Robots learn dexterous manipulation from human play and videos · 10 sources tracked
ByPulseAugur Editorial·[13 sources]·
Multiple research papers explore advancements in dexterous robotic manipulation, focusing on learning from human demonstrations and improving control strategies. Techniques like "Dexterity from Supervised Play" (DSP) and "DexMimic" leverage supervised learning and value-guided resets to enhance robot interaction data collection and skill generalization. "Whole-Body Dexterity" (WBD) introduces a framework for robots to use their entire bodies for manipulation, outperforming end-effector-only approaches. Other methods such as "DexFLEX" and "DexCast" focus on contact-aware control and learning from human videos, respectively, while "GOTT" proposes a reusable cross-embodiment primitive for robust contact acquisition. "EgoJEPA" and "P3-WAM" aim to learn embodiment-agnostic representations from egocentric videos for more efficient and effective robot control.
AI
IMPACT
These advancements in dexterous manipulation could lead to more capable robots in manufacturing, logistics, and household tasks.
RANK_REASON
Multiple research papers published on arXiv detailing new methods for dexterous robotic manipulation.
arXiv:2610.19089v2 Announce Type: replace-cross Abstract: Achieving human-like dexterous manipulation remains a central challenge in robotics. Learning from human hand-object interaction (HOI) datasets offers a promising pathway to diverse dexterous skills, yet existing physics-b…
arXiv:2610.22340v1 Announce Type: cross Abstract: How can robots autonomously collect the contact-rich experience needed to learn dexterous manipulation? We present DSP: Dexterity from supervised play, a framework in which a robot learns a play policy from minutes of human play t…
arXiv cs.AI
TIER_1English(EN)·Conor Mc Gartoll, An Dang, Pranay Thangeda, Arjun Gupta, Carolina Higuera, Nima Fazeli, Mustafa Mukadam, Manikantan Nambi·
arXiv:2610.22184v1 Announce Type: cross Abstract: Current humanoid dexterous manipulation remains largely confined to end-effectors and simple grasps, falling short of humans' ability to engage the entire body by bracing and sliding bulky objects against the arms and torso. We in…
How can robots autonomously collect the contact-rich experience needed to learn dexterous manipulation? We present DSP: Dexterity from supervised play, a framework in which a robot learns a play policy from minutes of human play through behavior cloning, then deploys it autonomou…
arXiv cs.AI
TIER_1English(EN)·Weikang Wan, Zeyu Lai, Jianyu Hou, Yuchen Zhou, Jiawei Fu, Henrik I Christensen, Hao Su·
arXiv:2610.14372v1 Announce Type: cross Abstract: Dexterous robot hands can receive useful motion intent from teleoperation or learned policies, but successful execution also depends on contact decisions that these commands rarely specify. We introduce DexFLEX, a contact-aware fo…
arXiv:2610.14691v2 Announce Type: cross Abstract: Recent advances in 3D hand-object interaction (HOI) perception have opened new opportunities for learning dexterous robotic skills from human videos. However, existing dexterous imitation learning methods often rely on costly and …
arXiv cs.AI
TIER_1English(EN)·Ying Yuan, David Held, Jitendra Malik, Haozhi Qi·
arXiv:2610.17161v1 Announce Type: cross Abstract: Teleoperation is widely used to collect high-quality demonstrations for robot learning. Dexterous manipulation, however, requires coordinated control of motion and contact forces. Operators can specify motion, but struggle to repr…
arXiv cs.AI
TIER_1English(EN)·Yulin Liu, Lai Wei, Yen-Jen Wang, Akash Sharma, Pieter Abbeel, Henrik I. Christensen, Haozhi Qi·
arXiv:2610.03861v2 Announce Type: replace-cross Abstract: Foundation models and large-scale human data provide rich sources of manipulation intent, but translating this intent into multi-fingered robot behavior remains difficult. Dexterous hands still lack a reusable low-level pr…
While learning dexterous manipulation from a single human video offers a promising alternative to costly robot demonstrations, many recent methods predominantly imitate demonstrated motions. Such strict motion matching often limits generalization to initial object poses, goal pos…
arXiv cs.CV
TIER_1English(EN)·Zirui Wu, Chen Guo, Tianjian Jiang, Zeren Jiang, Zeming Li, Lingchen Yang, Longteng Duan, Ying Xue, Handi Yin, Shenyuan Gao, Martin R. Oswald, Jun Zhang, Ping Tan, Jie Song·
arXiv:2610.22363v1 Announce Type: new Abstract: We introduce P$^{3}$-WAM, a unified world action model for egocentric dexterous manipulation that learns 3D perception, action-conditioned prediction, and goal-directed planning through three complementary modes: Perceive, which rec…
arXiv:2610.20904v1 Announce Type: cross Abstract: Egocentric human videos offer a scalable alternative to costly robot demonstra- tions, yet transferring their manipulation priors to robot control remains chal- lenging. The commonly adopted retargeting-based approaches convert hu…
arXiv cs.CV
TIER_1English(EN)·Hyeonwoo Kim, Jeonghwan Kim, Kyungwon Cho, Junyoung Lee, Hanbyul Joo·
arXiv:2604.20841v2 Announce Type: replace Abstract: Physics-based character control for dexterous Human-Object Interaction (HOI) relies heavily on motion-capture data, which limits its applicability beyond captured scenarios. Meanwhile, video generative models can synthesize real…
