Multi-Access Edge Computing for Real-Time AI and IoT Applications

Abstract

Multi-Access Edge Computing (MEC) places computation, storage, and service capabilities close to users and connected devices. This proximity supports latency-sensitive Artificial Intelligence (AI) inference and Internet of Things (IoT) services while reducing backhaul traffic and dependence on distant clouds. This paper presents a structured review of MEC architecture, enabling technologies, application domains, evaluation metrics, and deployment challenges. The review follows a transparent source-selection and thematic-coding process and compares representative studies across IoT integration, automated deployment, mobility-aware handover, security, dependability, and performance. The synthesis shows that MEC is most effective as part of a cloud-edge-device continuum: devices perform sensing and lightweight preprocessing, MEC nodes execute time-critical inference and local control, and cloud platforms support large-scale training, storage, and coordination. The main unresolved issues are resource-aware orchestration, seamless mobility, interoperability, secure multi-tenant operation, dependable service delivery, energy efficiency, and distributed AI model management.