Select before Act: Spatially Decoupled Action Repetition for Continuous Control

Buqing Nie, Yangqing Fu, Yue Gao · 2025

Read Select before Act: Spatially Decoupled Action Repetition for Continuous Control by Buqing Nie, Yangqing Fu, Yue Gao (2025) 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-2502-06919.txt.

Description

Buqing Nie, Yangqing Fu, and Yue Gao present a 2025 paper on teaching robots smoother control through action repetition. Repeating an action for several steps can make reinforcement learning faster and steadier. Prior methods, though, repeat every action dimension as one block, which stiffens the policy and wastes the idea's promise.

Their framework, SDAR, decides act-or-repeat separately for each action dimension in a closed loop. One joint can hold its command while another updates, giving a finer mix of persistence and change. The method proved more sample-efficient than earlier repetition schemes. It also reached higher policy scores with less action jitter. Tests spanned several continuous control tasks, from manipulation to locomotion. The design adds flexibility without new sensors or models.

Catalog record

CREATORBuqing Nie, Yangqing Fu, Yue Gao
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE43 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2502-06919

Browse the full library · About the Tower of Babel