Yi Yang (@_yiyang04) header
Yi Yang (@_yiyang04) profile picture

Robotics & embodied AI. Robotics researcher, Undergrad @sjtu1896 · Visiting student @CMU_Robotics

Joined October 2025
Photo by Yi Yang (@_yiyang04)Photo by Yi Yang (@_yiyang04)Photo by Yi Yang (@_yiyang04)Photo by Yi Yang (@_yiyang04)Photo by Yi Yang (@_yiyang04)Photo by Yi Yang (@_yiyang04)Photo by Yi Yang (@_yiyang04)Photo by Yi Yang (@_yiyang04)Photo by Yi Yang (@_yiyang04)Photo by Yi Yang (@_yiyang04)
A robot whips a rope to knock an apple off someone's head, using a policy trained only in simulation. Introducing DeformX (Oral @ IROS 2026): Cosserat rod physics co-simulated with NVIDIA Isaac Sim, so cables bend, twist and collide like the real thing. deformx.github.io
5
8
9
99
32,645
Anyway, here's Wonderwall. DeformX GPU is out: an open-source CUDA solver for cables and ropes, coupled with Isaac Lab. Every string on this guitar is simulated, bending and vibrating under the robot's pick. 🔊 Sound on.
4
4
1
43
2,924
Why GPU: rods are cheap in bulk. One RTX 4090 advances 1,024 ropes at 406 frames/s, 6.8× real time. The solver was introduced with Point It, Strike It, where it ran more than 20,000× faster than CPU DeformX: deformx.github.io/DeformY
1
83
Really like this direction. Contact and force could be a more transferable language for dexterity across embodiments.
Can you believe these are open-loop rollouts? Excited to introduce CARD 🃏 — Contact-Anchored Retargeting for Dexterous Manipulation. Almost every method today is limited to kinematics alone as the 𝐢𝐧𝐭𝐞𝐫𝐟𝐚𝐜𝐞 between the human and the robot. With CARD, we track real world contact and forces alongside kinematics to perform unprecedented dexterous tasks through tactile reactive replay! CARD is scalable: 📹 Multiple input sources 🤖 Multiple robot hands ⚡ needs no object models, no simulator, 10 Hz
4
139
A robot swings a rope and hits a falling balloon mid-air. Motion capture predicts where it will be; the policy strikes there. Point It, Strike It: send a rope tip to any 3D point from a chosen direction. Across 25 targets on three real ropes, 87% of strikes land within 5 cm.
2
4
2
14
1,770
Step 3, a new rope. RECAP fits 11 rope and rig parameters from 8 calibration swings, so the simulator matches the real rope. For each new goal, a correction policy trained in sim proposes candidates, the fitted sim picks the best, and the robot strikes once, open-loop.
1
1
56
Paper: arxiv.org/abs/2610.09573 Site: deformx.github.io/DeformY Built on DeformX (IROS '26 Oral): teapawt.org/_yiyang04/status/21043… With @ShawnFei331, @LehongW, @ZilinDai, Ruogu Li, Jiting Cai, Liyao Chang, Xinyi Yang, Henry Kou, Ruijie Fu, Lu Li & @howiechoset @cmubiorobotics @CMU_Robotics
A robot whips a rope to knock an apple off someone's head, using a policy trained only in simulation. Introducing DeformX (Oral @ IROS 2026): Cosserat rod physics co-simulated with NVIDIA Isaac Sim, so cables bend, twist and collide like the real thing. deformx.github.io
1
50
Application 2: dynamic rope swinging, where small modeling errors blow up at the tip. For a fair test, we first calibrate DeformX and Isaac Sim's linked-capsule rope on the same real rope, driven by a UR5e and tracked with mocap. Both end up with small, comparable errors.
1
215
Left: our policy's swing in DeformX. Right: the same motion replayed open-loop on a real UR5e. The real rope follows almost the same path to the goal (★). Real-robot min tip-to-goal distance (3 goals × 10 runs): DeformX: 6.6 / 7.3 / 5.8 cm Capsule baseline: 15.1 / 25.9 / 30.4 cm
1
177