| 초록 |
The Kalman filter has been considered as a powerful tool for many tracking and prediction tasks in signal processing. The algorithm of Kalman filter calculates more optimal estimates by combining predictions of a system model and measurements, compared to a low-pass filter or a moving average filter. However, the Kalman filter is so sensitive to model uncertainties that it is impractical to know the system model completely. When the model uncertainty is significant, the estimation errors from the inaccurate system model cannot be removed. Therefore, in this research, a new Kalman filter structure using an incremental proportional-integral-derivative (PID) form is proposed to compensate the differences between the measurements and the predictions. Then the proposed structure is applied to SISO (single-input single-output) and MIMO (multi-input multi-output) system for case studies. The case studies show that the proposed Kalman filter shows more accurate and robust output estimation performance regardless of model uncertainties than the previous approaches. |
