Articulated-Body Dynamics Network: Dynamics-Grounded Prior for Robot Learning

Sangwoo Shin, Kunzhao Ren, Xiaobin Xiong, Josiah P. Hanna · 2026

Read Articulated-Body Dynamics Network: Dynamics-Grounded Prior for Robot Learning by Sangwoo Shin, Kunzhao Ren, Xiaobin Xiong, Josiah P. Hanna (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2603-19078.txt.

Description

Shin and co-authors bake robot body dynamics into the network that learns robot control. Past work used link structure as a prior, but force flow through the body was left out. Their graph network copies the way the classic articulated body method passes inertia from child links to parent links. Learned values stand in for physical ones. Policies built on it learn faster and cope better with shifted dynamics. Tests cover simulated humanoid, four-legged, and hopper robots. Learned policies also run on two real robots.

Catalog record

CREATORSangwoo Shin, Kunzhao Ren, Xiaobin Xiong, Josiah P. Hanna
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE38068
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2603.19078

Browse the full library · About the Tower of Babel