Simulation can be a powerful tool for understanding machine learning systems and designing methods to solve real-world problems. Training and evaluating methods purely in simulation is often "doomed to succeed" at the desired task in a simulated environment, but the resulting models are incapable of operation in the real world. Here we present and evaluate a method for transferring a vision-based lane following driving policy from simulation to operation on a rural road without any real-world labels. Our approach leverages recent advances in image-to-image translation to achieve domain transfer while jointly learning a single-camera control policy from simulation control labels. We assess the driving performance of this method using both open-loop regression metrics, and closed-loop performance operating an autonomous vehicle on rural and urban roads.
@article{arxiv.1812.03823,
title = {Learning to Drive from Simulation without Real World Labels},
author = {Alex Bewley and Jessica Rigley and Yuxuan Liu and Jeffrey Hawke and Richard Shen and Vinh-Dieu Lam and Alex Kendall},
journal= {arXiv preprint arXiv:1812.03823},
year = {2018}
}