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

Learning-based vehicle planning is receiving increasing attention with theemergence of diverse driving simulators and large-scale driving datasets. Whileoffline reinforcement learning (RL) is well suited for these safety-criticaltasks, it still struggles to plan over extended periods. In this work, wepresent a skill-based framework that enhances offline RL to overcome thelong-horizon vehicle planning challenge. Specifically, we design a variationalautoencoder (VAE) to learn skills from offline demonstrations. To mitigateposterior collapse of common VAEs, we introduce a two-branch sequence encoderto capture both discrete options and continuous variations of the complexdriving skills. The final policy treats learned skills as actions and can betrained by any off-the-shelf offline RL algorithms. This facilitates a shift infocus from per-step actions to temporally extended skills, thereby enablinglong-term reasoning into the future. Extensive results on CARLA prove that ourmodel consistently outperforms strong baselines at both training and newscenarios. Additional visualizations and experiments demonstrate theinterpretability and transferability of extracted skills.

Boosting Offline Reinforcement Learning for Autonomous Driving with Hierarchical Latent Skills