Robot Self-Improvement via Human-Video Dynamics Models

• URL: http://arxiv.org/abs/2606.21406v1
• Date: Fri, 19 Jun 2026 13:17:27 GMT
• Title: Robot Self-Improvement via Human-Video Dynamics Models
• Abstract summary: We show that human videos can be used to learn embodiment-agnostic action, dynamics, and value representations that transfer across robot embodiments.<n>We introduce Dynamics-Guided Action Correction (DGAC), a training-free approach that leverages these adapted models to repair failed states.<n>Results show that human priors and robot failures can be combined to enable scalable autonomous policy improvement.
• Score: 52.14862052988773
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: A central question in robot learning is how to acquire skills from the kinds of data that humans learn from: passive observation, embodied practice, and the experience of failure. Human videos provide the first of these in abundance, and prior work has shown they can initialize useful policies. Far less clear is whether they can support the second and third: whether priors extracted from human videos can ground a robot's own attempts well enough to evaluate them, correct them, and improve from them. In this work, we show that human videos can be used to learn embodiment-agnostic action, dynamics, and value representations that transfer across robot embodiments, providing the predictive foundation required for robots to autonomously improve from their own rollouts and failures. We introduce Dynamics-Guided Action Correction (DGAC), a training-free approach that leverages these adapted models to repair failed states: each failure becomes a query for which the learned models propose and rank corrective actions, turning failures into supervision for the next policy update. Across seven real-world manipulation tasks spanning both a mobile manipulator and a static manipulator arm, our approach improves success rates from 40% to 81% across multiple policy backbones, demonstrating cross-embodiment robot self-improvement from human-video priors. These results show that human priors and robot failures can be combined to enable scalable autonomous policy improvement. Project page: https://ethz-mrl.github.io/robot-self-improvement-website/.

Related papers

• Ego2Robot: Scalable Robot Data Synthesis from Egocentric Human Data [60.94418315274988]
  Ego2Robot supports both curated and in-the-wild videos, producing 18,561 hours of robot training data spanning 15 robot morphologies.<n>Joint pretraining on Ego2Robot-synthesized and robot data consistently improves out-of-distribution generalization across multiple types.
  arXiv  Detail & Related papers  (2026-08-03T17:52:26Z)
• Human Preference Modeling Using Visual Motion Prediction Improves Robot Skill Learning from Egocentric Human Video [5.158632635415882]
  We present an approach to robot learning from egocentric human videos by modeling human preferences in a reward function and optimizing robot behavior to maximize this reward.<n>Our approach is capable of learning on a real robot, and we show that policies learned with our reward model match or outperform prior work across multiple tasks in both simulation and on the real robot.
  arXiv  Detail & Related papers  (2026-02-11T21:56:47Z)
• H2R-Grounder: A Paired-Data-Free Paradigm for Translating Human Interaction Videos into Physically Grounded Robot Videos [58.006918399913665]
  We propose a video-to-video translation framework that converts ordinary human-object interaction videos into motion-consistent robot manipulation videos.<n>Our approach does not require any paired human-robot videos for training only a set of unpaired robot videos, making the system easy to scale.<n>At test time, we apply the same process to human videos (inpainting the person and overlaying human pose cues) and generate high-quality robot videos that mimic the human's actions.
  arXiv  Detail & Related papers  (2025-12-10T07:59:45Z)
• EgoZero: Robot Learning from Smart Glasses [54.6168258133554]
  EgoZero learns robust manipulation policies from human demonstrations captured with Project Aria smart glasses.<n>We deploy EgoZero policies on a Franka Panda robot and demonstrate zero-shot transfer with 70% success rate over 7 manipulation tasks.<n>Our results suggest that in-the-wild human data can serve as a scalable foundation for real-world robot learning.
  arXiv  Detail & Related papers  (2025-05-26T17:59:17Z)
• Vid2Robot: End-to-end Video-conditioned Policy Learning with Cross-Attention Transformers [36.497624484863785]
  We introduce Vid2Robot, an end-to-end video-conditioned policy that takes human videos demonstrating manipulation tasks as input and produces robot actions. Our model is trained with a large dataset of prompt video-robot trajectory pairs to learn unified representations of human and robot actions from videos. We evaluate Vid2Robot on real-world robots and observe over 20% improvement over BC-Z when using human prompt videos.
  arXiv  Detail & Related papers  (2024-03-19T17:47:37Z)
• Interactive Robot Learning from Verbal Correction [42.37176329867376]
  OLAF allows users to teach a robot using verbal corrections when the robot makes mistakes. A key feature of OLAF is its ability to update the robot's visuomotor neural policy based on the verbal feedback. We demonstrate the efficacy of our design in experiments where a user teaches a robot to perform long-horizon manipulation tasks.
  arXiv  Detail & Related papers  (2023-10-26T16:46:12Z)
• Quality-Diversity Optimisation on a Physical Robot Through Dynamics-Aware and Reset-Free Learning [4.260312058817663]
  We build upon the Reset-Free QD (RF-QD) algorithm to learn controllers directly on a physical robot. This method uses a dynamics model, learned from interactions between the robot and the environment, to predict the robot's behaviour. RF-QD also includes a recovery policy that returns the robot to a safe zone when it has walked outside of it, allowing continuous learning.
  arXiv  Detail & Related papers  (2023-04-24T13:24:00Z)
• Learning a Universal Human Prior for Dexterous Manipulation from Human Preference [35.54663426598218]
  We propose a framework that learns a universal human prior using direct human preference feedback over videos. A task-agnostic reward model is trained through iteratively generating diverse polices and collecting human preference over the trajectories. Our method empirically demonstrates more human-like behaviors on robot hands in diverse tasks including even unseen tasks.
  arXiv  Detail & Related papers  (2023-04-10T14:17:33Z)
• Self-Improving Robots: End-to-End Autonomous Visuomotor Reinforcement Learning [54.636562516974884]
  In imitation and reinforcement learning, the cost of human supervision limits the amount of data that robots can be trained on. In this work, we propose MEDAL++, a novel design for self-improving robotic systems. The robot autonomously practices the task by learning to both do and undo the task, simultaneously inferring the reward function from the demonstrations.
  arXiv  Detail & Related papers  (2023-03-02T18:51:38Z)
• Learning Reward Functions for Robotic Manipulation by Observing Humans [92.30657414416527]
  We use unlabeled videos of humans solving a wide range of manipulation tasks to learn a task-agnostic reward function for robotic manipulation policies. The learned rewards are based on distances to a goal in an embedding space learned using a time-contrastive objective.
  arXiv  Detail & Related papers  (2022-11-16T16:26:48Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.