K-Means Spreading Factor Allocation for Large-Scale LoRa Networks

Low-power wide-area networks (LPWANs) are emerging rapidly as a fundamental Internet of Things (IoT) technology because of their low-power consumption, long-range connectivity, and ability to support massive numbers of users. With its high growth rate, Long-Range (LoRa) is becoming the most adopted LPWAN technology. This research work contributes to the problem of LoRa spreading factor (SF) allocation by proposing an algorithm on the basis of K-means clustering. We assess the network performance considering the outage probabilities of a large-scale unconfirmed-mode class-A LoRa Wide Area Network (LoRaWAN) model, without retransmissions. The proposed algorithm allows for different user distribution over SFs, thus rendering SF allocation flexible. Such distribution translates into network parameters that are application dependent. Simulation results consider different network scenarios and realistic parameters to illustrate how the distance from the gateway and the number of nodes in each SF affects transmission reliability. Theoretical and simulation results show that our SF allocation approach improves the network’s average coverage probability up to 5 percentage points when compared to the baseline model. Moreover, our results show a fairer network operation where the performance difference between the best- and worst-case nodes is significantly reduced. This happens because our method seeks to equalize the usage of each SF. We show that the worst-case performance in one deployment scenario can be enhanced by [Formula: see text] times.