CARLA Connect: A Connected Autonomous Vehicle (CAV) Driving Simulator

Abstract

Emerging technologies in the transportation field benefit greatly from vehicle simulators. These simulations help developers to validate the algorithms, test various dangerous edge-case scenarios, increase the speed at which the problems can be solved, reduce high costs by original equipment manufacturers (OEMs), etc. One such emerging technology is connected vehicles (CVs) which enable communication with neighboring vehicles and road infrastructure. This paper presents ‘CARLA Connect’, a Connected Autonomous Vehicle (CAV) driving simulator framework. For this, we have used CARLA, an open-source Autonomous Vehicle (AV) only driving simulator framework, and extended it to support AV, CV, and CAV features. CARLA Connect allows continuous development and testing of CV and CAV algorithms to visually validate them in real-time simulation. First, we developed two custom sensors needed in CV technology, a connectivity agent that supports both vehicle-to-vehicle (V2V) and vehicle-to-infrastructure (V2I) communication, and modular CV and CAV agents which contain the algorithms. These developed sensors and agents can now be used by developers to create and test their own scenarios. Using CARLA Connect, we have developed and validated six V2V CV algorithms, three V2I CV algorithms, and two V2V CAV algorithms. CARLA provides a pygame window where warnings can be displayed on the screen, simulating a human-machine interface (HMI). We use this to visually verify the warnings given to the driver and validate the underlying algorithms.