Latency-Optimized Data Harvesting in LoRa Networks Using a Mobile Sink
Conference
Saifullah, A, Ahmed, N, Jain, A et al. (2026). Latency-Optimized Data Harvesting in LoRa Networks Using a Mobile Sink
. Proceedings of the IEEE Real-Time and Embedded Technology and Applications Symposium, RTAS, 360-372. 10.1109/RTAS68450.2026.00039
Saifullah, A, Ahmed, N, Jain, A et al. (2026). Latency-Optimized Data Harvesting in LoRa Networks Using a Mobile Sink
. Proceedings of the IEEE Real-Time and Embedded Technology and Applications Symposium, RTAS, 360-372. 10.1109/RTAS68450.2026.00039
Remote IoT deployments in challenging terrains increasingly rely on LoRa and mobile sinks (e.g., drones) for data harvesting. Nodes produce heterogeneous, time-varying, high-volume data that must be uploaded within a brief contact window as the mobile sink traverses the network. Therefore, a critical requirement for data harvesting in remote LoRa deployments is to optimize latency in collecting data when a mobile sink arrives. Existing work mostly focuses on LoRa networks with in situ gateways; optimizing data harvesting latency with mobile sinks remains an unaddressed challenge. In this paper, we propose a novel data harvesting protocol that achieves near-optimal latency in remote LoRa deployments using a mobile sink equipped with a LoRa gateway. Our protocol leverages the parallel reception capabilities of a LoRa gateway by grouping nodes to confine collisions within each group, each group being assigned a unique channel-spreading factor pair. To optimize network latency under varying data volumes across nodes, we formulate node grouping as an optimization problem, prove its NP-hardness, and propose a novel 2-approximation algorithm, which the protocol adopts on the fly for concurrent transmission scheduling across groups. This work presents the first provable constant approximation bound on latency for a LoRa network. Our small-scale experiments show that the latency achieved through our approach can be as close as 1.07 times the optimal latency. Additionally, large-scale simulations in NS-3 demonstrate that our approach reduces latency by up to 80% compared to a state-of-the-art approach, while also decreasing energy consumption by 30%.
