Artificial Intelligence in Smart Cities A Systematic Review of Applications Technologies and Deployment Patterns

This study conducts a systematic literature review of artificial intelligence (AI) applications in smart cities, covering publications from 2015 to 2024. Following the PRISMA protocol, a total of 1260 documents were initially retrieved from Web of Science, Scopus, IEEE Xplore, and SpringerLink, of which 180 peer-reviewed papers were retained for analysis after multi-stage screening and quality control (Cohen's Kappa = 0.86). The reviewed studies were categorized across six application domains-traffic optimization, energy management, public safety, environmental monitoring, waste management, and citizen engagement-and mapped to four major AI technologies, namely machine learning, deep learning, computer vision, and natural language processing. The results show that traffic and energy account for more than 40% of the selected studies, with deep learning and graph neural networks achieving an average improvement of 12-18% in predictive accuracy compared to baseline models. Computer vision dominates public safety applications, while natural language processing is primarily adopted in citizen engagement. Environmental monitoring and waste management, though emerging, exhibit a 25% annual growth rate in publications since 2020, reflecting their increasing relevance. Despite these advances, challenges remain regarding cross-domain integration, data heterogeneity, model interpretability, privacy, and governance. This review contributes a structured synthesis of application trends, methodological choices, and deployment patterns, offering both a foundation for future academic research and practical insights for policy and decision-making in smart city development.