Robust grid-based road detection for ADAS and autonomous vehicles in urban environments

Abstract

For future advanced driver assistant systems a detailed knowledge about the road network in the immediate surroundings rises in importance for several reasons. It increases the robustness of scene interpretation especially in urban environments on the one hand, on the other hand it can be used for map-matching approaches to ensure a lane-accurate matching within an a priori map. In many inner-city side roads there are no lane markings available at all and curbs might be occluded by parking cars. Thus, the only information on the road course can be extracted from the stationary environment. In this paper we present an approach for road course detection in urban environments which is robust against a big variety of urban scenes. Our approach works with sensor data obtained from the road surface as well as from raised buildings. The measurements are obtained from a close-to-production laser-scanner and are accumulated in an occupancy grid. The algorithm runs online in real-time on a standard PC and has been evaluated in real urban environments.