Beyond Traditional Simulators: Adopting Videogames for Autonomous Vehicle Testing

Abstract

This article presents an ongoing research study investigating the potential of using Grand Theft Auto V (GTAV) as a simulation environment for testing edge cases in selfdriving vehicles. Our hypothesis is that GTAV can offer a more accessible and cost-effective solution for testing edge cases compared to established simulation environments such as CARLA. In this work-in-progress study, we aim to assess the feasibility, scalability, and effectiveness of utilizing GTAV by simulating complex driving scenarios. We selected Openpilot, an opensource Advanced Driver-Assistance System (ADAS), to test these scenarios. Compared to CARLA, GTAV proved to be a better choice for testing realistic urban environments, dynamic weather, complex intersections, and detailed nighttime settings. Moreover, these simulations offered valuable insights into how Openpilot reacts to various driving conditions and how it could be enhanced to handle edge cases more efficiently.