Efficient Mathematical and Data-driven Models for Closed-loop Adaptive Optics Systems for Ground-based Astronomy

Project: Research

View graph of relations



Many recent discoveries about the universe relied heavily on the development of modern telescopes—for instance, Profs. Genzel and Ghez were awarded the 2020 Nobel Prize in Physics for their works on supermassive black holes conducted using the European Southern Observatory telescopes. To look deeper into the universe and obtain clearer images, people are building larger and larger telescopes. However, one major complication when observing space objects on Earth is that the objects “twinkle.” This twinkling effect is due to atmospheric turbulence. The atmosphere is constantly moving, and light from space objects is bent randomly in many ways just before reaching Earth’s surface. The bending crumples the wavefront of arriving light at ground-based telescopes, resulting in a blurring effect that distorts the view of the astronomical objects. Our project aims at developing mathematical and data-driven methods to remove this distortion. One option is to estimate the blurring effect of the atmospheric turbulence and then obtain a clearer image by solving a deblurring problem. In the ’90s, the PI and his collaborators were one of the first research teams to deblur man-made satellite images using variational methods. The blurring effect was estimated using the light from a very distant isolated star or a bright spot in the atmosphere created by a laser beam, the so-called “artificial guide star”. However, most modern telescopes use adaptive optics (AO) systems to reduce the effect of the incoming wavefront distortion and thereby the blurring of the images. These systems first use wavefront sensors (WFS) to measure the gradient of the distortion. Then a reconstruction of the distortion is computed. The reconstruction is used to reshape the surface of a deformable mirror to compensate for the wavefront distortion, hence minimizing the atmosphere’s blurring effect. Most AO systems run in closed-loop where the compensated wavefront is measured and fed back into the system to update the system itself for the next time step. Because of various physical constraints (e.g., not enough photons reaching the WFS), WFS measurements are of very low resolution, yet massive data is collected on the telescopes. Therefore, efficient and accurate methods to reconstruct the wavefront from the WFS measurements are needed. Our goal is to develop fast and accurate variational models and deep neural network approaches to recover the wavefront in closed-loop AO systems. We will comprehensively compare our methods in terms of speed, accuracy, and robustness with other state-of-the-art methods. 


Project number9043373
Grant typeGRF
StatusNot started
Effective start/end date1/01/23 → …