Artificial intelligence based routing in PLC networks

With innovation in network design, a strong rise in the specific performance for the end-to-end communication (capacity, latency, connectivity, etc) is expected. This paper tackles the possibility of using an artificial-intelligence-based algorithm to encompass smart routing in a network with a relatively low computation cost. The AI aspect of our method lies in the adaptability of the proposed algorithm: with the need of users and network characteristics changing in time and space, a smart, dynamic approach is required. The application of this AI is done in the specific framework of powerline communication (PLC) used as a front-hauling solution for different communication technologies. Training of the AI processor is done with synthetic data. A huge dataset is generated by using the transmission-line theory based approach to compute the capacity of a powerline link and its behavior with respect to topological information, such as disposition of nodes, their mutual distance and their overall density on the geographical territory. From this bottom-up-generated data, it is eventually possible to infer the capacity of each link in a network from its topological information, which is the only input the AI processor needs.