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 Tools annual

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 Tools monthly

Global: Zendy Plus monthly

Presents the access of premium content as premium feature

Premium Content

Global: Zendy Plus annual

Global: Zendy Free Plan

Robotic reinforcement learning (RL) holds the promise of enabling robots tolearn complex behaviors through experience. However, realizing this promise forlong-horizon tasks in the real world requires mechanisms to reduce human burdenin terms of defining the task and scaffolding the learning process. In thispaper, we study how these challenges can be alleviated with an automatedrobotic learning framework, in which multi-stage tasks are defined simply byproviding videos of a human demonstrator and then learned autonomously by therobot from raw image observations. A central challenge in imitating humanvideos is the difference in appearance between the human and robot, whichtypically requires manual correspondence. We instead take an automated approachand perform pixel-level image translation via CycleGAN to convert the humandemonstration into a video of a robot, which can then be used to construct areward function for a model-based RL algorithm. The robot then learns the taskone stage at a time, automatically learning how to reset each stage to retry itmultiple times without human-provided resets. This makes the learning processlargely automatic, from intuitive task specification via a video to automatedtraining with minimal human intervention. We demonstrate that our approach iscapable of learning complex tasks, such as operating a coffee machine, directlyfrom raw image observations, requiring only 20 minutes to provide humandemonstrations and about 180 minutes of robot interaction.

AVID: Learning Multi-Stage Tasks via Pixel-Level Translation of Human Videos