How Accurate is LoRa Positioning in Realistic Conditions?

The LoRa communication technology has been developed for data transmission, but recent research papers showed that it could also be used to estimate the transmitting devices' location. The location can be found with reasonable accuracy, with median error as low as tens of meters; however, in many cases, it is evaluated in very optimistic transmission conditions, e.g. using line-of-sight communication and on a small data set. We estimate how LoRa localization performs in a realistic scenario, using an extensive data set of a telemetric network of a few thousand devices to calculate the location using the trilateration and estimate the error of the calculation. Within the analysed dataset the true-range multilateration provides highly inaccurate location estimation, with a median error of 588.64 m and an average error of 808.57 m. We argue that the LoRa localization in real-life conditions is ineffective, as the estimation of the distance between the node and the gateway changes heavily depending on the location of the node and radio channel attenuation.