Lateral Vehicle State and Environment Estimation Using Temporally Previewed Mapped Lane Features

This paper proposes a model-based method to estimate lateral planar vehicle states using a forward-looking monocular camera, a yaw rate gyroscope, and an a priori map of road superelevation and temporally previewed lane geometry. Theoretical estimator performance from a steady-state Kalman-filter implementation of the estimation framework is calculated for various look-ahead distances and vehicle speeds. The application of this filter structure to real driving data is also explored, along with error characteristics of the filter on straight and curved roads, with both superelevated and flat profiles. The effect of superelevation on estimator performance is found to be significant. Experimental and theoretical analysis both show that the benefits of state estimation using previewed lane geometry improve with increasing lane preview, but this improvement diminishes due to increased lane tracking errors at distances beyond 20 m ahead of the vehicle.