CMU's DeformX Couples Isaac Sim with Cosserat Engine for Cable-Robot Sim2Real
rsasaki0109 · x · 2026-10-11
A CMU-led team released DeformX, a versatile co-simulation framework for deformable linear objects (DLOs), winning an IROS 2026 Oral slot and Best Short Paper at the CVPR 2026 SynData4CV Workshop.
- Approach: couples NVIDIA Isaac Sim with a dedicated Cosserat rod engine to deliver both visual and physical realism for wires, cables, and ropes — covering sim-to-real fidelity, CAD-grade data generation, and robot learning. Existing visual methods rely on physically ungrounded geometric primitives, while physics methods approximate DLOs as rigid chains or generic soft bodies.
- Results: WireSeg-36k synthetic images with depth and instance masks; SAM3 + LoRA boosts segmentation mAP@75 on a real held-out set; real-robot error of 0cm on a UR5e rope-swing task vs a 15.1cm baseline.
- Highlight demo: a UR5e whips a rope to knock an apple off a person's head, extending the hit-target pipeline to dynamic striking.
More from Embodied
- KingKong's open-source Jumper crab robot packs 22 servos, jumps over 400mm — 4310sy · 2026-10-11
- AI PASSPORT turned into pocket console with 27 games, 90% works offline — ezshine · 2026-10-11
- LeaP from Southeast University, accepted at CoRL 2026, makes diffusion action priors state-aware — jiqizhixin · 2026-10-11
- DeformX co-simulation framework for deformable objects lands IROS 2026 oral — rsasaki0109 · 2026-10-11
- Eric Jang resurfaces 2021 essay: bet on compute pressure sparking spontaneous intelligence — ericjang11 · 2026-10-11
- Wuji opensources mjlab PPO stack for dexterous hand: in-hand reorientation and pen spinning — rohanpaul_ai · 2026-10-11