RAMA: A new agile AO bench for the telescope FEELINGS

Abstract

The ground-based observation of extended objects such as satellites suffer from severe atmospheric propagation constraints. Specifically, high tracking velocities and low-elevation lines of sight generate non-stationary turbulence alongside strong scintillation. Developing robust wavefront control strategies is therefore critical to maintain stable observations. In this context, we present RAMA, an adaptive optics (AO) testbench deployed on ONERA's 60\,cm FEELINGS telescope. Designed as a pathfinder for future systems like the PROVIDENCE ground station, RAMA evaluates a visible, non-modulated Pyramid Wavefront Sensor (PWFS). The hardware baseline also includes two pupil-conjugated deformable mirrors (DM97 and DM192) driven by the DAO Real-Time Computer (RTC). Inheriting the modular and evolving philosophy of the PAPYRUS project, the bench provides a flexible environment to test new components on-sky and allows for direct comparisons between classical controllers and advanced, data-driven strategies. By implementing Convolutional Neural Networks (CNN) for phase reconstruction and Reinforcement Learning (RL) for loop control, RAMA aims to overcome the specific limitations associated with scintillation and extended-object observations. This paper details the opto-mechanical design, numerical simulations of the bench expected wavefont control performance, preliminary laboratory closed-loop results and the first on-sky optical coupling with the telescope.