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

A popular paradigm in robotic learning is to train a policy from scratch forevery new robot. This is not only inefficient but also often impractical forcomplex robots. In this work, we consider the problem of transferring a policyacross two different robots with significantly different parameters such askinematics and morphology. Existing approaches that train a new policy bymatching the action or state transition distribution, including imitationlearning methods, fail due to optimal action and/or state distribution beingmismatched in different robots. In this paper, we propose a novel method named$REvolveR$ of using continuous evolutionary models for robotic policy transferimplemented in a physics simulator. We interpolate between the source robot andthe target robot by finding a continuous evolutionary change of robotparameters. An expert policy on the source robot is transferred throughtraining on a sequence of intermediate robots that gradually evolve into thetarget robot. Experiments on a physics simulator show that the proposedcontinuous evolutionary model can effectively transfer the policy across robotsand achieve superior sample efficiency on new robots. The proposed method isespecially advantageous in sparse reward settings where exploration can besignificantly reduced. Code is released at https://github.com/xingyul/revolver.

REvolveR: Continuous Evolutionary Models for Robot-to-robot Policy Transfer