DextrAH-G: Pixels-to-Action Dexterous Arm-Hand Grasping with Geometric Fabrics

TL;DR

DextrAH-G is a simulation-trained, depth-based robotic grasping policy that combines reinforcement learning and geometric fabrics to enable fast, safe, and generalizable arm-hand grasping across diverse objects.

Contribution

It introduces a novel simulation-trained policy using geometric fabrics and teacher-student distillation for dexterous robotic grasping.

Findings

Successfully grasped and transported various objects at high speed.

Generalized across diverse object geometries.

Ensured safe and continuous operation with hardware constraints.

Abstract

A pivotal challenge in robotics is achieving fast, safe, and robust dexterous grasping across a diverse range of objects, an important goal within industrial applications. However, existing methods often have very limited speed, dexterity, and generality, along with limited or no hardware safety guarantees. In this work, we introduce DextrAH-G, a depth-based dexterous grasping policy trained entirely in simulation that combines reinforcement learning, geometric fabrics, and teacher-student distillation. We address key challenges in joint arm-hand policy learning, such as high-dimensional observation and action spaces, the sim2real gap, collision avoidance, and hardware constraints. DextrAH-G enables a 23 motor arm-hand robot to safely and continuously grasp and transport a large variety of objects at high speed using multi-modal inputs including depth images, allowing generalization across object geometry. Videos at https://sites.google.com/view/dextrah-g.