Mitigation of Pointing Errors in FSO Systems Through AI-Driven Spatial Light Modulation

Free-space optical (FSO) communication is an emerging technology that enables higher data rates. While FSO offers several advantages, it faces significant challenges due to the atmospheric channel, which can introduce pointing errors (PE) between transceivers and this phenomenon can be exacerbated by mechanical instabilities. Current pointing error compensation (PEC) mechanisms are primarily based on mechatronic devices; however, these systems may require redundancy to ensure reliability. This work presents an experimental study investigating the integration of a spatial light modulator (SLM) with a generative convolutional neural network (CNN) to mitigate residual PE in FSO communication systems. In this context, the proposed method demonstrated improved coupling efficiency into a single-mode fiber (SMF) by compensating mechanical-induced PE, introducing a novel application for SLMs and CNNs. Experimental results show gains up to 34 dB for a 0.5mm beam misalignment, with sustained improvements in optical power for displacements up to 4.8 mm.