Performance Evaluation of LoRaWAN

Abstract

The uprising of Long Range (LoRa) technology for Low-Power Wide Area Network (LPWAN) has gained popularity in IoT applications. As LoRa utilizes the license-free ISM-bands, many network providers have deployed their LoRa networks that adhere to the LoRaWAN specification. With so many deployed networks, it is likely that devices belonging to neighboring networks will interfere and lead to increased packet loss due to collisions. However, not all colliding frames attribute to packet loss, because of the capture effect in LoRa.

In this thesis, we investigate the performance of LoRaWAN, particularly on frame collision and packet loss, through small-scale measurements, which is then followed by the evaluation in a large-scale setup involving multiple gateways. It is expected that adding more gateways can reduce packet loss. By estimating the Data Extraction Rate (DER) from the measurements and validating the result with simulations, we found that adding more gateways, to some extent, does improve the DER when the capture effect is considered. However, downlink traffic can significantly decrease DER of a single gateway as the currently available gateways operate in half-duplex mode. The impact of different spreading factor parameters to the network performance is also evaluated.

We also characterize and evaluate the two major LoRaWAN networks in the Netherlands, namely The Things Network (TTN) and KPN, which respectively represent the crowdsourced and commercial usage of LoRaWAN networks. The influence of traffic from end-devices belonging to other networks on the channel utilization is evaluated. The results can be used as an empirical basis for developing more accurate models and simulations to better understand the capabilities of LoRaWAN.