Sim-to-Real via Sim-to-Seg: End-to-end Off-road Autonomous Driving Without Real Data

TL;DR

This paper introduces Sim2Seg, a novel simulation-to-real transfer method for off-road autonomous driving that trains end-to-end RL policies without using any real-world data, bridging the visual reality gap effectively.

Contribution

It presents a new approach that translates simulation images into segmentation and depth maps, enabling direct deployment of RL policies in real-world off-road driving without real data.

Findings

Achieves real-world autonomous driving performance comparable to traditional methods.

Trains in 48 hours on a single GPU, significantly reducing development time.

Effectively bridges the visual reality gap for off-road environments.

Abstract

Autonomous driving is complex, requiring sophisticated 3D scene understanding, localization, mapping, and control. Rather than explicitly modelling and fusing each of these components, we instead consider an end-to-end approach via reinforcement learning (RL). However, collecting exploration driving data in the real world is impractical and dangerous. While training in simulation and deploying visual sim-to-real techniques has worked well for robot manipulation, deploying beyond controlled workspace viewpoints remains a challenge. In this paper, we address this challenge by presenting Sim2Seg, a re-imagining of RCAN that crosses the visual reality gap for off-road autonomous driving, without using any real-world data. This is done by learning to translate randomized simulation images into simulated segmentation and depth maps, subsequently enabling real-world images to also be translated. This allows us to train an end-to-end RL policy in simulation, and directly deploy in the real-world. Our approach, which can be trained in 48 hours on 1 GPU, can perform equally as well as a classical perception and control stack that took thousands of engineering hours over several months to build. We hope this work motivates future end-to-end autonomous driving research.