A 3D Simulation Εnvironment for Αssessing Interactions with Autonomous Vehicles

Abstract

Self-driving passenger cars are already on public roads to support the vision of a sustainable city. The research community and academia should further support the human-machine symbiosis of autonomous vehicles, manually driven vehicles, and even pedestrians. At the time being, most vehicles operate at level 2 of the Society of Engineers (SAE) autonomy scale as they only have partial automation. However, they are expected to escalate due to cost reduction of essential hardware components (sensors and actuators) and software advancements in Deep Neural Networks (DNN). The main challenge is thoroughly testing self-driving behaviors and human-machine symbiosis scenarios for improving safety conditions and engaging citizens in this concept. Video game simulators could provide a basis for testing and showcasing the innovative features of autonomous driving. Unreal Engine and Unity are two prominent examples of software tools for creating photorealistic graphical environments. In addition, the scientific community and academia elaborate on open-source simulators due to their interoperability features for data exchange and their inherent connections with robotic frameworks. The proposed research work presents an integration framework for the CARLA simulator. The framework uses an integrated Graphical User Interface (GUI) to facilitate the deployment of a fully functional simulation that enables users to include an environment for the passenger cars and several Non-Playable Characters (NPCs) that will participate in the simulation. The purpose of this study is to exploit the interactions of manually driven and autonomous vehicles in a 3D simulation environment and provide evidence that further supports human-machine symbiosis on public roads.