Frame Loss Characterisation andData Erasure Coding for LoRaWAN

Internet of Things (IoT) solutions are increasingly being deployed for smart applications. With the rise of smart cities there is a need for low cost and long range Low Power Wide Area Network (LPWAN) technologies. LoRaWAN is an energy efficient and cheap LPWAN solution. It is rapidly being adopted all around the world. LoRaWAN however does not provide communication robustness in its basic configuration. Transmitted frames can be lost. In this study we perform extensive measurements on a brand new LoRaWAN network to characterise spatial and temporal properties of frame loss. The empirical outage probability for the farthest measured distance from the closest gateway of 7.5 km in our deployment is as low as 0.004, but the frame loss at this distance is significantly high with 70% frame loss. Furthermore, we show that burstiness in frame loss can be expected for both mobile and stationary situations. Frame loss results in data loss, since in the basic configuration frames are only transmitted once. To reduce data loss in packet-based transmission like in LoRaWAN, we design a novel data erasure coding scheme called DaRe. This coding scheme extends frames with redundant information that is calculated from the data from previous frames. We use a fountain code approach with a sliding window. We also develop an implementation for DaRe in LoRaWAN. We show that 99% of the data can be recovered with a code rate of 1/2 when frame loss is up to 40%. Compared to a simple repetition coding method DaRe provides 21% more data recovery, and can save up to 42% energy consumption on transmission for 10 byte data units. DaRe also provides better resilience against bursty frame loss. This study provides useful results to both LoRaWAN network operators as well as developers of LoRaWAN applications. Network operators can use the results to identify possible weaknesses in the network, and application developers are offered a tool to prevent possible data loss.