Otimização de parâmetros de uma rede LoRa através de algoritmos genéticos

Abstract

The LoRa technology stands out for allowing the connection of a large number of devices allocated several kilometers away from the base station, with low energy consumption and maintenance costs. To do so, each device has a set of settings that define how it will work, like carrier frequency, bandwidth, and so on. Choosing the right configuration is essential for a well working network. In the LoRa technology, the communication is usually controlled by the LoRaWAN protocol, that defines how the network will exchange messages. In this protocol, there are instructions that change some configurations of the connected devices, optimizing its performance by either enlarging the transmission’s data rate or lowering the energy consumption. However, this optimization considers just the device itself, and not how it will affect the rest of the network. This project compares the performance of a network that is configured by the LoRaWAN protocol with the performance of a network configured by a genetic algorithm, which is capable of evaluating the network as a whole and search solutions tending to the optimum. The genetic algorithm was implemented in Python and the network performance was measured using simulations based on existing works. It was possible to see that, for a small amount of devices connected on the network, the LoRaWAN protocol has a good enough performance. Still, when the network complexity gets larger, solutions generated by the genetic algorithm are superior than the ones generated by the protocol.