Learning Differentiable Reachability Maps for Optimization-based Humanoid Motion Generation

Masaki Murooka, Iori Kumagai, Mitsuharu Morisawa, Fumio Kanehiro · 2025

Read Learning Differentiable Reachability Maps for Optimization-based Humanoid Motion Generation by Masaki Murooka, Iori Kumagai, Mitsuharu Morisawa, Fumio Kanehiro (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-2508-11275.txt.

Description

Masaki Murooka and co-authors teach humanoid robots a smooth map of where their hands can reach. Their 2025 paper tackles the cost of planning motion for two-legged robots. Reach checks are usually slow grids or all-or-nothing tests that planners cannot lean on.

Their map is a single smooth function over the space around the robot. It scores points as reachable or not, and its slope can guide an optimizer. The map is learned from poses made by the robot's own joint model, using a neural net or a support vector machine. With the map as a rule, planners solved footstep, contact, and lifting tasks with ease in the tests shown.

Catalog record

CREATORMasaki Murooka, Iori Kumagai, Mitsuharu Morisawa, Fumio Kanehiro
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE43 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2508-11275

Browse the full library · About the Tower of Babel