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.

Low-cost teleop systems have democratized robot data collection, but they lack any force feedback, making it challenging to teleoperate contact-rich tasks. Many robot arms provide force information — a critical yet underutilized modality in robot learning. We introduce: 1. 🦾A low-cost, force-feedback-enabled teleop system. 2. 🥊Force-Attending Curriculum Training (FACTR)... show more

Jason Liu

2,154 subscribers

152,482 views • 1 year ago •via X (Twitter)

Science & Technology Education

Anya Rossi• Live Now

Private livecam show

11 Comments

Jason Liu1 year ago

🦾Our teleop system (~$1K per leader arm) lets users intuitively feel forces experienced by robots, without additional sensor hardware. This can be done for many arms. We also integrate gravity comp, redundancy resolution, etc. in the teleop leader arm, augmenting the teleop experience. (2/N)

Jason Liu1 year ago

🥊 Naively adding force to policy learning does not guarantee performance improvement. BC policies often overfit to vision input and ignore force when added naively, limiting performance in contact-rich tasks. (3/N)

Jason Liu1 year ago

🥊 We propose FACTR, a curriculum that corrupts vision with decreasing intensity during training. This prevents overfitting to vision and guides the policy to properly attend to force. (4/N)

Jason Liu1 year ago

🥊 FACTR policies generalize to unseen objects significantly better than baseline policies (vision+force without FACTR). (5/N)

Jason Liu1 year ago

🥊 We visualize cross attention of action tokens to force (blue) and vision (orange) tokens. [Left] Without FACTR, policy does not pay much attention to force. [Right] FACTR policies properly attend to force, even learning mode-switches, indicated by force attention outweighing vision. (6/N)

Jason Liu1 year ago

🥊 By attending to force, FACTR policies also exhibit emergent recovery behavior (not in training data), even for unseen objects. We do not observe this in baseline policies. (7/N)

Jason Liu1 year ago

Both our teleop system and autonomous policies significantly outperform baselines. FACTR is easy to set up on existing systems. All components will be open-sourced soon! w/ @yulongli42 @kenny__shaw @_tonytao_ @rsalakhu @pathak2206 (8/N)

ARK Electronics2 years ago

Excited about the latest tech for your drone product? Our NDAA-compliant, US-made flight controllers are designed to accelerate your path to market and provide a solid platform for developing your autonomous software. Check them out! #Drones #UAV #UAS #Robotics #MadeInUSA

Masato Kobayashi @るっと🐺1 year ago

Excellent work🎉! I believe that force information/control is crucial for robot manipulation. If you are interested in my research, please feel free to visit my post and paper, highlighting the importance of position and force information and control !

Klajd Lika`1 year ago

Did you use the internal force estimates from Franka? You can check also our force sensing platform that converts any robot to a multi axis force sensor. We are also working in something. Let’s keep in touch

Jason Liu1 year ago

We used the built-in external force estimates from Franka. What you have here is super cool! Let’s keep in touch

Related Videos

💥Introducing FACTR 2, learning external force sensing on commodity robot arms without needing dedicated sensors. We show that learned force signals enable force-feedback teleop on low-cost arms and improve BC policies. FACTR 2 consists of: 1. Neural External Torque (NEXT): learns external forces without needing dedicated force sensors. 2. Force-Informed Re-Sampling Training (FIRST): uses the learned force signal to identify task-critical regions and upsample them during training. w/ Steven Oh Tony Tao 🧵(1/N)

💥Introducing FACTR 2, learning external force sensing on commodity robot arms without needing dedicated sensors. We show that learned force signals enable force-feedback teleop on low-cost arms and improve BC policies. FACTR 2 consists of: 1. Neural External Torque (NEXT): learns external forces without needing dedicated force sensors. 2. Force-Informed Re-Sampling Training (FIRST): uses the learned force signal to identify task-critical regions and upsample them during training. w/ Steven Oh Tony Tao 🧵(1/N)

Jason Liu

113,061 views • 3 months ago

One less-known fact about glove-based data collection: it produces higher quality data than teleop on contact-rich tasks. Remote teleop can’t provide good force feedback, but gloves do naturally, making tasks like sock folding, which rely on feel, far easier to capture.

One less-known fact about glove-based data collection: it produces higher quality data than teleop on contact-rich tasks. Remote teleop can’t provide good force feedback, but gloves do naturally, making tasks like sock folding, which rely on feel, far easier to capture.

Tony Zhao

189,503 views • 10 months ago

Force is arguably the most overlooked ingredient in modern robot learning. Introducing FACTR 2: it turns *any* commodity robot into a force-aware system with no force sensors required. Train a tiny force network in <1min with <10mins of data and drop it into any existing teleop pipelines: ✅ Free force sensing for both the robot and the operator arm ✅ Makes demos higher-quality → fewer of them needed. ✅ A new force-aware learning algorithm (FIRST) uses those recovered forces to figure out which parts of a demo actually matter, making learning data-efficient. ✅ Strong performance on complex tasks with fewer demos and even no pretraining! More details below.

Force is arguably the most overlooked ingredient in modern robot learning. Introducing FACTR 2: it turns any commodity robot into a force-aware system with no force sensors required. Train a tiny force network in <1min with <10mins of data and drop it into any existing teleop pipelines: ✅ Free force sensing for both the robot and the operator arm ✅ Makes demos higher-quality → fewer of them needed. ✅ A new force-aware learning algorithm (FIRST) uses those recovered forces to figure out which parts of a demo actually matter, making learning data-efficient. ✅ Strong performance on complex tasks with fewer demos and even no pretraining! More details below.

Deepak Pathak

41,030 views • 3 months ago

1/🧠Humans are the best robot data source — but video alone misses one thing: force. 2/🙁Tactile gloves capture force — but they're costly and block the real touch manipulation depends on. 3/💪Maybe the future of touch lives on your wrist: surface EMG reads the muscles that cause force — tactile sensing without ever touching a tactile sensor. 4/🔥Want a fully open-source framework — hardware + software — to train your own force-aware learn-from-human-data robot policy? 🚀We introduce ForceBand: Learning Forceful Manipulation with sEMG -- bring force into human videos with sEMG, for force-aware manipulation ⬇️ ✦ Zero-Shot Human-to-Robot Transfer ✦ Force Beyond Vision ✦ Free-Hand Force Sensing ✦ Collect by Anyone, Anytime, Anywhere ✦ Deploy on Any Robot, Any Camera, Any Environment ✦ Open-Source & Low-Cost & Easy-to-Implement Let's squeeze every bit of signal out of human data, and let robots feel the force! 🌐 Website: 📄 Paper: 💻 Code: 🎥 Video: 🧵 1/n

1/🧠Humans are the best robot data source — but video alone misses one thing: force. 2/🙁Tactile gloves capture force — but they're costly and block the real touch manipulation depends on. 3/💪Maybe the future of touch lives on your wrist: surface EMG reads the muscles that cause force — tactile sensing without ever touching a tactile sensor. 4/🔥Want a fully open-source framework — hardware + software — to train your own force-aware learn-from-human-data robot policy? 🚀We introduce ForceBand: Learning Forceful Manipulation with sEMG -- bring force into human videos with sEMG, for force-aware manipulation ⬇️ ✦ Zero-Shot Human-to-Robot Transfer ✦ Force Beyond Vision ✦ Free-Hand Force Sensing ✦ Collect by Anyone, Anytime, Anywhere ✦ Deploy on Any Robot, Any Camera, Any Environment ✦ Open-Source & Low-Cost & Easy-to-Implement Let's squeeze every bit of signal out of human data, and let robots feel the force! 🌐 Website: 📄 Paper: 💻 Code: 🎥 Video: 🧵 1/n

Zhi (Leo) Wang

52,455 views • 3 months ago

So I heard we need more data for robot learning :) Purely real world teleop is expensive and slow, making large scale data collection challenging. I’ve been excited about getting more data into robot learning, going beyond just real-world teleop data. To this end, we’ve been scaling up data generation with RL in realistic simulations generated on the fly from crowdsourced videos. Enables realistic data collection, much more cheaply than purely real world teleop. Importantly, data collection becomes even*cheaper* with more environments, allowing training with over 100x more data. Transfers to real robots for generalizable manipulation. A 🧵 (1/N)

So I heard we need more data for robot learning :) Purely real world teleop is expensive and slow, making large scale data collection challenging. I’ve been excited about getting more data into robot learning, going beyond just real-world teleop data. To this end, we’ve been scaling up data generation with RL in realistic simulations generated on the fly from crowdsourced videos. Enables realistic data collection, much more cheaply than purely real world teleop. Importantly, data collection becomes evencheaper with more environments, allowing training with over 100x more data. Transfers to real robots for generalizable manipulation. A 🧵 (1/N)

Abhishek Gupta

13,358 views • 1 year ago

What if some parts of a robot demonstration are more important than others? Most of a trajectory is free-space motion. But success or failure is often determined by a few critical moments around contact. In FACTR 2, we use force to find these moments and prioritize them for training. We find this helps policies learn better alignment and recovery behaviors, like the example below. w/ Steven Oh Jason Liu 🧵(1/N)

What if some parts of a robot demonstration are more important than others? Most of a trajectory is free-space motion. But success or failure is often determined by a few critical moments around contact. In FACTR 2, we use force to find these moments and prioritize them for training. We find this helps policies learn better alignment and recovery behaviors, like the example below. w/ Steven Oh Jason Liu 🧵(1/N)

Tony Tao

24,319 views • 3 months ago

Gemini Robotics 2 by Google DeepMind uses sim-only post-training on a mixed-data base model to handle automotive kitting. Contact-rich tasks like these have long been considered one of the hardest challenges in Sim2Real. Gemini Robotics 2 shows simulation can work even here. General robot intelligence needs much more data than real-robot collection alone can provide. Simulation has proven to be an effective way to fill that gap.

Gemini Robotics 2 by Google DeepMind uses sim-only post-training on a mixed-data base model to handle automotive kitting. Contact-rich tasks like these have long been considered one of the hardest challenges in Sim2Real. Gemini Robotics 2 shows simulation can work even here. General robot intelligence needs much more data than real-robot collection alone can provide. Simulation has proven to be an effective way to fill that gap.

Axis Robotics

16,857 views • 1 month ago

Force-sensing fingers! 🧤 Stanford researchers just released UMI-FT, a handheld data collection platform that puts compact six-axis force/torque sensors on each finger, enabling finger-level wrench measurements alongside RGB, depth, and pose data. Many manipulation tasks require careful force modulation: too little force and the task fails, too much and you cause damage. But commercial force/torque sensors are expensive, bulky, and fragile, which has limited large-scale force-aware policy learning. UMI-FT changes the economics. The platform uses an iPhone for RGB vision, ultrawide RGB, depth, and pose via ARKit, with each finger sensorized using a CoinFT sensor to capture per-finger wrench information during manipulation. This multimodal data trains an adaptive compliance policy that predicts position targets, grasp force, and stiffness for execution on standard compliance controllers. The learned policy runs slowest and generates reference targets, while model-based compliance and force controllers provide delicate 6D compliance control and real-time force modulation. They tested on three contact-rich, force-sensitive tasks: whiteboard wiping (locate eraser, grasp, wipe until clean), skewering zucchini (grasp slice firmly, push onto stick until punctured), and lightbulb insertion (grasp bulb, align bayonet pin with socket slit, insert while overcoming spring force, rotate to light up). The results are clear. Policies without compliance struggle to modulate contact force and trigger safety faults from excessive force. Policies without force sensing fail to grasp unseen objects or resist reaction forces, causing slippage. Here's the project page: ~~ ♻️ Join the weekly robotics newsletter, and never miss any news →

Force-sensing fingers! 🧤 Stanford researchers just released UMI-FT, a handheld data collection platform that puts compact six-axis force/torque sensors on each finger, enabling finger-level wrench measurements alongside RGB, depth, and pose data. Many manipulation tasks require careful force modulation: too little force and the task fails, too much and you cause damage. But commercial force/torque sensors are expensive, bulky, and fragile, which has limited large-scale force-aware policy learning. UMI-FT changes the economics. The platform uses an iPhone for RGB vision, ultrawide RGB, depth, and pose via ARKit, with each finger sensorized using a CoinFT sensor to capture per-finger wrench information during manipulation. This multimodal data trains an adaptive compliance policy that predicts position targets, grasp force, and stiffness for execution on standard compliance controllers. The learned policy runs slowest and generates reference targets, while model-based compliance and force controllers provide delicate 6D compliance control and real-time force modulation. They tested on three contact-rich, force-sensitive tasks: whiteboard wiping (locate eraser, grasp, wipe until clean), skewering zucchini (grasp slice firmly, push onto stick until punctured), and lightbulb insertion (grasp bulb, align bayonet pin with socket slit, insert while overcoming spring force, rotate to light up). The results are clear. Policies without compliance struggle to modulate contact force and trigger safety faults from excessive force. Policies without force sensing fail to grasp unseen objects or resist reaction forces, causing slippage. Here's the project page: ~~ ♻️ Join the weekly robotics newsletter, and never miss any news →

Lukas Ziegler

12,868 views • 8 months ago

1/ 🧠Humans are the best robot data source! 2/ 👓Human egocentric video is rich in quantity, but poor in quality. 3/ Beyond scaling data, smarter representation and architecture matter just as much. 4/ Want an open-source framework to train your own learn-from-human-data robot policy? 🚀We introduce HumanEgo: Zero-Shot Robot Learning from Minutes of Human Egocentric Videos⬇️ ✦ Zero-Shot Human-to-Robot Transfer ✦ Robot-Data-Free ✦ Just 30 min of data per task ✦ Collect by Anyone, Anytime, Anywhere ✦ Deploy on Any Robot, Any Camera, Any Environment ✦ Open-Source & Easy-to-Implement Let's squeeze every bit of signal out of human data! 🌐 Website: 📄 Paper: 💻 Code: 📹 Video: 🧵 1/n

1/ 🧠Humans are the best robot data source! 2/ 👓Human egocentric video is rich in quantity, but poor in quality. 3/ Beyond scaling data, smarter representation and architecture matter just as much. 4/ Want an open-source framework to train your own learn-from-human-data robot policy? 🚀We introduce HumanEgo: Zero-Shot Robot Learning from Minutes of Human Egocentric Videos⬇️ ✦ Zero-Shot Human-to-Robot Transfer ✦ Robot-Data-Free ✦ Just 30 min of data per task ✦ Collect by Anyone, Anytime, Anywhere ✦ Deploy on Any Robot, Any Camera, Any Environment ✦ Open-Source & Easy-to-Implement Let's squeeze every bit of signal out of human data! 🌐 Website: 📄 Paper: 💻 Code: 📹 Video: 🧵 1/n

Zhi (Leo) Wang

111,277 views • 4 months ago

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

35,121 views • 2 months ago

Reward AI just came out of stealth and unveiled OM-1, a robot policy learned entirely from humans. No teleoperation, no on-robot data. One Model, Any Body. 1. Capture. A wearable 7-DoF hand (Omnibody Hand) lets people demonstrate manipulation naturally while working, cooking, or sorting. Sensors record tactile, proximity, force and hand pose at full human speed. 2. Learn. OM-1 trains on this human data alone. No teleop, no on-robot experience. Then it runs across bodies from industrial arms to humanoids, and scaling is just adding more human data. 3. Control. A high-frequency control layer, trained with RL in sim, executes those actions on any robot, holding steady through unknown loads like a stuck fridge door. The result: OM-1 picks up a brand-new task from under 30 minutes of data.

The Humanoid Hub

35,064 views • 10 days ago

What if a dexterous robot could learn a 20+ step chemistry experiment without a single on-robot training demo? Meet TwinDEX: a pair of co-designed, three-finger, nine-DoF dexterous manipulation interface: one wearable for data collection, one for robot deployment. The twinned design shares identical kinematics, contact surfaces, visual appearance, and sensors across collection and deployment — keeping observations and actions aligned end to end. Trained from scratch on only a few hundred wearable demonstrations - with zero on-robot training or intervention data - TwinDEX completed a standardized chemistry experiment involving tool switches, fine force control, and bimanual coordination. Robot-free data showed comparable learning efficiency on the multi-task evaluation, TwinDEX delivered 5.3 times effective throughput than on-robot teleoperation. TwinDEX demonstrates that high-quality robot-free data can fully substitute for on-robot teleoperation data on challenging dexterous tasks — removing the dependency on real-robot hardware that has been the central bottleneck to scaling dexterous manipulation data. This was the proof-it phase. Now comes scale: what emerges at tens of thousands, or millions, of episodes? Watch the demo and read the technical blog: #TwinDEX #Robotics #EmbodiedAI #DexterousManipulation

What if a dexterous robot could learn a 20+ step chemistry experiment without a single on-robot training demo? Meet TwinDEX: a pair of co-designed, three-finger, nine-DoF dexterous manipulation interface: one wearable for data collection, one for robot deployment. The twinned design shares identical kinematics, contact surfaces, visual appearance, and sensors across collection and deployment — keeping observations and actions aligned end to end. Trained from scratch on only a few hundred wearable demonstrations - with zero on-robot training or intervention data - TwinDEX completed a standardized chemistry experiment involving tool switches, fine force control, and bimanual coordination. Robot-free data showed comparable learning efficiency on the multi-task evaluation, TwinDEX delivered 5.3 times effective throughput than on-robot teleoperation. TwinDEX demonstrates that high-quality robot-free data can fully substitute for on-robot teleoperation data on challenging dexterous tasks — removing the dependency on real-robot hardware that has been the central bottleneck to scaling dexterous manipulation data. This was the proof-it phase. Now comes scale: what emerges at tens of thousands, or millions, of episodes? Watch the demo and read the technical blog: #TwinDEX #Robotics #EmbodiedAI #DexterousManipulation

X Square Robot

1,345,872 views • 22 days ago

Feeling what a robot feels is becoming real! 🫰 Fluid Reality just ran its first full end-to-end touch teleoperation system, and the demo shows why high-resolution haptics might be the missing piece in robot control. A 22 mm fingertip display with 32 independently actuated “bubbles” was mapped directly to tactile sensors on a robot hand. 🫧 Three fingers streamed real-time touch data robot → human, giving the operator actual contact feedback instead of waving their hands blind, like most teleop systems today. It’s still early, but this is a real step toward dexterous ops and high-quality data for training Physical AI systems. Companies involved: Samsung Electronics, Alt-Bionics, Inc. (the robot hand), MANUS™ (hand tracking), and Sensible Robotics (fingertip sensors) Touch isn’t a nice-to-have in robotics. It’s the difference between manipulation that works in the real world P.S. Guess how many touch receptors we humans, have in our hand! :) ~~ ♻ Join the weekly robotics newsletter, and never miss any news →

Feeling what a robot feels is becoming real! 🫰 Fluid Reality just ran its first full end-to-end touch teleoperation system, and the demo shows why high-resolution haptics might be the missing piece in robot control. A 22 mm fingertip display with 32 independently actuated “bubbles” was mapped directly to tactile sensors on a robot hand. 🫧 Three fingers streamed real-time touch data robot → human, giving the operator actual contact feedback instead of waving their hands blind, like most teleop systems today. It’s still early, but this is a real step toward dexterous ops and high-quality data for training Physical AI systems. Companies involved: Samsung Electronics, Alt-Bionics, Inc. (the robot hand), MANUS™ (hand tracking), and Sensible Robotics (fingertip sensors) Touch isn’t a nice-to-have in robotics. It’s the difference between manipulation that works in the real world P.S. Guess how many touch receptors we humans, have in our hand! :) ~~ ♻ Join the weekly robotics newsletter, and never miss any news →

Lukas Ziegler

46,927 views • 9 months ago

In my experience, robot 'generalists' are often jacks of all trades but masters of none. In training across multiple tasks and environments, robot policies fail to generalize robustly and effectively to each particular test setting. What if at test time, we non-parametrically *retrieved* “relevant” data from the training set and used it to significantly improve the performance of few-shot imitation learning to be robust to various test time scenes. Notably, we are *not* collecting lots of new data, just training more on sub-components of the same training data! Now, we’re certainly not the first to suggest retrieval, but in our new work - STRAP, we show how retrieving relevant *sub-trajectories* from offline datasets can significantly increase data reuse across tasks, when paired with an appropriate metric space. A 🧵 (1/7)

In my experience, robot 'generalists' are often jacks of all trades but masters of none. In training across multiple tasks and environments, robot policies fail to generalize robustly and effectively to each particular test setting. What if at test time, we non-parametrically retrieved “relevant” data from the training set and used it to significantly improve the performance of few-shot imitation learning to be robust to various test time scenes. Notably, we are not collecting lots of new data, just training more on sub-components of the same training data! Now, we’re certainly not the first to suggest retrieval, but in our new work - STRAP, we show how retrieving relevant sub-trajectories from offline datasets can significantly increase data reuse across tasks, when paired with an appropriate metric space. A 🧵 (1/7)

Abhishek Gupta

12,045 views • 1 year ago

A lot of robot learning pipelines treat embodiment mismatch as a software problem. Collect data with one device, retarget the motion to another robot, and try to deal with the differences later. TwinDEX flips that idea around. This system is from a Chinese robotics company called X Square. Instead of starting with a generic data-collection device, the team co-designs a wearable three-finger end-effector and a robotic hand as two sides of the same learning system. They align the things that actually matter during manipulation: joint structure, link proportions, contact geometry, surface properties, visual appearance, sensing, and timing. That matters because dexterous manipulation is much more than matching hand poses. A few millimeters of contact error or a different force relationship can completely change how a task is executed. In the collection evaluation, TwinDEX delivered up to 5.3 times the effective throughput of on-robot teleoperation. It’s a useful reminder that scaling robot learning may not only be about bigger models and more data. Sometimes the hardware collecting the data is part of the data problem itself.

A lot of robot learning pipelines treat embodiment mismatch as a software problem. Collect data with one device, retarget the motion to another robot, and try to deal with the differences later. TwinDEX flips that idea around. This system is from a Chinese robotics company called X Square. Instead of starting with a generic data-collection device, the team co-designs a wearable three-finger end-effector and a robotic hand as two sides of the same learning system. They align the things that actually matter during manipulation: joint structure, link proportions, contact geometry, surface properties, visual appearance, sensing, and timing. That matters because dexterous manipulation is much more than matching hand poses. A few millimeters of contact error or a different force relationship can completely change how a task is executed. In the collection evaluation, TwinDEX delivered up to 5.3 times the effective throughput of on-robot teleoperation. It’s a useful reminder that scaling robot learning may not only be about bigger models and more data. Sometimes the hardware collecting the data is part of the data problem itself.

Alamin

48,112 views • 22 days ago

I was really impressed by the UMI gripper (Cheng Chi et al.), but a key limitation is that **force-related data wasn’t captured**: humans feel haptic feedback through the mechanical springs, but the robot couldn’t leverage that info, limiting the data’s value for fine-grained manipulation tasks. Led by my amazing students Yolanda Zhu and Binghao Huang, we designed a **portable visuo-tactile gripper** by integrating our dense, flexible tactile arrays with the UMI gripper to enable large-scale in-the-wild data collection. 🔗 We demonstrate **cross-modal representation learning** and **downstream policy learning** on tasks requiring in-hand state estimation (e.g., test tube reorientation) and fine-grained force sensing (e.g., pipette fluid transfer). Key takeaways: - Our flexible tactile arrays store the rich haptic information humans perceive as dense tactile signals. - Portability and robustness are key for in-the-wild data collection; our portable gripper is compact, lightweight, and durable. - Touch provides precise, robust measurements of in-hand object pose, invariant to lighting and viewpoint. - Cross-modal pretraining on large-scale in-the-wild data significantly improves policy robustness and sample efficiency (as shown many times before — and verified again here!). Also check out our previous investigations of dense, flexible tactile grids for understanding human-robot-environment interactions: - Dense tactile glove (Nature ’19): - 3D-ViTac (CoRL ’24):

I was really impressed by the UMI gripper (Cheng Chi et al.), but a key limitation is that force-related data wasn’t captured: humans feel haptic feedback through the mechanical springs, but the robot couldn’t leverage that info, limiting the data’s value for fine-grained manipulation tasks. Led by my amazing students Yolanda Zhu and Binghao Huang, we designed a portable visuo-tactile gripper by integrating our dense, flexible tactile arrays with the UMI gripper to enable large-scale in-the-wild data collection. 🔗 We demonstrate cross-modal representation learning and downstream policy learning on tasks requiring in-hand state estimation (e.g., test tube reorientation) and fine-grained force sensing (e.g., pipette fluid transfer). Key takeaways: - Our flexible tactile arrays store the rich haptic information humans perceive as dense tactile signals. - Portability and robustness are key for in-the-wild data collection; our portable gripper is compact, lightweight, and durable. - Touch provides precise, robust measurements of in-hand object pose, invariant to lighting and viewpoint. - Cross-modal pretraining on large-scale in-the-wild data significantly improves policy robustness and sample efficiency (as shown many times before — and verified again here!). Also check out our previous investigations of dense, flexible tactile grids for understanding human-robot-environment interactions: - Dense tactile glove (Nature ’19): - 3D-ViTac (CoRL ’24):

Yunzhu Li

13,510 views • 1 year ago

This is how ALOHA's "teleoperation" system works - a fancy word for "remote control". Training robots will be more and more like playing games in the physical world. A human operates a "joystick++" to perform tasks and collect data, or intervene if there's any safety concern. There's actually a learning curve to master the controller, much like practicing gaming skills. Teleoperation can be done in many different ways. ALOHA is an impressive custom-built system with very low cost. Here're a few alternatives: (1) Motion Capture (MoCap): apply the MoCap systems used for Hollywood movies to capture the fine-grained motions of hand joints. There would be no "embodiment gap" if the robot hand has 5 fingers. For instance, a demonstrator can wear a CyberGlove ( and manipulate the objects. CyberGlove will capture the motion signals & haptic feedback in real-time, which can be re-targeted onto the humanoid. (2) Wearing gloves & markers can be clumsy. An alternative way to do MoCap is through computer vision. DexPilot from NVIDIA enables marker-less and glove-free data collection. The human operator simply uses their bare hands to perform the tasks. 4 Intel RealSense depth cameras and 2 NVIDIA Titan XP GPUs (yeah, 2019 work) translate the pixels to precise motion signals for robot learning. (3) VR Headset: turn the training room into a VR game and "role play" the robot. This has the advantage of scalable remote data collection - annotators from around the world can contribute without coming onsite. VR demonstration technique appeared in research projects like the iGibson home robot simulator, an initiative that I participated in at Stanford: Behind-the-scene video by Litian Liang

This is how ALOHA's "teleoperation" system works - a fancy word for "remote control". Training robots will be more and more like playing games in the physical world. A human operates a "joystick++" to perform tasks and collect data, or intervene if there's any safety concern. There's actually a learning curve to master the controller, much like practicing gaming skills. Teleoperation can be done in many different ways. ALOHA is an impressive custom-built system with very low cost. Here're a few alternatives: (1) Motion Capture (MoCap): apply the MoCap systems used for Hollywood movies to capture the fine-grained motions of hand joints. There would be no "embodiment gap" if the robot hand has 5 fingers. For instance, a demonstrator can wear a CyberGlove ( and manipulate the objects. CyberGlove will capture the motion signals & haptic feedback in real-time, which can be re-targeted onto the humanoid. (2) Wearing gloves & markers can be clumsy. An alternative way to do MoCap is through computer vision. DexPilot from NVIDIA enables marker-less and glove-free data collection. The human operator simply uses their bare hands to perform the tasks. 4 Intel RealSense depth cameras and 2 NVIDIA Titan XP GPUs (yeah, 2019 work) translate the pixels to precise motion signals for robot learning. (3) VR Headset: turn the training room into a VR game and "role play" the robot. This has the advantage of scalable remote data collection - annotators from around the world can contribute without coming onsite. VR demonstration technique appeared in research projects like the iGibson home robot simulator, an initiative that I participated in at Stanford: Behind-the-scene video by Litian Liang

Jim Fan

124,783 views • 2 years ago

Open-source dexterous hands with fingertip sensors! 🪬 ORCA Dexterity just released three dexterous hand models. Their mission: is to democratize robotic hand dexterity. They're sharing progress on exceptional, low-cost hardware and a software layer from low-level control to robotic hand learning. This is the open-source approach to dexterous manipulation. It has 83 taxels per finger with 0.1 N force detection which pretty is impressive for an open-source design. Tactile sensing is critical for dexterous manipulation, knowing contact forces enables gentle grasping, slip detection, and force-controlled assembly. Also 700g weight for the lite version makes it practical for mounting on robot arms without exceeding payload limits. Lower weight means faster movements and lower torque requirements. Open hardware accelerates robotics by letting researchers and builders modify designs for their specific needs without starting from scratch! ~~ ♻️ Join the weekly robotics newsletter, and never miss any news →

Open-source dexterous hands with fingertip sensors! 🪬 ORCA Dexterity just released three dexterous hand models. Their mission: is to democratize robotic hand dexterity. They're sharing progress on exceptional, low-cost hardware and a software layer from low-level control to robotic hand learning. This is the open-source approach to dexterous manipulation. It has 83 taxels per finger with 0.1 N force detection which pretty is impressive for an open-source design. Tactile sensing is critical for dexterous manipulation, knowing contact forces enables gentle grasping, slip detection, and force-controlled assembly. Also 700g weight for the lite version makes it practical for mounting on robot arms without exceeding payload limits. Lower weight means faster movements and lower torque requirements. Open hardware accelerates robotics by letting researchers and builders modify designs for their specific needs without starting from scratch! ~~ ♻️ Join the weekly robotics newsletter, and never miss any news →

Lukas Ziegler

38,249 views • 6 months ago