Development of advanced control strategies for Adaptive Optics systems

Atmospheric turbulence is a fast disturbance that requires high control frequency. At the same time, celestial objects are faint sources of light and thus WFSs often work in a low photon count regime. These two conditions require a trade-off between high closed-loop control frequency to improve the disturbance rejection performance, and large WFS exposure time to gather enough photons for the integrated signal to increase the Signal-to-Noise ratio (SNR), making the control a delicate yet fundamental aspect for AO systems. The AO plant and atmospheric turbulence were formalized as state-space linear time-invariant systems. The full AO system model is the ground upon which a model-based control can be designed. A Shack-Hartmann wavefront sensor was used to measure the horizontal atmospheric turbulence. The experimental measurements yielded to the Cn2 atmospheric structure parameter, which is key to describe the turbulence statistics, and the Zernike terms time-series. Experimental validation shows that the centroid extraction algorithm implemented on the Jetson GPU outperforms (i.e. is faster) than the CPU implementation on the same hardware. In fact, due to the construction of the Shack-Hartmann wavefront sensor, the intensity image captured from its camera is partitioned into several sub-images, each related to a point of the incoming wavefront. Such sub-images are independent each-other and can be computed concurrently. The AO model is exploited to automatically design an advanced linear-quadratic Gaussian controller with integral action. Experimental evidence shows that the system augmentation approach outperforms the simple integrator and the integrator filtered with the Kalman predictor, and that it requires less parameters to tune.