Vision-Closed-Loop Robot Solves Transparent Substrate Handling

bravo_abad · x · 2026-07-14

Automated labs often struggle with machine vision identifying and automatically replacing fragile, transparent glass substrates. Researchers solved this by using side blue lighting to illuminate the transparent edges, combined with a two-stage detector utilizing geometric models and convolutional neural networks for precise localization. Employing a "detect and retry" closed-loop system, they achieved a 100% placement success rate in 200 autonomous material-changing tests. This demonstrates that in research automation for materials and batteries, fault-tolerance and error-correction mechanisms are more effective at boosting throughput than simply improving precision.

Original post →

More from Embodied

Embodied channel →