Can adaptive strategies sustain bidirectional LoRaWAN traffic?

Low Power Wide Area Networks (LPWANs) are a viable solution for addressing the requirements of Internet of Things (IoT) applications that can tolerate sporadic updates, possibly only in the uplink (from the field towards the cloud and the user) direction. LoRaWAN is currently the most diffused example of LPWAN, exploiting LoRa radios. LoRa, an enhanced version of chirp spread spectrum modulation, trades off the radio sensitivity for the data rate, permitting extensive area coverage. However, the price is a relatively long frame duration, leading to high collision probability in dense, asynchronous scenarios; fortunately, polarity inversion minimizes collisions between uplink and downlink traffic. Therefore, adaptive strategies (based on some feedback from the backend) have also been suggested to adapt the data rate depending on the environment condition dynamically. This work compares the ADR (Adaptive Data Rate) and the ADP (Adaptive Data Payload) strategies when bidirectional traffic exists, using a purposely designed simulator. Results considering a different number of gateways show that ADP tries to maximize the amount of data effectively received in both direction but suffer from the adoption of longer frame duration, leading to higher collision probability. In contrast, ADR tries to force the adoption of the fastest data rate, limiting the overall amount of transferred data.