Urban traffic congestion remains a critical challenge to mobility efficiency and environmental sustainability. This paper presents a Digital Twin framework for real-time urban traffic management in Issy-les-Moulineaux, France. The framework integrates heterogeneous traffic data sources, including historical, real-time, and predictive data, into a unified architecture for scenario-based analytics and decision support. For traffic forecasting, deep learning models, including Long Short-Term Memory (LSTM) and Transformer-based approaches, are evaluated alongside baseline models such as SARIMA and Neural Prophet. The LSTM model is selected based on its stable and consistent performance across heterogeneous traffic conditions, achieving mean absolute error (MAE) values ranging from approximately 2 to 33 vehicles/hour across different road segments. The system also supports dynamic event handling, including full and time-based road closures, with capacity-constrained and equal-allocation diversion strategies for first-order impact estimation. Quantitative evaluations demonstrate that the framework enables proactive congestion mitigation, comparative emission impact assessment, and improved interoperability with city platforms through NGSI-LD standardization. By combining predictive modeling, real-time monitoring, and scenario-driven analytics, the proposed Digital Twin delivers actionable insights for urban planners, reducing short-term disruptions and supporting long-term sustainable mobility management.

The increasing demand for smart cities calls for advanced solutions to enhance urban sustainability. Digital twin technology offers transformative potential by synchronizing virtual and physical environments in real time. However, existing approaches struggle with scalability due to the reliance on numerous specialized prediction models for individual urban components, the lack of a unified framework for different use cases limits generalization, and high latencies in real-time synchronization. To address these, this paper presents a comprehensive software architecture for smart city DT and integrates correlation-aware model reduction and dynamic adaptive forecasting to support diverse urban applications to improve generalizability and scalability. This is done while adapting a smart distribution of DT software components between edge and core servers to ensure a low-latency performance. Validated using real-world traffic and air quality data, the system demonstrates significant improvements in traffic flow, emissions reduction, and public transportation efficiency, and enhances air quality monitoring, forecasting, and pollutant management. Key contributions include a scalable and generalizable DT architecture, AI-driven adaptability, edge-core deployment, and extensive validation through predictive analytics. This work establishes a replicable blueprint for metropolitan-scale DTs, balancing computational efficiency with responsive urban analytics.

The growing need for intelligent tools to support urban planning and resource management has positioned Digital Twin (DT) technology as a cornerstone of smart city development. DTs, as dynamic virtual replicas of physical systems, offer capabilities that extend beyond mere representation, enabling monitoring, diagnostics, forecasting, and optimization. In the context of urban traffic management, DTs provide a robust solution for real-time traffic monitoring and predictive analytics. However, existing approaches often lack a systematic design methodology, leading to challenges in scalability and adaptability, particularly in heterogeneous environments. This paper presents a novel methodology for developing scalable and adaptive smart city DT architectures, with a focus on real-time traffic management. A modular and unified software framework is proposed, leveraging AI-driven approaches to address the complexity of managing diverse traffic data sources. A sequential learning model is integrated into the architecture to enhance the DT's adaptability to evolving traffic conditions and congestion patterns. The proposed framework is validated using real-world traffic data from an IoT network deployed in Madrid, demonstrating its scalability and low-latency performance. Experimental results highlight the effectiveness of the framework in handling heterogeneous traffic scenarios and its ability to deliver accurate predictions while minimizing resource overhead.

Network Digital Twin (NDT) collects data from physical, virtual, and software components and supports real-time network performance analysis, emulation, and intelligent physical network control. This paper surveys the current state of NDT specifications and explores NDT benefits for Network Operators (NOs) and its possible roles in future network management. It discusses the NDT key components, architecture, and integration of Machine Learning and Artificial Intelligence models in the NDT. Further, it covers virtualization technology management, suitability of Software-Defined Networking capabilities, and simulation tools to empower NDT. Two perspectives make the position of this survey different from existing studies; first, it highlights NDT limitations regarding Edge–Cloud Continuum (ECC) contextualization. ECC is a purposeful trending integration of Edge and Cloud Computing , involving multiple stakeholders like Service Providers, Customers, and Platform or Infrastructure Providers. However, current NDT specifications have not mentioned the ways to benefit stakeholders other than NOs. We also discuss notable computing and communication technologies transformations necessary to consider during NDT modeling, the existing data models, and reusable vocabularies that can be extended to achieve a detailed ECC representation for all stakeholders, essentially for Service Providers and Customers. Secondly, a data model is proposed that covers descriptive and prescriptive features and aims to provide a granular representation of ECC components to meet stakeholders’ requirements and render particular user information views. Different explored NDT perspectives, and proposed data model reduces the impact of existing NDT limitations in ECC representation.

Complex urban systems can be difficult to monitor, diagnose and manage because the complete states of such systems are only partially observable with sensors. State estimation techniques can be used to determine the underlying dynamic behavior of such complex systems with their highly non-linear processes and external time-variant influences. States can be estimated by clustering observed sensor readings. However, clustering performance degrades as the number of sensors and readings (i.e. feature dimension) increases. To address this problem, we propose a framework that learns a feature-centric lower dimensional representation of data for clustering to support analysis of system dynamics. We propose Unsupervised Feature Attention with Compact Representation (UFACR) to rank features contributing to a cluster assignment. These weighted features are then used to learn a reduced-dimension temporal representation of the data with a deep-learning model. The resulting low-dimensional representation can be effectively clustered into states. UFACR is evaluated on real-world and synthetic wastewater treatment plant data sets, and feature ranking outcomes were validated by Wastewater treatment domain experts. Our quantitative and qualitative experimental analyses demonstrate the effectiveness of UFACR for uncovering system dynamics in an automated and unsupervised manner to offer guidance to wastewater engineers to enhance industrial productivity and treatment efficiency.

This paper presents a modular Digital Twin software architecture designed for smart cities, leveraging Edge-Cloud Continuum to enable the development of flexible and scalable DT-based solutions. Digital Twin technology provides a powerful framework for simulating, analyzing, and optimizing urban environments by integrating real-time and historical data from various city sensors through IoT, AI, and cloud computing. The proposed architecture addresses the limitations of existing DT frameworks by focusing on smart city-specific requirements such as dynamic resource management, real-time data processing, and autonomous decision-making. The viability of the proposed framework is demonstrated through a case study on autonomous traffic management in the city of Issy-les-Moulineaux. It shows how the proposed framework predicts traffic patterns and manages network resource allocation by adjusting the data sampling frequency to balance prediction accuracy and communication costs. The architecture’s modular design supports seamless integration and adaptability, making it suitable for various smart city applications, thereby advancing the development of more efficient, sustainable, and resilient urban environments.

Specifications for Network Digital Twin (NDT) from Standardization Development Organizations (SDOs), such as the Internet Engineering Task Force (IETF), and academic contributions focus primarily on benefiting network operators. However, they often overlook the needs of stakeholders in the Edge-Cloud Continuum (ECC), such as Service Providers, customers, and Platform or Infrastructure Providers. In ECC, resource heterogeneity, agile software component integration, and quality of service requirements are challenges. To address these challenges, continuous and granular monitoring of software and physical resources is required. In this paper, we present the design and ongoing implementation of a data model. It captures and characterizes the physical and software properties, i.e., Key Performance Indicators (KPIs), of Kubernetes-managed components in the ECC. Collected data is structured in NGSI-LD-compliant format and managed through interoperable context brokers for authenticating the requests of various stakeholder applications. We also demonstrate sample data curation from a lab-configured platform, its integration into the context broker, and how it responds to the queries concerning a particular component information, thereby representing a partial implementation of proposed data model to render the views for particular stakeholders.

Deep learning is being widely utilized in industrial process monitoring, control, and optimization. However, in the wastewater industry, its applications are still underexplored. This is because deep learning requires a large amount of labeled training data to induce effective predictive models. Owing to the high cost of sensors and frequency and delay in sampling and laboratory analytics, wastewater treatment process data can be sparse with varying frequencies. One option to address training data limitations is to use transfer learning. However, owing to the large covariate shift between the commonly adopted source domains for transfer learning and the target domain of wastewater processes, this approach leads to unacceptable performance. We address this issue by proposing a novel synthetic data generation method for deep predictive modeling of wastewater plants. Employing a Markov process that utilizes random walk, our technique enables the generation of abundant annotated data for our target domain. The method preserves the temporal dynamics and distribution of the original data, thereby closely mimicking the potential original samples of the domain. We extensively evaluate our method over two different high-rate-algae-based treatment datasets, demonstrating considerable performance gains over existing transfer learning. Our proposed algorithm can assist plant operators to deploy responsive supportive models with limited data.