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

Efficient traffic signal control is critical for reducing traffic congestionand improving overall transportation efficiency. The dynamic nature of trafficflow has prompted researchers to explore Reinforcement Learning (RL) fortraffic signal control (TSC). Compared with traditional methods, RL-basedsolutions have shown preferable performance. However, the application ofRL-based traffic signal controllers in the real world is limited by the lowsample efficiency and high computational requirements of these solutions. Inthis work, we propose DTLight, a simple yet powerful lightweight DecisionTransformer-based TSC method that can learn policy from easily accessibleoffline datasets. DTLight novelly leverages knowledge distillation to learn alightweight controller from a well-trained larger teacher model to reduceimplementation computation. Additionally, it integrates adapter modules tomitigate the expenses associated with fine-tuning, which makes DTLightpractical for online adaptation with minimal computation and only a fewfine-tuning steps during real deployment. Moreover, DTLight is further enhancedto be more applicable to real-world TSC problems. Extensive experiments onsynthetic and real-world scenarios show that DTLight pre-trained purely onoffline datasets can outperform state-of-the-art online RL-based methods inmost scenarios. Experiment results also show that online fine-tuning furtherimproves the performance of DTLight by up to 42.6% over the best online RLbaseline methods. In this work, we also introduce Datasets specificallydesigned for TSC with offline RL (referred to as DTRL). Our datasets and codeare publicly available.

Traffic Signal Control Using Lightweight Transformers: An Offline-to-Online RL Approach