Loading video...

Video Failed to Load

There was a problem loading this video. This could be due to a temporary network issue or the video might be unavailable.

How can we move beyond static-arm lab setups and learn robot policies in our messy homes? We introduce HoMeR, an imitation learning agent for in-the-wild mobile manipulation. 🧵1/8

Priya Sundaresan

1,922 subscribers

42,482 views • 1 year ago •via X (Twitter)

Anya Rossi• Live Now

Private livecam show

12 Comments

Priya Sundaresan1 year ago

In real homes, both data collection and policy learning are hard — workspaces are large, task horizons are long, and scenes are too diverse to fully cover. Many tasks, like wiping a table or opening a door, also require coordinated whole-body movement, not just arm motion. 🧵2/8

Priya Sundaresan1 year ago

We address these challenges with HoMeR, a hybrid IL agent that leverages: • point clouds + images as input And outputs • absolute poses for long-range movement • relative poses for fine-grained manip all executed via a whole-body controller. 🧵3/8

Priya Sundaresan1 year ago

Specifically, HoMeR builds on our prior work SPHINX: extending it to @jimmyyhwu‘s Tidybot++: with a whole-body controller based on @kevin_zakka‘s mink! 🧵4/8

Priya Sundaresan1 year ago

Across 6 simulated and real-world tasks, HoMeR achieves 79% success, outperforming baselines that lack hybrid actions or whole-body control by 29% 🧵5/8

Priya Sundaresan1 year ago

HoMER also can be conditioned on keypoints derived from VLMs, enabling generalization to unseen visual appearances and clutter. 🧵6/8

Priya Sundaresan1 year ago

We also release an iphone-based interface for intuitive whole-body teleoperation, with self-collision avoidance between the arm, base, and external camera mounts. 🧵7/8

Priya Sundaresan1 year ago

Most fun I’ve had working from home :P Big thanks to the entire team! Rhea Malhotra, Phillip Miao, @yjy0625 , @jimmyyhwu , @HengyuanH, @contactrika, @FrancisEngelman, @DorsaSadigh, @leto__jean paper📄: website + code🔗: 🧵8/8

Rainmaker2 years ago

Which Machine Learning model can beat the market? Check out this post on my free Substack where I share code and commentary for several strategies that beat passively holding a tech stock.

Gianluca Bencomo1 year ago

Really cool work!

Priya Sundaresan1 year ago

Thank you!!

Neil Nie1 year ago

This is awesome! Congrats Priya!

Priya Sundaresan1 year ago

Thanks Neil :))

Related Videos

Policies trained on real robot data via imitation can be surprisingly capable. But for domains like dexterous manipulation, they are often not quite good enough: they move slowly, miss grasps, make unreliable contact, and fail under small perturbations. Can we improve them without any additional data collection on the real robot? In SCORE, we show that we can improve real-world diffusion/flow policies cheaply by using simulation to simply learn how to steer them on deployment. This leads to large gains in real-world success and speed across a variety of tasks, without requiring additional real-world experience: 🧵 (1/10)

Policies trained on real robot data via imitation can be surprisingly capable. But for domains like dexterous manipulation, they are often not quite good enough: they move slowly, miss grasps, make unreliable contact, and fail under small perturbations. Can we improve them without any additional data collection on the real robot? In SCORE, we show that we can improve real-world diffusion/flow policies cheaply by using simulation to simply learn how to steer them on deployment. This leads to large gains in real-world success and speed across a variety of tasks, without requiring additional real-world experience: 🧵 (1/10)

Abhishek Gupta

33,950 views • 24 days ago

🤖 How can robot policies zero-shot generalize to any new environment and any new object? Introducing our new project: 🚀Data Scaling Laws in Imitation Learning for Robotic Manipulation🚀—bringing us closer to the dream of having robots work as waiters in hot pot restaurants! 🍲

🤖 How can robot policies zero-shot generalize to any new environment and any new object? Introducing our new project: 🚀Data Scaling Laws in Imitation Learning for Robotic Manipulation🚀—bringing us closer to the dream of having robots work as waiters in hot pot restaurants! 🍲

Yang Gao

125,057 views • 1 year ago

How to harness foundation models for *generalization in the wild* in robot manipulation? Introducing VoxPoser: use LLM+VLM to label affordances and constraints directly in 3D perceptual space for zero-shot robot manipulation in the real world! 🌐 🧵👇

How to harness foundation models for generalization in the wild in robot manipulation? Introducing VoxPoser: use LLM+VLM to label affordances and constraints directly in 3D perceptual space for zero-shot robot manipulation in the real world! 🌐 🧵👇

Wenlong Huang

293,910 views • 3 years ago

Over the last few months, we’ve been thinking about how to learn from “off-domain” data - data from non-robot sources like video or simulation. These data sources are not quite good enough to learn policies (even monolithic VLA models) directly, but they still contain lots of information that can be useful for generalizable robot control. How can we develop robot learning models that are able to make use of this type of data for generalizable control? In new work, that we call HAMSTER, we show that VLMs can be useful for enabling robotic learning from off-domain data, but specifically when used through hierarchical VLA architectures. We show that this class of models can learn generalizable robot policies for the real world from large-scale, off-domain data. A 🧵 (1/10)

Over the last few months, we’ve been thinking about how to learn from “off-domain” data - data from non-robot sources like video or simulation. These data sources are not quite good enough to learn policies (even monolithic VLA models) directly, but they still contain lots of information that can be useful for generalizable robot control. How can we develop robot learning models that are able to make use of this type of data for generalizable control? In new work, that we call HAMSTER, we show that VLMs can be useful for enabling robotic learning from off-domain data, but specifically when used through hierarchical VLA architectures. We show that this class of models can learn generalizable robot policies for the real world from large-scale, off-domain data. A 🧵 (1/10)

Abhishek Gupta

11,994 views • 1 year ago

Tesla Optimus can arrange batteries in their factories, ours can do skincare (on Yuzhe Qin)! We opensource Bunny-VisionPro, a teleoperation system for bimanual hand manipulation. The users can control the robot hands in real time using VisionPro, flexible like a bunny. 🐇 We also have kitchen tasks, playing Rubik's Cube, and dynamic motion tasks. Imitation learning policies are trained on sweeping with a broom, serving a drink, and wiping glasses. Check our website for more details: The project is led by Runyu Ding Runyu Ding, Yuzhe Qin Yuzhe Qin , and Jiyue Zhu

Tesla Optimus can arrange batteries in their factories, ours can do skincare (on Yuzhe Qin)! We opensource Bunny-VisionPro, a teleoperation system for bimanual hand manipulation. The users can control the robot hands in real time using VisionPro, flexible like a bunny. 🐇 We also have kitchen tasks, playing Rubik's Cube, and dynamic motion tasks. Imitation learning policies are trained on sweeping with a broom, serving a drink, and wiping glasses. Check our website for more details: The project is led by Runyu Ding Runyu Ding, Yuzhe Qin Yuzhe Qin , and Jiyue Zhu

Xiaolong Wang

90,902 views • 2 years ago

Cross-embodied robot policies hold the promise of one policy to control all robots. But how far does transfer go? In new work, we study positive transfer between *manipulation* & *navigation* and show that nav data helps manipulation, and vice versa! 🧵 👇

Cross-embodied robot policies hold the promise of one policy to control all robots. But how far does transfer go? In new work, we study positive transfer between manipulation & navigation and show that nav data helps manipulation, and vice versa! 🧵 👇

Sergey Levine

42,721 views • 2 years ago

When will robots help us with our household chores? TidyBot++ brings us closer to that future. Our new open-source mobile manipulator makes it more accessible and practical to do robot learning research outside the lab, in real homes!

When will robots help us with our household chores? TidyBot++ brings us closer to that future. Our new open-source mobile manipulator makes it more accessible and practical to do robot learning research outside the lab, in real homes!

Jimmy Wu

188,242 views • 1 year ago

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 views • 2 years ago

✨ Introducing Keypoint Action Tokens. 🤖 We translate visual observations and robot actions into a "language" that off-the-shelf LLMs can ingest and output. This transforms LLMs into *in-context, low-level imitation learning machines*. 🚀 Let me explain. 👇🧵

✨ Introducing Keypoint Action Tokens. 🤖 We translate visual observations and robot actions into a "language" that off-the-shelf LLMs can ingest and output. This transforms LLMs into in-context, low-level imitation learning machines. 🚀 Let me explain. 👇🧵

Norman Di Palo

23,097 views • 2 years ago

Existing 3D human manipulation datasets are valuable, but are limited in scale and diversity. At #CVPR2025, we will introduce GigaHands👐 which, to our knowledge, is the most extensive 3D bimanual manipulation, interaction, and gesture dataset.🧵👇(1/9)

Existing 3D human manipulation datasets are valuable, but are limited in scale and diversity. At #CVPR2025, we will introduce GigaHands👐 which, to our knowledge, is the most extensive 3D bimanual manipulation, interaction, and gesture dataset.🧵👇(1/9)

Srinath Sridhar

13,581 views • 1 year ago

Is humanoid robot the home service robot of future? What if ppl are afraid of having that in their homes? What if we have something like Dreame Cyber X that can move up & down stairs + clean? What if you add an arm to it? Thread on how vacuum robot becomes primary home robot 👇

Is humanoid robot the home service robot of future? What if ppl are afraid of having that in their homes? What if we have something like Dreame Cyber X that can move up & down stairs + clean? What if you add an arm to it? Thread on how vacuum robot becomes primary home robot 👇

tphuang

21,376 views • 9 months ago

YouTube is a LARGE dataset of demonstration videos to train Generalist robot agents, but lacks action data. How can we learn DEXTEROUS skills from them? In #CoRL2024, we explore the problem of learning a Generalist Piano Playing agent from YouTube videos.

YouTube is a LARGE dataset of demonstration videos to train Generalist robot agents, but lacks action data. How can we learn DEXTEROUS skills from them? In #CoRL2024, we explore the problem of learning a Generalist Piano Playing agent from YouTube videos.

Julen Urain

42,451 views • 1 year ago

We made a physical AI agent that navigates in your home Our robot starts in a minute, can talk and listen, and use an arm This runs on Innate MARS, our open-source agentic platform–here with a Google Gemini based agent ⬇️ Closed beta available right-now

We made a physical AI agent that navigates in your home Our robot starts in a minute, can talk and listen, and use an arm This runs on Innate MARS, our open-source agentic platform–here with a Google Gemini based agent ⬇️ Closed beta available right-now

Axel

39,167 views • 4 months ago

A big part of scaling robot learning to solve real-world problems is that we somehow need to get enough diverse, high-quality data to train our robots to perform useful things. GPT and its fellow large language models were bootstrapped and proved out on a massive dataset of real-world language data. Unfortunately, despite our best efforts, similarly massive datasets don’t really exist for robotics — so, in our unending pursuit of high-quality, useful data, we turn to simulation. I compared a couple recent works on sim-to-real robot manipulation, which discuss how to train perception-driven manipulation policies in simulation, in such a way that they’re useful in the real world. - DextraH-RGB, from NVIDIA - Sim-and-Real Co-Training: A Simple Recipe for Vision-Based Robotic Manipulation, also from NVIDIA — specifically the GEAR lab - Sim-to-Real Reinforcement Learning for Vision-Based Dexterous Manipulation on Humanoids, another GEAR lab paper - Local Policies Enable Zero-shot Long-Horizon Manipulation, from CMU (video from DextrAH-RGB)

A big part of scaling robot learning to solve real-world problems is that we somehow need to get enough diverse, high-quality data to train our robots to perform useful things. GPT and its fellow large language models were bootstrapped and proved out on a massive dataset of real-world language data. Unfortunately, despite our best efforts, similarly massive datasets don’t really exist for robotics — so, in our unending pursuit of high-quality, useful data, we turn to simulation. I compared a couple recent works on sim-to-real robot manipulation, which discuss how to train perception-driven manipulation policies in simulation, in such a way that they’re useful in the real world. - DextraH-RGB, from NVIDIA - Sim-and-Real Co-Training: A Simple Recipe for Vision-Based Robotic Manipulation, also from NVIDIA — specifically the GEAR lab - Sim-to-Real Reinforcement Learning for Vision-Based Dexterous Manipulation on Humanoids, another GEAR lab paper - Local Policies Enable Zero-shot Long-Horizon Manipulation, from CMU (video from DextrAH-RGB)

Chris Paxton

20,486 views • 1 year ago

Robot Learning needs 4D world models! Robot Learning needs 4D world models! Robot Learning needs 4D world models! We introduce TesserAct, a 4D embodied world model that can simulate how agents interact with the 3D world over time! We achieve this by simply extending a pre-trained 2D video generation model to jointly predict RGB, depth, and surface normals. It enables: 1️⃣ Much better policy learning in the wild 2️⃣ Temporal + spatial coherence in 4D dynamic prediction 3️⃣ Novel view synthesis for embodied scenes Code: Paper Link: Project page:

Robot Learning needs 4D world models! Robot Learning needs 4D world models! Robot Learning needs 4D world models! We introduce TesserAct, a 4D embodied world model that can simulate how agents interact with the 3D world over time! We achieve this by simply extending a pre-trained 2D video generation model to jointly predict RGB, depth, and surface normals. It enables: 1️⃣ Much better policy learning in the wild 2️⃣ Temporal + spatial coherence in 4D dynamic prediction 3️⃣ Novel view synthesis for embodied scenes Code: Paper Link: Project page:

Chuang Gan

43,265 views • 1 year ago

We have seen a lot of legged robots doing navigation in the wild. But how about mobile manipulation in the wild? I have been pushing the direction of learning a unified, efficient, and dynamic 3D representation of scenes (for navigation) and objects (for manipulation) for the past two years. And now we have GeFF --- our large-scale, generalizable feature field, that combines the speed of a feed-forward neural network with the rich semantics from Foundation Models, to handle dynamically changing scenes, and enable open-ended, language-grounded scene and object understanding.

We have seen a lot of legged robots doing navigation in the wild. But how about mobile manipulation in the wild? I have been pushing the direction of learning a unified, efficient, and dynamic 3D representation of scenes (for navigation) and objects (for manipulation) for the past two years. And now we have GeFF --- our large-scale, generalizable feature field, that combines the speed of a feed-forward neural network with the rich semantics from Foundation Models, to handle dynamically changing scenes, and enable open-ended, language-grounded scene and object understanding.

Xiaolong Wang

42,767 views • 2 years ago

Current robot learning methods are good at imitating tasks seen during training, but struggle to compose behaviors in new ways. When training imitation policies, we found something surprising—using temporally-aligned task representations enabled compositional generalization. 1/

Current robot learning methods are good at imitating tasks seen during training, but struggle to compose behaviors in new ways. When training imitation policies, we found something surprising—using temporally-aligned task representations enabled compositional generalization. 1/

Vivek Myers

39,477 views • 1 year ago

Excited to release OK-Robot, an open-vocabulary mobile-manipulator for homes. Simply tell the robot what to pick and where to drop it in natural language, and it will do it. Like: Me: "OK Robot, move the Takis from the desk to the nightstand" Robot: ⬇️

Excited to release OK-Robot, an open-vocabulary mobile-manipulator for homes. Simply tell the robot what to pick and where to drop it in natural language, and it will do it. Like: Me: "OK Robot, move the Takis from the desk to the nightstand" Robot: ⬇️

Lerrel Pinto

152,292 views • 2 years ago

How to use simulation data for real-world robot manipulation? We present sim-and-real co-training, a simple recipe for manipulation. We demonstrate that sim data can significantly enhance real-world performance, even with notable differences between the sim and the real. (1/n)

How to use simulation data for real-world robot manipulation? We present sim-and-real co-training, a simple recipe for manipulation. We demonstrate that sim data can significantly enhance real-world performance, even with notable differences between the sim and the real. (1/n)

Zhenyu Jiang

44,310 views • 1 year ago

Introducing Learning WalkThrus: How We Learn - a short guide for students. A taste of the content coming soon to all our members via the website and covered in our Learning Walkthrus book out now:

Introducing Learning WalkThrus: How We Learn - a short guide for students. A taste of the content coming soon to all our members via the website and covered in our Learning Walkthrus book out now:

Walkthrus

81,423 views • 2 years ago