Researcher: Astra Robot Shows Step-Level Image-Action Understanding Unlike Prior VLAs
YuXiang_IRVL · x · 2026-09-14
Yu Xiang of IRVL Lab shares a 2-minute condensed clip of the Astra robot executing a task and offers a technical observation: Astra appears to understand each robot action step by analyzing the captured images. He notes he hasn't seen this level of understanding from previous VLAs or WAMs, suggesting Astra may be discovering visual-servo-like correspondences between images and actions. More demos are promised soon.
More from Embodied
- Open-source companion robot Lamp starts shipping to 20+ countries at $969 — dee_hw · 2026-10-30
- SCAMPER demo: virtual rat trots and gallops with gaits it never saw in training — tyrell_turing · 2026-10-30
- DIY humanoid build, day 7: creator gives up fixing physics, adds 1-DOF ankles — _Stocko_ · 2026-10-30
- Lambda Surgical launches first-in-human study of LiDAR-based LambdaGuide at PINE — ditzikow · 2026-10-30
- Gaussian splats double as shared GPS: one city map localizes drones, robots and phones — bilawalsidhu · 2026-10-30
- SimReady dataset offers 2,677 simulation-ready 3D assets to build virtual test worlds — victormustar · 2026-10-30