In the framework of ground-based astronomical observation with large telescopes, the performance obtained in terms of image quality is strongly dependent on the quality of the adaptive optics (AO) system used. Among the recent work in this field, the development of new regulators (which compute the command applied to the deformable mirror to correct the incoming wavefront) is an approach that could be used to create the next generation of high-performance AO systems as well as a relatively easy way to upgrade existing systems. A particular example of such a controller is the Linear Quadratic Gaussian regulator that uses a model-based predictive approach to compute the DM commands. In the framework of the European H2020 ORP project, we propose to develop a fully auto-tuned and unsupervised LQG regulator to be implemented later on an existing AO system. The unsupervised procedure eliminates the need to manually do the calibration and embeds the auto-tuned part that updates the identified model and regulator parameters during observation. The goal is to allow astronomers to benefit for optimal AO control without the need for a control expert.