# Learning to Drive from Simulation without Real World Labels

## TL;DR

This paper introduces a method for transferring vision-based driving policies learned in simulation to real-world rural roads without using real-world labels, leveraging image-to-image translation for domain adaptation.

## Contribution

It presents a novel approach combining domain transfer with joint learning of control policies directly from simulation data, eliminating the need for real-world labels.

## Key findings

- Effective domain transfer achieved with image-to-image translation.
- Successful real-world deployment on rural and urban roads.
- Comparable performance to methods using real-world labels.

## Abstract

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.