正在加载视频...

视频加载失败

Can we bring human-like Touch to robots🤖? Introducing our CoRL work on 3D-ViTac. Humans rely on both vision 👁️ and touch 🫳 for complex tasks. With combined visual-tactile sensing, robots can now tackle challenging tasks, like precise in-hand reorientation, fragile objects grasping. Website: #Robotics #CoRL2024 #Touch #tactile #AI #ML

49,500 次观看 • 1 年前 •via X (Twitter)

10 条评论

Binghao Huang 的头像
Binghao Huang1 年前

An open-source tactile sensing system from @Columbia designed to democratize touch to robot data collection. Please check out our website for tutorials ( The sensor is equipped with dense sensing units, each covering an area of 3 millimeter square. These sensors are low-cost and flexible, providing detailed and extensive coverage of physical contacts, effectively complementing visual information. (2/9)

Binghao Huang 的头像
Binghao Huang1 年前

We install flexible tactile sensors on soft gripper to get touch information and use multi-view RGBD cameras to get visual information. To integrate tactile and visual data, we fuse them into a unified 3D representation space that preserves their 3D structures and spatial relationships.(3/9)

Binghao Huang 的头像
Binghao Huang1 年前

With 3D-ViTac, robots can now handle delicate objects, like eggs, with ease! 🥚 In this Egg Steaming task, the robot first uses its right hand to open the egg tray, then carefully grasps and places an egg into a egg cooker (narrow space). (4/9)

Binghao Huang 的头像
Binghao Huang1 年前

The robot keeps retrying until it successfully grasps the grapes even under huge visual occlusions. (5/9)

Binghao Huang 的头像
Binghao Huang1 年前

We observe three key benefits of integrating touch. (1) Tactile sensors provide critical feedback on the presence of contact and the appropriate amount of force to apply. (2) Our policy leverages detailed contact patterns provided by touch to address visual occlusions effectively. (6/9)

Binghao Huang 的头像
Binghao Huang1 年前

This project is led by @binghao_huang and would have been impossible without the hard work from co-authors: @YXWangBot , Xinyi Yang, @LuoYiyue , @YunzhuLiYZ. (7/9)

Tony Kam 的头像
Tony Kam1 年前

Awesome!

Jason 的头像
Jason1 年前

Nice work. I’ve seen other tactile sensing work emphasize improvements picking and placing deformable objects. But this is the first one I’ve seen that emphasizes visual occlusion performance. The grape demo is really cool! Are there plans for your lab to continue iterating on the hardware for the tactile sensing system?

Binghao Huang 的头像
Binghao Huang1 年前

Thanks, Jason! Yes, we’re continuing to iterate on our hardware. Over the next few weeks, we'll release an easy-to-follow guide for reproducing our flexible sensor. In future iterations, we're planning to develop flexible PCBs to produce sensors directly from the factory, allowing for easy scaling.

Stormcatch ⚡️ ($SCATCH) 的头像
Stormcatch ⚡️ ($SCATCH)1 年前

Yes the human-like touch will come!

相关视频

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

13,188 次观看 • 11 个月前