Sponsored by: NVidia

Abstract

Testing autonomous vehicles on the roads can be expensive, time-consuming, and dangerous; therefore NVIDIA is developing DRIVE Sim simulation software to virtually test autonomous vehicle behavior reproduced from real traffic conditions. The goal of our project is to create a tool that generates scenarios to be replayed in DRIVE Sim from logs collected during real test drives. We utilized existing analytics tools to create scenario files by extracting the events of interest and developed a program that recreates the starting conditions of these events, such as the initial location and speed of the autonomous vehicle under test and cars surrounding it. These scenarios are then played in DRIVE SIM to assist in the RoadRunner software testing and development process. The evaluation of our tool's accuracy demonstrates that we were able to successfully reproduce the initial conditions of the cars in DRIVE Sim given the current fidelity of the software and log data.

Download

• Final report: pdf