Reproduce “Uranus: Building the Next-Generation Simulation Infrastructure for Embodied AI” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Wenkang Qin, Yukun Zhou, Noah Shen, Jisong Cai, Dongxiao Mao, Baicheng Li, Yue Zhang, Wei Sui
Uranus is a learned robot simulator that predicts short visual observations from supplied joint-position trajectories, using each prediction as context for the next. In a matched 100-example G1 comparison, it reaches 12.29 dB PSNR and 0.45 SSIM at 140–160 seconds, versus 7.07 and 0.20 for GE-Sim 2.0. These results measure visual fidelity, not verified contact physics.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.