Opportunistic LoRa-based gateways for delay-tolerant sensor data collection in urban settings

There are several ways that sensor nodes in a smart city setting can get data to a base station for processing. Sensor nodes that need not report their data to the base station in real time can opt for a delay-tolerant networking or data milling setup. If the data collector itself would send the data to the internet through its own (mobile) backhaul link, we can consider the data collector to be a mobile gateway. The choice of radio technology for the link between the sensor or end device and the mobile gateway is important it would affect the number of mobile gateways needed, the required mobility pattern for them to provide a certain level of delay guarantees to the data, and the overall efficiency of the scheme. LoRa, with its long range and relatively low power consumption, can potentially decrease the number of gateways needed in the system, while decreasing delays. This study explores opportunistic, mobile LoRa-based gateways using a pull-based transfer of data to implement a low cost system in terms of communication and the hardware used. We present a MAC layer protocol, for use with the LoRa radio/physical layer, with three gateway operation modes: Independent, Pure Pass Through, and Instruction-based. Comparison with a more traditional baseline system with push-based transfer of data reveals that our pull-based system performs better in terms of percentage of data points that are successfully sent by the end devices to the network server. We also evaluate the performance and scalability of the three modes using a small campus-wide deployment and discrete-event simulations. Experiments reveal that of the three modes, the Instruction-based mode is the most balanced in terms of computation power requirement and communication costs. In this mode, the network server handles the computation load, while the gateway handles the generation of the beacons. This shows a better result compared to the other modes wherein the computation load and beacon generation is condensed in the gateway or the network server. This also follows that the communication cost is high in the latter setup since it is dependent on the connection between the gateway and the network server. Our results also show that while utilizing multiple gateways can result in lower delays, they have to be used carefully, since the existence of multiple gateways in the system can easily result to problems with interference, possibly degrading, instead of improving, the performance of the system.