AI-Powered Edge Analytics for Smart Cities Enabling Real-Time Data Processing, Privacy Preservation, and Sustainable Urban Intelligence

Abstract

Smart cities generate massive volumes of sensor, video, and telemetry data that must be processed with low latency, privacy guarantees, and limited bandwidth. AI-powered edge analytics — running machine learning models on devices at or near the data source — enables real-time decision making, bandwidth reduction, and improved privacy compared with cloud-only systems. This paper surveys recent advances in edge AI and federated learning for urban applications, analyzes architectural and deployment trade-offs, describes representative existing systems, and proposes a modular, privacy-aware edge analytics framework for smart cities. We present an experimental setup and simulated results demonstrating latency, bandwidth, and accuracy trade-offs, discuss open challenges (security, model updates, fairness, regulation), and conclude with recommended research directions.