IoT, Edge, and Cloud Computing in Smart Cities

1. Introduction

The Internet of Things (IoT), edge computing, and cloud computing are essential drivers of the smart city concept, enabling urban-scale sensing, immediate decision-making, and data-driven services. Their combined deployment is increasingly impacting city-wide cyber–physical systems, demanding the seamless operation of a variety of devices, networks, and computing infrastructures, frequently under demanding constraints of latency, reliability, privacy, and security. This Special Issue emphasizes the integration and mutual enhancement of these technologies within contemporary urban settings, underscoring both fundamental advancements and practical applications.

This collection, aligned with Future Internet’s “Network Virtualization and Edge/Fog Computing” Section, encompasses topics such as cybersecurity, edge-to-cloud integration, data analytics, intelligent transportation systems, federated learning, cyber–physical systems, and sustainable urban development.

The published contributions incorporate all aspects of the field, from resource management and distributed intelligence at the peripheral level, to secure and privacy-preserving mobility services, to computer vision-based monitoring of urban infrastructure, while also providing wider architectural insights through survey and review studies.

2. Contributions

In the first paper, Rakin et al. [1] explore AI-assisted infrastructure monitoring by employing a YOLOv11-based object detection pipeline to detect urban infrastructure defects from images. A curated dataset and benchmarks against previous YOLO versions are used to illustrate how vision-based automation can enhance and partially scale inspection processes in urban environments, thereby fostering proactive maintenance and enhancing public safety.

Wang et al. [2] introduce CPCROK, a communication-efficient pseudonym-changing scheme designed for low-density VANET environments where conventional mix-zone strategies may be ineffective. The solution integrates Kalman filter-based coarse trajectory estimation, RNN-augmented prediction, and trajectory generation to create realistic fake trajectories while minimizing transmission overhead, thereby facilitating privacy-preserving intelligent transportation services.

Manolov et al. [3] provide a pragmatic comparison of Azure DevOps and GitHub as CI/CD platforms, analyzing automation workflows, scalability, security and compliance attributes, and financial implications. As smart city applications increasingly depend on intricate, perpetually advancing cloud–edge software architectures, such engineering-focused analyses offer valuable guidance to organizations for choosing toolchains that facilitate dependable and secure implementations.

Trigka and Dritsas [4] examine the function of edge and cloud computing in smart cities, addressing architectural frameworks, enabling technologies, application domains, and unresolved research challenges. Their survey delineates essential trade-offs, such as minimizing latency via localized processing versus extensive analytics through centralized resources, thus providing a framework for developing resilient, distributed, and sustainable smart city services.

Souza et al. [5] present a review of collaborative computing architectures in the Internet of Everything epoch, contrasting edge computing, cloud computing, fog computing, and blockchain/web services via the 3C collaboration model (communication, cooperation, and coordination). In the review, the authors emphasize the complementary roles of edge and cloud computing in facilitating low-latency processing and scalable coordination, insights particularly pertinent to the design of human-in-the-loop and multi-stakeholder smart city systems.

Abdelmoniem et al. [6] propose a decentralized, client-driven, and model-centric framework for scalable edge AI, driven by privacy considerations and the practical limitations of centralized and federated methods in heterogeneous environments. The architecture prioritizes model discovery and distillation to facilitate efficient model sharing among entities, providing a different approach to enhancing edge intelligence in IoT-dense environments.

Khanafer et al. [7] introduce an enhanced adaptive backoff algorithm (I-ABA) for IEEE 802.15.4 networks, motivated by performance degradation in extensive deployments when contention windows are limited. The proposed method employs collision-probability-aware adaptation and curve-fitted modeling of contention window dynamics to enhance scalability in dense sensor network environments relevant to health and wearable applications in smart city ecosystems.

Xie et al. [8] introduce DyGPP for predicting passenger behavior, representing passenger–station interactions as a heterogeneous dynamic graph to capture correlations frequently overlooked by sequential models. DyGPP enhances predictive accuracy on real-world datasets by analyzing temporal patterns of both passengers and stations and correlating their dynamics, which enables more intelligent transit planning and risk-aware operations in urban mobility systems.

Mahbub et al. [9] present FIVADMI, a framework for detecting anomalies in in-vehicle communications, tailored to the limitations of automotive ECUs and networks. The framework facilitates real-time monitoring and integrates system-level isolation by using Trusted Execution Environments, while addressing side-channel vulnerabilities and conforming to AUTOSAR design principles for interoperability and reuse.

Šatkauskas and Venčkauskas [10] address the volatility and heterogeneity of fog environments through a multi-agent fog service placement orchestrator that executes two-stage optimization while facilitating dynamic service discovery and resource monitoring. The authors also address operational robustness by facilitating signed and encrypted messaging between fog nodes, thereby ensuring reliable orchestration for latency-sensitive urban services.

Jamal et al. [11] propose a hybrid multi-agent reinforcement learning framework for spectrum sharing in vehicular networks, integrating centralized and decentralized learning through DQN-based agents coordinated by QMIX. The methodology is focused on practical V2V/V2I spectrum reuse scenarios and is corroborated by simulations, demonstrating enhancements in critical communication performance metrics.