Tronis® is an Unreal-Engine based software for the virtual prototyping of driver assistance systems for highly automated and autonomous driving, with which traffic scenarios can be efficiently modeled and simulated for testing. Given the high level of detail present in real-world urban road networks, a completely manual 3D modeling process for this purpose is often not feasible. Instead, a more abstract domain-specific description is preferable, as it is specified in the ASAM OpenDRIVE and OpenSCENARIO standards. However, as OpenDRIVE does not directly describe the road geometry as meshes, it always requires a suitable automatic mesh generation mechanism for rendering the traffic scene in a 3D engine. Algorithmic generation of road meshes comes with significant challenges given the diversity of environments due to the complexity and context-specific nature of real-world scenarios. Simplistic approaches might not fully capture irregularities or be able to prevent undesirable artifacts, and thereby may yield results that fail to correctly represent the true geometry of the road network. In this paper, partially funded by AI-NET-ANTILLAS, we describe our adaption of the widely known Marching Squares algorithm for the generation of road junctions from OpenDRIVE descriptions, which often lead to challenging geometric problems during their generation. We compare the result to other state-of-the-art implementations and discuss possible future improvements.
Session:
Poster
|