Won Kyung Do

Robotics | Stanford ME PhD | Tactile sensing

[Won Kyung Do @wkdo74374]

@MattStr10050461 Thank you Matt!!!

[Won Kyung Do @wkdo74374]

@SwannAiden Thank you Aiden!!

[Won Kyung Do @wkdo74374]

That’s what we’re building with OM-1 and Omnibody Hand: much of that tacit know-how and useful dexterity are captured in one compact model and robotic hand. Excited to finally share what we’ve been building!

[Reward AI @RewardAI_]

Introducing OM-1, our first robot foundation model, zero-shot generalizing to any robot: table-top arms, industrial arms and humanoids.
- learned directly from human manipulation data
- no teleop/robot data
- close to human-level dexterity and efficiency
- multi-robot collab

[Chen Wang @chenwang_j]

The speed is what you notice first. What excites me most is how we were able to transfer skills learned from human motion across different robots, without robot data or robot-specific fine-tuning. Really proud of what our team has built!

[Yifeng Zhu @yifengzhu_ut]

During my PhD and time at big tech labs, I approached robot manipulation one robot and one task at a time. We could teach robots new behaviors, but they were slow, limited to one or a few robots, and rarely made it beyond the lab.

At Reward AI, we see cross-embodiment and efficiency as first-class citizens to deploy millions of robots in the real world. And they won't all share a single form factor. With our OM model series, we are bringing human-level efficiency and generalization to any robot.

[Won Kyung Do @wkdo74374]

If a wearable forces you to change your grip, it changes the demonstration. Omnibody Hand is designed to fit different hands and let people grasp and reorient objects naturally, preserving the behavior we want robots to learn.

[Reward AI @RewardAI_]

OM-1 exhibits several emergent behaviors. Each arm learns to compensate for the other’s mistakes. The model knows when to retry, when to adapt to adversarial perturbations, and when to stop given the environment changes too drastically.

[Won Kyung Do @wkdo74374]

Spent the past six years creating DenseTact sensors, thinking about what happens at our fingertips. Humans make dexterous manipulation look almost unfairly easy: we can pick up and reorient objects without thinking, yet explaining exactly how we do it is surprisingly hard.

[Reward AI @RewardAI_]

Introducing OM-1, our first robot foundation model, zero-shot generalizing to any robot: table-top arms, industrial arms and humanoids.
- learned directly from human manipulation data
- no teleop/robot data
- close to human-level dexterity and efficiency
- multi-robot collab

[Won Kyung Do @wkdo74374]

@RewardAI_ Humans make the dexterous task look unfairly easy. And that's what OM-1 and Omnibody Hand aims to capture.

[Matt Strong @MattStr10050461]

Happy to share a recent work I've been a part of -- TensorTouch! 🖐️

TensorTouch enables the calibration of optical tactile sensors for dense stress tensor AND deformation for dexterous manipulation.

website: https://t.co/GKGkazgPzP
code: https://t.co/pr25wmyYCE

1/4

[MrNeRF @janusch_patas]

Touch-GS: Visual-Tactile Supervised 3D Gaussian Splatting
Project: https://t.co/4IyCXiQa8G
Paper: https://t.co/g2TincE97I