Linear recursive target state and time-to-collision estimator for automotive collision warning system

This paper proposes a new linear recursive target state and time-to-collision estimator for the development of the automotive collision warning system. The addressed problem can be cast into the representative nonlinear state estimation under cluttered environment. To prevent the tracking performance degradation due to the inherent nonlinearity between the polar coordinates measurements of the automotive radar and the target state, a practical linear filter design scheme employing the estimated line-of-sight (LOS) Cartesian coordinate system (ELCCS) is proposed. ELCCS is redefined by using a priori LOS estimates in every update period in order to ensure the unbiasedness of the proposed linear tracking filter. Moreover, in order to effectively cope with the cluttered environment and to enhance the target tracking performance, a modified probabilistic data association filter (MPDAF) is newly proposed. Finally, using the most probable closing velocity and range measurements within the validation region, a quasi-optimal linear robust time-to-collision (TTC) estimator is designed. For its linear recursive filter structure, the proposed method is more suitable for the development of the performed and reliable collision warning system. The performance of the proposed scheme is demonstrated by computer simulations.