Raw GIS to 3D road modeling for real-time traffic simulation

In this work, we propose a new approach to road modeling and 3D traffic simulation. Based on the raw geographic information system (GIS) data laid out as sparse polylines with attributes, we compute a more adequate functional description for real-time simulation of on-road vehicle animation. The proposed approach begins with a filtering/subdivision module where the raw polylines are transformed into a graph of functional road segments as arcs and the nodes as intersections. Then, the vehicle speed profile is computed based on its dynamics, its neighborhood and the curvature profile of the road. Afterward, a multi-agent system is proposed in order to handle a large number of simulated vehicle/driver couples. Finally, we deploy a 3D rendering engine to display the computed 3D simulation on screen. The resulting model satisfies most of the real road features for traffic simulation including road interchanges, roundabouts, intersections, lanes, etc. More importantly, the simulated driving qualitatively mimics the real behavior of the drivers/vehicles on the road as can be seen in the accompanying video (RTSP video). We also validate our findings with a technical assessment based on macroscopic and microscopic traffic simulation metrics in several road traffic scenarios.