Seminarios de Instrumentación
Focal Plane Wavefront Sensing via Deep Neural Networks

Resumen

Using only the speckle pattern of an unresolved source in the image-plane, it is possible to reconstruct the pupil-plane wavefront for use in an adaptive optics loop using Deep Neural Networks.  My thesis has involved exploring and refining this technique with applications from real time correction of the atmosphere for ELF telescopes, to sensing residual, non-common-path aberrations in coronagraphic instruments -- where dedicated WFS hardware becomes much more challenging.  I will present some background and results of both simulated as well as successful on-telescope testing.  Further, I may spend a bit of time on my recent exploration into using Deep Reinforcement Learning: its uses in bridging the sim-to-real gap, and for solving problems where exact solutions aren't available.

Seminarios Investigacion le está invitando a una reunión de Zoom programada.