Abstract
This paper presents a current-sensorless control architecture for a buck DC–DC converter that integrates input–output feedback linearization (IOFL) with an extended Kalman filter (EKF). IOFL is applied to the averaged converter model to obtain a linear input–output relation, allowing the proportional–integral–derivative (PID) gains to be selected directly from a prescribed damping ratio and natural frequency. To eliminate direct inductor-current feedback, an augmented EKF jointly estimates the inductor current and load resistance using the measured input and output voltages together with the applied duty ratio. A common lossy averaged model is employed in both the IOFL cancellation law and the EKF prediction step, thereby maintaining model consistency between the controller and estimator. The proposed architecture is experimentally evaluated under start-up conditions and abrupt changes in load resistance, input voltage, and reference voltage. Compared with a fixed-gain proportional–integral (PI) controller, the proposed method reduces the inductor-current overshoot at start-up from to and the start-up time from to . For a -to- input-voltage step, the output-voltage overshoot is reduced from to . The EKF yields maximum absolute relative estimation errors of for both the inductor current and load resistance. These experimental results demonstrate the feasibility of real-time current-sensorless nonlinear control with online load-resistance estimation on an embedded buck-converter platform.