-varying disturbance, and it could effectively suppress the influence of the disturbance and improve the robustness of the system combining with LQR. Finally, a co-simulation model of high-power tractor driveline was built based on Matlab/Simulink and AMESim, and proportional integral observer (PIO) was introduced as the comparison to verify the effectiveness of the starting control strategy. Taking 60% throttle as the example, the results showed that the maximum starting jerk by HDO were 12.25m/s 3 and 11.32m/s 3 in transportation and ploughing conditions, respectively, which were 21.92% and 22.20% lower than that by LQR, respectively. Besides, the slipping friction work by HDO were 0.49kJ and 0.11kJ lower than that by LQR in transportation and ploughing conditions, respectively. In different starting intentions and working conditions, the proposed method could effectively reduce the starting jerk and friction work of the power shifting tractor during the starting process, and had better starting quality and control robustness.