正在加载视频...

视频加载失败

加载此视频时出现问题。这可能是由于临时网络问题，或视频可能不可用。

Wouldn't it be great if we could train robots without any teleoperation! In our latest paper, we train robots to mimic a human video of the task by simply matching the object features using RL. We only need one video and under an hour of robot training.

Lerrel Pinto

9,377 subscribers

46,221 次观看 • 1 年前 •via X (Twitter)

科学技术教育

Anya Rossi• Live Now

Private livecam show

10 条评论

Lerrel Pinto 的头像

Lerrel Pinto1 年前

HuDOR generates rewards from human videos by tracking points on manipulable objects using off-the-shelf trackers. Then during online learning, the robot optimizes its policy to match the trajectory of the manipulated objects to those in the human video.

Lerrel Pinto 的头像

Lerrel Pinto1 年前

HuDOR also generalizes to new objects! Instead of training policies from scratch, we use different text prompts to get object point movements and apply pre-trained HuDOR policies. HuDOR reasonably generalizes, with failures mainly due to weight.

Lerrel Pinto 的头像

Lerrel Pinto1 年前

To generalize to larger areas, we decouple reaching and manipulation. Using object detection and relative transforms, we locate objects in both robot and human demos, calculate offsets, and integrate them into HuDOR-trained policies, achieving ~60% success within a 40x50 cm area.

Lerrel Pinto 的头像

Lerrel Pinto1 年前

This work was led by @irmakkguzey with @wxdyl0915, @georgysavva and @Raunaqmb. For more details, including additional robot videos, visit our website:

Homanga Bharadhwaj 的头像

Homanga Bharadhwaj1 年前

This is very cool work, Lerrel. And congrats @irmakkguzey ! I was thinking about something very similar recently, doing real-world RL based on a human demo in the same scene (and was concerned about resets) It's great to see you have validated this already :)

Thanos Variant 的头像

Thanos Variant1 年前

I wonder if robots will ever be able to improvise if you remove their hands. Can they still perform the task in a different way? How would you even train that?🤔

Sergey Golubev 的头像

Sergey Golubev1 年前

Impressive advancement! Using object feature matching and RL to train robots without teleoperation opens new possibilities in human-robot interaction.

Crynet 的头像

Crynet1 年前

Training robots from a single human video using RL could revolutionize robotics. Reducing data needs and training time is a significant step towards more adaptable AI systems.

Dave Robertson 的头像

Dave Robertson1 年前

@ylecun Too slow

Chris Thompson 的头像

Chris Thompson1 年前

Love the idea of ditching teleoperation! What's the most challenging part of implementing this in real-world scenarios?

相关视频

Robots has a data problem, in that robotics data is rare. While human video is quite common, it’s not usually directly usable for robots for a variety of reasons, most significantly that it’s missing explicit, accurate robot actions. Instead, Jeremy Collins proposes that we predict keypoint trajectories — basically, how any given point in an object will move as a robot performs a task. This lets us use action-free human video to train robot skills. Learn more by watching Episode #37 of RoboPapers with Michael Cho - Rbt/Acc and Chris Paxton now.

Robots has a data problem, in that robotics data is rare. While human video is quite common, it’s not usually directly usable for robots for a variety of reasons, most significantly that it’s missing explicit, accurate robot actions. Instead, Jeremy Collins proposes that we predict keypoint trajectories — basically, how any given point in an object will move as a robot performs a task. This lets us use action-free human video to train robot skills. Learn more by watching Episode #37 of RoboPapers with Michael Cho - Rbt/Acc and Chris Paxton now.

RoboPapers

89,011 次观看 • 9 个月前

🚨Is it possible to devise an intuitive approach for crowdsourcing trainable data for robots without requiring a physical robot🤖? Can we democratize robot learning for all?🧑‍🤝‍🧑 Check out our latest #CoRL2023 paper-> AR2-D2: Training a Robot Without a Robot

🚨Is it possible to devise an intuitive approach for crowdsourcing trainable data for robots without requiring a physical robot🤖? Can we democratize robot learning for all?🧑‍🤝‍🧑 Check out our latest #CoRL2023 paper-> AR2-D2: Training a Robot Without a Robot

Jiafei Duan

38,871 次观看 • 2 年前

Using computer vision techniques, the robot analyzed each frame of video and was able to identify the different objects and features. In the future, robots will be able to learn human behaviors just by watching videos.

Using computer vision techniques, the robot analyzed each frame of video and was able to identify the different objects and features. In the future, robots will be able to learn human behaviors just by watching videos.

Aakash Gupta

55,257 次观看 • 3 年前

Our robots now run autonomously >80% of the time. We still have human teleoperators in-the-loop, who take control when the robot gets stuck or something goes wrong. As autonomy improves, we'll grow our robot-to-operator efficiency until we have one person overseeing 10+ robots. Useful, AI-powered robots are here. And it's only going to get better.

Our robots now run autonomously >80% of the time. We still have human teleoperators in-the-loop, who take control when the robot gets stuck or something goes wrong. As autonomy improves, we'll grow our robot-to-operator efficiency until we have one person overseeing 10+ robots. Useful, AI-powered robots are here. And it's only going to get better.

Ultra

26,315 次观看 • 10 个月前

HDMI (HumanoiD iMitation for Interaction) is a framework enabling humanoid robots to learn whole-body object interaction skills from monocular RGB human videos. It extracts and retargets human poses and object trajectories using GVHMR and LocoMujoco, building reference datasets with contact annotations. The data is used to train an RL policy via robot-object co-tracking. HDMI achieved 67 consecutive door traversals.

HDMI (HumanoiD iMitation for Interaction) is a framework enabling humanoid robots to learn whole-body object interaction skills from monocular RGB human videos. It extracts and retargets human poses and object trajectories using GVHMR and LocoMujoco, building reference datasets with contact annotations. The data is used to train an RL policy via robot-object co-tracking. HDMI achieved 67 consecutive door traversals.

The Humanoid Hub

17,395 次观看 • 10 个月前

Brain-controlled exoskeletons to train humanoid robots! 🧠 Fourier just presented human tele-operators using brain control interfaces and exoskeletal arms to train humanoid robots on home tasks. The brain control interface is the interesting part. Instead of using a controller or joystick to teleoperate, the operator's movements and intentions are captured more naturally through the exoskeleton and BCI. This means the demonstrations are more fluid, more human-like, and better suited for training robots to perform delicate home tasks. Multiple tele-operators are simultaneously generating training data across multiple robots. This is how you build the dataset needed for eventual full autonomy, without waiting years for it to arrive. This might be the bridge between "robots that work in controlled environments" and "robots that work in homes." Not full autonomy right away, but trusted human intelligence operating through a robot body, getting better with every task completed. ~~ ♻️ Join the weekly robotics newsletter, and never miss any news →

Brain-controlled exoskeletons to train humanoid robots! 🧠 Fourier just presented human tele-operators using brain control interfaces and exoskeletal arms to train humanoid robots on home tasks. The brain control interface is the interesting part. Instead of using a controller or joystick to teleoperate, the operator's movements and intentions are captured more naturally through the exoskeleton and BCI. This means the demonstrations are more fluid, more human-like, and better suited for training robots to perform delicate home tasks. Multiple tele-operators are simultaneously generating training data across multiple robots. This is how you build the dataset needed for eventual full autonomy, without waiting years for it to arrive. This might be the bridge between "robots that work in controlled environments" and "robots that work in homes." Not full autonomy right away, but trusted human intelligence operating through a robot body, getting better with every task completed. ~~ ♻️ Join the weekly robotics newsletter, and never miss any news →

Lukas Ziegler

20,898 次观看 • 5 个月前

We discovered an emergent property of VLAs like π0/π0.5/π0.6: as we scale up pre-training, the model learns to align human videos and robot data! This gives us a simple way to leverage human videos. Once π0.5 knows how to control robots, it can naturally learn from human video.

We discovered an emergent property of VLAs like π0/π0.5/π0.6: as we scale up pre-training, the model learns to align human videos and robot data! This gives us a simple way to leverage human videos. Once π0.5 knows how to control robots, it can naturally learn from human video.

Physical Intelligence

1,183,757 次观看 • 7 个月前

TWIST: a real-time teleoperation system for humanoid robots to mimic whole-body motions. Reference motion data is generated by retargeting human motion-capture data to the robot. Then, the controller is trained in simulation using reinforcement learning and behavior cloning.

TWIST: a real-time teleoperation system for humanoid robots to mimic whole-body motions. Reference motion data is generated by retargeting human motion-capture data to the robot. Then, the controller is trained in simulation using reinforcement learning and behavior cloning.

The Humanoid Hub

46,715 次观看 • 1 年前

Tired of teleoperating your robots? We built a way to scale robot datasets without teleop, dynamic simulation, or even robot hardware. Just one smartphone scan + one human hand demo video → thousands of diverse robot trajectories. Trainable by diffusion policy and VLA models as-is. Introducing: Real2Render2Real 👉

Tired of teleoperating your robots? We built a way to scale robot datasets without teleop, dynamic simulation, or even robot hardware. Just one smartphone scan + one human hand demo video → thousands of diverse robot trajectories. Trainable by diffusion policy and VLA models as-is. Introducing: Real2Render2Real 👉

Max Fu

69,381 次观看 • 1 年前

We tend to blame the government for everything! - We blame the government if a train is late. - We blame the government when a train is stopped in the middle of nowhere. - We blame the government if there is an attack on a train. - We blame the government if a train is not clean. But we often ignore the root causes behind these issues. People like the one in the video stop trains by coming onto the tracks, break train windows by throwing stones, pull the emergency chain without a valid reason, and litter inside coaches. I'm sorry to say this, but many Indians lack basic civic sense.

We tend to blame the government for everything! - We blame the government if a train is late. - We blame the government when a train is stopped in the middle of nowhere. - We blame the government if there is an attack on a train. - We blame the government if a train is not clean. But we often ignore the root causes behind these issues. People like the one in the video stop trains by coming onto the tracks, break train windows by throwing stones, pull the emergency chain without a valid reason, and litter inside coaches. I'm sorry to say this, but many Indians lack basic civic sense.

Amit Kumar Sindhi

120,656 次观看 • 21 天前

We trained a humanoid with 22-DoF dexterous hands to assemble model cars, operate syringes, sort poker cards, fold/roll shirts, all learned primarily from 20,000+ hours of egocentric human video with no robot in the loop. Humans are the most scalable embodiment on the planet. We discovered a near-perfect log-linear scaling law (R² = 0.998) between human video volume and action prediction loss, and this loss directly predicts real-robot success rate. Humanoid robots will be the end game, because they are the practical form factor with minimal embodiment gap from humans. Call it the Bitter Lesson of robot hardware: the kinematic similarity lets us simply retarget human finger motion onto dexterous robot hand joints. No learned embeddings, no fancy transfer algorithms needed. Relative wrist motion + retargeted 22-DoF finger actions serve as a unified action space that carries through from pre-training to robot execution. Our recipe is called "EgoScale": - Pre-train GR00T N1.5 on 20K hours of human video, mid-train with only 4 hours (!) of robot play data with Sharpa hands. 54% gains over training from scratch across 5 highly dexterous tasks. - Most surprising result: a *single* teleop demo is sufficient to learn a never-before-seen task. Our recipe enables extreme data efficiency. - Although we pre-train in 22-DoF hand joint space, the policy transfers to a Unitree G1 with 7-DoF tri-finger hands. 30%+ gains over training on G1 data alone. The scalable path to robot dexterity was never more robots. It was always us. Deep dives in thread:

We trained a humanoid with 22-DoF dexterous hands to assemble model cars, operate syringes, sort poker cards, fold/roll shirts, all learned primarily from 20,000+ hours of egocentric human video with no robot in the loop. Humans are the most scalable embodiment on the planet. We discovered a near-perfect log-linear scaling law (R² = 0.998) between human video volume and action prediction loss, and this loss directly predicts real-robot success rate. Humanoid robots will be the end game, because they are the practical form factor with minimal embodiment gap from humans. Call it the Bitter Lesson of robot hardware: the kinematic similarity lets us simply retarget human finger motion onto dexterous robot hand joints. No learned embeddings, no fancy transfer algorithms needed. Relative wrist motion + retargeted 22-DoF finger actions serve as a unified action space that carries through from pre-training to robot execution. Our recipe is called "EgoScale": - Pre-train GR00T N1.5 on 20K hours of human video, mid-train with only 4 hours (!) of robot play data with Sharpa hands. 54% gains over training from scratch across 5 highly dexterous tasks. - Most surprising result: a single teleop demo is sufficient to learn a never-before-seen task. Our recipe enables extreme data efficiency. - Although we pre-train in 22-DoF hand joint space, the policy transfers to a Unitree G1 with 7-DoF tri-finger hands. 30%+ gains over training on G1 data alone. The scalable path to robot dexterity was never more robots. It was always us. Deep dives in thread:

Jim Fan

293,585 次观看 • 5 个月前

NVIDIA has published a paper on DREAMGEN – a powerful 4-step pipeline for generating synthetic data for humanoids that enables task and environment generalization. - Step 1: Fine-tune a video generation model using a small number of human teleoperation videos - Step 2: Prompt the fine-tuned model to turn a single real image into new AI-imagined videos - Step 3: Automatically label actions in the generated videos - Step 4: Train a robot AI model with the labeled synthetic dataset This enabled humanoid robots to perform 22 novel behaviors – such as pouring, opening/closing articulated objects, and manipulating a variety of tools. The original teleoperation dataset only included pick-and-place tasks. This takes task extensibility to another level without requiring human teleoperation for every single task. The pipeline will be made open-source soon. Project page:

NVIDIA has published a paper on DREAMGEN – a powerful 4-step pipeline for generating synthetic data for humanoids that enables task and environment generalization. - Step 1: Fine-tune a video generation model using a small number of human teleoperation videos - Step 2: Prompt the fine-tuned model to turn a single real image into new AI-imagined videos - Step 3: Automatically label actions in the generated videos - Step 4: Train a robot AI model with the labeled synthetic dataset This enabled humanoid robots to perform 22 novel behaviors – such as pouring, opening/closing articulated objects, and manipulating a variety of tools. The original teleoperation dataset only included pick-and-place tasks. This takes task extensibility to another level without requiring human teleoperation for every single task. The pipeline will be made open-source soon. Project page:

The Humanoid Hub

12,074 次观看 • 1 年前

SkildAI, which is developing a foundational AI model for robots, has released a new video of their humanoid robot walking over objects and exploring environments it had never seen before. The robot is using vision & adapts in real time. SkildAI: "From raw images and joint feedback, the model directly outputs low-level motor commands. This single neural network enables humanoid robots to seamlessly walk across flat ground, climb stairs, and step over obstacles without any planning, mapping, or manual switching between behaviors. For all of our testing in this video, we deployed the robot in each new environment without any prior planning or mapping. We took off the humanoid’s shoes so you can hear it.

SkildAI, which is developing a foundational AI model for robots, has released a new video of their humanoid robot walking over objects and exploring environments it had never seen before. The robot is using vision & adapts in real time. SkildAI: "From raw images and joint feedback, the model directly outputs low-level motor commands. This single neural network enables humanoid robots to seamlessly walk across flat ground, climb stairs, and step over obstacles without any planning, mapping, or manual switching between behaviors. For all of our testing in this video, we deployed the robot in each new environment without any prior planning or mapping. We took off the humanoid’s shoes so you can hear it.

Sawyer Merritt

69,297 次观看 • 11 个月前

1958, The Robot Guest. This Robbie The Robot influenced many ideas of how we see and react to robots even today. It is also one reason anthropomorphic bipedal humanoid robots are easier psychologically.

1958, The Robot Guest. This Robbie The Robot influenced many ideas of how we see and react to robots even today. It is also one reason anthropomorphic bipedal humanoid robots are easier psychologically.

Brian Roemmele

19,785 次观看 • 7 个月前

Introducing Open-TeleVision: with Fully Autonomous policy video👇. We can conduct a long-horizon task with inserting 12 cans nonstop without any interruptions. We offer: 🤖 Highly precise and smooth bimanual manipulation. 📺 Active egocentric vision (with a moving neck) feedback. It is achieved by imitation learning from teleoperation: We propose a VR-based REAL-TIME teleoperation that streams the stereo video observation from the robot camera to the VR device. The robot neck moves as the human head moves, the robot hands move as the human hands move, offering the operator an intuitive experience as the human herself becomes the robot. The devils are all in the details, and how to implement things right: ✅How to perform IK/retargeting for smooth and precise control. ✅How to do all these and also stream stereo video without no latency, all in real time. We released our code here: Active head hardware design: 1/n

Introducing Open-TeleVision: with Fully Autonomous policy video👇. We can conduct a long-horizon task with inserting 12 cans nonstop without any interruptions. We offer: 🤖 Highly precise and smooth bimanual manipulation. 📺 Active egocentric vision (with a moving neck) feedback. It is achieved by imitation learning from teleoperation: We propose a VR-based REAL-TIME teleoperation that streams the stereo video observation from the robot camera to the VR device. The robot neck moves as the human head moves, the robot hands move as the human hands move, offering the operator an intuitive experience as the human herself becomes the robot. The devils are all in the details, and how to implement things right: ✅How to perform IK/retargeting for smooth and precise control. ✅How to do all these and also stream stereo video without no latency, all in real time. We released our code here: Active head hardware design: 1/n

Xiaolong Wang

25,572 次观看 • 2 年前

Cathie Wood says if she could only pick one stock to invest in, she would choose $TSLA. “It is a convergence of our three major platforms, Robots, Energy Storage, AI. It’s not stopping with Robotaxis, there’s a story beyond that with Humanoid Robots, and our $2600 PT (2030) has nothing to do with humanoid robots — we just thought it would be an investment period.”

Cathie Wood says if she could only pick one stock to invest in, she would choose $TSLA. “It is a convergence of our three major platforms, Robots, Energy Storage, AI. It’s not stopping with Robotaxis, there’s a story beyond that with Humanoid Robots, and our $2600 PT (2030) has nothing to do with humanoid robots — we just thought it would be an investment period.”

Nic Cruz Patane

3,607,315 次观看 • 1 年前

Scientists developed HumanX, a system that lets humanoid robots learn new skills by watching just one short human video. It learns directly from demonstrations instead of large datasets or complex training rewards. Using this, the Unitree G1 humanoid robot learned 10 different real-world skills without extra training.

Scientists developed HumanX, a system that lets humanoid robots learn new skills by watching just one short human video. It learns directly from demonstrations instead of large datasets or complex training rewards. Using this, the Unitree G1 humanoid robot learned 10 different real-world skills without extra training.

Space and Technology

27,686 次观看 • 3 个月前

Tracking Anything with Decoupled Video Segmentation paper page: Training data for video segmentation are expensive to annotate. This impedes extensions of end-to-end algorithms to new video segmentation tasks, especially in large-vocabulary settings. To 'track anything' without training on video data for every individual task, we develop a decoupled video segmentation approach (DEVA), composed of task-specific image-level segmentation and class/task-agnostic bi-directional temporal propagation. Due to this design, we only need an image-level model for the target task (which is cheaper to train) and a universal temporal propagation model which is trained once and generalizes across tasks. To effectively combine these two modules, we use bi-directional propagation for (semi-)online fusion of segmentation hypotheses from different frames to generate a coherent segmentation. We show that this decoupled formulation compares favorably to end-to-end approaches in several data-scarce tasks including large-vocabulary video panoptic segmentation, open-world video segmentation, referring video segmentation, and unsupervised video object segmentation.

Tracking Anything with Decoupled Video Segmentation paper page: Training data for video segmentation are expensive to annotate. This impedes extensions of end-to-end algorithms to new video segmentation tasks, especially in large-vocabulary settings. To 'track anything' without training on video data for every individual task, we develop a decoupled video segmentation approach (DEVA), composed of task-specific image-level segmentation and class/task-agnostic bi-directional temporal propagation. Due to this design, we only need an image-level model for the target task (which is cheaper to train) and a universal temporal propagation model which is trained once and generalizes across tasks. To effectively combine these two modules, we use bi-directional propagation for (semi-)online fusion of segmentation hypotheses from different frames to generate a coherent segmentation. We show that this decoupled formulation compares favorably to end-to-end approaches in several data-scarce tasks including large-vocabulary video panoptic segmentation, open-world video segmentation, referring video segmentation, and unsupervised video object segmentation.

AK

305,667 次观看 • 2 年前

Teaching robots to learn only from RGB human videos is hard! In Feel The Force (FTF), we teach robots to mimic the tactile feedback humans experience when handling objects. This allows for delicate, touch-sensitive tasks—like picking up a raw egg without breaking it. 🧵👇

Teaching robots to learn only from RGB human videos is hard! In Feel The Force (FTF), we teach robots to mimic the tactile feedback humans experience when handling objects. This allows for delicate, touch-sensitive tasks—like picking up a raw egg without breaking it. 🧵👇

Lerrel Pinto

70,328 次观看 • 1 年前