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

Data efficiency in robotic skill acquisition is crucial for operating robotsin varied small-batch assembly settings. To operate in such environments,robots must have robust obstacle avoidance and versatile goal conditioningacquired from only a few simple demonstrations. Existing approaches, however,fall short of these requirements. Deep reinforcement learning (RL) enables arobot to learn complex manipulation tasks but is often limited to small taskspaces in the real world due to sample inefficiency and safety concerns. Motionplanning (MP) can generate collision-free paths in obstructed environments, butcannot solve complex manipulation tasks and requires goal states oftenspecified by a user or object-specific pose estimator. In this work, we proposea system for efficient skill acquisition that leverages an object-centricgenerative model (OCGM) for versatile goal identification to specify a goal forMP combined with RL to solve complex manipulation tasks in obstructedenvironments. Specifically, OCGM enables one-shot target object identificationand re-identification in new scenes, allowing MP to guide the robot to thetarget object while avoiding obstacles. This is combined with a skilltransition network, which bridges the gap between terminal states of MP andfeasible start states of a sample-efficient RL policy. The experimentsdemonstrate that our OCGM-based one-shot goal identification providescompetitive accuracy to other baseline approaches and that our modularframework outperforms competitive baselines, including a state-of-the-art RLalgorithm, by a significant margin for complex manipulation tasks in obstructedenvironments.

Efficient Skill Acquisition for Complex Manipulation Tasks in Obstructed Environments