Abstract : Edge intelligence is one of the key paradigms related to the efficient implementation of future wireless networks. In this context, the surrounding telecommunication infrastructure is intended to support the functioning of the wireless system. Thus, it is necessary to offload some computational efforts to the system edge, allowing the infrastructure to learn and make prospective decisions. However, for the reliable realization of the edge intelligence concept, the wireless system architecture should be designed properly to minimize both storage cost and induced latency. In this paper, we compare three architecture proposals (centralized, distributed, and hybrid) tailored to the autonomous driving use case, being one of the key vertical scenarios in contemporary and future wireless networks. We evaluate the performance of an autonomous platooning system, where the operating frequency is selected dynamically with the support of infrastructure to minimize the overall interference level in the whole band. Extensive computer simulations have been carried out to analyze the impact of the induced delay in signal processing and storage efficiency.
Index terms : Context awareness, edge intelligence for vehicular spectrum access, spectrum sharing, system architecture, V2X communications and autonomous driving.