A new method for robust far-distance road course estimation in advanced driver assistance systems

Abstract

An advanced method for road course estimation is presented. It is based on the state-of-the-art Kalman filter lane detection and allows for a robust sensor-based estimation of road courses in great distances. Only the parameters for the road course are estimated which results in a reduced parameter space and therewith more robustness. Instead of laterally displaced single feature points tangential structures are used as measurements in the filter model. Therefore the method is translation-invariant and applicable for all continuous differentiable road course models. As shown with video and radar input examples it is also sensor-independent and particularly suitable for sensor fusion approaches. For accuracy estimations an advanced method based on inertial navigation is used which is independent of lateral movements of the host vehicle and the road model.