Optimizing Energy Efficiency in Edge Computing: A Machine Learning Approach

Abstract

Edge computing has become an essential paradigm for processing data near its source, minimizing latency, and improving real-time decision-making in various applications, including the Internet of Things (IoT), autonomous systems, and smart settings. Nonetheless, the energy requirements of edge computing are daunting considering that edge devices are resource-limited. It discusses the application of machine learning techniques for energy efficiency optimization in edge computing by utilizing intelligent resource management, dynamic task offloading, adaptive power management, and predictive energy modeling. Leading state approaches in energy-aware machine learning, including deep reinforcement learning, regression-based optimization, and federated learning, show great potential to reduce power consumption while ensuring adequate system performance. They have great potential, but extending machine learning models to edge environments brings challenges such as computational overhead, real-time processing constraints, as well as security concerns. Addressing these limitations calls for lightweight, hardware-efficient algorithms, cross-domain collaboration, and integration with emerging technologies like 5G and IoT. This study reviews the recent state-of-the-art developments and the main challenges to facilitate energy efficient edge computing through machine learning. This research helps contribute to sustainable edge infrastructures capable of hosting scalable, elastic, high-performance applications through the synergy of intelligent computing and energy management.