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

Robotic manipulation requires accurate perception of the environment, whichposes a significant challenge due to its inherent complexity and constantlychanging nature. In this context, RGB image and point-cloud observations aretwo commonly used modalities in visual-based robotic manipulation, but each ofthese modalities have their own limitations. Commercial point-cloudobservations often suffer from issues like sparse sampling and noisy output dueto the limits of the emission-reception imaging principle. On the other hand,RGB images, while rich in texture information, lack essential depth and 3Dinformation crucial for robotic manipulation. To mitigate these challenges, wepropose an image-only robotic manipulation framework that leverages aneye-on-hand monocular camera installed on the robot's parallel gripper. Bymoving with the robot gripper, this camera gains the ability to activelyperceive object from multiple perspectives during the manipulation process.This enables the estimation of 6D object poses, which can be utilized formanipulation. While, obtaining images from more and diverse viewpointstypically improves pose estimation, it also increases the manipulation time. Toaddress this trade-off, we employ a reinforcement learning policy tosynchronize the manipulation strategy with active perception, achieving abalance between 6D pose accuracy and manipulation efficiency. Our experimentalresults in both simulated and real-world environments showcase thestate-of-the-art effectiveness of our approach. %, which, to the best of ourknowledge, is the first to achieve robust real-world robotic manipulationthrough active pose estimation. We believe that our method will inspire furtherresearch on real-world-oriented robotic manipulation.

RGBManip: Monocular Image-based Robotic Manipulation through Active Object Pose Estimation