Research on Data-Driven LoRa Link Quality Estimation

Long Rang (LoRa) link quality is critical to the data link layer protocol. The appropriate LoRa parameters can be decided by link quality to improve the reliability of data transmission and reduce the number of retransmissions. This paper utilizes statistics of the Signal-to-Noise Ratio (SNR) and the Received Signal Strength Indication (RSSI) on received packets in the current window as the link quality feature, and the data-driven based link quality estimation model is constructed by Genetic Algorithm optimization Back-Propagation Neural Network (GABPNN). The packets reception rate (PRR) within the time window is the output of the model. The link quality is divided by the PRR. In the Not Line of Sight (NLOS) scene, this paper will compare the link quality estimation results of BPNN, naive bayes, and support vector machine models in different reception window sizes and different packets transmission rate conditions. Experiments show that the accuracy of the GA-BPNN model link quality estimation result can reach over 93.5%.