Experimental performance evaluation of VAR LQG predictive control in adaptive optics for LEO satellites.

This article focuses on a laboratory demonstration of a predictive control law in an adaptive optics-assisted free-space communication link from a LEO satellite to the ground. The considered predictive control law is a linear quadratic Gaussian (LQG) law, used with an enhanced turbulent phase evolution model, a Zernike-based vector auto-regressive (VAR) model. Previous works in our team at ONERA have shown that VAR LQG control brings significant improvement in terms of coupling efficiency (CE) performance in a single-mode fiber (SMF), in numerical simulations. Additionally, to our best knowledge, no experimental validation of predictive control for LEO-to-ground free-space communication links has been performed yet. The goal of this work is thus twofold. First, we assess the experimental performance in controlled conditions of VAR LQG control and compare it to the output of an end-to-end simulation developed in parallel, with the aim of controlling and understanding the different error terms that can lead to discrepancies between both results. The CE performance with VAR LQG control is also compared to the CE performance with other classic and predictive control laws. To obtain a global understanding of the gain in performance with VAR LQG, we also include robustness studies of the input parameters. Second, the validated numerical tool allows simulations of more representative systems with realistic turbulence profiles, and their performance with VAR LQG. This is applied to telecom links as well as satellite observation. Ultimately, this work paves the way to on-sky validations on optical ground stations (OGS) such as ONERA's FEELINGS OGS.