Experimental throughput models for LoRa networks with capture effect

The Long Range (LoRa) modulation keeps gaining relevance in the landscape of low-power sensor networks. Most models used to evaluate the performances of LoRa deployments are based on the assumption that two colliding frames are necessarily lost. Recent findings have shown that the capture effect occurs in these networks, allowing the receiver to sometimes demodulate the frame featured with the highest signal power. This finding notably improves the overall throughput compared to expectations, but in turn decreases the network fairness. In this paper, we analyze the benefits and drawbacks of such an effect. We therefore provide new throughput models for LoRa networks operating Pure and Slotted ALOHA access schemes. For this purpose, an experimental testbed has been setup and used to measure the occurrence probabilities of capture events in several transmission scenarios. The resulting models are validated with real-life data gathered on the same setup. We additionally analyze the fairness in our deployment, showing that the devices featured with the highest average power at the receiver benefit from a higher success rate than others. By computing Jain's index, we show that this unfairness gets more pronounced as the traffic load increases.