DeAlmeida, Jonathan M.DaSilva, Luiz A.Both, Cristiano BonatoRalha, Celia G.Marotta, Marcelo A.2022-01-042022-01-042021-06-071556-6072http://hdl.handle.net/10919/107371Cloud and fog-based networks are promising paradigms for vehicular and mobile networks. Fog Radio Access Networks (F-RANs), in particular, can offload computation tasks to the network edge and reduce the latency. Artificial Intelligence (AI) techniques can be used in F-RANs to achieve, for example, enhanced energy efficiency and increased throughput. Nonetheless, the appropriate technique selection must consider the different time granularities at which decision-making occurs in F-RANs. We discuss the benefits and challenges of implementing an AI-driven F-RAN considering different timescales, highlighting key Machine Learning (ML) techniques for each granularity. Finally, we discuss the challenges and opportunities to integrate different ML solutions in F-RANs.Pages 137-148application/pdfenIn CopyrightNetworking & TelecommunicationsArticle - Refereed2022-01-04https://doi.org/10.1109/MVT.2021.30784171631556-6080