NDO-MPC: A Robust Servo-Motor Model Predictive Position Control Method Based on Noise-Suppressed Disturbance Observer

Model predictive control (MPC) of servo motors is sensitive to parameter mismatch and load disturbance, as the prediction law relies on an accurate plant model. To address this issue, we propose a robust position MPC scheme with a noise-suppressed disturbance observer. Starting from the servo-motor dynamics, we establish a lumped disturbance model of the current and speed loops by a truncated Taylor expansion, which accounts for electrical-parameter variations, mechanical-parameter variations and load changes. The resulting disturbance model is used to reconstruct both the position-servo prediction model and the associated linear constraints. For the current and speed disturbance channels, generalized proportional–integral observers with embedded first-order filtering are designed so that the disturbance and its first derivative can be estimated with improved high-frequency noise attenuation. The estimated quantities are employed online to compensate the prediction model and to correct the constraint bounds. A constrained convex program, comprising an integrated position–speed–current cost and the corrected inequality constraints, is then solved by the Hildreth algorithm and applied to the inverter through space-vector pulse-width modulation. Comparative experiments on a surface-mounted permanent-magnet servo platform demonstrate that the proposed method improves positioning performance under parameter mismatch and load disturbance.