Densification leveraging mobility: An IoT architecture based on mesh networking and vehicles

Disruptive changes are underway in the automotive industry as large-scale platforms based on vehicular fleets are deployed to deliver ride sharing and delivery services. Such platforms can also be leveraged to deliver wireless connectivity services, e.g., largescale connectivity for the Internet of Things (IoT). This paper examines a network architecture based on a mesh of IoT devices, roadside repositories and vehicular mobile gateways - referred to as mesh+vehicular. We propose a system-level model to study its relative merits versus conventional infrastructure-based IoT architectures-referred to as mesh+cellular. The model reflects the salient properties of the architectures including the key interplay among the variability in the network geometries, routing trees, wireless capacity and eventually IoT queue stability. The paper provides an initial study of the scaling of the IoT sensing capacity of the routing trees per repository and base station respectively for the two architectures: i.e., the scaling the maximum common traffic rate the trees' IoT devices can generate while maintaining the stability of its queues. We then define the harvesting capacity per mobile gateway and base station in the two architectures, i.e., the average aggregate IoT rate each can extract assuming IoT devices are limited to their sensing capacity in each tree. Perhaps surprisingly, we show that as the spatial density λ_s of IoT devices and corresponding density of repositories along roads scale up, the proposed mesh+vehicular architecture has a gain in its harvesting capacity of order at least λ_s ^γ/4 where γ is the wireless path loss exponent. Underlying such gains is a fundamental shift in network geometry and information flows: in mesh+cellular systems IoT data is routed toward cells' sinks (zero-dimensional objects) while in mesh+vehicular data is routed to road induced cell edges (one-dimensional objects). Detailed system-level simulations validate the obtained scaling results.