METHOD FOR NEURAL NETWORK CLUSTERING OF SPATIALLY DISTRIBUTED WIRELESS SENSOR NETWORK Russian patent published in 2025 - IPC H04W40/32 H04W84/18 G06N3/88 

FIELD: wireless communication network.

SUBSTANCE: invention relates to wireless networks and telecommunications and can be used in hierarchical routing protocols of spatially distributed self-organizing wireless sensor networks (WSN). Method involves hierarchical division of structural elements of the WSN into head and "slave" cluster nodes using a self-organizing artificial neural network (ANN) of Kohonen, which is trained using normalized values of the training WSN matrix. Training WSN matrix is formed by calculating the Hadamard product of the connectivity matrix and the utility matrix of the network. Structure of the entire WSN is described based on a utility matrix represented by a weighted directed graph of the network. WSN clustering data are used when forming communication information directions to provide inter-cluster and intra-cluster interaction between WSN nodes.

EFFECT: high stability of information interaction between WSN nodes by taking into account additional parameters of functioning of network nodes and longer time of relevance of results of clustering of WSN nodes in conditions of high dynamics of WSN topology.

1 cl, 11 dwg