AI-Computing Through Low-Consumption VCSEL-based Analog Radio-over-Fiber System

Next generation networks, as the 6G, aim at developing high energy efficiency systems providing unprecedented performances in terms of data transmission and latency. At the same time, Artificial Intelligence (AI) is expected to be a fundamental resource for achieving these goals, even though its use in large scale can bring to an exponential undesired huge power consumption. This poses important challenges for which regards the infrastructure realization which has to be investigated. Among the different technologies, Analog Radio-over-Fiber (ARoF) based on Vertical Cavity Surface Emitting Lasers (VCSELs) is one of the most promising ones, especially in the front-haul connection to provide lower latency, higher bandwidth with low energy consumptions where the nonlinear behavior represents one of the major challenges for its development. In this work it is shown how such non-linear behavior can be actually exploited and turn into an advantage for integrating AI features within the ARoF infrastructure, which is directly used for computation, performing in this way data processing through the communication system itself.