Machine Learning based modulation format classification framework for inter-satellite optical wireless communication system (IsOWCS)

Abstract The exponential growth in demand for high-capacity optical systems has driven the advancement of advanced modulation formats to upgrade transmission capacity and transmission quality. Effective fault diagnosis and self-configuration in inter-satellite optical wireless communication systems (IsOWCS) depend intensely on the generated data. Machine learning (ML) approaches offer promising solutions in evaluating the execution of these networks. In this study, a dataset was created using OptiSystem 18.0. The dataset was composed of various modulation formats such as duobinary, return-to-zero (RZ), non-return-to-zero (NRZ), 33 % RZ, chirped NRZ, vestigial sideband (VSB) NRZ, carrier-suppressed return-to-zero (CSRZ), and VSB CSRZ. The classification of modulation formats has been presented in this study using ML. The dataset was created by varying input power from 0 to 20 dBm and evaluating parameters such as Q factor, input/output signal-to-noise ratio (SNR), power, range, eye closure, amplitude, height, eye opening, output OSNR. Four ML classifiers were used to predict the classification of different modulation formats. Random forest (RF) classifier performed exceptionally well and achieved 100 % accuracy. Moreover, an interactive user-friendly web page was also developed using Anvil for modulation format classification. The proposed research underscores the significance of selecting the appropriate modulation format to optimize the performance and transmission distance of IsOWCS, subsequently enhancing the operation of high-speed optical communication systems.