Federated learning for IoT and AIoT: applications, challenges and perspectives

Kh.M. Eleev

Abstract. This paper discusses the concept of federated learning (FL), a distributed collaborative approach to artificial intelligence (AI) that enables AI training on distributed IoT devices without need for data sharing. Approaches and methods for implementing FL for AIoT devices have been classified into three types of federated learning architecture for organizing interactions between learning participants, centralized, decentralized, and hybrid. Approaches based on different technologies such as Knowledge Distillation, blockchain, wireless networks like Mesh, Hybrid-IoT, DHA-FL are considered. For each technology considered, the main advantages, problems and challenges are outlined. The paper sums up with conclusions about the prospects of FL development for IoT and AIoT.

Keywords: Internet of things (IoT), federated learning (FL), artificial intelligence of things (AIoT), blockchain, architecture

For  citation.  Eleev Kh.M.  Federated  learning  for  IoT  and  AIoT:  applications,  challenges  and  perspectives. News  of  the  Kabardino-Balkarian  Scientific  Center  of  RAS. 2024. Vol. 26. No. 2. Pp. 26–33.  DOI: 10.35330/ 1991-6639-2024-26-2-26-33