Sliding-Mode Speed-Then-Position-Based Sensorless Control Strategy With Convergence Effect Mitigation for Robotic SPMSMs Over Low-Speed Range

This paper proposes a sliding-mode speed-then position-based (SM-STP) sensorless control strategy for the low speed range. Firstly, by extracting speed directly from the current model rather than relying on high-frequency injection, the SM-STP achieves noise-free operation and applicability to SPMSMs without relying on magnetic saliency. Secondly, the paper systematically analyzes the influence of the convergence effect (CE) on speed and estimated position errors, identifying the inherent time delays that lead to performance degradation under large load transients. To mitigate the CE, an error compensation manifold (ECM) based on deep symbolic regression (DSR) and data filling (DF) is employed, in which the DSR is innovatively used to generate large-load transient data, that cannot be obtained accurately. The DF method is further applied to generate compensation data covering the entire low-speed operating range. Finally, the feasibility and robustness of the SM STP is verified, and the position error remains below 0.02 rad under operating conditions at 100 rpm and below. Meanwhile, the effectiveness of the compensation method is verified. It not only resolves the stalling problem under large load transients, but also limits the position error at to only 0.012 rad.