Intelligent Payload and Energy Balancing in Optical Transport Network Using Graph Neural Network

The algorithm of intelligent flow routing control in optical transport networks with the use of graph neural networks is presented in the work. The use of graph neural networks to solve telecommunications problems is due to the fact that the input parameters for training are the characteristics of the network itself. That is, the matrix of adjacencies, characteristics of nodes and edges at a certain point in time are the input parameters for training graph neural networks. Based on these data, you can predict the state in which the node is. This approach allows a more flexible approach to the process of determining the state of the node because it does not require rewriting the software but only administratively changing the parameters of the neural network. The proposed algorithm allowed to reduce the delay during peak hours of the network by 18%.