State Estimation By Observer Using Median Operation for Observed Output with Outliers

A design method for a state estimation observer for observed outputs with outliers is proposed in this paper. Outliers are one of the most important problems to be solved in actual systems control, such as engine valve control in automobiles and motion control using camera devices. When outliers occur in output signals, the output data for the control system becomes invalid. The state estimation observer does not work well because of this invalid value. Therefore, the effect of the outliers needs to be overcome for implementing state feedback control using state estimators. Accordingly, we propose a method to remove the effect of the outliers for the estimated state using median operation. The effectiveness of the observer based on the median of candidate vectors is verified by numerical simulation.