DexNinja: Learning Robust Dexterous Cutting Policy with a Real-to-Sim-to-Real Data Engine

Key points

• Cutting a piece of food looks effortless in human hands.1
• For a robot, however, every cut is a high-contact interaction with a deformable, elasto-plastic object whose behavior changes with pose, contact mode, and cutting stage.1
• DexNinja (i) reconstructs object instances from real captures, (ii) randomizes physically meaningful parameters under category-level constraints, and (iii) produces augmented dexterous manipulation episodes using a custom differentiable simulator that couples robot dynamics, tactile contact, and deformable cutting.1
• We evaluate DexNinja on a dexterous food-slicing task and show that (1) with tens of real demonstrations and hundreds of augmented simulator data, the success rate could reach more than 60%.1

Sources (1)

• [1]DexNinja: Learning Robust Dexterous Cutting Policy with a Real-to-Sim-to-Real Data Engine
  arXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 12, 05:59 PM

  We propose DexNinja, a real2sim2real framework that turns a handful of real demonstrations into a dense and realistic training distribution.

  Cutting a piece of food looks effortless in human hands.

Extractive summary: sentences quoted from the sources.