Integrating supervised and reinforcement learning for predictive control with an unmodulated pyramid wavefront sensor for adaptive optics

TL;DR

This paper introduces a hybrid control method combining supervised and reinforcement learning to improve adaptive optics systems with pyramid wavefront sensors, enhancing performance in challenging observational conditions.

Contribution

It presents a novel integrated control framework that leverages offline supervised learning and online reinforcement learning for better phase reconstruction and predictive control in adaptive optics.

Findings

Outperforms traditional control methods in simulations

Shows robustness under faint star and poor seeing conditions

Improves adaptive optics performance significantly

Abstract

We propose a novel control approach that combines offline supervised learning to address the challenges posed by non-linear phase reconstruction using unmodulated pyramid wavefront sensors (P-WFS) and online reinforcement learning for predictive control. The control approach uses a high-order P-WFS to drive a tip-tilt stage and a high-dimensional mirror concurrently. Simulation results demonstrate that our method outperforms traditional control techniques, showing significant improvements in performance under challenging conditions such as faint stars and poor seeing, and exhibits robustness against variations in atmospheric conditions.