Image-based position estimation of UAV using Kalman Filter

This paper deals with the position estimation problem by using the Kalman Filter with compensations for unexpected observations. In the position estimation problem, robot observations sometimes yield unexpected values, resulting in the deterioration of the estimation accuracy. For example, visual observation with an unmanned aerial vehicle often yields unexpected results because of blurred images. In this paper, we propose a method to assigns weights to the observations in order to remove the effects of unexpected observations. In the proposed method, unexpected observations are detected by comparing the observation values with its estimates; the weights of these observations are then determined. On the basis of simulation and experimental results, we demonstrate that a robot´s position can be estimated by the proposed method.