Abstract: The rotor speed and the position of a permanent magnet synchronous motor (PMSM) are estimated for building sensorless vector control system with speed and current double closed loops based on an extended Kalman filter (EKF) algorithm. To solve the problems of the biased estimation and the lack of robustness against the errors of the model in the EKF, a sliding mode speed controller is proposed based on the exponent reaching law. To improve the capacity of anti-load-torque-disturbance of speed loop, a load torque observer is designed and the result of observation is imported into the input of current controller functioning as the controlling input of the feed-forward compensation. The simulation results show that the proposed control strategy is superior to conventional PI (proportional-integral) speed controller for its smaller speed tracking errors, faster response, no overshoot, stronger ability of anti-load-torque-disturbance, etc.