Evaluation of a mapping strategy based on smooth arc splines for different road types

Abstract

Digital maps enrich advanced driver assistance systems by providing information on the local environment of a vehicle. This paper presents various results of a mapping strategy which uses smooth arc splines as geometric model. For any given tolerance, the curve approximation method (SMAP) generates a smooth arc spline with the minimally possible number of curve segments. The evaluation shows the performance of this method regarding the accuracy, the data volume and significant curvature characteristics on both rural and highway roads. It can be stated that the arc spline approximation generally outperforms polygonal representations regarding the information content and the number of segments which has a direct influence on the computational complexity of map calculations and the required data volume for the map storage. These properties are beneficial for many driver assistance system applications like autonomous driving. Furthermore, it is shown that the curvature estimation in the curve apexes is widely stable for a broad range of approximation tolerance values which is crucial for curve speed warnings.