Estimation of a Robot Vehicle Position by Kalman Filtering

Abstract

Extended Kalman filter is applied to the estimation of robot vehicle position with optical and wheel measurements. Measurement of wheel rotation is used to compute its nominal trajectory. A model that a contact point of a wheel with floor changes randomly is introduced to obtain a state equation with respect to the deviation. A fast slit-pattern measurement using a TV camera is formulated as an observation equation. Then, a Kalman filter and a fixed point smoother are constructed. Several experiments are conducted to evaluate the accuracy of estimated trajectory.