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

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). Additionally, 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 these challenges by presenting a reimagining 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.

sim2seg_new_title.mp4

Using Sim2Seg , we successfully learn avoidance behaviors which transfer zero-shot to real-world.

In simulation, we train a goal-conditioned policy to reach waypoints while avoiding environmental collisions.

We train the policy entirely using semantically and visually diverse simulations, with no guarantees of similarity between simulated and real world scenes. During deployment, we deploy the trained policy zero-shot.

Three simulated scenes—"Meadow", "Landscapes", and "Canyon"

Real world deployment.

In simulation, Sim2Seg proposes trajectories around obstacles.

Using real-world images, Sim2Seg's trajectory proposals are affected by detection of perceived obstacles (such as rocks and shrubs).

Poster