Dexterity from Supervised Play
AuthorsBotao He, Yiming Dou, Yiannis Aloimonos, Jitendra Malik, Akash Sharma, Ruoshi Liu
AffiliationsAmazon Frontier AI & Robotics (FAR) · University of Maryland, College Park
A robot learns to play from brief human guidance, then uses its own experience to practice and master dexterous manipulation.
Key results
DSP success rate in simulated cube reorientation.
DSP uses about 490 times less counted-frame interaction.
Play-guided MPC success rate.
Reduction in 50-step rotation MAE as autonomous-play data grows from 0.54 to 5.11 hours.
Milliseconds per run.
DSP success rate on hardware.
What the paper found
DSP, or Dexterity from Supervised Play, turns 15 minutes of goal-free teleoperation into a diffusion-transformer play policy that autonomously gathers contact-rich experience. The system trains an action-conditioned world model with Diffusion Forcing; at test time, the play policy proposes action chunks and model-predictive control ranks their predicted outcomes against a requested goal. In simulated cube reorientation, DSP succeeds within 0.4 radians on 72.7% of trials and within 0.6 radians on 91.4%, compared with 25.8% and 53.9% for hindsight behavior cloning. On the real robot, it reaches 90.0% success within 0.6 radians. DSP uses 701K counted frames, while performance-matched PPO requires 344M—about 490 times more interaction. On Push-T, it achieves 98% success at the 0.90 threshold. Increasing autonomous-play data from 0.54 to 5.11 hours cuts 50-step world-model rotation error by 18.8%. A policy distilled from model-based self-play runs in 21.8 milliseconds, about seven times faster than replanning; the system’s real-time planning runs on an NVIDIA RTX 5090. The results present supervised play as a reusable source of interaction and action proposals, rather than task-specific demonstrations or reward-engineered exploration.
Original 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 through behavior cloning, then deploys it autonomously to collect hours of interaction data to train a world model. During inference, the same play policy then guides model-predictive control (MPC) by proposing candidate action sequences for evaluation under the learned world model. To accelerate inference, we also introduce a method for distilling a closed-loop, goal-conditioned policy through self-play within the learned world model. We evaluate this framework on underactuated pushing and any-to-any in-hand pose reorientation with a multifingered hand, in simulation and real world. In real world experiment, our method outperforms hindsight behavior cloning by 69.6% in success rate and achieves comparable success to reinforcement learning with 490 times less interaction data. These results demonstrate that supervised play can autonomously collect interaction data, guide model-based planning, and enable policy learning without any task-specific demonstrations or sim-to-real transfer.
Read the original paperMore in Robotics
Browse all 56 papers →Blind Precise Assembly: Proprioceptive Sim-to-Real Peg-in-Hole for Diverse Parts
Antonia Bronars, Younghyo Park, Kushal Kedia, Pulkit Agrawal
A robot learns to insert, mesh, and thread parts by feeling contact through its own joints, without vision or tactile sensors.
Dream What I Can: Sparse Visual Planning for Long-Horizon Robot Manipulation
Chongkai Gao, Wenlong Huang, Haonan Chen, Lin Shao, Yilun Du, Jiajun Wu, Ruohan Zhang, Li Fei-Fei
DreamCan turns imagined video futures into sparse, achievable visual goals to help robots complete long and complex manipulation tasks.
Dex-One2Many: Learning Dexterous Manipulation from a Single Human Demonstration
Jusuk Lee, Sungha Kim, Yeonsoo Park, Jonguk Cheon, Yoonkyo Jung, Yongjun You, H. Jin Kim, Jia-Bin Huang, Furong Huang, Youngseok Jang, Seungjae Lee
A single human video is transformed into scene-graph guidance that helps a robot hand learn dexterous skills and adapt to situations it never saw.