Using artificial neural networks for open-loop tomography.

Area of Science:

• Astronomy
• Optical Engineering
• Atmospheric Physics

Background:

• Large telescopes require adaptive optics (AO) to correct atmospheric turbulence.
• Tomographic techniques are essential for reconstructing phase aberrations using off-axis guide stars in multi-object adaptive optics (MOAO).

Purpose of the Study:

• To present a novel artificial neural network (ANN) method for phase reconstruction in MOAO systems.
• To evaluate the ANN method's performance against standard least squares and existing MOAO techniques.
• To develop a more robust and less condition-dependent phase reconstruction method.

Main Methods:

• An artificial neural network (ANN) was trained using a wide range of turbulent layer configurations.
• The ANN reconstructs target phase aberrations using off-axis reference sources.
• Performance was compared to least squares matrix multiplication and the CANARY instrument's learn and apply method.

Main Results:

• The ANN method demonstrates robustness to varying turbulence profiles without requiring prior knowledge.
• The ANN's non-linear response enhances its resilience to noisy centroid measurements compared to linear methods.
• The ANN approach shows comparable or improved performance against established MOAO reconstruction techniques.

Conclusions:

• Artificial neural networks offer a promising approach for phase reconstruction in MOAO systems.
• The developed ANN method provides a robust and adaptable solution for astronomical adaptive optics.
• This technique enhances the capabilities of large telescopes by improving atmospheric aberration correction.