Use of a neural network to control an adaptive optics system for an astronomical telescope

Abstract

ANGEL et al.1 recently showed how an artificial neural network could be used to measure optical phase distortion induced by atmospheric turbulence, and demonstrated by numerical simulation that such a system could be used to control the six 1.8-m mirrors of the Multiple Mirror Telescope by constantly adjusting them to compensate for atmospheric distortion of the image. The neural network estimates the phase distortion using two images of a reference star, or of a laser-produced guide star2, one image being at the best focus of the telescope while the other is intentionally out of focus. Here we report the successful test of a neural network with a real star. We applied a neural network to in- and out-of-focus images of Vega obtained with the 1.5-m single-mirror telescope at the Starfire Optical Range of the Air Force Phillips Laboratory near Albuquerque, New Mexico. The experimental results agree well with phase reconstructions obtained simultaneously with a conventional wave-front sensor.