Reproduce “Control-Ready Uncertainty for Trajectory Diffusion” 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
Zhiwei Xue, Jia Yue Kam, Jinhang Qiu, Yifeng Cheng, Ege Gursoy, Jiaming Wang, Vincent Bonnet, Harold Soh
SCOPE turns a diffusion model’s learned local density into per-timestep uncertainty tubes, estimating spread around proposed trajectories without repeated sampling while leaving trajectory generation and controller rollouts intact. In Maze2D, with 1,024 MPPI rollouts, it achieves 99.25% success versus 96.35% for the same-prior fixed-covariance baseline; this benchmark result is not a general safety guarantee.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.