Leveraging cloud intelligence for hybrid vehicular communications

Vehicular networks are required to meet strict performance criteria to reliably support safety-critical applications at scale. However, wireless communication technologies that are currently available for vehicles usually have unique sets of limitations in terms of latency, throughput, and/or signal propagation characteristics. A promising solution to break through these limitations is hybrid vehicular networking, in which on-board units in vehicles concurrently make use of multiple types of radios (e.g., DSRC, Wi-Fi, LTE, etc.), dynamically selecting the most suitable network interface(s) according to the performance requirements and network conditions. In this paper, we explore the potential of a quasi-centralized approach to hybrid vehicular networking. Vehicles opportunistically report their positions and channel load measurements to a cloud-based central server when they have access to network infrastructure (e.g., roadside units). Analyzing the wide-scale statistics on road and network conditions, the server derives and distributes recommended network selection strategy for each geographical region, allowing vehicles to make globally-optimized network selection. Through a case study and proof-of-concept simulations, we show the basic feasibility of the concept of cloud-assisted hybrid vehicular networking.