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

Large-scale training have propelled significant progress in varioussub-fields of AI such as computer vision and natural language processing.However, building robot learning systems at a comparable scale remainschallenging. To develop robots that can perform a wide range of skills andadapt to new scenarios, efficient methods for collecting vast and diverseamounts of data on physical robot systems are required, as well as thecapability to train high-capacity policies using such datasets. In this work,we propose a framework for scaling robot learning, with specific focus onmulti-task and multi-scene manipulation in kitchen environments, both insimulation and in the real world. Our proposed framework, CACTI, comprises fourstages that separately handle: data collection, data augmentation, visualrepresentation learning, and imitation policy training, to enable scalabilityin robot learning . We make use of state-of-the-art generative models as partof the data augmentation stage, and use pre-trained out-of-domain visualrepresentations to improve training efficiency. Experimental resultsdemonstrate the effectiveness of our approach. On a real robot setup, CACTIenables efficient training of a single policy that can perform 10 manipulationtasks involving kitchen objects, and is robust to varying layouts ofdistractors. In a simulated kitchen environment, CACTI trains a single policyto perform 18 semantic tasks across 100 layout variations for each individualtask. We will release the simulation task benchmark and augmented datasets inboth real and simulated environments to facilitate future research.

CACTI: A Framework for Scalable Multi-Task Multi-Scene Visual Imitation Learning