Exploiting Policy Idling for Dexterous Manipulation - 2025-08-21 - Research Publication - The Pause-Induced Perturbations study evaluates policy improvement on a real precision connector-insertion task using a KUKA iiwa 14 with a three-finger Shadow DEX-EE hand and fingertip tactile sensing. The method improves… - source: arxiv.org
ExoStart: Efficient learning for dexterous manipulation with sensorized exoskeleton demonstrations - 2025-06-13 - Research Publication - ExoStart uses a sensorized exoskeleton matched to the Shadow DEX-EE kinematics, dynamics filtering and auto-curriculum reinforcement learning to train vision policies for a physical DEX-EE on a KUKA iiwa. Seven real tasks… - source: arxiv.org
DemoStart: Demonstration-led auto-curriculum applied to sim-to-real with multi-fingered robots - 2024-09-10 - Research Publication - DemoStart trains multi-finger manipulation policies from sparse rewards and a small set of simulated demonstrations, then transfers pixel-based policies zero-shot to physical KUKA iiwa systems fitted with Shadow DEX-EE… - source: arxiv.org