gym_AO

This work presents a reinforcement learning (RL) environment designed for wavefront sensorless adaptive optics (AO) in optical satellite communication downlinks. The environment, developed according to the OpenAI Gymnasium framework, is used to train RL models for controlling a deformable mirror (DM) to focus optical beams on a single-mode fiber (SMF) without the need for traditional wavefront sensors. The simulation offers different atmospheric turbulence conditions (quasi-static, semi-dynamic, and dynamic) and allows flexibility in configuring observation space, action space, and reward functions. This enables the RL agent to adapt and optimize the optical link in varying turbulence scenarios, with the primary goal of improving satellite data transfer efficiency by reducing latency and system costs.