| title | Reflex: Visually Grounded Reactive Humanoid Control |
|---|---|
| authors | Taoyang Jia, Weikai Huang, Linxin Song, Jared Darlington, Jieyu Zhao, Yue Wang, Jiafei Duan, Zhongzheng Ren, Ranjay Krishna |
| date | 2026-10-08 |
| venue | arXiv (preprint, not peer reviewed) |
| id | arXiv:2610.12772 |
| affiliation | University of Washington |
| upvotes | 57 (Hugging Face Daily Papers, 2026-11-01) |
| links | abs pdf hf code |
| abstract | Humans effortlessly react to the world; they adeptly return a tennis shot, catch a falling cup, step aside from a passing cart, or pass a box tossed during truck unloading. These everyday actions require anticipating a moving target and reflexively responding with the whole body within about a second. Such visual reactive whole-body control remains challenging for humanoids, which must (1) coordinate whole-body control, (2) react to object dynamics in real time, and (3) ground perception in egocentric vision. We study catching thrown boxes as a representative task that demands all three capabilities; existing systems cover only part of it, relying on end-effectors, motion capture, or marked objects. We present Reflex, a three-stage framework that learns these capabilities stage by stage while connecting them through a compact representation of object dynamics. First, a state-based policy learns whole-body catching with reinforcement learning. Second, the policy learns to infer the dynamics of a moving box from delayed and incomplete observations. Third, an egocentric RGB-D encoder learns to recover this dynamics representation directly from visual observations while keeping the whole-body controller fixed. This decomposition provides each capability with a direct training signal and makes visual learning substantially more scalable. In simulation, Reflex achieves 85.7% catch success from egocentric RGB-D, compared with 87.9% when given privileged box observations. Temporal visual history is critical: removing it reduces success by 35.6 and 39.8 points on centered and laterally offset throws. On a real Unitree G1, Reflex catches human-thrown boxes with 65% and 40% success at 2.5 and 5 m, reaching 76% and 59% of human catchers' success at the same distances. |
cite
@misc{jia2026reflex,
title={Reflex: Visually Grounded Reactive Humanoid Control},
author={Taoyang Jia and Weikai Huang and Linxin Song and Jared Darlington and Jieyu Zhao and Yue Wang and Jiafei Duan and Zhongzheng Ren and Ranjay Krishna},
year={2026},
eprint={2610.12772},
archivePrefix={arXiv},
url={https://arxiv.org/abs/2610.12772}
}