A Sensorless Permanent Magnet Synchronous Machine Servo Drive Assisted by the Visual Information

The advanced sensorless technology of the permanent magnet synchronous machine (PMSM) can replace the position sensor on the machine side, however, the encoder is still the necessary feedback on the load side for a full-closed-loop servo system. This article proposes a sensorless drive scheme assisted by the computer vision to further replace the load side encoders, so to form a quasi-sensorless servo system. Although the vision technology can easily provide the load position for the closed-loop motion control, the feedback continuity cannot be satisfied especially for the multi-object scenarios, where the vision capture device and the computation ability cannot provide an ideal image processing. The issue of spatial-temporal discontinuity is discussed and the solution on basis of fusing the electrical and kinetic models is proposed. With this idea, the only feedback of the full-closed-loop servo drive is the computer vision, the motor side and load side encoders are both cancelled, and the reliable control performance with respect to the heavy load and high precision positioning is maintained. The scheme is validated on a heavy-load full-closed-loop servo bench driven by a sensorless PMSM with only feedback captured from a regular camera.