Point-to-multipoint coherent PON system modeling based on the diffusion model

Modeling point-to-multipoint (P2MP) coherent passive optical network (PON) systems based on digital subcarrier multiplexing (DSCM) plays a crucial role in global system optimization and overall performance monitoring. The subcarriers could support any rate and modulation format. Besides, subcarriers reach different optical network units (ONUs) with different fiber link distances and received optical power. Therefore, it is necessary to construct a generalization model of the communication system to support arbitrary configuration of link parameters. In this paper, for what we believe to be the first time, we use the diffusion model in machine learning to establish a generalized model of the P2MP system. The SNR error of receiving signals between the modeling channel and the original channel is less than 0.1 dB under a multi-ONU channel with different configurations. It provides a new idea, to our knowledge, for P2MP coherent PON modeling.