GoalDex: Learning Image-Goal-Conditioned Policies for Dexterous In-Hand Manipulation
Overview
This research introduces GoalDex, an end-to-end visuomotor framework for dexterous in-hand manipulation that operates on RGB observations and proprioception without intermediate pose estimation. The approach jointly learns visual representations and control policies through behavior distillation from privileged state teachers and auxiliary reconstruction objectives. It enables image-goal-conditioned reorientation, allowing desired object configurations to be specified directly through images. Evaluated across diverse object geometries in simulation, the method consistently outperforms pose-based methods and policies built on pretrained visual encoders. The authors also demonstrate direct sim-to-real transfer of a single image-goal-conditioned policy across five real-world objects, achieving accurate dexterous reorientation without explicit pose estimation.
Generated from attributed reports · Updated 1 hours ago
Event evidence and corrections
0 attributed source owners. Ownership does not establish independent confirmation. Quantities are reported separately and are never added together.
Artifact availability · code: reported pending
“dexterous reorientation can be achieved without explicit pose estimation. For videos and code release, please refer to the project website: https://rffr.leggedrobotics.com/works/goaldex/”
Exact source · revision 1Source owner not reported
Report timeline
Follow attributed reports and material updates.
- arXiv Robotics — research abstractsSignalGoalDex: Learning Image-Goal-Conditioned Policies for Dexterous In-Hand Manipulation
This research introduces GoalDex, an end-to-end visuomotor framework for dexterous in-hand manipulation that operates on RGB observations and proprioception without intermediate pose estimation. The approach jointly learns visual representations and control policies through behavior distillation from privileged state teachers and auxiliary reconstruction objectives. It enables image-goal-conditioned reorientation, allowing desired object configurations to be specified directly through images. Evaluated across diverse object geometries in simulation, the method consistently outperforms pose-based methods and policies built on pretrained visual encoders. The authors also demonstrate direct sim-to-real transfer of a single image-goal-conditioned policy across five real-world objects, achieving accurate dexterous reorientation without explicit pose estimation.
Event coverage history
There is not enough continuous observation data to show a trend.
Timezone · UTC
Article dates follow your selected timezone. Briefing editions use Hong Kong time (UTC+8).