Raunaq Bhirangi

Building something new in robotics.
I make robots that feel touch!

[Raunaq Bhirangi @Raunaqmb]

Tactile sensing is gaining traction, but slowly. Why? Because integration remains difficult. But what if adding touch sensors to your robot was as easy as hitting “print”?

Introducing eFlesh: a 3D-printable, customizable tactile sensor.

Shape it. Size it. Print it. 🧶👇

[Max Mullen @Max]

Some personal news! Recently, I stepped back from the day-to-day at @Instacart to focus on investing full-time. I sat down with my friend @jaltma on Uncapped to talk about building Instacart and my philosophy on early-stage investing — check it out here!

[Raunaq Bhirangi @Raunaqmb]

Huge thanks to @anyazorin, @zhazhali01, @WowRobo and @LerrelPinto for putting this release together.
Buy one here: https://t.co/Q11f9PQORq
@zhazhali01's controller: https://t.co/k5qmjgrcQH
Of course, this would not be possible without the RUKA project: https://t.co/IwWo17aeVC

[Raunaq Bhirangi @Raunaqmb]

When @anyazorin and @irmakkguzey open-sourced the RUKA Hand (a low-cost robotic hand) earlier this year, people kept asking us how to get one.
Open hardware isn’t as easy to share as code.
So we’re releasing an off-the-shelf RUKA, in collaboration with @WowRobo and @zhazhali01.

[Raunaq Bhirangi @Raunaqmb]

This release also features a new controller developed by Ziqian Li (@zhazhali01), making the hand much more springy and lifelike, a big step towards more dexterous hands.
Open source (link below), affordable, and improving fast.

[Lerrel Pinto @LerrelPinto]

Thanks @IlirAliu_ for a detailed thread on our lab's work! I have been so fortunate to work with spectacular students and collaborators over the past decade.

Now prepping for our next big release 😉

[Raunaq Bhirangi @Raunaqmb]

Generalist robots don’t make sense if they can’t perceive and respond to their environment. Environment-aware stair climbing is a big step forward. Incredible work from my PhD friends at Skild!

[Skild AI @SkildAI]

We’ve all seen humanoid robots doing backflips and dance routines for years.

But if you ask them to climb a few stairs in the real world, they stumble!

We took our robot on a walk around town to environments that it hadn’t seen before. Here’s how it works🧵⬇️

[Lerrel Pinto @LerrelPinto]

I had a wonderful time talking to @robotknower and Niko Ciminelli for @SemiAnalysis_'s article on Robotics Levels of Autonomy.

The frontier of robotics (level 4) is indeed force-dependent dexterity, which our team is actively pushing on. Stay tuned😉

https://t.co/9QRq5mBsoB

[Lerrel Pinto @LerrelPinto]

It is difficult to get robots to be both precise and general.

We just released a new technique for precise manipulation that achieves millimeter-level precision while being robust to large visual variations. The key is a careful combination of visuo-tactile learning and RL. 🧵👇

[Raunaq Bhirangi @Raunaqmb]

@the_carlosdp Yes, I think the key takeaway from ViTaL is that it significantly improves the generalizability of real-world RL. Policies trained in the lab can now be directly deployed in new settings like homes, factories and charging stations!

[Raunaq Bhirangi @Raunaqmb]

Generalization needs data. But data collection is hard for precise tasks like plugging USBs, swiping cards, inserting plugs, and keying locks. Introducing robust, precise VisuoTactile Local (ViTaL) policies: >90% success rates from just 30 demos and 45 min of real-world RL.🧶⬇️

[Zifan Zhao @Zifan_Zhao_2718]

🚀 With minimal data and a straightforward training setup, our VisualTactile Local Policy (ViTaL) fuses egocentric vision + tactile feedback to achieve millimeter-level precision & zero-shot generalization! 🤖✨

Details ▶️ https://t.co/OOEjKqisEZ

[Siddhant Haldar @haldar_siddhant]

Current robot policies often face a tradeoff: they're either precise (but brittle) or generalizable (but imprecise).

We present ViTaL, a framework that lets robots generalize precise, contact-rich manipulation skills across unseen environments with millimeter-level precision. 🧵

[Raunaq Bhirangi @Raunaqmb]

This work wouldn't be possible without @Zifan_Zhao_2718's relentless pursuit of robust policies that actually generalize. It was a pleasure collaborating with @haldar_siddhant, @cui_jinda_hri and @LerrelPinto.
Videos/details: https://t.co/SqkHjTEGXc
Paper: https://t.co/vAWMHo9xrW

[Raunaq Bhirangi @Raunaqmb]

At inference, ViTaL stays modular: we use MolMo to locate the object and move close. Then, ViTaL takes over for precise, contact-rich control.

Since ViTaL only uses egocentric vision and AnySkin tactile sensing, we can deploy learned policies on new robots in new environments.

[Raunaq Bhirangi @Raunaqmb]

This might be the fastest we've gone from releasing a sensor to someone making their OWN copy of it! Hope eFlesh continues lowering the barrier for roboticists to use tactile sensors.

[Stone Tao @Stone_Tao]

this is really cool

and readily popular / reproducible enough to the point i’m finding people making the sensor in the depths of chinese wechat groups

the attention to reproducibility from Lerrel’s lab is incredible and something I often strive to achieve

[Lerrel Pinto @LerrelPinto]

Thanks @Stone_Tao ! Glad to see this sensor being made by the community!