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

Assembly of multi-part physical structures is both a valuable end product forautonomous robotics, as well as a valuable diagnostic task for open-endedtraining of embodied intelligent agents. We introduce a naturalisticphysics-based environment with a set of connectable magnet blocks inspired bychildren's toy kits. The objective is to assemble blocks into a succession oftarget blueprints. Despite the simplicity of this objective, the compositionalnature of building diverse blueprints from a set of blocks leads to anexplosion of complexity in structures that agents encounter. Furthermore,assembly stresses agents' multi-step planning, physical reasoning, and bimanualcoordination. We find that the combination of large-scale reinforcementlearning and graph-based policies -- surprisingly without any additionalcomplexity -- is an effective recipe for training agents that not onlygeneralize to complex unseen blueprints in a zero-shot manner, but even operatein a reset-free setting without being trained to do so. Through extensiveexperiments, we highlight the importance of large-scale training, structuredrepresentations, contributions of multi-task vs. single-task learning, as wellas the effects of curriculums, and discuss qualitative behaviors of trainedagents.

Blocks Assemble! Learning to Assemble with Large-Scale Structured Reinforcement Learning