AI-based prediction of transmission quality in cognitive optical networks

Abstract Quality of Transmission (QoT) prediction is done by a proposed method in optical networks. It uses a Radial Basis Function Network (RBFN) model trained with data from a comprehensive optical model. The RBFN model is enhanced with several techniques to improve its accuracy. The key objective is to enhance hardware utilization by significantly reducing the required system margin, potentially up to the order of dBs. To achieve this, the study employs the Radial Basis Function Network (RBFN) model, capitalizing on input data related to connectivity and signal characteristics for QoT prediction. The proposed method achieves good performance (MSE: 0.802, MAE: 0.2) but is slower than some existing methods. However, compared to these existing methods, the proposed method has 1.54 %, 5.32 %, and 5.46 % higher performance than SOM-RBF, AHFSE, and Wavelet-chaos NN. This research also contributes to the field by introducing a new cognitive-based QoT model that uses deep learning techniques. The study showcases the potential for practical implementation and optimization in relevant applications, emphasizing the intersection of artificial intelligence and optical network resource utilization.