AI-Augmented Resource Management in Edge, Fog, and Cloud Computing Systems

Abstract

Artificial Intelligence (AI) is redefining resource management across edge, fog, and cloud computing systems by enabling dynamic, predictive, and autonomous decision-making. This paper explores emerging AI-augmented strategies designed to optimize latency, energy consumption, workload distribution, and quality of service (QoS). Traditional heuristic-based algorithms, while foundational, often fall short in handling heterogeneous, dynamic environments characterized by variable loads and tight latency constraints. AI models—ranging from Support Vector Machines (SVM) and reinforcement learning (RL) to clustering and regression techniques—have demonstrated superior adaptability through workload prediction, anomaly detection, and optimized resource provisioning. For instance, dynamic resource allocation using ML approaches like k-means clustering for anomaly detection and RL for task placement have shown promising results in fog/edge contexts ScienceDirect. Frameworks such as ENORM have tackled auto-scaling edge resources, reducing latency by 20–80% and network traffic by up to 95% arXiv. Survey studies underscore the evolution of fog/resource management solutions and AI-driven enhancements across computing layers arXiv+2arXiv+2. While AI methods enable real-time scaling, predictive scheduling, and QoS-aware load balancing, challenges persist—most notably regarding scalability, interpretability, and computational overhead. This paper synthesizes state-of-the-art approaches, outlines architectural workflows, evaluates benefits and constraints, discusses implementation results, and proposes future directions to harness AI for next-generation, energy-efficient, and low-latency distributed resource management.