Vehicular Networks Simulation with Realistic Physics

Abstract

Evaluation of cooperative automated driving applications requires the capability of simulating the vehicle and traffic dynamics as well as the communications with a level of accuracy that most of the current tools still lack. In this paper, we explore the use of game engines in hybrid traffic-network simulators. We describe and validate a novel framework based on this approach: Veneris. Our framework is made of a traffic simulator, implemented on the top of the Unity game engine, which includes a realistic vehicle model and a set of driving and lane change behaviors adapted to a 3D environment that reproduces real-world traffic dynamics; a ray-launching propagation simulator on graphics-processing-unit (GPU), called Opal, and a set of modules, which enable bidirectional coupling with the OMNET++ network simulator. The more relevant and novel mechanisms of Veneris are introduced, but further implementation details can be checked on the source code provided in our repository. We discuss the validation tests we have performed and show how it provides accurate results in three key areas: 1) the fidelity of the vehicle dynamics; 2) the recreation of realistic traffic flows, and; 3) the accuracy of the propagation simulation. In addition, the general results of the expected performance are provided.