Abstract: To achieve high tracking and robustness performance in a servo system, considering the time-varying nature of friction and load, a control scheme based on a diagonal recurrent neural network controller is proposed to improve the system's tracking performance and disturbance rejection. Simulation results show that the method using the diagonal recurrent neural network controller has good dynamic performance and good robustness to load disturbances and motor parameter changes. Keywords: Permanent-magnet synchronous motor (PMSM); Diagonal recurrent neural network; Disturbance observer Abstract: In order to improve the tracking and robust performance of a servo system, due to the effects of nonlinear friction and time-varying load, a diagonal recurrent neural network controller is used to improve the tracking and anti-interference. The simulation test results indicate that this scheme has good dynamic performance and strong robustness to both load disturbance and the parametric variation of motor. Keywords: permanent-magnet synchronous motor (PMSM); diagonal recurrent neural network; disturbance observer 1 Introduction In servo systems, disturbance is the main factor causing the degradation of system servo performance. Therefore, for a servo system, disturbance immunity and tracking performance are important indicators for measuring its servo performance. The proportional-integral position controller can achieve the tracking performance of the system when the parameters do not change significantly. The disturbance observer works well at low frequencies and with small disturbances, but its disturbance immunity becomes unsatisfactory when the disturbance is large or the frequency is high. In addition, the proportional-integral position controller cannot effectively track periodic sinusoidal input signals. Therefore, to improve the performance of permanent magnet synchronous motor (PMSM) drive systems under different periodic signal inputs, this paper proposes a recursive neural network controller to control the position system of the PMSM drive. Simulation results show that this control scheme has good real-time performance and strong robustness, speed, and accuracy. For details, please click to download: AC Servo System Controlled by Diagonal Recursive Neural Network.