Pointing prediction during the tracking process in free-space optical communication

In free-space optical communication systems, stable beam tracking is a prerequisite for establishing laser communication. Among existing beam tracking methods, the pointing, acquisition, and tracking (PAT) technique typically relies on sensors to detect the deviation between the incident beam and the optical axis, followed by driving units to correct the orientation of the optical antenna. Considering that in practical applications, the PAT correction process lags behind spot detection, we propose to incorporate predictive pointing estimation into the tracking process. First, the terminal pointing is decomposed into azimuth and elevation angles of the servo motors. Angle values at different time instances are recorded to reconstruct historical pointing trajectories. Then, a long short-term memory (LSTM) neural network is employed to achieve short-term prediction of the terminal pointing based on the constructed sequential data. Simulations demonstrate the effectiveness of the proposed method.