Synthetic Training of a Machine Learnin...

This paper presents an AI-inspired approach to analyze the performance of a PLC network based on its topology and geometrical parameters. Transmission Line Theory allows the performance analysis of cabled links based on their physical properties. In this case it is used as a basis to synthetically train an AI to infer a link's capacity based on geometrical properties of the network, such as number of nodes, average distance between nodes, standard deviation of the distances and actual distance between a communicating pair. Eventually, the inferred capacity is used to optimize the polling procedure in an Automatic Meter Reading application by using the AI to choose which nodes should act as repeaters.