Abstract: To address the speed control problem of the permanent magnet synchronous motor (PMSM), we present a sliding mode control strategy with anti-windup strategy. The strategy has advantages such as quick response, weak chattering, and overshoot. The main work involves the following:First, to weaken the chattering and reduce the adjust time of sliding mode control, we replace switch control by using radial basis function neural network, which greatly weakens the chattering and decreases the settling time. Second, considering the saturation effect on system overshoot given that the motor is a physical limited object, we introduce the attractor exponential coefficient to the design of an integral part of the sliding mode surface and feedback inhibition. We introduce an anti-windup strategy with separation type of proportional integral to the system. Simulation results show that the proposed method can effectively weaken the overshoot and the chattering in sliding mode control to address the speed control problem of PMSM.