SIM1: Physics-Aligned Simulator as Zero-Shot Data Scaler in Deformable Worlds
2026-04-09 • Robotics
RoboticsArtificial IntelligenceComputer Vision and Pattern Recognition
AI summaryⓘ
The authors studied how robots can learn to manipulate soft, bendy objects like cloth, which is much harder than dealing with hard, rigid objects. They found that current simulations aren't realistic enough because they don't match the real-world physics closely. To fix this, the authors created a system called SIM1 that takes real-world data, builds accurate physics-based simulations, and generates lots of realistic robot behaviors from limited examples. They showed that training robots with this synthetic data works just as well as using real data, and even helps robots succeed more often in new situations. This approach makes learning to handle deformable objects more efficient and practical.
robotic manipulationdeformable objectsphysics simulationsim-to-real transferelastic modelingtrajectory generationdiffusion modelsdata-efficient learningpolicy trainingzero-shot success
Authors
Yunsong Zhou, Hangxu Liu, Xuekun Jiang, Xing Shen, Yuanzhen Zhou, Hui Wang, Baole Fang, Yang Tian, Mulin Yu, Qiaojun Yu, Li Ma, Hengjie Li, Hanqing Wang, Jia Zeng, Jiangmiao Pang
Abstract
Robotic manipulation with deformable objects represents a data-intensive regime in embodied learning, where shape, contact, and topology co-evolve in ways that far exceed the variability of rigids. Although simulation promises relief from the cost of real-world data acquisition, prevailing sim-to-real pipelines remain rooted in rigid-body abstractions, producing mismatched geometry, fragile soft dynamics, and motion primitives poorly suited for cloth interaction. We posit that simulation fails not for being synthetic, but for being ungrounded. To address this, we introduce SIM1, a physics-aligned real-to-sim-to-real data engine that grounds simulation in the physical world. Given limited demonstrations, the system digitizes scenes into metric-consistent twins, calibrates deformable dynamics through elastic modeling, and expands behaviors via diffusion-based trajectory generation with quality filtering. This pipeline transforms sparse observations into scaled synthetic supervision with near-demonstration fidelity. Experiments show that policies trained on purely synthetic data achieve parity with real-data baselines at a 1:15 equivalence ratio, while delivering 90% zero-shot success and 50% generalization gains in real-world deployment. These results validate physics-aligned simulation as scalable supervision for deformable manipulation and a practical pathway for data-efficient policy learning.