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

The latent world model provides a promising way to learn policies in acompact latent space for tasks with high-dimensional observations, however, itsgeneralization across diverse environments with unseen dynamics remainschallenging. Although the recurrent structure utilized in current advanceshelps to capture local dynamics, modeling only state transitions without anexplicit understanding of environmental context limits the generalizationability of the dynamics model. To address this issue, we propose a PrototypicalContext-Aware Dynamics (ProtoCAD) model, which captures the local dynamics bytime consistent latent context and enables dynamics generalization inhigh-dimensional control tasks. ProtoCAD extracts useful contextual informationwith the help of the prototypes clustered over batch and benefits model-basedRL in two folds: 1) It utilizes a temporally consistent prototypicalregularizer that encourages the prototype assignments produced for differenttime parts of the same latent trajectory to be temporally consistent instead ofcomparing the features; 2) A context representation is designed which combinesboth the projection embedding of latent states and aggregated prototypes andcan significantly improve the dynamics generalization ability. Extensiveexperiments show that ProtoCAD surpasses existing methods in terms of dynamicsgeneralization. Compared with the recurrent-based model RSSM, ProtoCAD delivers13.2% and 26.7% better mean and median performance across all dynamicsgeneralization tasks.

Prototypical context-aware dynamics generalization for high-dimensional model-based reinforcement learning