Rapid Procedural Generation of Real World Environments for Autonomous Vehicle Testing

Anonymous

Keywords: Autonomous vehicle, virtual testing, safety evaluation

TL;DR: An automated pipeline to create digital twin model of real world location for AV testing

Abstract: This study proposes an automated virtual test creation pipeline that replicates real-world environments in simulation for autonomous vehicle development. By integrating publicly available tools, the pipeline accelerates AV testing with minimal human intervention. Our approach develops real-world maps for open-source simulators like CARLA and LGSVL, facilitating testing and development of most exist- ing autonomous driving controllers, including Autoware and Apollo. This study outlines the methodology and results of our approach, highlighting its potential to significantly enhance the scalability and efficiency of autonomous vehicle testing and development.