Paper 13094-111
Paper 13094-111
Machine learning compensator for active optics control of the SOAR primary mirror
18 June 2024 • 17:30 - 19:00 Japan Standard Time | Room G5, North - 1F
Abstract
A machine learning compensator for the SOAR active optics control system is proposed to reduce the convergence time. The compensator uses a deep neural network to predict the mirror shape from sensor measurements and generate actuator commands to correct the mirror shape. Results of a simulated SOAR telescope show that the compensator could reduce the convergence time by up to 90%.
Presenter
Association of Universities for Research in Astronomy (Chile), NSF's National Optical-Infrared Astronomy Research Lab. (Chile)
Engineer with over 8 years of experience in ground-based telescope engineering. Graduated from Pontificia Universidad Católica de Valparaíso in Chile with a degree in Electronic Engineering, with a specialization in machine learning and a minor in automatic control and instrumentation. Passionate about exploring ideas and alternatives for applications where AI and control systems intersect. Enthusiastic about applied research of the AI solution for improving telescope systems and mentoring new engineers in this field.
Currently part of the teamwork of electronics engineering in the MSO division of USA NSF's NOIRLab, with valued contributions to extend the life cycle of SOAR mount and to improve the performance of the active optics system in both telescopes, 4m-Victor Blanco and SOAR.