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

Deep reinforcement learning (DRL) has been proven to be a powerful paradigmfor learning complex control policy autonomously. Numerous recent applicationsof DRL in robotic grasping have successfully trained DRL robotic agentsend-to-end, mapping visual inputs into control instructions directly, but theamount of training data required may hinder these applications in practice. Inthis paper, we propose a DRL based robotic visual grasping framework, in whichvisual perception and control policy are trained separately rather thanend-to-end. The visual perception produces physical descriptions of graspedobjects and the policy takes use of them to decide optimal actions based onDRL. Benefiting from the explicit representation of objects, the policy isexpected to be endowed with more generalization power over new objects andenvironments. In addition, the policy can be trained in simulation andtransferred in real robotic system without any further training. We evaluateour framework in a real world robotic system on a number of robotic graspingtasks, such as semantic grasping, clustered object grasping, moving objectgrasping. The results show impressive robustness and generalization of oursystem.

Towards Generalization and Data Efficient Learning of Deep Robotic Grasping