Modulation format identification in elastic optical networks using integrated photonic reservoir computing and untrained K-nearest neighbors algorithm

Abstract

In the next generation of Elastic Optical Networks, various modulation formats exhibit varying degrees of sensitivity to channel impairments during transmission. To adopt appropriate channel equalization schemes at the receiver, it is essential to perform modulation format identification prior to the receiver, followed by the adjustment of receiver parameters and types based on the recognition results. A system based on a 52-node integrated photonic reservoir chip and untrained K-nearest neighbors (KNN) algorithm is proposed for the recognition of OOK, PAM4, QPSK, and BPSK modulation formats in optical channel transmission. Its performance is validated across optical signal-to-noise ratios ranging from 8 to 23 dB, taking into account the dispersion damage of 20 km single-mode fiber transmission. In all tested scenarios, the recognition accuracy consistently surpasses 96.25%, showcasing a 14.93% improvement over prior works and an 82.81% enhancement over traditional algorithmic methods under identical conditions. The study explores the impact of different waveguide delay amounts, random phases, and algorithm K values on recognition accuracy.