GeoMatch
google-deepmind
Geometry-matching research for grasp synthesis across different robot embodiments.
What you can explore
Inspect geometric representations that associate object surfaces with robotic end effectors.
Before you use it
A geometry match is not a physical stability guarantee. Reproduction depends on the supplied embodiment definitions and object assets.
Original contributors
Google DeepMind / GeoMatch contributors
License / access: Code: Apache 2.0 · other repository material: CC BY 4.0; check third-party assets
Keep exploring
GenDexGrasp
Contact-map-based grasp synthesis designed to generalize across robotic hand embodiments.
RobotFingerPrint
A common-coordinate representation for transferring grasp information across different robot hands.
Dexonomy
Contact templates and a grasp taxonomy for synthesizing human-like dexterous grasp configurations.
Editorial reference · checked 2026-09-22