Low-Latency In-Band Integration of Multiple Low-Power Wide-Area Networks

Today, industrial and agricultural Internet of Things (IoT) are emerging in very large-scale and wide-area applications (e.g., oil-field management, smart farming) that may spread over hundreds of square miles (e.g., $45\mathrm{m}\mathrm{i}\times 12\mathrm{m}\mathrm{i}$ East Texas Oil-field). Although a single Low-Power Wide-Area Network (LPWAN) covers several miles, it faces coverage challenge in such extremely large-area IoT applications, specially in rural or remote areas with no/limited infrastructure, requiring an in-band integration of multiple LPWANs. To avoid the crowd in the limited ISM band and the cost of licensed band and infrastructure, SNOW (Sensor Network Over White spaces) is an LPWAN architecture over the TV white spaces. It offers high scalability through concurrent and bi-directional communication between a base station and numerous nodes. We consider a seamless integration of multiple SNOWs. Existing approach does not consider minimizing network latency and is less suitable for delay-sensitive or real-time applications. We propose the first scalable in-band integration of multiple SNOWs that minimizes network latency. By taking into account the impact of bandwidth on latency and base station power dissipation, we formulate lowlatency integration of multiple SNOWs as a constrained spectrum allocation problem. It is solved through a greedy algorithm by analyzing network latency and by adopting a latency- and traffic- aware bandwidth allocation along the links to achieve an integrated network. We have implemented the proposed integration both on SNOW hardware and in NS-3 simulator. Both physical experiments and simulations show a significant reduction (44% and 97%, resp.) in network latency under our approach compared to existing approach.