Edge Artificial Intelligence (Edge AI) and Federated Learning (FL) have emerged as the two dominant methodological trends reshaping how AI and Internet of Things (IoT) systems are deployed across urban engineering domains, yet existing reviews continue to organize the field by application sector rather than by these transversal deployment paradigms. This review addresses the gap through a cross-domain synthesis of Edge AI and FL implementations spanning transportation, environmental, and building engineering, drawing on peer-reviewed literature published between January 2021 and February 2026. A structured search protocol applied across IEEE Xplore, Scopus, ScienceDirect, ACM Digital Library, and Web of Science identified an initial corpus that, after screening against pre-registered inclusion criteria and a PRISMA 2020 selection flow, yielded over 200 studies for inclusion. The review makes four contributions: (1) a transversal taxonomy that classifies smart city AI implementations along axes of computation locality (cloud, fog, edge, on-device) and learning paradigm (centralized, federated, hybrid), exposing methodological convergence patterns invisible in domain-siloed reviews; (2) a five-layer reference architecture explicitly annotated with Edge AI and FL placement decisions; (3) a comparative synthesis of deployment outcomes across more than thirty international case studies, contrasting infrastructure-intensive and resource-constrained settings; and (4) an evidence-based research agenda identifying eight open problems where Edge AI and FL maturity remains insufficient for production-grade urban deployment. The findings indicate that Edge AI delivers consistent latency and energy benefits in transportation and building applications where on-device inference is feasible, while FL is most impactful where data sovereignty, regulatory constraint, or cross-jurisdictional coordination dominate the problem geometry, particularly in environmental monitoring and multi-stakeholder energy systems. Implementation challenges persist in model drift under non-IID urban data, communication efficiency under intermittent connectivity, and the absence of standardized fairness and explainability tooling for distributed urban AI. This work provides actionable guidance for civil engineers, urban planners, technology providers, and policymakers selecting between centralized, edge, and federated deployment strategies for next-generation urban infrastructure.

Soiling, or dust accumulation, is a significant challenge affecting the performance and reliability of photovoltaic (PV) systems in arid regions. This paper introduces Fusion for Understanding Soiling Events (FUSE), a data-driven framework for classifying atmospheric dust-loading conditions associated with potential PV soiling risk. The framework employs a dual-level fusion architecture. The first level integrates environmental information from globally available datasets, including solar radiation, meteorological variables, and atmospheric particulate matter, for two Egyptian locations, Cairo and Aswan. The second level combines machine learning classifiers through weighted soft voting, with weights determined by cross-validation F1-scores. The framework was trained on observations from 2019 to 2022 and evaluated on an independent 2023 temporal holdout from the same two locations. For same-day environmental risk assessment, FUSE achieved a weighted F1-score of 91.65% and a ROC-AUC of 0.9324, outperforming Random Forest in weighted F1 while showing comparable performance to XGBoost and LightGBM. Wind and particulate matter variables were among the most informative predictor groups, while solar variables contributed relatively little. At the default classification threshold, the ensemble identified 53% of the higher-dust-loading proxy events, increasing to 68% using a decision threshold selected from training data. The primary framework provides retrospective same-day assessments after the required environmental observations become available. A separate next-day prediction experiment using preceding-day environmental features achieved a weighted F1-score of up to 90.77% on the 2023 test set. The resulting classifications represent AOD-derived atmospheric dust-loading proxy labels rather than independently measured PV module soiling. Although the framework may be adaptable to other regions with comparable data availability, geographic transferability beyond Cairo and Aswan has not been evaluated. These findings suggest that environmental data fusion may contribute to future condition-based PV maintenance and Prognostics and Health Management applications, subject to independent module-soiling validation and techno-economic assessment.