Neural networks and PCA coefficients to identify and correct aberrations in adaptive optics

Context. Static and quasi-static aberrations represent a great limit for high-contrast imaging in large telescopes. Among them the most important ones are all not corrected by adaptive optics (AO) system, which called non-common path (NCPA). Several techniques have been proposed to mitigate it. The typical approach is set an offset on AO system with exactly opposite sign of NCPA order correct introduced optical components downstream wave-front sensor (WFS) up science camera. An estimate can be obtained trial-and-error or more sophisticated focal-plane sensing. Aims. In cases, fast procedure desirable telescope downtime repeat, if needed, correction cope temporal variation NCPA. Very recently, new approaches based neural networks (NNs) also as alternative. Methods. this work, through simulated images, we test application supervised NN mitigation NCPAs at visible wavelengths and, particular, investigate possibility applying method imagers such SHARK-VIS, forthcoming visible-band imager Large Binocular Telescope (LBT). Results. Preliminary results show measurement accuracy 2 nm root mean square (RMS) each sensed Zernike mode turbulence-free conditions, 5 RMS per when residual turbulence has error (WFE) approximately 42.5 RMS, value during LBT calibration. This sufficient guarantee that, after correction, residuals negligible compared WFE > 100 best systems Conclusions. Our simulations robust even presence turbulence-induced that labelled training phase NN. could thus used real-world setting offloading corrective static