Deep-Learning-Enabled Autonomous Alignment of Multipass Herriott Cells for Field-Deployable Optical Sensors

Read full text Summarise with VIRAT
Authors: Evgenii Venediktov, Shuda Zhong, Guangyin Zhang, Zach Splain, Majid Chauhdry, Emeka V. Ikpeazu, zhi-hong mao, Ruishu Feng Wright, Nageswara R. Lalam, Kevin Chen
Cited by: 0

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

Herriott cells enable sensitive trace gas detection through extended optical path lengths, but their multiple-reflection configuration is highly sensitive to mirror alignment. Despite advances in ML-based optical alignment, automated correction of Herriott cell mirrors, where reflection patterns are highly scattered and configuration-sensitive, has remained an open problem. We present a closed-loop edge-deployable autonomous alignment system using a camera and lightweight convolutional neural network to infer angular corrections directly from reflection patterns. Deployed on a Raspberry Pi with low-cost stepper motors, the system achieved 100% alignment success across the full operational range, converging in fewer than two steps on average with a mean error of 0.045°. This approach enables rapid, in-situ recalibration for field-deployable gas sensors in demanding environments without human intervention.

Fields of Study & Subjects

Spectroscopy and Laser Applications Advanced Measurement and Detection Methods Air Quality Monitoring and Forecasting

Publication details

Publication date
19 March 2026
Language
EN
Document type
Preprint