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

Humans are remarkably proficient at controlling their limbs and tools from awide range of viewpoints and angles, even in the presence of opticaldistortions. In robotics, this ability is referred to as visual servoing:moving a tool or end-point to a desired location using primarily visualfeedback. In this paper, we study how viewpoint-invariant visual servoingskills can be learned automatically in a robotic manipulation scenario. To thisend, we train a deep recurrent controller that can automatically determinewhich actions move the end-point of a robotic arm to a desired object. Theproblem that must be solved by this controller is fundamentally ambiguous:under severe variation in viewpoint, it may be impossible to determine theactions in a single feedforward operation. Instead, our visual servoing systemmust use its memory of past movements to understand how the actions affect therobot motion from the current viewpoint, correcting mistakes and graduallymoving closer to the target. This ability is in stark contrast to most visualservoing methods, which either assume known dynamics or require a calibrationphase. We show how we can learn this recurrent controller using simulated dataand a reinforcement learning objective. We then describe how the resultingmodel can be transferred to a real-world robot by disentangling perception fromcontrol and only adapting the visual layers. The adapted model can servo topreviously unseen objects from novel viewpoints on a real-world Kuka IIWArobotic arm. For supplementary videos, see:https://fsadeghi.github.io/Sim2RealViewInvariantServo

Sim2Real View Invariant Visual Servoing by Recurrent Control