Efficient minimization of servo lag error in adaptive optics using data stream mining

Abstract

Prediction of the wavefronts helps in reducing the servo lag error in adaptive optics caused by finite time delays (~ 1-5 ms) before wavefront correction. Piecewise linear segmentation based prediction is not suitable in cases where the turbulence statistics of the atmosphere are fluctuating. In this paper, we address this problem by real time control of the prediction parameters through the application of data stream mining on wavefront sensor data obtained in real-time. Numerical experiments suggest that pixel-wise prediction of phase screens and slope extrapolation techniques lead to similar improvement while modal prediction is sensitive to the number of moments used and can yield better results with optimum number of modes.