Performance evaluation of nonlinear Kalman filtering techniques in low speed brushless DC motors driven sensor-less positioning systems

Appropriate position estimation of electric actuator is one of the most important stages in the control and automation processes. Considering various applications including robotics and automotive systems, Brushless DC (BLDC) motors sensor-less positioning is substantial in two major aspects. Electric commutation on the electric side should be performed precisely as well as motion control on the mechanical side. The mathematical models which befit precise applications are inherently nonlinear and require specific control techniques to deal with. Nonlinear Kalman filtering methods are considered as suitable solutions to estimation problems where uncertainty and noise exist. Three major and basic algorithms are Extended (EKF), Unscented (UKF) and Cubature Kalman filtering (CKF). In this paper, the application of these methods in estimation of rotor angular position with emphasis on low speed state is presented. Performance measures are compared using experimental setup. A typical 3-phase low voltage BLDC motor is implemented in the setup so that system noise could deteriorate quality of Back-EMF signal in low speed mode. It is shown that UKF and CKF techniques yield better results and performance in comparison to EKF according to measures. Estimated model states diagrams indicate the superior performance of Unscented and Cubature types regarding both accuracy and convergence.