Development of a New Statistical-Based Estimation Algorithm for Hybrid Control of Position, Speed, and Torque of Permanent Magnet Motors in Discontinuous and Continuous Operation

Accurate rotor position estimation is essential for reliable and efficient control of Permanent Magnet Synchronous Machines (PMSMs), particularly in sensorless drive applications where mechanical sensors are unavailable. This paper introduces a hybrid estimation framework that combines Maximum Likelihood Estimation based position estimation (MLEBPE), Particle Filtering (PF), and Dempster-Shafer Theory (DST) with sector-adaptive covariance modeling. The MLEBPE stage provides computationally efficient sector-based angle estimates, while the PF offers robustness against nonlinearities and measurement noise by propagating probabilistic state hypotheses. The DST framework resolves conflicts between estimators and manages uncertainty by fusing interval-based evidence. Sector-wise covariance adaptation further enhances performance near sector boundaries where uniform noise assumptions fail. Evaluation results show that the proposed hybrid method significantly improves accuracy, reducing root-mean-square error by approximately 35% compared to standalone MLEBPE. These findings demonstrate the effectiveness of combining probabilistic and evidential approaches, making the method suitable for robust and noise-resilient rotor position estimation in real-time PMSM sensorless drives.