Official websites use .gov
A .gov website belongs to an official government organization in the United States.

Secure .gov websites use HTTPS
A lock ( ) or https:// means you’ve safely connected to the .gov website. Share sensitive information only on official, secure websites.

PROJECTS/PROGRAMS

Machine Learning Driven Self-correcting Autonomous Metrology Systems (SAMS)

Summary

As part of the NIST on a Chip program, the autonomous metrology project is developing machine learning driven systems to remove barriers to adoption of advance metrology systems. Towards this end the team is working to apply physics-informed machine learning (ML) to NIST on a Chip (NOAC) sensor measurements to deliver laboratory-grade measurement reliability and accuracy in complex operational environments, effectively putting the expert knowledge of NIST scientists behind every NOAC measurement made in the field by non-experts.

Description

The NIST on a Chip program has led to the development of a range of disruptive measurement solutions that leverage nanophotonics, quantum optomechanics and/or spectroscopic tools to enable traceable measurements that either eliminate or greatly diminish the need for frequent calibrations. Transition of these technologies from lab to marketplace success of NOAC technologies, however, will depend on range complex variables including the user-community’s ability to analyze and interpret data from these novel technologies employing physics beyond their present expertise.

This collaborative effort is focused on the development of machine learning-driven autonomous metrology research systems, with the goal of accelerating the development of self-correcting photonic and quantum sensor networks. Our goal is to remove the “PhD-in-the-loop” to democratize advance metrology technologies and make them accessible to all users regardless of expertise. To accomplish this goal our approach relies on combining machine learning with machine-controlled measurement tools for closed loop experiment design, execution, and analysis, where experiment design is guided by active learning, Bayesian optimization, and similar methods. A key challenge is the integration of prior knowledge into the data analysis, including both device physics and material properties.

We are primarily interested in photonic (e.g. silicon ring resonators) and quantum (NV diamond) sensor networks for thermodynamic metrology (temperature, pressure, and humidity). On going projects include developing self-correcting models for Fiber Bragg grating based temperature sensors, SiN membrane based optomechanical sensors and machine-controlled NV diamond spectrometer for pressure and dosimetry measurements.