VALIDATION OF PARAMETRIC OPENDRIVE ROAD SPACE MODELS

Abstract

The standard OpenDRIVE is widely used for exchanging road space models in order to simulate the traffic of a city or individual driving situations. For modeling continuous road courses at lane level, OpenDRIVE utilizes its own parametric geometry model. However, violations of continuity requirements due to geometric leaps or kinks can cause the vehicle dynamics simulation to fail when testing vehicle components. But also defective lane predecessor and successor relations can result in an OpenDRIVE dataset not being usable as a reference map for vehicle navigation. Since these geometric, topological, and semantic constraints go beyond the rules encoded in the schema, this article presents a framework and a first implementation for validating OpenDRIVE datasets. As the lane widths are defined parametrically relative to the reference line of the respective road, lane connectivities at road transitions are evaluated using explicit geometries derived from the parametric geometry representations. Moreover, a derived CityGML representation enables a visual inspection of the parametric models to identify unexpected but visible defects. The implemented framework is applied to examine a total of 99 OpenDRIVE datasets, where significant lane gaps were detected in the explicit representation for about 20% of the datasets.