Machine learning for optics and controls

Synopsis

Optical cavities are widely used in physics and precision measurement.  This project will explore the use of modern machine learning methods for the control of suspended optical cavities.  

Research fields

Photonics, Lasers and Nonlinear Optics

Required background

Some knowledge of python or another imperative programming language, image processing experience would be useful.  

Description

Modern gravitational wave detectors such as Advanced LIGO, which recently detected gravitational waves, are the most sensitive measurement devices ever constructed.  They are based around multiple nested optical cavities, with each mirror in the system suspended to limit the impact of local seismic motion.  This leaves the mirrors free to move, but for the detectors to function, the mirrors positions and orientations must be precisely sensed and delicately controlled.  This yields a rich controls problem with many relevant degrees-of-freedom.  

We want to apply the methods of machine learning and artificial intelligence to this controls problem, to find optimal solutions that cannot be obtained by humans in a reasonable time.  This project will develop a machine learning system for alignment control of a single optical cavity.  The machine learning system will be trained and tested with a numerical simulation of a cavity system, and then demonstrated in a table-top optical experiment.