Methodology for Creating Urban Environment of Dubai in CARLA Simulator for Automated Driving Training

The CARLA is a simulator platform providing training, prototyping, and validation of models for automated driving using open-source code, protocols and digital assets, such as buildings, vehicles, urban layouts, etc. It helps researchers a faster and safer way to accumulate automated driving experience. However, unavailability of utility vehicles and limited geographical environments hinders the simulation of autonomous utility vehicles or in fact the scenarios of environment interaction with the utility vehicle. Further availability of local environment helps in better learning of automated driving models. The focus in the current work is to develop UAE relevant environment in the CARLA simulator, creating and importing commercial utility vehicles, and validating the same. For the current scope, Dubai International Academic City (DIAC), Dubai is the chosen location. The DIAC environment is created using an open-source 3D computer graphics software Blender and imported into CARLA. Further, a utility vehicle is modeled, rigged, and imported into CARLA. Finally, the simulation of autonomous operation of the utility vehicle on the DIAC road inside CARLA replicates the real scenario virtually, thus providing a virtual asset to future self-driving vehicle operators to train their automated driving software in the Dubai scenario.