查看文章

Edge and fog computing, prominent parts of the upcoming 5G mobile networks and future 6G technologies, promise to reduce applications’ latencies, improve controls on privacy, and reduce network bandwidth usage. The promises are delivered by pulling computations from the remote cloud to close to the devices, where data is generated and applications are used. In stark contrast, current artificial intelligence (AI) and in particular machine learning (ML) methods mostly assume all computations are conducted in a homogeneous cloud with ample computational and data transmission resources available. The integration of edge computing with AI/ML methods–an endeavour we call EdgeAI–promises to improve both fields in a variety of aspects. Our research aims to identify the challenges and detail the potential benefits of EdgeAI, building a coherent and overarching vision of what distributed artificial intelligence means in the context of edge and fog computing. Further, we aim to find the methods of realizing those benefits, testing our hypotheses in a real-world setting on the edge-based computational platform we’re building upon the 5G test network (http://5gtn. fi). The vision will be realized during the 8-year time span of the 6Genesis Flagsip research program. Bringing edge computing and AI/ML together is challenging due to the fundamental difference in the premises of AI and edge computing. While edge computing is by design distributed and fog leans towards decentralization, modern AI/ML methods are only beginning to allow for distributed, let alone decentralized, computations. Further, intermittent connectivity may corrupt or slow down …