Abstract
In the lane recognition system of driving support devices typically known as Lane Keeping Assist System and Lane Keeping System, it is necessary that coordinates of lane markers detected by image processing of forward road image are converted into practical values for steering control, i.e. lateral distance, heading angle to the lane and curvature of the lane. To realize state steering control, the Kalman Filter is widely applied for this conversion process which is called estimation process of lane state. In this paper, in addition to forward lane points usually used for estimation of lane state, backward lane points, which are obtained from vehicle motion without using backward camera are newly introduced as observables of the Kalman Filter. Also, lane shape is approximated by cubic curve to consider clothoid parameter of the lane. Conducting estimation of lane state using the developed Kalman Filter, it is indicated that noises of estimated values are reduced in comparison with estimated values obtained from a previous model without degradation of estimation performance of maximum curvature.
