Artificial-Intelligence-Based Performan...

Powerline Communications (PLC) is a popular technology providing infrastructure for applications related to IoT, smart grids, smart cities, in-home networking and has been experimentally considered for broadband access. Sensor networks and Automatic Meter Reading applications are closely related to this technology, as it provides free infrastructure and sustains the data rate requirements. The application here considered consists in the implementation of the G3-PLC LOADng routing protocol in the nodes of a sensor/meter network, where the nodes share all the same medium. G3-PLC is a powerline communication standard, employing OFDM at the physical layer and oriented at enabling the smart grid vision. The Medium Access Control implements CSMA/CA, while the Logical Link Control implements LOADng routing, which is the ITU-T G.9903 recommended specification for Lossy and Low-power Networks (LLNs). In this paper, we consider the mapping phase of the routing protocol, in which the central element of the network establishes the routes to reach any node. By simulating this process via a physical simulation tool, it is possible to synthetically train an Artificial Neural Network and teach it how the optimally established routes correlate to the topological and geometrical properties of the network. Eventually, we discuss how, by employing this AI approach, it is possible to speed-up the routing mapping process.