Binglin Andre Wang (@AndreBinglin) profile picture

BA/MS @ NYU | Tactile sensing & dexterous manipulation

Joined October 2022
Binglin Andre Wang retweeted
Robots are getting a sense of touch for as little as $9 per fingertip! 🤯 Researchers from NYU, CUNY, and Columbia University have developed PanoTac, an open-source tactile sensor that helps robots understand what happens when they touch and manipulate objects. And the results are pretty impressive. In their experiments, robots using both vision and tactile sensing achieved a 92% success rate, compared to just 45% when relying on vision alone. The interesting part is what the sensor actually measures. PanoTac combines two different sensing technologies inside a soft robotic fingertip. One detects where an object makes contact, while the other measures how the fingertip deforms under pressure and sideways forces. Together, they produce a detailed 12 × 32 × 3 map of physical contact, updated up to 100 times per second. Why does this matter? Imagine a robot picking up a bottle without knowing what's inside. A camera can recognize the bottle, but it cannot reliably tell whether it's empty or full. With PanoTac, the robot can use the forces acting on its fingertip to estimate the weight and decide what to do next. The researchers also demonstrated robots removing stoppers from tubes, preparing samples, and handling granular materials. These are exactly the kinds of tasks where vision alone reaches its limits. And with a reported hardware cost of roughly $9 to $50 per fingertip, the technology could make sophisticated tactile sensing accessible to far more robotics researchers and developers. I find the combination of low-cost hardware and improved manipulation performance particularly interesting. The original X thread by @binghao_huang shows the hardware, tactile signal visualizations, and real robot demos. 📌 Project: panotac.github.io/ Paper: arxiv.org/abs/2611.03520 👉 Follow along, if you are interested in robotics & AI! —— Weekly robotics and AI insights. Subscribe free: 22astronauts.com
8
20
2
121
5,736
Binglin Andre Wang retweeted
Excited to announce Robocap-10k. 10,000 hours of ego with metric actions. This is the high-fidelity dataset I wished I had in grad school but at industrial scale. It’s time we make robotics model scaling a science.
Introducing RoboCap-10k. 10,000 hours of open egocentric data with metric hand / head pose on every frame. Captured on RoboCap across 11 countries and 21 industries. GroundedAPI produces 3D enrichment giving robot policies high-fidelity action grounding. with @BitRobotNetwork @huggingface @rerundotio ↓
8
6
40
3,216
Binglin Andre Wang retweeted
Loved being part of this project!
As we get closer to a world where we live among robots, it’s becoming more and more valuable for robots to learn quickly with and from people. I’m very excited to share Steertuning, a novel approach that adapts robots to new tasks from 5 human demonstrations in under 5 minutes, from demonstration to deployment. Website: steertuning.github.io/ Code: github.com/billy-yibo-yan/st…
2
8
537
Binglin Andre Wang retweeted
As we get closer to a world where we live among robots, it’s becoming more and more valuable for robots to learn quickly with and from people. I’m very excited to share Steertuning, a novel approach that adapts robots to new tasks from 5 human demonstrations in under 5 minutes, from demonstration to deployment. Website: steertuning.github.io/ Code: github.com/billy-yibo-yan/st…
9
29
10
110
11,594
Binglin Andre Wang retweeted
Wanted to get into manipulation. Decided to first build a full stack that scales. Plan: copy @GeneralistAI's gripper. Fun weekend project. It was not. So @t_k_233 and I documented every design choice and lesson along the way. Meet Specialist: fully open-source hardware, data pipeline, training, and deployment code. specialist-ai.github.io/
46
112
37
904
102,711
Binglin Andre Wang retweeted
Robots learn by imitating human teleop demos. But some tasks are too hard to demonstrate Raw VR teleop success for hard tasks: 0% 🍅, 10% 🖍️, 0% 🍷 Our CoRL work, GLIDE, can bump this to 90%! How? By smartly editing teleop code. A refreshingly robotics-first way to use LLMs!
2
20
6
97
6,865
Binglin Andre Wang retweeted
In behavior cloning, we mastered how to learn robotics from "robot-complete" data – but what happens to data that is not so perfect? Intelligence is often about learning from incomplete references; to come up with a "story" given disparate datapoints. Our Pick Anything demo shows an ability of this sort: the policy learned human-like dexterity from absolutely no human-robot interactions, no teleop or handheld tool data, just letting humans be humans and robots be robots. How did we do this? Learning from humans without actions is often reduced to a semantics problem, a fancy way to prompt the robot. What, if anything, can robots learn about the dynamics from them? RL for manipulation has the opposite problem – often we are stuck with "technically correct" solution that are nevertheless aesthetically unsatisfying. In Pick Anything, we found a way to unite the two in a way for dexterous manipulation that I personally find quite beautiful (and I can't wait to share the paper with you!) I believe what we have here is the key to human level dexterity for robots, and all of that without needing any real robot data whatsoever. For this Monday/Tuesday, you can visit our demo booth at CoRL (Griffin hall, demo 12) and ask me more questions! It's a labor of love from my past year of work, and I'm very excited to be finally able to talk about it.
8
14
3
81
5,292
Binglin Andre Wang retweeted
I will be at CoRL, presenting Do as I Do, and a yet unreleased demo ✋🤖 My opinion on the future of robotics has shifted drastically over the past year. I believe we're headed towards an inflection point: soon, majority of our robot data will be created without robots (or even humans) in the loop. Why? Because scaling robotics at the speed of robots or humans is almost always going to be slower than scaling at the speed of compute. We want robots that work in the real world. How do we stop making the real world our bottleneck? Looking forward to sharing my thoughts, and would love to hear yours if you are going to be in Austin next week.
Robots are the bottleneck in scaling robotics, and learning from human video promises to solve it. But how can chaotic human data ever measure up to sanitized, lab-made teleoperation data? Introducing Do as I Do: establishing a much needed correspondence between human videos and dexterous robot data. Some fun insights below: 🧵
6
6
1
93
5,979
1/5 FlexiTac has been accepted to the CoRL 2026 Demo Track, and today we’re releasing FlexiTac-Gen! 🧵 Try it: flexitac.github.io/flexitac-… Over the past two years, we’ve received many requests to customize FlexiTac for different robots. With help from AI agents, we built a tool that makes it easy to design your own sensor and export the files for fabrication. A simple piece of tactile hardware to help move your research forward. See you at CoRL!
8
34
5
208
13,487
Binglin Andre Wang retweeted
Big jump for me from physics to robotics! The project is super fun, one interesting thing is seeing the lore of "wrist cam ≈ tactile" in person. We noticed sometimes the wrist camera can learn tactile cues from the deformable fingertip. (1/2)
❓I keep coming back to this: what tactile representation does manipulation need? We introduce PanoTac: A dense three-axis Tactile Sensor. In the demo, the robot reorients and inserts a tube, then wiggles the stopper free while the tube sits loosely in the rack. The right side show the tactile signals that helps robot detect when the stopper releases and when to switch from wiggling to lifting. (1/7)
2
2
5
454
Binglin Andre Wang retweeted
Introducing ID-WM: a hierarchical world model that decouples "where to go" from "how to get there." Finally, we can plan over hundreds to thousands of steps with WMs. It beats the strongest baseline by 38.5%, single-level WMs by 135%. #NeurIPS2026 🧵 id-wm.github.io
16
46
3
237
26,363
Binglin Andre Wang retweeted
Robots need to feel where they touch and how forces change. Introducing PanoTac: an open-source, dense 3D tactile sensor for <$10 USD. Watch as the robot racks a test tube and pulls the stopper: PanoTac tells the tube orientation, and feels when the stopper comes free. 🧵
6
15
1
112
5,916
Binglin Andre Wang retweeted
Really excited to share Catch It All!, my first paper as last author! In this work led by @Kehlani_Fay @arth_shukla we develop a system for catching diverse objects using a single egocentric camera 🧵👇
4
24
2
102
5,553
Thanks Kevin!! You’ve seen it evolve from the early ideas to bring the two sensing mechanisms together!
It was so cool seeing the progress throughout the last year!
132
After about a year of building and experimenting, I’m excited to share PanoTac, my first research project! 🤖 We built a compliant tactile fingertip that costs under $50 and takes ~30 min to assemble. By combining piezoresistive and magnetic sensing, it captures contact geometry and provides normal and shear feedback in a dense 12×32×3 tactile field. I’m especially excited about how these sensing modalities complement each other: even simple fusion broadens what either can do alone. We use touch every day to guide our actions, often without thinking about it. Seeing the robot use tactile feedback to decide how to move and when to change actions has been incredibly rewarding. This project has also shaped how I think about hardware–learning co-design: what should we build into the hardware to give a policy the information it needs? I see tactile sensor fusion as a promising direction for exploring this connection.💡 Grateful to my coauthors, advisors, and everyone who shared ideas and feedback along the way! Check out the demos below 👇 panotac.github.io/
❓I keep coming back to this: what tactile representation does manipulation need? We introduce PanoTac: A dense three-axis Tactile Sensor. In the demo, the robot reorients and inserts a tube, then wiggles the stopper free while the tube sits loosely in the rack. The right side show the tactile signals that helps robot detect when the stopper releases and when to switch from wiggling to lifting. (1/7)
2
8
2
24
1,157
Hardware looks scary until you break it down. We’ve been developing Learn FlexiTac: Building Your Robot’s First Tactile Skin. This interactive course walks you through the full stack, one hands-on lesson at a time: the sensor and its readout board, then robot integration, then simulation. Try it here 👇 flexitac.github.io/learn/ I’m still working on it and would love your feedback! Shoutout to @wheresryan22 for Anatomy!
Introducing Anatomy Describe a machine in one line. Claude brings it to life. - Interactive isometric wireframes you can rotate, inspect and explore - WebGL shaders for fire, water, plasma and more - Real 3D motion, with individual parts that rotate and move Install it as an agent skill. Fully open source.
4
17
103
3,976
An elegant way to add shear sensitivity to FlexiTac! TacGB uses passive bumps to encode shear into pressure patterns that robot policies can learn from directly, without changing the electronics. Great to see @jeffwli and team build on our open-source hardware!
We gave tactile sensors goosebumps and helped robots feel friction. 🤖🦾 👉 Using ordinary cabinet bumpers. Meet TacGooseBumps (TacGB), a passive add-on that sticks directly onto existing normal-only tactile sensors, with no changes to their electronics. How does it work? Under tangential loads, the bumps tilt and redistribute pressure, creating patterns that a robot policy can learn from directly, with no explicit force reconstruction. Across four manipulation tasks, TacGB improves insertion success by up to 36 percentage points, enables gentler egg placement, and produces more continuous, straighter drawing. 🌐 Project: jeffwli.github.io/tacgb/ 📄 Paper: arxiv.org/abs/2609.34006
4
4
50
2,647
Binglin Andre Wang retweeted
🎉 Excited to share HuRo, accepted to CoRL 2026! Can robotized human videos provide scalable robot learning supervision? HuRo turns human videos into robot aligned observations and actions. In VLA pretraining, OOD completion rises from 34.9% to 72.2%. 🧵 More in replies.
4
26
3
223
18,691
Binglin Andre Wang retweeted
Strong multimodal models = strong robot models. Check out some early glimpses of Meta's Muse Spark 1.2 on robots here!
1/ muse spark 1.2 is a very strong multimodal model—it can do visual coding, robotics planning, and audio-visual understanding that all come together through agentic tools.
5
12
87
19,161