ActiveScale: teaching robots active perception via model-data-hardware co-design
Haoyu_Xiong_ · x · 2026-09-19
A CMU + HKUST(GZ) team introduced ActiveScale, a framework advancing robotic active perception through coordinated model, data, and hardware design:
- Model: augments a VLA with historical video observations and explicit camera-pose supervision, using per-frame pose tokens and a lightweight prediction head to associate observations across viewpoints.
- Data: a scalable human–robot mid-training recipe using 1,000 hours of egocentric and robotic data, exploiting natural camera motion in human activity.
- Hardware: the Active Perception Mobile Manipulation Platform (AMP) supports single-operator teleoperation for scalable demonstration collection coordinating viewpoint changes with manipulation.
Experiments show improved success rates on active-perception tasks, with ablations confirming camera-pose-aware modeling and egocentric data. Code, dataset, and models are released.
Related event: ActiveScale: Co-Designed Framework Teaches Robots Active Perception(2 posts)→
More from Embodied
- Baidu's Apollo Go wraps two-week public robotaxi test rides in Tung Chung, Hong Kong — Baidu_Inc · 2026-11-04
- imec neural probe integrates photonics directly into 130nm CMOS backend — jwt0625 · 2026-11-04
- Chris Paxton: interactive robot sim is the future but glaringly data-limited — Scobleizer · 2026-11-04
- DreamX: video diffusion models as universal latent action learners for robots — kaist-ai · 2026-11-04
- RoboTryOn turns human videos into robot demos, lifting success rate from 2% to 60% — kaist-ai · 2026-11-04
- INMO Future glasses: dual 1080p displays, 125g, $799 early-bird on Kickstarter Dec 10 — Scobleizer · 2026-11-04