Video yükleniyor...

Video Yüklenemedi

Ana Sayfaya Dön

New project! Flow Policy Gradients for Robot Control tldr; a simple online RL recipe for training and fine-tuning flow policies for robots co-led w/ Hongsuk Benjamin Choi:

brenthyi's profile picture

Brent Yi

1,773 subscribers

92,555 görüntüleme • 6 ay önce •via X (Twitter)

LIVE

Anya Rossi• Live Now

Private livecam show

0 Yorum

Yorum bulunmuyor

Orijinal gönderinin yorumları burada görünecek

Benzer Videolar

🤔 How to fine-tune an Imitation Learning policy (e.g., Diffusion Policy, ACT) with RL? As an RL practitioner, I’ve been struggling with this problem for a while. Here’s why it’s tough: 1️⃣ Special designs (usually for multimodal action distributions) in modern IL models make them non-trivial to fine-tune by RL. 2️⃣ Large policy models + RL's poor sample efficiency = a nightmare But finally, we figured out a simple solution that works for any model architecture! 🌟 Check out our #ICLR2025 paper: “Policy Decorator: Model-Agnostic Online Refinement for Large Policy Models”, led by my amazing mentee Xiu Yuan. 🔗 🧵 Read more below!
0:16

🤔 How to fine-tune an Imitation Learning policy (e.g., Diffusion Policy, ACT) with RL? As an RL practitioner, I’ve been struggling with this problem for a while. Here’s why it’s tough: 1️⃣ Special designs (usually for multimodal action distributions) in modern IL models make them non-trivial to fine-tune by RL. 2️⃣ Large policy models + RL's poor sample efficiency = a nightmare But finally, we figured out a simple solution that works for any model architecture! 🌟 Check out our #ICLR2025 paper: “Policy Decorator: Model-Agnostic Online Refinement for Large Policy Models”, led by my amazing mentee Xiu Yuan. 🔗 🧵 Read more below!

Tongzhou Mu 🤖🦾🦿

17,018 görüntüleme • 1 yıl önce

Reinforcement Learning (RL) has long been the dominant method for fine-tuning, powering many state-of-the-art LLMs. Methods like PPO and GRPO explore in action space. But can we instead explore directly in parameter space? YES we can. We propose a scalable framework for full-parameter fine-tuning using Evolution Strategies (ES). By skipping gradients and optimizing directly in parameter space, ES achieves more accurate, efficient, and stable fine-tuning. Paper: Code:
0:16

Reinforcement Learning (RL) has long been the dominant method for fine-tuning, powering many state-of-the-art LLMs. Methods like PPO and GRPO explore in action space. But can we instead explore directly in parameter space? YES we can. We propose a scalable framework for full-parameter fine-tuning using Evolution Strategies (ES). By skipping gradients and optimizing directly in parameter space, ES achieves more accurate, efficient, and stable fine-tuning. Paper: Code:

Yulu Gan

415,417 görüntüleme • 10 ay önce

How does high-fidelity tactile simulation help robots nail the last millimeter? We’re releasing VT-Refine, accepted to CoRL: a real-to-sim-to-real visuo-tactile policy using a GPU-parallel tactile sim for our piezoresistive skin FlexiTac. Then fine-tuning a diffusion policy with large-scale RL in simulation. Website: #CoRL2025 #RobotLearning #Sim2Real
0:17

How does high-fidelity tactile simulation help robots nail the last millimeter? We’re releasing VT-Refine, accepted to CoRL: a real-to-sim-to-real visuo-tactile policy using a GPU-parallel tactile sim for our piezoresistive skin FlexiTac. Then fine-tuning a diffusion policy with large-scale RL in simulation. Website: #CoRL2025 #RobotLearning #Sim2Real

Binghao Huang

46,982 görüntüleme • 10 ay önce

MolmoAct2 is landing in LeRobot! Ai2's open Action Reasoning Model combines a Molmo2-ER vision-language backbone with a flow-matching continuous action expert to predict robot action chunks from images, language instructions, and proprioceptive state. An open robot foundation model built for real-world control, with strong out-of-the-box performance and easy fine-tuning in LeRobot. Pick-and-place inference running on NVIDIA DGX Spark! Blog: Paper: Thanks to Ai2 Jiafei Duan Haoquan Fang
0:11

MolmoAct2 is landing in LeRobot! Ai2's open Action Reasoning Model combines a Molmo2-ER vision-language backbone with a flow-matching continuous action expert to predict robot action chunks from images, language instructions, and proprioceptive state. An open robot foundation model built for real-world control, with strong out-of-the-box performance and easy fine-tuning in LeRobot. Pick-and-place inference running on NVIDIA DGX Spark! Blog: Paper: Thanks to Ai2 Jiafei Duan Haoquan Fang

LeRobot

24,727 görüntüleme • 2 ay önce

I just spent weeks handwriting a 200+ page guide on the entirety of Robotics foundations from scratch. Featuring: - Kinematics (Forward, Inverse Kin, Jacobians, DoF) - Controls (PID, Force/Impedance Control, Feedforward) - SLAM (Bayes Filters, Kalman, Particle Filters, Mapping) - Motion Planning (Dijkstra, A*, RRT, RRT*) - RL (MDP, Policy Gradients, PPO, Sim2Real, Value Func) - VLA (Q/K/V, ViT, SigLIP, Diffusion/Flow-Matching) - ACT, Diffusion Policy, UMI, LeRobot tooling - World Model (Dreamer, VideoGen, Gaussian Splatting) Everything to get started, for free.
0:19

I just spent weeks handwriting a 200+ page guide on the entirety of Robotics foundations from scratch. Featuring: - Kinematics (Forward, Inverse Kin, Jacobians, DoF) - Controls (PID, Force/Impedance Control, Feedforward) - SLAM (Bayes Filters, Kalman, Particle Filters, Mapping) - Motion Planning (Dijkstra, A*, RRT, RRT*) - RL (MDP, Policy Gradients, PPO, Sim2Real, Value Func) - VLA (Q/K/V, ViT, SigLIP, Diffusion/Flow-Matching) - ACT, Diffusion Policy, UMI, LeRobot tooling - World Model (Dreamer, VideoGen, Gaussian Splatting) Everything to get started, for free.

Arjun Virk

64,134 görüntüleme • 1 ay önce

Success! 🍎🤖 I can now control Unitree robots on a Mac Mini M4 running macOS with a simple tweak! The unitree_sdk2_python ( thread implementation isn’t essential—getting the DDS subscriber and publisher to work is enough. I replaced Unitree’s Linux thread logic with a simple Python while loop for sensor reading and robot control. Here’s a video of me controlling a Go2 robot on macOS (M4 Mac Mini) using a Python script for low-level control and sensor reading.
0:11

Success! 🍎🤖 I can now control Unitree robots on a Mac Mini M4 running macOS with a simple tweak! The unitree_sdk2_python ( thread implementation isn’t essential—getting the DDS subscriber and publisher to work is enough. I replaced Unitree’s Linux thread logic with a simple Python while loop for sensor reading and robot control. Here’s a video of me controlling a Go2 robot on macOS (M4 Mac Mini) using a Python script for low-level control and sensor reading.

Tairan He

57,603 görüntüleme • 1 yıl önce

Generalist robot policies need a benchmark that works across any robot and any policy. 🦾 Introducing RoboLab, a high‑fidelity simulation benchmark built on NVIDIA Isaac and Omniverse to evaluate generalist robot policies in diverse, photoreal, physics‑based environments. Coming soon to the NVIDIA Isaac Lab‑Arena roadmap for large‑scale, robotic policy evaluation. 📖 #NationalRoboticsWeek
0:15

Generalist robot policies need a benchmark that works across any robot and any policy. 🦾 Introducing RoboLab, a high‑fidelity simulation benchmark built on NVIDIA Isaac and Omniverse to evaluate generalist robot policies in diverse, photoreal, physics‑based environments. Coming soon to the NVIDIA Isaac Lab‑Arena roadmap for large‑scale, robotic policy evaluation. 📖 #NationalRoboticsWeek

NVIDIA Robotics

23,872 görüntüleme • 4 ay önce

Introducing CQN: Coarse-to-fine Q-Network, a value-based RL algorithm for continuous control🦾Initialized with 20~50 demonstrations, it learns to solve real-world robotic tasks within 10 mins of training, without any pre-training and shaped rewards! (1/4)
0:10

Introducing CQN: Coarse-to-fine Q-Network, a value-based RL algorithm for continuous control🦾Initialized with 20~50 demonstrations, it learns to solve real-world robotic tasks within 10 mins of training, without any pre-training and shaped rewards! (1/4)

Younggyo Seo

16,413 görüntüleme • 2 yıl önce

New research from Databricks: LLMs Can Learn to Reason via Off-Policy RL Optimal Advantage-based Policy Optimization with Lagged Inference policy (OAPL) shows you don’t need strict on-policy training to improve reasoning. It matches or beats Group Relative Policy Optimization (GRPO), stays stable with large policy lag, and uses ~3× fewer training generations. For Databricks customers, it’s a simpler, practical, and equally powerful approach to RL that Databricks is pioneering internally — and bringing directly to Databricks customers, so enterprises can improve agents using the same methods we use for our in-house agents, without complex infrastructure changes.
0:13

New research from Databricks: LLMs Can Learn to Reason via Off-Policy RL Optimal Advantage-based Policy Optimization with Lagged Inference policy (OAPL) shows you don’t need strict on-policy training to improve reasoning. It matches or beats Group Relative Policy Optimization (GRPO), stays stable with large policy lag, and uses ~3× fewer training generations. For Databricks customers, it’s a simpler, practical, and equally powerful approach to RL that Databricks is pioneering internally — and bringing directly to Databricks customers, so enterprises can improve agents using the same methods we use for our in-house agents, without complex infrastructure changes.

Databricks AI Research

12,699 görüntüleme • 5 ay önce

D4RL is a great benchmark, but is saturated. Introducing OGBench, a new benchmark for offline goal-conditioned RL and offline RL! Tasks include HumanoidMaze, Puzzle, Drawing, and more 🙂 Project page: GitHub: 🧵↓
0:16

D4RL is a great benchmark, but is saturated. Introducing OGBench, a new benchmark for offline goal-conditioned RL and offline RL! Tasks include HumanoidMaze, Puzzle, Drawing, and more 🙂 Project page: GitHub: 🧵↓

Seohong Park

36,447 görüntüleme • 1 yıl önce

Disney’s testing a new crowd control system for the first day of Flower & Garden today! Three lines: attractions, pins, and merch — they’ll be released at different stages to help crowd flow!
0:16

Disney’s testing a new crowd control system for the first day of Flower & Garden today! Three lines: attractions, pins, and merch — they’ll be released at different stages to help crowd flow!

Disney Food Blog

31,751 görüntüleme • 5 ay önce

Can robots learn without training❓ [𝗜𝘁'𝘀 𝗼𝗽𝗲𝗻 𝘀𝗼𝘂𝗿𝗰𝗲𝗱 ⬇ ] Teaching robots to do complex tasks WITHOUT spending hours training them. Sounds cool, right? That's exactly what DIAL-MPC does! The first training-free method for whole-body torque control using full-order dynamics: ✅ Instantly checks if a robot's moves are right or wrong ✅ Adapts quickly to new tasks without needing extra training ✅ Could work hand-in-hand with other robot learning methods Robots are getting smarter AND faster without the need for long training sessions. Website: Paper: Code: Saw this first Haoru Xue ✈️ CVPR 🙏
0:10

Can robots learn without training❓ [𝗜𝘁'𝘀 𝗼𝗽𝗲𝗻 𝘀𝗼𝘂𝗿𝗰𝗲𝗱 ⬇ ] Teaching robots to do complex tasks WITHOUT spending hours training them. Sounds cool, right? That's exactly what DIAL-MPC does! The first training-free method for whole-body torque control using full-order dynamics: ✅ Instantly checks if a robot's moves are right or wrong ✅ Adapts quickly to new tasks without needing extra training ✅ Could work hand-in-hand with other robot learning methods Robots are getting smarter AND faster without the need for long training sessions. Website: Paper: Code: Saw this first Haoru Xue ✈️ CVPR 🙏

Ilir Aliu

71,502 görüntüleme • 1 yıl önce

🚀 Introducing Frank 🤖—a whole-body robot control system for day-to-day household chores. Frank has been in works for the past year—a tightly coupled hardware and a remote teleoperation interface co-led by Michelle, Nandini, and Pulkit Agrawal. Thanks to Aditya Agarwal and Shankara Narayanan for the awesome help with recording the videos! 🎥 🧵[1/6]
0:14

🚀 Introducing Frank 🤖—a whole-body robot control system for day-to-day household chores. Frank has been in works for the past year—a tightly coupled hardware and a remote teleoperation interface co-led by Michelle, Nandini, and Pulkit Agrawal. Thanks to Aditya Agarwal and Shankara Narayanan for the awesome help with recording the videos! 🎥 🧵[1/6]

Bipasha Sen

51,137 görüntüleme • 1 yıl önce

Announcing our commercial partnership with Booster Robotics Booster builds humanoid robot hardware, OS, and developer tools to make humanoid robots more affordable, reliable, and practical. The partnership centers on using simulation to multiply the value of real-robot data—expanding teleoperation demonstrations into scalable training data across diverse tasks and scenes. This joint effort powers sim-real co-training and foundation-model development, accelerating progress from hardware iteration to deployable robot policies. Together, we're building simulation‑powered data infrastructure for Physical AI — making scalable training data accessible to model developers and the broader robotics ecosystem.
0:12

Announcing our commercial partnership with Booster Robotics Booster builds humanoid robot hardware, OS, and developer tools to make humanoid robots more affordable, reliable, and practical. The partnership centers on using simulation to multiply the value of real-robot data—expanding teleoperation demonstrations into scalable training data across diverse tasks and scenes. This joint effort powers sim-real co-training and foundation-model development, accelerating progress from hardware iteration to deployable robot policies. Together, we're building simulation‑powered data infrastructure for Physical AI — making scalable training data accessible to model developers and the broader robotics ecosystem.

Axis Robotics

65,424 görüntüleme • 13 gün önce

𝗗𝗼𝗻'𝘁 𝗳𝗶𝗻𝗲-𝘁𝘂𝗻𝗲 𝗿𝗼𝗯𝗼𝘁 𝗳𝗼𝘂𝗻𝗱𝗮𝘁𝗶𝗼𝗻 𝗺𝗼𝗱𝗲𝗹𝘀. 𝗦𝘁𝗲𝗲𝗿 𝘁𝗵𝗲𝗺 𝘄𝗶𝘁𝗵 𝗵𝘂𝗺𝗮𝗻 𝗰𝗼𝗿𝗿𝗲𝗰𝘁𝗶𝗼𝗻𝘀 𝗶𝗻𝘀𝘁𝗲𝗮𝗱, 𝘄𝗶𝘁𝗵𝗼𝘂𝘁 𝗰𝗵𝗮𝗻𝗴𝗶𝗻𝗴 𝘁𝗵𝗲 𝗯𝗮𝘀𝗲 𝗽𝗼𝗹𝗶𝗰𝘆 Modern VLAs and world-action models can perform impressive manipulation skills, but adapting them reliably to new robots and tasks remains challenging. A natural solution is DAgger-style online imitation learning: deploy the robot, collect human corrections, and update the policy. Yet foundation models are fragile in the low-data regime, fine-tuning on a handful of interventions can improve one behavior while degrading others. Online post-training or reinforcement learning can require costly data collection and exploration, making real-world learning expensive and potentially unsafe. In our new paper, 𝗙𝗹𝗼𝘄𝗗𝗔𝗴𝗴𝗲𝗿, we take a different approach: 𝗜𝗻𝘀𝘁𝗲𝗮𝗱 𝗼𝗳 𝗰𝗵𝗮𝗻𝗴𝗶𝗻𝗴 𝘁𝗵𝗲 𝗳𝗼𝘂𝗻𝗱𝗮𝘁𝗶𝗼𝗻 𝗺𝗼𝗱𝗲𝗹, 𝘄𝗲 𝗹𝗲𝗮𝗿𝗻 𝗵𝗼𝘄 𝘁𝗼 𝘀𝘁𝗲𝗲𝗿 𝗶𝘁 𝗳𝗿𝗼𝗺 𝗵𝘂𝗺𝗮𝗻 𝗰𝗼𝗿𝗿𝗲𝗰𝘁𝗶𝗼𝗻𝘀. The key idea is 𝗮𝗰𝘁𝗶𝗼𝗻 𝗶𝗻𝘃𝗲𝗿𝘀𝗶𝗼𝗻: we map human corrective actions back into the latent noise space of the frozen generative policy. These latent targets train a lightweight controller that adapts the robot while preserving the original model's capabilities. Across simulation and real robots, FlowDAgger: 📈 Learns from only 5–20 human intervention episodes 🏆 Outperforms supervised fine-tuning and latent-space reinforcement learning 🤖 Works across VLAs, diffusion policies, and world-action models ✔️ Provides reliable improvements without modifying the pretrained policy We believe this offers a practical path toward making robot foundation models improve during deployment, learning from the way humans naturally teach: through corrections. 📄 Paper: 🌐 Project: 💻 Code: This project was led by my amazing colleague Michael Murray with help from Daphne Chen, Simran Bagaria, Dean Fortier, Tess Hellebrekers, Harshavardhan Reddy Gajarla, Galen Mullins and Andrey Kolobov at Microsoft Research and Maya Cakmak at University of Washington
0:11

𝗗𝗼𝗻'𝘁 𝗳𝗶𝗻𝗲-𝘁𝘂𝗻𝗲 𝗿𝗼𝗯𝗼𝘁 𝗳𝗼𝘂𝗻𝗱𝗮𝘁𝗶𝗼𝗻 𝗺𝗼𝗱𝗲𝗹𝘀. 𝗦𝘁𝗲𝗲𝗿 𝘁𝗵𝗲𝗺 𝘄𝗶𝘁𝗵 𝗵𝘂𝗺𝗮𝗻 𝗰𝗼𝗿𝗿𝗲𝗰𝘁𝗶𝗼𝗻𝘀 𝗶𝗻𝘀𝘁𝗲𝗮𝗱, 𝘄𝗶𝘁𝗵𝗼𝘂𝘁 𝗰𝗵𝗮𝗻𝗴𝗶𝗻𝗴 𝘁𝗵𝗲 𝗯𝗮𝘀𝗲 𝗽𝗼𝗹𝗶𝗰𝘆 Modern VLAs and world-action models can perform impressive manipulation skills, but adapting them reliably to new robots and tasks remains challenging. A natural solution is DAgger-style online imitation learning: deploy the robot, collect human corrections, and update the policy. Yet foundation models are fragile in the low-data regime, fine-tuning on a handful of interventions can improve one behavior while degrading others. Online post-training or reinforcement learning can require costly data collection and exploration, making real-world learning expensive and potentially unsafe. In our new paper, 𝗙𝗹𝗼𝘄𝗗𝗔𝗴𝗴𝗲𝗿, we take a different approach: 𝗜𝗻𝘀𝘁𝗲𝗮𝗱 𝗼𝗳 𝗰𝗵𝗮𝗻𝗴𝗶𝗻𝗴 𝘁𝗵𝗲 𝗳𝗼𝘂𝗻𝗱𝗮𝘁𝗶𝗼𝗻 𝗺𝗼𝗱𝗲𝗹, 𝘄𝗲 𝗹𝗲𝗮𝗿𝗻 𝗵𝗼𝘄 𝘁𝗼 𝘀𝘁𝗲𝗲𝗿 𝗶𝘁 𝗳𝗿𝗼𝗺 𝗵𝘂𝗺𝗮𝗻 𝗰𝗼𝗿𝗿𝗲𝗰𝘁𝗶𝗼𝗻𝘀. The key idea is 𝗮𝗰𝘁𝗶𝗼𝗻 𝗶𝗻𝘃𝗲𝗿𝘀𝗶𝗼𝗻: we map human corrective actions back into the latent noise space of the frozen generative policy. These latent targets train a lightweight controller that adapts the robot while preserving the original model's capabilities. Across simulation and real robots, FlowDAgger: 📈 Learns from only 5–20 human intervention episodes 🏆 Outperforms supervised fine-tuning and latent-space reinforcement learning 🤖 Works across VLAs, diffusion policies, and world-action models ✔️ Provides reliable improvements without modifying the pretrained policy We believe this offers a practical path toward making robot foundation models improve during deployment, learning from the way humans naturally teach: through corrections. 📄 Paper: 🌐 Project: 💻 Code: This project was led by my amazing colleague Michael Murray with help from Daphne Chen, Simran Bagaria, Dean Fortier, Tess Hellebrekers, Harshavardhan Reddy Gajarla, Galen Mullins and Andrey Kolobov at Microsoft Research and Maya Cakmak at University of Washington

Oier Mees

13,032 görüntüleme • 1 ay önce

🤖 Another zero-shot reward model is now in LeRobot: ROBOMETER. A general-purpose, zero-shot video-language reward model from University of South Carolina, UT Dallas, Massachusetts Institute of Technology (MIT), University of Washington, Ai2, and NVIDIA that predicts frame-level task progress. Trained on 1M+ trajectories from 21 robot embodiments, generalizes zero-shot to unseen tasks, scenes, and robots. 2.4–4.5x better downstream success rates across online RL, offline RL, data filtering, failure detection, and data retrieval for IL. Project: Paper:
0:14

🤖 Another zero-shot reward model is now in LeRobot: ROBOMETER. A general-purpose, zero-shot video-language reward model from University of South Carolina, UT Dallas, Massachusetts Institute of Technology (MIT), University of Washington, Ai2, and NVIDIA that predicts frame-level task progress. Trained on 1M+ trajectories from 21 robot embodiments, generalizes zero-shot to unseen tasks, scenes, and robots. 2.4–4.5x better downstream success rates across online RL, offline RL, data filtering, failure detection, and data retrieval for IL. Project: Paper:

LeRobot

32,625 görüntüleme • 2 ay önce

We built a bipedal robot for about $2,500. A real, mostly 3D-printed robot you can build, repair, simulate, train, and control. Today we’re releasing LeRobot Humanoid: an open robot-learning platform with hardware, runtime, identification tools, and training environments. Blog post: Repo:
0:16

We built a bipedal robot for about $2,500. A real, mostly 3D-printed robot you can build, repair, simulate, train, and control. Today we’re releasing LeRobot Humanoid: an open robot-learning platform with hardware, runtime, identification tools, and training environments. Blog post: Repo:

LeRobot

199,750 görüntüleme • 3 ay önce