Smart Cities: A Systematic Review of Emerging Technologies

Highlights

What are the main findings?

• Smart cities are increasingly enabled by the convergence of the IoT, the computing continuum, and artificial intelligence.
• The computing continuum (edge–fog–cloud–HPC) plays a central role in handling data locality, scalability, and responsiveness.

What is the implication of the main finding?

• The integration of emerging paradigms into smart cities enables real-time, citizen-centric, and sustainable urban services.
• Policies and governance must evolve alongside technology to address ethical, privacy, and digital divide concerns.

Abstract

In the 21st century, rapid urbanisation has brought both challenges and opportunities. Smart cities have emerged as innovative solutions to meet the complex demands of urban life. Information and Communication Technology (ICT) serves as the backbone of this transformation, integrating infrastructure, public services, and environmental sustainability. Within ICT, the computing continuum has become a key paradigm for efficient resource management, while Artificial Intelligence (AI) and the Internet of Things (IoT) enhance urban planning, optimise resource use, and strengthen governance. This paper systematically reviews smart city developments from January 2020 to June 2025, focusing on technological advances and sustainability goals in databases such as Scopus, IEEE Xplore, and Web of Science. By synthesising the literature, it identifies common challenges, implementation strategies, and future directions. The review highlights the central role of the computing continuum and AI, covering enabling technologies, applications, case studies, and deployment challenges. Our findings indicate that the IoT, AI, and data analytics are currently dominant approaches, yet significant gaps remain in citizen participation, equitable access, and long-term governance. Overall, smart cities aim to integrate data, digital technologies, and intelligent infrastructure to improve the quality of life while promoting sustainable, resilient, and inclusive services.

1. Introduction

The world is undergoing a profound urban transformation in the 21st century. Rapid urbanisation brings both challenges and opportunities, and smart cities have emerged as sustainable, innovative responses to these demands. This trend is accompanied by the growing complexity of managing urban environments. Information and Communication Technology (ICT) has become a key enabler, transforming traditional cities into smart, connected, and responsive ecosystems [1].

ICT serves as the backbone of this revolution. As urban populations continue to expand, cities face increasing pressure on their infrastructure, public services, and environmental sustainability. Smart cities utilise ICT to interconnect their infrastructure, government, and citizens, thereby creating data-informed and efficient systems.

Within ICT, the computing continuum has become a cornerstone for building large-scale systems that support the deployment of smart cities [2]. In this paradigm, infrastructure and services are organised in a multi-layered, hierarchical architecture to handle the constant growth of data and users, while bringing computation closer to data sources. Typically, these layers encompass the edge, fog, and cloud [3,4]. Through this structure, smart cities can establish hyper-distributed environments that support large-scale data processing and storage.

The transformation into a smart city yields numerous benefits [5]: improved public services through automated systems that monitor utilities and allocate resources efficiently (e.g., optimising electricity distribution with smart grids); enhanced mobility through traffic management systems that reduce congestion and provide real-time transport tracking; greater environmental sustainability through smart energy management, waste reduction, and pollution monitoring; economic growth as technology-driven hubs attract investment and foster entrepreneurship (e.g., start-ups in the IoT, clean technology, and AI); and stronger governance and civic engagement, as digital platforms enable participatory governance by allowing citizens to report issues, contribute to policymaking, and access services online.

Figure 1 shows the main application domains of smart cities, which leverage these benefits [6,7]. The domains include the following:

• Smart Governance: Transparent, participatory, and data-driven decision-making. Digital platforms streamline citizen interaction with government services. ICT tools enable e-voting, digital licences, and online grievance redressal systems, promoting transparency and efficiency.
• Smart Mobility: Efficient, clean, and integrated transportation systems. ICT-based solutions encompass intelligent transport systems (ITS), GPS vehicle tracking, and app-based integration of public transport. Real-time data help manage traffic and provide commuters with live updates.
• Smart Environment: Sustainable resource management and environmental protection. ICT systems monitor and manage energy grids, water distribution, and waste disposal. Smart meters and grid automation reduce losses and ensure efficient use of resources.
• Smart Healthcare: High-quality health services supported by technology. Telemedicine, remote diagnostics, and ICT-powered digital platforms improve access to healthcare, especially during emergencies such as pandemics.
• Smart Public Safety: ICT-based video surveillance, facial recognition, and emergency response systems enhance urban safety. AI-driven analytics support law enforcement and disaster management.
• Smart Society: Educated, creative, and digitally empowered citizens. Innovation-driven economic development is fostered through the ICT resources of smart cities.

There is no universally accepted definition of a smart city. Broadly, smart cities are urban areas that integrate digital technology and data-driven solutions to enhance performance, improve well-being, and reduce resource consumption [8]. Most interpretations converge on the integration of digital technologies into urban management. According to the International Telecommunication Union (ITU), a smart city is “an innovative city that uses ICTs and other means to improve quality of life, efficiency of urban operations and services, and competitiveness, while ensuring that it meets the needs of present and future generations with respect to economic, social, and environmental aspects” [9].

In this context, sustainability is a primary goal of smart cities. In line with the United Nations Sustainable Development Goals (UN SDGs) [10], sustainability is defined as “meeting the needs of the present without compromising the ability of future generations to meet their own needs” [11]. For smart cities, this means examining how digital technologies help transition towards more inclusive, resilient, and environmentally sustainable urban systems, while preparing for next-generation challenges [12].

At the core of smart cities lies the integration of ICT systems, which support the achievement of UN SDGs. These include cloud computing, the computing continuum, big-data analytics, broadband networks, sensor networks, and mobile communication. ICT enables real-time data collection, analysis, and decision-making to enhance services such as traffic control, energy management, and public safety [13]. More recently, Artificial Intelligence has taken on a central role in developing smart city tools and exploiting massive datasets across multiple sectors [14].

This paper systematically reviews the literature published from January 2020 to June 2025, focusing on smart city technologies and sustainability. This period was chosen based on publication trends in smart city research between 2014 and 2024, during which about 60% of papers appeared between 2020 and 2024. This scope was extended to include the first half of 2025 to capture emerging research. Our review examines the evolution of smart cities over the last five years, highlighting key technological advances, policy frameworks, sustainability goals, and societal impacts. Particular emphasis is placed on ICT-centred components, including enabling technologies, applications, case studies, and implementation challenges [15]. By synthesising the peer-reviewed literature, the review identifies common challenges, implementation strategies, and future directions, with a focus on the technologies driving smart city development, the benefits they provide, the challenges they pose, and the trends they are likely to follow.

The remainder of this paper is organised as follows. Section 2 outlines the methodology used to conduct this systematic review. Section 3 examines the main technologies and advances in smart cities from a computing continuum perspective. Section 4 presents case studies of real-world deployments. Section 5 provides a discussion of the findings. Finally, Section 6 summarises the main conclusions and highlights avenues for future work.

2. Materials and Methods

We conducted this systematic review in accordance with Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines [16], with the completed PRISMA 2020 Checklist provided in the Supplementary Materials (27-item checklist), and the PRISMA 2020 extension for abstracts provided in the Supplementary Materials (12-item checklist). We chose PRISMA over other methodologies (e.g., the Cochrane Framework) because it aligns with the interdisciplinary scope of smart city research and emphasises transparency, reproducibility, and replicability.

To develop this survey, we conducted a Systematic Literature Review (SLR) on technologies enabling smart cities, with a particular focus on the computing continuum. The literature encompasses numerous studies that present various technologies applied to smart city initiatives, each addressing distinct challenges. Our review synthesises these contributions to identify the main limitations and challenges of commonly used technologies, as well as to highlight opportunities that could mitigate them.

The first eligibility criterion required peer-reviewed articles written in English and published within the last five years (2020–2025). We prioritised articles published in Q1 and Q2 journals, as well as papers presented at CORE Conference Rankings A and B conferences. We excluded articles without a publicly accessible full text. We included conference papers because they often present cutting-edge innovations and emerging trends relevant to our study. During screening, we also considered external records (websites and reports) of relevant tools and frameworks. This allowed us to distinguish two corpora: the core academic corpus and the contextual corpus.

We retrieved the academic corpus from major databases, including Scopus, Web of Science, IEEE Xplore, and the websites of relevant journals. All articles were selected using the queries in

Table 1. Queries performed and number of results obtained.

ID	Query	Number of Publications (2020–2025)
		Scopus	IEEE Xplore	WoS
Q1	(TITLE-ABS-KEY("smart city") OR TITLE-ABS-KEY("smart cities")) AND PUBYEAR > 2019 AND (LIMIT-TO(SRCTYPE, "j") OR LIMIT-TO(SRCTYPE, "p")) AND (LIMIT-TO(DOCTYPE, "ar") OR LIMIT-TO(DOCTYPE, "cp")) AND LIMIT-TO(PUBSTAGE, "final") AND LIMIT-TO(LANGUAGE, "English")	25,443	19,452	31,317
Q2	(TITLE-ABS-KEY("smart city") OR TITLE-ABS-KEY("smart cities")) AND (TITLE-ABS-KEY(technology) OR TITLE-ABS-KEY(applications) OR TITLE-ABS-KEY(analytics) OR TITLE-ABS-KEY("artificial intelligence") OR TITLE-ABS-KEY("computing continuum")) AND PUBYEAR > 2019 AND (LIMIT-TO(SRCTYPE, "j") OR LIMIT-TO(SRCTYPE, "p")) AND (LIMIT-TO(DOCTYPE, "ar") OR LIMIT-TO(DOCTYPE, "cp")) AND LIMIT-TO(PUBSTAGE, "final") AND LIMIT-TO(LANGUAGE, "English")	16,288	15,170	26,225

WoS: Web of Science.

. Query Q1 searched for all papers containing the terms smart city or smart cities, while Query Q2 refined the search by requiring one of the following: technology, applications, analytics, artificial intelligence, or computing continuum. On average, Q1 yielded 23,228 results and Q2 yielded 18,920.

Figure 2 shows the number of entries obtained from both queries. For this analysis, we extended the time period to 2014–2024 in order to illustrate how smart cities have evolved as a research topic over the last decade. In 2014, the number of publications was relatively low across the three databases consulted (Scopus, IEEE Xplore, and Web of Science). Between 2015 and 2016, the number of papers published increased by 48%. By 2018, the number of publications in the Web of Science exceeded 8000, and since 2020, it has remained above 9000 annually. Almost 60% of the papers were published between 2020 and 2025, which is the period chosen for this review—specifically, January 2020 to June 2025.

We formulated three research questions to guide the review:

RQ1. 

What are the core technologies supporting data collection, processing, and integration in smart cities, and how can they be classified?

RQ2. 

How do emerging paradigms such as the computing continuum and artificial intelligence contribute to the development of smart city applications?

RQ3. 

What are the main challenges in the design and implementation of smart city systems, and what technologies are currently used to address them?

To address these questions, we structured the review around smart cities from the perspective of emerging paradigms such as the computing continuum and machine learning. These paradigms are crucial for managing the vast volumes of data generated by sensors and IoT devices, supporting the development of diverse smart city applications, and addressing real-world adoption challenges.

Figure 3 presents the PRISMA flow diagram [17], which outlines the process of collecting sources for this review. We retrieved records using Query Q2 from three databases: Scopus, IEEE Xplore, and Web of Science. As shown in Table 1, each database returned more than 15,000 articles. To ensure feasibility, we limited the sample to the first 1000 records per database ranked by relevance, yielding a total of 3000 records. These records formed the core academic corpus. In addition, we included a contextual corpus comprising seven reports from external sources that referenced relevant frameworks and tools, as well as four records identified through citation chaining during the full-text review.

From the 3000 records initially identified, we removed 1091 duplicates using the Zotero 7.0.24 (64 bit) software. The remaining 1909 records were screened independently by three reviewers based on titles and abstracts, and 1483 were excluded at this stage. Next, we assessed the full text of 426 articles, resulting in the exclusion of 343. The primary reasons for exclusion were that the papers were outside the scope of this review (e.g., non-technological) or did not provide evidence relevant to the research questions. In total, 83 studies were included.

3. Smart Cities: Main Technologies from an Information and Communication Technology Perspective

Smart cities are enabled by a combination of foundational technologies organised across different layers—from edge and fog infrastructures to cloud and high-performance computing (HPC) systems. This layered architecture, known as the computing continuum, is a keystone of modern Information and Communication Technology (ICT). In the literature, multiple authors describe various architectures for organising components and participants when using the continuum as the basis for smart cities. For example, Lea et al. [18] define three layers, the IoT, edge, and cloud, which collaborate to manage and transform data across devices. Similarly, Rosmaninho et al. [19] propose a multi-level architecture with a cloud layer and multiple cloud and edge nodes, supporting service deployment across locations. Kumar et al. [20] present another three-layer design (edge, fog, and cloud), each with distinct computational characteristics, services, and requirements.

Despite differences, these architectures share an n-layer design, with variations in naming and the number of layers depending on the problem addressed and the available infrastructure. These architectures are applied to smart city challenges such as extracting behavioural insights [21], identifying energy usage patterns [22], and analysing traffic dynamics [23]. The goal of multi-tier architectures is to decentralise data processing by executing tasks closer to data sources, thereby reducing transfer overhead and improving response times [19]. Figure 4 summarises these architectures. Our model employs a three-layer architecture (edge, fog, and cloud), which supports smart city services on top.

Each layer of the continuum exhibits distinct characteristics relevant to smart city design [24]. In terms of data control, edge and fog layers provide greater flexibility in where data are processed and stored. By contrast, the cloud reduces such control, as service providers typically manage placement under paradigms such as Platform-as-a-Service (PaaS) and Software-as-a-Service (SaaS). Regarding computational resources, the edge consists of vast numbers of low-capacity devices. At the same time, the cloud comprises fewer but far more powerful machines in large-scale data centres, offering virtually unlimited virtualised resources. However, the cloud usually introduces higher latency, especially relative to IoT data sources.

Across all layers in Figure 4, HPC capabilities enable the processing of large-scale data. Accelerators at the edge support real-time processing, while clusters and supercomputers in fog and cloud layers enable large-scale analytics [25]. In HPC environments, ad hoc file systems mitigate I/O bottlenecks common in data-intensive applications [26,27].

In smart cities, the computing continuum is complemented by other paradigms that enable effective data management and the generation of insights. For example, the HYPER-AI [28] and EMPYREAN projects [29,30] use the continuum as a testbed for large-scale data processing, leveraging AI for infrastructure coordination. AI is also applied to solutions for mobility, healthcare, and Industry 4.0. Flight [31], a hierarchical federated learning tool, also uses the continuum for training models organised in layers and deployed through a function-as-a-service (FaaS) model. Solutions such as HYPER-AI and Flight rely on data management tools like Globus [32], which enable ubiquitous access to data in distributed environments [24]. As shown in Figure 5, we identify five key pillars of the computing continuum in smart cities: the Internet of Things (IoT), the computing continuum itself, AI, data storage, and sustainability, which acts as a transversal layer for meeting smart city requirements and goals [33].

At the top of Figure 5, the IoT encompasses a wide range of devices (e.g., wearables, traffic cameras, and environmental sensors) together with the protocols, tools, and standards that interconnect physical objects via networks [18]. These devices integrate multiple sensors to capture diverse types of data [34]. For example, wearables record vital signs [35], while vehicles collect proximity and location information [36]. IoT devices can be integrated into smart city systems to enhance decision-making, such as coordinating traffic lights to reduce congestion, monitoring energy consumption for demand–response strategies, or tracking air quality [37].

The third pillar in Figure 5 is AI. Advances in the computing continuum and HPC have accelerated the adoption of AI techniques [38]. AI systems perform prediction, classification, optimisation, and decision-making tasks, reducing the need for human intervention [39]. Common approaches include supervised learning, unsupervised learning, and reinforcement learning [40]. Supervised learning trains models on labelled data; unsupervised methods discover patterns in unlabelled data; and reinforcement learning optimises decisions by interacting with an environment to maximise rewards. More recently, federated learning has emerged as a paradigm for distributed model training across devices, addressing scalability and privacy concerns associated with centralised learning [41].

Given the vast quantity of data produced in smart cities, data storage technologies are critical for efficient management [42]. Therefore, we consider storage as a key pillar in Figure 5. It includes local storage on IoT and edge devices, as well as long-term preservation across fog and cloud infrastructures [24]. Storage systems must support availability through replication, resilience via erasure coding, performance through load balancing, and low-latency access via caching strategies [43]. They must also comply with security and privacy regulations by incorporating access control, encryption, and policy enforcement [44].

Finally, sustainability acts transversally across all smart city technologies (Figure 5). Its requirements include scalability, efficiency in energy and resources, security, resilience, and accessibility. Meeting these requirements allows technologies to contribute to multiple goals. For instance, energy and resource efficiency support responsible consumption and production by optimising continuum resources and reducing energy use. Similarly, advances in AI, the computing continuum, and data storage foster resilient digital infrastructures and promote innovation across distributed systems.

In the following sections, we review representative works related to the five pillars: the IoT, the computing continuum, AI, and data storage.

3.1. Internet of Things

The rapid development of the Internet of Things (IoT) has enabled large-scale smart city initiatives. IoT devices collect data remotely through sensors embedded in smartphones, wearables, vehicles, webcams, and traffic lights [45]. Cisco estimated that over 25 billion devices were connected to the Internet before 2020, with projections of more than 50 billion permanent and 200 billion intermittent connections [46]. Strategy Analytics further predicted 38 billion devices by 2025 and 50 billion by 2030 [47].

Figure 6 presents a taxonomy of IoT technologies for smart cities, including standards and protocols for managing connectivity and data exchange. Communication typically occurs over networks such as 5G, Wi-Fi, and Low-Power Wide Area Network (LPWAN), enabling interconnection and device-to-device messaging. The IoT also encompasses sensing devices that collect diverse types of data. While IoT focuses on data collection and device interconnection, smart cities exploit these data to address urban challenges such as reducing congestion, improving weather forecasting, and optimising public transport routes [37,48].

In mobility, IoT devices collect GPS data to optimise traffic flows and reduce energy consumption [49]. Ciuchi et al. [50] highlighted the complexity of large-scale deployments in the SmartSantander project, which integrated more than 12,000 devices for environmental sensing, parking, augmented reality, and mobility. These sensors measured parameters including temperature, humidity, rainfall, noise, air quality, and GPS data.

The IoT also contributes to infrastructure resilience and energy access. For instance, Rehman et al. [51] proposed a solar-powered portable charging system with IoT-based monitoring. Their design integrates high-efficiency solar panels, intelligent controllers, and sensor modules for real-time analytics. The system supports a wide range of scenarios, including emergency response, healthcare, outdoor recreation, and public infrastructure.

The value of the IoT lies not only in collecting data but also in transforming them into actionable insights. This transformation requires integration with complementary technologies such as cloud and edge computing. Cloud platforms offer large-scale storage and analytics, while edge computing enables near-real-time processing at the data-source level. Increasingly, sensors integrate lightweight processing to filter or pre-analyse data locally before transmission. For example, a smart temperature sensor may detect anomalies and trigger an alert immediately, without relying on cloud processing.

The combination of intelligent sensors, advanced communication protocols (e.g., 5G), and energy-efficient designs has expanded IoT applications across various domains, ranging from agriculture to densely populated urban areas. However, challenges remain in managing massive data flows and ensuring interoperability among heterogeneous devices. To address this, researchers are developing standardised communication frameworks and improved integration mechanisms [52]. Collectively, these advances demonstrate the evolution of the IoT from simple sensing into a cornerstone of intelligent systems capable of learning, acting, and improving over time [53].

The IoT provides the sensing layer for smart cities, enabling the collection of real-time data across various domains, including mobility, environment, and safety. Its effectiveness, however, depends on overcoming persistent issues of interoperability, energy efficiency, and device security. Integration with complementary technologies such as continuum infrastructures, blockchain for trust, and digital twins can help create more adaptive and resilient ecosystems.

A single system cannot manage the vast and heterogeneous data generated by IoT devices. To process these data at scale while minimising latency, smart cities increasingly rely on the computing continuum spanning edge, fog, and cloud infrastructures.

3.2. Computing Continuum: Edge, Fog, Cloud

The computing continuum—an architectural paradigm spanning from edge devices to centralised cloud infrastructures—has become a cornerstone of intelligent and responsive smart city ecosystems. By integrating computational resources across edge, fog, and cloud layers, the continuum supports the diverse requirements of urban services, including ultra-low latency, scalable data processing, and contextual adaptability. Recent research shows how this paradigm optimises service delivery, enhances decision-making, and enables applications in domains such as mobility, healthcare, logistics, and public administration.

Figure 7 presents a taxonomy of the main components of the computing continuum. Edge computing processes data directly on collection devices, enabling real-time decision-making and analysis. Fog computing, composed of micro data centres near edge devices, aggregates and stores local data. The cloud provides centralised storage and large-scale analytics thanks to virtually unlimited resources. Transversally, orchestration technologies coordinate workloads and data across environments. Depending on organisational requirements, additional layers can be added to this model.

The continuum enables dynamic allocation of computational workloads across heterogeneous environments. Unlike centralised models, it executes latency-sensitive tasks—such as video analytics or sensor fusion—close to the source, while offloading resource-intensive, non-time-critical processing to the cloud. This flexibility is crucial in smart cities, where devices generate continuous streams of data that must be processed in real time.

Belcastro et al. [2] present a representative three-tier architecture for processing mobility data. IoT-equipped vehicles transmit information to edge servers for immediate tasks such as collision detection or traffic sign recognition, while large-scale analytics (e.g., route optimisation, fleet demand prediction) run in the cloud. This hybrid approach proves effective as the number of connected vehicles grows, showing the scalability of continuum-aware designs.

Vincent et al. [54] investigate orchestration strategies for latency-sensitive workloads in smart transportation. Using a 5G-based vehicular perception case study, they demonstrate how real-time video streams can be processed at the edge and redistributed to nearby vehicles. Their orchestration system dynamically adjusts deployments based on cost, performance, and resource availability, introducing adaptive CPU tuning to optimise performance under constraints.

In logistics and sensing, Bassolillo et al. [55] conceptualise drones as mobile edge nodes within a continuum. Their system employs Ant Colony Optimisation for routing under capacity and autonomy constraints, coupled with path planning and predictive control for obstacle avoidance. Integration of drone clusters with IoT sensors and fog platforms enables real-time urban delivery, infrastructure inspection, and crowd monitoring.

Finally, Nezami et al. [56] propose a decentralised optimisation framework for service placement and migration across the edge-to-cloud continuum. By adapting to spatio-temporal mobility patterns, the system activates or deactivates services in real time to avoid deadline violations and minimise resource underutilization. Both real-world experiments and large-scale simulations confirm its ability to balance performance and cost-effectiveness in dynamic urban contexts.

3.2.1. Scalable Execution for Smart City Services

Complementing the distributed infrastructure of the computing continuum, serverless computing introduces an event-driven model that abstracts infrastructure management and enables elastic, stateless services. In smart cities, it is particularly suitable for applications with unpredictable workloads, frequent bursts of activity, and short execution times—such as public alerts, sensor data processing, and on-demand analytics.

De Maio et al. [57] show how serverless functions can be dynamically deployed across edge nodes to support latency-constrained services such as traffic alerts and augmented reality tourism. Their approach leverages mobility data and system load information to guide function placement, adapting to user location and infrastructure conditions.

Risco et al. [58] illustrate a hybrid strategy where edge devices perform anonymisation tasks (e.g., face blurring). At the same time, cloud-hosted serverless functions run deep learning models for mask detection in a public health scenario. This design combines the responsiveness of edge processing with the scalability of cloud services.

In healthcare, Cassel et al. [59] present an architecture where vital signs from wearable sensors are prioritised using a triage-inspired offloading algorithm. High-priority alerts are processed immediately by serverless functions across fog and cloud layers, ensuring timely intervention. The system incorporates dynamic provisioning to maintain service quality during demand surges, thereby contributing to life-saving outcomes in densely populated urban environments.

3.2.2. Distributed Trust for Smart City Services

While the computing continuum and serverless computing provide flexibility and responsiveness, smart cities also require mechanisms for trust, verifiability, and decentralised governance. Blockchain technologies address these needs by providing tamper-proof ledgers and programmable smart contracts, enabling transparent interactions across multiple stakeholders and systems.

Simonet-Boulogne et al. [60] integrate blockchain into a Platform-as-a-Service (PaaS) framework for automating application lifecycle management across the continuum. Their system employs a Domain-Specific Language (DSL) to specify high-level requirements, automatically selects resources, deploys services, and adapts to environmental changes. The blockchain-enhanced PaaS is validated in two industrial use cases—marketing data pipelines and digital health—demonstrating secure and adaptive service delivery.

Ullah et al. [61] examine the role of smart contracts in urban governance and service automation. They identify ten key dimensions—ranging from legal compliance to trust management—and organise them into a six-layer framework (network, transaction, blockchain, trust, application, and security/management). A decentralised application (DApp) for real-estate shows how smart contracts reduce administrative overhead and enhance transparency in transactions.

Blockchain’s civic potential is highlighted by Chentouf et al. [62], who developed a decentralised e-voting platform using Ethereum. Smart contracts written in Solidity interact with client-side applications via Ethereum wallets, eliminating vulnerabilities of centralised voting databases. This design ensures immutability, enforces single-vote casting, and provides verifiable tallies—aligning blockchain’s transparency, immutability, and decentralisation with democratic principles [63,64].

For cross-domain communication, Xu et al. [65] introduce a hierarchical blockchain system optimised for scalability and low overhead. Local sub-chains handle intra-domain communication and access control, while a global chain guarantees inter-domain consistency. A lightweight consensus protocol dynamically manages node credibility and mitigates malicious activity using a Byzantine fault-tolerant approach. Formal threat modelling confirms the system’s suitability for low-power devices interacting across distributed urban environments.

3.2.3. Security Across the Continuum

Securing the continuum requires end-to-end mechanisms spanning data privacy, secure computation, authentication, and fault tolerance. As urban systems grow in scale and complexity, the risks of data breaches, identity theft, and infrastructure manipulation increase proportionally. Robust protections must therefore be integrated from devices to the cloud, and from networks to applications.

Hou et al. [4] propose a distributed forecasting system for smart energy grids, in which raw data from smart meters remain local, and only protected outputs are shared. Techniques such as model randomisation, re-encryption, and shuffle schemes safeguard against data leakage and extraction attacks while preserving efficiency. Liang et al. [66] similarly employ federated learning to maintain data locality, utilising graph neural networks to recover spatial dependencies across road networks while preserving confidentiality.

Tregi et al. [67] explore advanced protections, combining Falcon signatures, dual-stage pseudonymisation, and BB84 quantum key distribution to defend against sophisticated, including quantum-enabled, attacks. For lightweight environments, Sarwar et al. [68] propose a privacy-aware distribution mechanism based on user-defined privacy levels, optimising storage and processing loads while ensuring confidentiality.

Overall, the computing continuum decentralises computation to improve scalability and responsiveness, positioning itself as the architectural backbone of smart cities. However, it also introduces orchestration complexity and security challenges. Addressing these trade-offs is essential for realising its potential as a reliable and adaptive foundation for urban services.

3.2.4. Fulfilment of Non-Functional Requirements in the Continuum

As discussed in the previous sections, managing non-functional requirements—such as confidentiality, integrity, scalability, efficiency, latency, and trust—is essential to ensuring the quality of service in the continuum.

Table 2. Comparative analysis of recent works addressing smart city applications across the computing continuum.

Work	Technologies Used	Containers/ Serverless	Continuum Layers				Primary Use Case	Non-Functional Req.
			IoT	Edge	Fog	Cloud		C	I	S	E	L	T
[2] (2023)	Traditional VMs/ distributed	-	✓	✓	-	✓	Urban mobility (taxis, traffic)	-	-	✓	✓	✓	-
[55] (2025)	Custom ACO, VG, MPC	-	✓	✓	✓	✓	Drone-based logistics and sensing	-	-	✓	✓	✓	-
[54] (2024)	5G, orchestration	✓(containers)	✓	✓	✓	✓	Vehicular cooperative perception	-	-	✓	✓	✓	-
[56] (2025)	Decentralized orchestration	-	✓	✓	✓	✓	Vehicular services in motion	-	-	✓	✓	✓	✓
[57] (2022)	Serverless, edge-cloud	✓(FaaS, workflows)	✓	✓	-	✓	Traffic safety, urban tourism	-	-	-	-	✓	✓
[58] (2021)	Serverless, cloud continuum	✓(containerized FaaS)	✓	✓	-	✓	Face mask detection with privacy	✓	-	-	✓	✓	-
[59] (2024)	Serverless, autoscaling	✓	✓	✓	✓	✓	Health emergency response	-	-	✓	-	✓	-
[60] (2022)	Blockchain, PaaS, DSL	✓(PaaS deployment)	✓	✓	✓	✓	Marketing pipelines, digital health	✓	✓	✓	-	-	✓
[61] (2023)	Smart contracts, DApp	-	-	-	-	✓	Real-estate transaction automation	-	✓	-	-	-	✓
[62] (2023)	Ethereum, smart contracts	-	-	-	-	✓	E-voting and civic engagement	✓	✓	-	-	-	✓
[65] (2021)	Lightweight blockchain, sub/global chains	-	✓	✓	✓	✓	Cross-domain communication	✓	✓	✓	-	-	✓
[4] (2020)	Distributed privacy, encryption	-	✓	✓	✓	✓	Smart energy demand prediction	✓	✓	-	✓	-	-
[66] (2024)	Federated learning, GCN, privacy	-	✓	✓	-	✓	Traffic pattern prediction	✓	-	✓	✓	-	-
[67] (2025)	Quantum cryptography, digital signatures	-	✓	✓	-	✓	Traffic data protection	✓	✓	-	-	-	✓
[68] (2021)	LoP-based encryption and division	-	✓	✓	✓	-	Aggregated sensor data processing	✓	-	✓	✓	-	-

C: Confidentiality; I: Integrity; S: Scalability; E: Efficiency; L: Latency; T: Trust; ✓: Supported; -: Not supported.

summarises the continuum tools presented, showing how each addresses these key non-functional demands.

Confidentiality is crucial for secure data management, as it prevents unlawful access. It is achieved through mechanisms such as encryption, federated learning, and quantum-safe protocols, as demonstrated in [4,66,67,68]. Integrity is equally important, ensuring that transmitted data remain authentic and consistent, typically through blockchain and distributed ledger technologies [60,62,65].

Across continuum applications, scalability and efficiency emerge as fundamental requirements. They ensure that distributed infrastructures can manage the immense growth of connected devices while optimising both computational and energy resources [2,54,56,59]. Low latency is another defining feature of continuum-based architectures, particularly in real-time domains such as traffic management, emergency response, and drone coordination [55,57,58].

Finally, trust mechanisms are critical for collaboration and accountability across participants, with smart contracts and decentralised governance playing an increasingly important role [60,61,62].

Overall, these studies demonstrate that addressing non-functional requirements holistically—not as isolated design constraints but as interdependent pillars—is essential to achieving the reliability, adaptability, and ethical governance expected from next-generation smart city infrastructures.

3.3. Artificial Intelligence

Artificial intelligence (AI), encompassing machine learning, expert systems, natural language processing, and computer vision, has shown strong potential in addressing smart city challenges [69]. Zekić-Sušac et al. [22] evaluate AI models for optimising energy consumption within MERIDA, an intelligent energy management system. Similarly, Ghazal et al. [70] review AI in healthcare, covering the analysis of Computed Tomography and X-ray images, patient monitoring, and disease diagnosis. While the continuum manages and processes distributed data, AI transforms them into actionable insights, enabling informed decision-making across domains.

Figure 8 presents a taxonomy of AI contributions across three dimensions. The first is knowledge acquisition and training, contrasting centralised development with distributed paradigms such as federated learning. The second concerns reasoning and inference, which can occur locally on edge devices for low-latency decisions or in the cloud for large-scale computation. The third highlights applications, including anomaly detection, mobility, public safety, and healthcare.

Centralised AI training faces two key limitations in smart cities:

1.

Confidentiality: Urban datasets often contain sensitive information, including geolocation, health, or behavioural records. Centralised approaches can expose these data during transmission or storage [71,72].

2.

Bandwidth and Latency: Aggregating data for centralised AI requires high-bandwidth transfers, leading to costs and delays that hinder real-time responsiveness [73,74].

To address these issues, federated learning has emerged as a privacy-preserving paradigm where edge or fog nodes train local models and only share parameters, preserving data locality and reducing network loads.

Federated learning has been applied across multiple smart city domains. In mobility, Badu-Marfo et al. [75] propose NATALIE, a federated ensemble framework analysing GPS trajectories with Long Short-Term Memory (LSTM), Gated Recurrent Unit (GRU), and Convolutional Neural Network (CNN) models. In transportation, Liu et al. [76] introduce FedGRU for traffic forecasting.

3.4. Data Storage and Movement

A persistent challenge in smart cities is the efficient movement and storage of data. Although paradigms such as federated learning and the computing continuum reduce the need to transfer raw data to the public cloud, I/O operations across networks continue to be a bottleneck. From the device perspective, connectivity relies on technologies such as 5G, Wi-Fi, ZigBee, Bluetooth, and Mobile Ad Hoc Networks [37]. However, these networks are often unreliable due to variability and intermittent connectivity. To mitigate this, robust application-layer protocols have been developed. For instance, MQTT [77] provides a lightweight publish/subscribe model, minimising bandwidth use and energy consumption—making it well suited for IoT and edge devices [78].

Beyond communication, storage limitations at the edge also affect system performance. Figure 9 illustrates a taxonomy that encompasses the dimensions involved in managing smart city data across the computing continuum. Again, each environment in the continuum provides advantages and disadvantages that must be considered when creating a solution. Local storage at the edge enables faster insights and reduces dependency on cloud infrastructures [3], but the dispersal of data across heterogeneous sites complicates management [79]. Traditional cloud-based storage services such as Amazon S3 [80] and Google Cloud Storage [81], often paired with streaming platforms like Apache Kafka [82], remain common. However, these architectures introduce latency and overhead due to the physical distance between edge devices and remote cloud nodes [83].

The effectiveness of AI and continuum smart city applications also depends on reliable storage and data movement strategies. Without scalable and resilient storage solutions across the continuum, the training, inference, and sharing of intelligent services cannot be sustained.

To address this gap, recent solutions exploit storage closer to data sources. For example, KubeEdge [84] extends Kubernetes to create persistent volumes on local edge devices, while EdgeMesh [85] interconnects nodes into a storage mesh. These approaches leverage under-utilised local resources, but still face challenges of durability, consistency, and capacity limits.

More advanced frameworks aim to unify storage across edge, fog, and cloud. Expand [86], a Portable Operating System Interface (POSIX)-based parallel file system initially designed for HPC, has been adapted to IoT scenarios by integrating Message Queuing Telemetry Transport (MQTT)-based clients. StructMesh [24] adopts a serverless design to manage uploads, downloads, and pipelines across the continuum, fulfilling non-functional requirements such as security, reliability, and cost-efficiency [87]. DynoStore [43] similarly provides a wide-area storage fabric, integrating existing backends into a resilient distributed system.

Finally, peer-to-peer storage models such as IPFS [88] enable decentralised data exchange among IoT and edge devices in smart city environments [89,90]. While removing central points of failure, they introduce privacy and traceability challenges, which some systems address by combining IPFS with blockchain and proxy re-encryption to enforce confidentiality and integrity [91].

In summary, storage and data movement remain central to the scalability of smart cities. While cloud services offer capacity and durability, they also introduce latency and cost, motivating the rise of edge- and fog-based storage solutions. Future systems must strike a balance between latency, resilience, and energy efficiency while supporting interoperability across heterogeneous backends and ensuring compliance with urban data governance and privacy regulations.

3.5. Convergence of Sustainability and ICT in Smart Cities

The development of sustainable smart cities relies on the integration of several core technological domains, each of which contributes uniquely to the UN SDGs. The computing continuum, including the IoT, AI, and storage and data management, acts as complementary pillars that support innovation, efficiency, and sustainability in urban environments.

The computing continuum establishes scalable and resilient digital infrastructures by distributing computational resources across cloud, edge, and device layers. That infrastructure and the services strongly help to satisfy several SDGs [92,93], as shown below.

The computing continuum has demonstrated strong applicability in contributing to SDG 6 (Clean Water and Sanitation), particularly through the development of advanced systems to monitor and control water quality. A pilot project for water quality monitoring in smart cities [94] presented a multi-parameter system in Bristol Floating Harbour, which successfully demonstrated the feasibility of collecting real-time high-frequency data and displaying them online. Smart water quality monitoring systems based on IoT technology detect water quality and pipe flow using sensors. They enhance web-based water monitoring systems and allow authorities to be alerted in case of health or quality problems [95,96].

In relation to SDG 9 (Industry, Innovation, and Infrastructure) [97], the continuum enables robust and adaptive infrastructures that can sustain innovation across sectors. Roman et al. [98] demonstrate how edge devices in patient monitoring ecosystems can dynamically adapt data processing pipelines, illustrating resource-efficient deployment across heterogeneous IoT sensors. Similarly, Yepes et al. [99] propose a tsunami early-warning system implemented along the continuum, leveraging real-time sensor-to-cloud computation for rapid detection of danger. Bassolillo et al. [55] conceptualise drones as mobile edge nodes within urban delivery and monitoring networks, combining Ant Colony Optimisation and fog computing for real-time routing and inspection. At the same time, its ability to provide low-latency processing and localised decision-making contributes to SDG 11 (Sustainable Cities and Communities) [100], supporting applications such as smart mobility, real-time traffic management, and efficient governance platforms. For instance, Al Harbi et al. [101] demonstrate an energy-efficient edge–fog–cloud architecture for smart agriculture, reducing carbon emissions by optimising data processing across layers, while De Maio et al. [57] illustrate the dynamic deployment of serverless functions on edge nodes for latency-sensitive urban services like traffic alerts and augmented reality tourism. By enhancing the scalability and responsiveness of urban systems, the computing continuum lays the groundwork for more resilient and inclusive cities.

The IoT plays a pivotal role in advancing SDG 11 through pervasive sensing and real-time monitoring, which improves service delivery in areas such as transportation, energy, and waste management. Banerjee et al. [102] introduce a reinforcement learning-based sleep-scheduling strategy for wireless sensor nodes, enhancing network stability and sustainable operations. Furthermore, the IoT contributes to SDG 12 (Responsible Consumption and Production) [103] by enabling the precise tracking and optimisation of resource use, reducing waste, and supporting circular economy initiatives. For instance, Kamal et al. [104] present an IoT-enabled bus-stop monitoring system to optimise energy usage and air quality management, demonstrating practical applications in public infrastructure. Syed et al. [105] highlight the integration of real-time sensors with digital twin models for improved water monitoring. In contrast, Rehman et al. [51] develop a solar-powered portable charging system with IoT analytics for emergency and public service applications. Collectively, these works demonstrate how the IoT serves as the connective tissue, linking physical infrastructure with digital intelligence, thereby promoting sustainable urban management.

AI technologies [106] provide advanced analytical and predictive capabilities that cut across multiple SDGs. For SDG 9, AI fosters innovation by enabling data-driven decision-making and optimising industrial and infrastructural processes. Mohsen et al. [107] analyse traffic flow management in emerging cities, showing that AI-enabled traffic cameras improve signal control and reduce congestion. Reddy et al. [108] demonstrate the use of generative AI for ride-hailing demand prediction, thereby enhancing urban transport efficiency and service reliability. Liu et al. [76] introduce FedGRU for traffic forecasting using federated learning, achieving high accuracy without compromising data privacy. AI also contributes directly to SDG 13 (Climate Action) [109] by powering predictive models for environmental monitoring, climate adaptation, and disaster response, thereby equipping cities with tools for risk mitigation and resilience. Roy et al. [110] focus on AI for marine ecosystem preservation, while Gupta et al. [111] employ AI-driven waste sorting and recycling prediction to reduce plastic impact. When combined with IoT and the computing continuum, AI becomes a central driver of intelligent, adaptive, and sustainable urban ecosystems.

Robust storage and data management infrastructures are fundamental to ensuring the interoperability, security, and accessibility of urban data. Supporting SDG 12, these systems facilitate the efficient handling and analysis of data generated by urban processes, helping to optimise resource consumption and reduce waste [87]. Savarimuthu et al. [112] define “digital data waste” and present machine learning models to detect and eliminate inefficiencies, reducing computational and environmental costs. They also underpin SDG 17 (Partnerships for the Goals) [113] by enabling data sharing across institutions and jurisdictions, fostering collaboration and collective innovation. For example, Izadifar et al. [114] demonstrate a biomanufacturing knowledge hub connecting multiple stakeholders through secure data sharing. In contrast, Gegenhuber et al. [115] develop a framework for distributed data governance in Open Social Innovation projects, highlighting the importance of openness, accountability, and power dynamics in collaborative systems. Moreover, by ensuring the availability and reliability of data, storage and management systems strengthen the integration of digital technologies in advancing sustainability objectives across the smart city ecosystem.

In summary, these domains—computing continuum, IoT, AI, and data management—constitute interdependent technological layers that collectively enable the alignment of smart city development with global sustainability objectives. Their convergence provides not only sector-specific contributions to individual SDGs but also a systemic foundation for innovation, resilience, and inclusivity in the urban context.

4. Case Studies of Smart Cities

In recent years, the rapid evolution of ICT [116] has driven a proliferation of smart city initiatives worldwide. These projects demonstrate how ICT facilitates urban data acquisition, analytics, knowledge generation, and sharing [117], thereby promoting smart governance [118], citizen participation [119], and innovative business models [120]. Below, we present representative examples from Europe, Asia, and beyond.

Table 3. Comparison of smart services in smart city use cases.

Smart City Services	Amsterdam	Bangalore	Santander	Seoul	Singapore
Smart Transportation	Smart mobility platform (MaaS), real-time public transit data	Intelligent traffic management system, smart signal lights	Sensor-based traffic monitoring, smart parking	T-money card, AI-powered traffic flow management	Autonomous vehicle trials, smart traffic lights, MaaS
E-Governance	Open data portal, digital citizen services	Digital city dashboard, mobile apps for municipal services	SmartSantander platform for municipal data access	Mobile government services (e-Gov), IoT-enabled governance	MyInfo e-services platform, SingPass for digital identity
Environmental Monitoring	Air and noise pollution sensors, smart waste bins	Smart water and energy meters, waste management tracking	20,000+ IoT sensors for air, noise, and light	Smart waste collection, IoT for air quality and noise	Smart Nation Sensor Platform, real-time air and water quality monitoring
Smart Energy	Smart grid integration, energy-positive neighbourhoods	Automatic meter reading for utilities, rooftop solar systems	Energy-efficient lighting, smart meters	Building energy management systems, national smart grid	Smart buildings, nationwide smart grid with real-time usage data
Healthcare	Telemedicine pilots, digital health data exchange	E-health records, remote consultation apps	Health apps integrated with city services	AI diagnosis assistance, remote health monitoring	HealthHub portal, telehealth, and elderly care tech
Citizen Engagement	Apps for public participation (e.g., FixMyStreet), urban innovation labs	Public grievance redressal apps, open feedback portals	Real-time issue reporting via app, citizen data crowdsourcing	Smart Seoul app, citizen co-creation projects	OneService app for reporting municipal issues, digital engagement via GovTech
Safety and Security	Smart street lighting, video surveillance in public spaces	AI-powered surveillance, emergency response systems	Video analytics for public safety, emergency alert systems	Closed-circuit televisions with AI facial recognition, real-time crime mapping	Predictive policing, integrated surveillance systems
Data and Innovation	Amsterdam smart city platform, open data collaborations	IoT-based city dashboards, data exchange platform (IUDX)	Urban lab testing IoT pilots, open research datasets	Big data hub for urban planning, public–private innovation labs	Open data portal, Smart Nation R&D initiatives

compares notable smart cities: Amsterdam [121], Bangalore [122], Santander [123,124], Seoul [125,126], and Singapore [127]. It summarises the main ICT-enabled services implemented in these case studies.

Smart transportation aims to improve mobility by reducing congestion and creating efficient routes [128]. Amsterdam, Seoul, and Singapore utilise Mobility-as-a-Service (MaaS) platforms that integrate multiple modes of transportation [129], while AI systems help control traffic by automatically managing signal lights.

E-governance enhances the accessibility and transparency of public services. Most case studies offer open data portals, allowing citizens to access government datasets. Traditional services have also been digitalised, such as electronic ID systems [127]. Open data platforms further promote collaboration between citizens, developers, and the public sector [130]. For instance, the Smart Seoul 2015 initiative offered city-wide Wi-Fi and e-government services [131], with its e-governance model ranked among the most advanced worldwide [132].

Environmental monitoring is supported by IoT sensors for air quality, noise, and light pollution [124,128]. Bangalore, for example, has deployed smart waste systems that track bin levels in real time, optimising collection routes and reducing costs. It also employs air quality monitoring to inform citizens and guide policy.

Energy solutions aim to improve efficiency through the use of big data and AI tools for resource management and infrastructure monitoring [124,130]. Healthcare services are increasingly adopting ICT for remote consultations, diagnosis assistance, digital records, and patient monitoring, thereby improving response times and reducing bottlenecks. Safety and security rely on IoT and AI for surveillance, video analytics, and emergency response systems, enhancing monitoring of public spaces and automating alerts [130].

All these initiatives also contribute to sustainability in line with the UN SDGs. For example, waste management, mobility planning, and real-time air monitoring support SDG 11 (sustainable cities), SDG 6 (clean water), and SDG 13 (climate action). Open data and e-governance platforms contribute to SDG 16 (strong institutions) and SDG 17 (partnerships). Amsterdam is widely regarded as a leading example, contributing to the achievement of SDGs 9, 11, and 17. Overall, smart cities show clear potential to advance sustainability goals as more projects are deployed.

5. Discussion

Current smart city approaches require the coordination of multiple devices that produce, process, and store data. This coordination must address both functional and non-functional requirements.

From a functional perspective, smart city systems aim to enhance urban services that directly benefit citizens. For example, analysing traffic patterns can reduce congestion, while environmental monitoring improves the accuracy of pollution and weather forecasts.

From a non-functional perspective, systems must ensure reliability, confidentiality, accessibility, and scalability. The computing continuum contributes to scalability, reliability, and accessibility by distributing processing and storage tasks, while federated learning enhances confidentiality by keeping data local during training. Most of the reviewed papers adopt such decentralised approaches.

Nevertheless, challenges remain. Continuum-based architectures reduce latency but introduce complexity in orchestration and reliability concerns in heterogeneous environments. Federated learning mitigates data-sharing risks but can sacrifice accuracy and remains vulnerable to poisoning attacks. These trade-offs suggest that future smart city architectures must balance decentralisation with strong orchestration, reliability, and security mechanisms.

Figure 10 illustrates a hierarchical smart city environment, from the IoT layer at the base to the cloud layer at the top. This builds on the n-tier architecture commonly used to describe the continuum, typically composed of edge, fog, and cloud layers [2,3,27]. Recent studies also add the IoT and services (including AI) as layers connecting data collection and end-user access. For each layer in Figure 10, we identify the main advantages and challenges.

• IoT layer: Devices such as industrial sensors, healthcare monitors, meteorological stations, and transport systems enable real-time data collection for automation and predictive tasks (e.g., preventive maintenance). Challenges include interoperability, limited power, and connectivity constraints.
• Edge layer: Provides local processing, reducing latency and improving resilience to connectivity loss, while shielding sensitive data. Challenges lie in ensuring data consistency, managing scarce resources, and handling failures.
• Fog layer: Acts as an intermediary, reducing bandwidth usage and supporting offline operation. However, fog infrastructures face issues of reliability, interoperability, and high deployment costs.
• High-End Computing layer: Data centres and supercomputers support intensive analytics, visualisation, and large-scale prediction. They offer scalability and broad access but raise concerns about latency, privacy, vendor lock-in, and limited infrastructure control.
• Services: Platforms deliver monitoring, cybersecurity, predictive analytics, and citizen services. However, greater connectivity expands the attack surface and raises concerns about privacy, surveillance, open standards, and unequal access (digital divide).

Integrating these layers supports analytics, monitoring, visualisation, and predictive services, driving the digital transformation of cities into smart, sustainable, and citizen-centric models. The review findings highlight the central role of technology and data in smart city development.

Despite these advances, major gaps persist in inclusivity and long-term data ethics. Policies must evolve in tandem with technologies to ensure equitable outcomes. As José et al. [33] note, smart city innovation lags behind other domains due to barriers such as the following ones:

• Privacy and surveillance: Mass data collection raises concerns about privacy, surveillance, and ethical AI. Strong legal frameworks and safeguards are needed.
• Digital divide: Unequal access to digital tools creates disparities. Limited access to databases and publication bias exacerbate this divide [133].
• High initial costs: Building smart infrastructure requires substantial investment. Scalable and modular architectures are necessary for sustainable growth.

These challenges explain why smart city deployments progress more slowly in developing regions compared to countries with advanced IT infrastructures and stronger economies. Addressing them requires context-specific strategies. For example, Bittencourt [134] examines the multifaceted dimensions of intelligent cities, from infrastructure design to citizen-centric decision-making, highlighting how poor alignment can exacerbate the digital divide. Similarly, Batmetan et al. [135] propose a strategy tailored to developing cities, structured around six smart city dimensions: smart governance, smart branding, smart economy, smart living, smart society, and smart environment. Their framework provides measurable guidance by prioritising efficiency and aligning initiatives with both internal and external contextual factors, as also emphasised in [136].

In summary, smart city progress depends not only on advances in ICT and paradigms such as the computing continuum and federated learning, but also on inclusive policies, ethical governance, and investment models that bridge the digital divide.

6. Conclusions

A range of technologies has been adopted to support the design and deployment of smart cities. In this review, we examined recent contributions that explore the adoption of emerging paradigms, including the computing continuum, AI, and the IoT, in urban environments. Through an analysis of practical implementations, we identified common use cases including surveillance, pollution monitoring, traffic control, and waste management. These scenarios share core challenges such as coordinating distributed devices, ensuring efficient data collection and processing, and enabling informed decision-making based on analytical results.

Our findings suggest that the computing continuum and the other technologies discussed in this paper can be aligned with sustainability goals. The computing continuum and AI contribute to the development of resilient digital infrastructures and the promotion of innovation across distributed systems. The continuum, IoT, and storage technologies facilitate the creation of scalable digital services and efficient urban management, including applications in e-governance and smart mobility. Furthermore, recent continuum and AI solutions promote more sustainable energy consumption, while storage solutions enable collaborative approaches to sustainability by supporting interoperability and data sharing across institutions. In combination, these technologies help develop systems for disaster response, climate adaptation, and environmental monitoring, among other applications. The findings of this review highlight the central role of technology and data in successful smart city development.

Overall, our analysis indicates that the computing continuum and AI are positioned to become foundational pillars in the evolution of smart cities. Future research should focus on integrating them seamlessly into robust, interoperable, and privacy-preserving solutions, accompanied by inclusive policies and sustainable investment models, to ensure equitable and transformative urban development. Furthermore, as future work, we will quantitatively evaluate different computing continuum and IoT solutions for smart cities. This will include designing a smart city framework using real-time and historical open data from the Community of Madrid in the domains of mobility and environmental management [137].