Abstract: The shift quality of automated manual transmission(AMT) is difficult to control, and its main reason is that the dry clutch has no oil film buffer, and there are uncertain factors like various friction coefficient, the diaphragm spring decline, etc. To improve the shift quality, a powertrain shift control mode is proposed, which keeps clutch slip process and the rise of driving torque under control parallelly, along with that a control strategy of "disturbance optimal controller + model error observer" for shift process is designed. The modeling errors and disturbances in clutch and the plant model of vehicle powertrain are obtained through the disturbance observer firstly, which regarded as the system disturbance. Then the shift optimal controller which could deal with the rapid time-varying disturbances is designed with the comprehensive optimization between the slipping friction loss and jerk in any working condition. The results of the co-simulation (AMESim & Matlab/Simulink) and experimental tests show that, the proposed methods not only improve the shift quality, but also have self-adaptability to different working conditions, which solve the industry problem that it is hard for AMT vehicles to ensure the consistency of quality control effects.