Nonlinear state-observer techniques for sensorless control of automotive PMSM's, including load-torque estimation and saliency

The paper investigates various non-linear observer-based rotor position estimation schemes for sensorless control of permanent magnet synchronous motors (PMSMs). Attributes of particular importance to the application of brushless motors in the automotive sector, are considered e.g. implementation cost, accuracy of predictions during load transients, the impact of motor saliency and algorithm complexity. Emphasis is given to techniques based on model linearisation during each sampling period (EKF); feedback-linearisation followed by Luenberger observer design based on the resulting �linear� motor characteristics; and direct design of non-linear observers. Although the benefits of sensorless commutation of PMSMs have been well expounded in the literature, an integrated approach to their design for application to salient machines subject to load torque transients remains outstanding. Furthermore, this paper shows that the inherent characteristics of some non-linear observer structures are particularly attractive since they provide a phase-locked-loop (PLL)-type of configuration that can encourage stable rotor position estimation, thereby enhancing the overall sensorless scheme. Moreover, experimental results show how operation through, and from, zero speed, is readily obtainable. Experimental results are also employed to demonstrate the attributes of each methodology, and provide dynamic and computational performance comparisons.