Optical stabilization for laser communication satellite systems through proportional-integral-derivative (PID) control and reinforcement learning approach.

Abstract

One of the main issues of the satellite-to-ground optical communication, including free-space satellite quantum key distribution (QKD), is an achievement of the reasonable accuracy of positioning, navigation, and optical stabilization. Proportional-integral-derivative (PID) controllers can handle various control tasks in optical systems. Recent research shows the promising results in the area of composite control systems including classical control via PID controllers and reinforcement learning (RL) approach. In this work, we apply a RL agent to an experimental stand of the optical stabilization system of the QKD terminal. We find via agent control history more precise PID parameters and also provide an effective combined RL-PID dynamic control approach for the optical stabilization of the satellite-to-ground communication system.