Abstract: Aiming at coping with the dual clutch transmission (DCT) shifting process control under different working conditions, a knowledge-based shifting control method is proposed in this paper. By implementing the ensemble learning algorithm, the clutch target torque generated based on the actual vehicle data is applied online to the DCT shift process. To ensure that the ensemble learning algorithm can accurately generate the target torque according to the speed measured by the sensor, model predictive control (MPC) is used to reduce the deviation between the real speed and the reference speed of the system. According to the state dynamic equation during DCT shifting process, the state prediction model of MPC controller is established. The objective function of shifting control is constructed with the constraints of state variables and control variables. Moreover, the rolling optimization of the shifting control is realized by MPC. The DCT upshift and downshift processes are simulated on the Matlab/Simulink co-simulation platform, respectively. The shifting quality of the proposed knowledge-based method is compared with the fuzzy control and the actual vehicle calibration control. The results show that the proposed knowledge-based shifting control method can apply the shifting control laws accurately from the actual vehicle data by data mining. Furthermore, the goal of fast and smooth shifting is achieved, which indicates that it is more intelligent shifting control and has better shifting performance.