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

It is desirable for future robots to quickly learn new tasks and adaptlearned skills to constantly changing environments. To this end, ProbabilisticMovement Primitives (ProMPs) have shown to be a promising framework to learngeneralizable trajectory generators from distributions over demonstratedtrajectories. However, in practical applications that require high precision inthe manipulation of objects, the accuracy of ProMPs is often insufficient, inparticular when they are learned in cartesian space from external observationsand executed with limited controller gains. Therefore, we propose to combineProMPs with recently introduced Residual Reinforcement Learning (RRL), toaccount for both, corrections in position and orientation during taskexecution. In particular, we learn a residual on top of a nominal ProMPtrajectory with Soft-Actor Critic and incorporate the variability in thedemonstrations as a decision variable to reduce the search space for RRL. As aproof of concept, we evaluate our proposed method on a 3D block insertion taskwith a 7-DoF Franka Emika Panda robot. Experimental results show that the robotsuccessfully learns to complete the insertion which was not possible beforewith using basic ProMPs.

Residual Robot Learning for Object-Centric Probabilistic Movement Primitives