For hybrid electric vehicles with traction motors installed behind transmission, traction motors can compensate the torque interrupt during gear shifting. Thus engine speed is adjusted to achieve the speed synchronization with AMT so that shifting smoothness is guaranteed without disengaging clutch. Quantitative feedback theory (QFT) combined with Smith predictor (SP) was adopted to achieve engine speed control. Firstly, a high realistic model of naturally aspirated gasoline engine was established in AMESim and the speed characteristic was validated to guarantee the reasonable features. Secondly, the engine nonlinear model was identified under piecewise working conditions so that several linear models with regions of parametric uncertainty were achieved. Thirdly, an optimized Smith predictor model was selected based on two criteria in frequency domain so as to deal with the intaketotorque uncertain time delay. Finally, based on quantitative feedback theory, an engine speed SP-QFT robust controller was designed to guarantee the robust stability and reference tracking. The simulation results showed that the step responses between the minimum and maximum speeds were of 0.75% overshoot, 0.7s settling time and 6r/min steady state error which met the demands of performance requirements of the synchronizer engagement during gear shifting of AMT and obviously improved the system dynamic characteristics compared with QFT controller and PID controller. Moreover, the simple form of SP-QFT controller is convenient to implement in engineering.