Federated Edge Computing for Edge-Assisted Consumer Electronics
Aparna Kumari, Pronaya Bhattacharya, Ashwin Verma, Zhu Zhu, Thippa Reddy Gadekallu
Abstract
The modern consumer electronics (CE) has shifted computation and data processing from centralized remote servers to decentralized local edge devices. Federated edge computing (FEC) is a recent gamechanger in edge-assisted CE, which has allowed predictive analysis on CE devices, such as wearables, smart watches, sensors, and gaming consoles at the edge network, which is specially 1–2 hops to the device itself. FEC collaboratively trains models using machine learning and deep learning, sharing only the updated model weights (gradients) instead of private consumer data, thus enhancing privacy on edge nodes. This article analyzes the applications of FEC in edge-assisted CE, its centralized and decentralized versions, the key components and technologies to support FEC, implementation strategies, and types of learning paradigms. Further, a key focus on the open challenges and future directions is presented. This article highlights FEC transformative potential in edge-assisted CE, paving the way for future innovations in the domain.