English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

Autonomous driving is complex, requiring sophisticated 3D sceneunderstanding, localization, mapping, and control. Rather than explicitlymodelling and fusing each of these components, we instead consider anend-to-end approach via reinforcement learning (RL). However, collectingexploration driving data in the real world is impractical and dangerous. Whiletraining in simulation and deploying visual sim-to-real techniques has workedwell for robot manipulation, deploying beyond controlled workspace viewpointsremains a challenge. In this paper, we address this challenge by presentingSim2Seg, a re-imagining of RCAN that crosses the visual reality gap foroff-road autonomous driving, without using any real-world data. This is done bylearning to translate randomized simulation images into simulated segmentationand 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 directlydeploy in the real-world. Our approach, which can be trained in 48 hours on 1GPU, can perform equally as well as a classical perception and control stackthat took thousands of engineering hours over several months to build. We hopethis work motivates future end-to-end autonomous driving research.

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