Laboratoire d'Ingénierie Circulation Transports

This study presents a comprehensive evaluation of the real challenges related to the calibration of Network Macroscopic Fundamental Diagrams (NMFDs), with a focus on two key aspects: observability and stability. Observability addresses the estimation of NMFDs using multimodal traffic data, including loop detectors (LDD), floating car data (FCD), and public transport data (PTD), and examines how data quality, spatial coverage, and aggregation intervals influence NMFD characteristics. Stability explores the temporal consistency of NMFDs through longitudinal analysis, investigating their sensitivity to loading and unloading phases, weekday versus weekend dynamics, and recurring demand patterns. Using case studies in Athens and Lyon demonstrates that fusing data sources improves the accuracy of multimodal NMFD estimation, while separating network loading/unloading phases and different days enables reliable NMFD estimation. Building on these insights, we propose a structured NMFD calibration pipeline that integrates bias correction, data fusion, and temporal clustering. This framework supports reproducible, data-driven NMFD estimation and offers practical guidance for urban traffic monitoring and multimodal network management.

As climate change accelerates, the frequency and intensity of flood events are rising, creating significant challenges for critical infrastructure systems worldwide. Transportation, energy, and communication networks are particularly vulnerable, and their resilience to such disastersis crucial for minimizing long-term impacts. This study examines five recent flood events—Germany (2021), Belgium (2021), Sydney (2022), Auckland (2023), and Italy (2023)—to explore the effects of these floods on critical infrastructure and identify best practices for enhancing resilience. The research focuses on answering the central question:How do recent flood events impact critical infrastructure, and what best practices can be identified for improving resilience?Due to the recent nature of these floods, data collection was a pivotal aspect of the study, with information sourced from public news reports, research journals, government reports, and interviews. A Multi-Criteria Decision Making (MCDM) method- the Vikor, was employed to rank hazards, vulnerabilities, and the resilience of critical infrastructure in each case study. This approach provided a systematic evaluation of shared vulnerabilities and region-specificdifferences in disaster response and infrastructure resilience.The findings highlight the importance of multi-stakeholder collaboration, early warning systems, and adaptive infrastructure solutions in mitigating flood impacts. Best practices were identified across all phases of disaster management—pre-disaster preparedness, immediate emergency response, and long-term recovery. These practices emphasize the need for innovative infrastructure adaptations, community engagement, and coordinatedgovernance to build more resilient systems.This research offers valuable insights for policymakers, urban planners, and disaster management professionals. By analyzing these five flood events, the study provides transferable lessons on how to enhance infrastructure resilience and integrate adaptive strategies into policy frameworks. Ultimately, this research contributes to the broader global discourse on climate adaptation and disaster risk reduction, aiming to strengthen preparednessfor future flood events.Keywords: Flood resilience, critical infrastructure, case study analysis, MCDM, disaster management, data collection, best practices

Online continuous action recognition has emerged as a critical research area due to its practical implications in real-world applications, such as human-computer interaction, healthcare, and robotics. Among various modalities, skeleton-based approaches have gained significant popularity, demonstrating their effectiveness in capturing 3D temporal data while ensuring robustness to environmental variations. However, most existing works focus on segment-based recognition, making them unsuitable for real-time, continuous recognition scenarios. In this paper, we propose a novel online recognition system designed for real-time skeleton sequence streaming. Our approach leverages a hybrid architecture combining Spatial Graph Convolutional Networks (S-GCN) for spatial feature extraction and a Transformer-based Graph Encoder (TGE) for capturing temporal dependencies across frames. Additionally, we introduce a continual learning mechanism to enhance model adaptability to evolving data distributions, ensuring robust recognition in dynamic environments. We evaluate our method on the SHREC'21 benchmark dataset, demonstrating its superior performance in online hand gesture recognition. Our approach not only achieves state-of-the-art accuracy but also significantly reduces false positive rates, making it a compelling solution for real-time applications. The proposed system can be seamlessly integrated into various domains, including human-robot collaboration and assistive technologies, where natural and intuitive interaction is crucial.

This study presents a flexible model for risk-taking behavior and accident injury severity. It is specifically designed to evaluate the impact of Via Sicura, a Swiss road safety program, on the severity of accident outcomes. Our proposed model treats the risk-taking behavior of each driver as a latent variable that depends on a number of socioeconomic and contextual factors, and whose manifestation can be measured by means of behavioral indicators. The aggregated risk, a central feature of our framework, represents the combined latent risk-taking behaviors among all drivers within an accident and is successfully identified as explanatory of the severity of injuries sustained by all individuals involved. Our findings reveal that Via Sicura’s repressive measures successfully deter risk-taking behavior among drivers, preventing an estimated 63 fatal, 876 major and 2’303 minor injuries over a ten-year period.

Health data is one of the most sensitive data for people, which attracts the attention of malicious activities. We propose an open-source health data management framework, that follows a patient-centric approach. The proposed framework implements the Self-Sovereign Identity paradigm with innovative technologies such as Decentralized Identifiers and Verifiable Credentials. The framework uses Blockchain technology to provide immutability, verifiable data registry, and auditability, as well as an agent-based model to provide protection and privacy for the patient data. We also define different use cases regarding the daily patient-practitioner-laboratory interactions and specific functions to cover patient data loss, data access revocation, and emergency cases where patients are unable to give consent and access to their data. To address this design, a proof of concept is created with an interaction between patient and doctor. The most feasible technologies are selected and the created design is validated. We discuss the differences and novelties of this framework, which includes the patient-centric approach also for data storage, the designed recovery and emergency plan, the defined backup procedure, and the selected blockchain platform.

Abstract The race for digitization created a real need to protect smart infrastructures. Environments are becoming highly connected and automated. Their growing complexity and connectivity make it hard to assure and assess their cyber resilience, i.e., protecting them from cyberattacks, failures, and errors. Traditional strategies for ensuring the cyber resilience of smart infrastructures suffer from a lack of holism. Indeed, since smart infrastructures are often structured in layers, traditional protection methods can lead to conflicting and competing goals. For instance, they may increase the resilience of specific layers at the expense of decreasing the performance of others. This chapter reviews existing methods aiming to address this problem. We focus on two leading methodological assessment families: quantitative and qualitative. The former includes numerical metrics to quantify and assist system-dependent decision-making processes. The latter builds upon symbolic modeling to offer a system-agnostic assessment. The chapter provides an in-depth exploration of quantitative and qualitative methodologies with significant potential to enhance the resilience of layered smart infrastructures. Our exploration covers classical technological aspects (e.g., cascading effects) and socio-technical factors (e.g., human-in-the-loop interaction).

This study aims to understand the phase transitions, between congested and free-flow states, by estimating the congestion boundary. Since the congestion transition process occurs rapidly, the probability density distributions of traffic flow parameters, i.e. speed and density, show bimodal shapes. The congestion boundary is defined as the threshold that decomposes the bimodal distribution. The dedicated short-range communication data and toll collection system data were used to estimate congestion boundaries on the highway section from Seoul tollgate to Shingal junction. The results show that the congestion boundaries for speed, density, and flow rate were estimated to be 66.9 km/h, 22.8 veh/km, and 341.3 veh/5 min, respectively. With the flow rate at the congestion boundary, the capacity was estimated to be between 71.4% and 82.7% of traditional capacities. The congestion boundary approach provides critical thresholds for each dynamic transition phase on highways, offering crucial insights for effective congestion management.

Data usage control enables data owners to enforce policies for their data, by defining authorizations, but also obligations, which are actions to be performed before, during or after being granted access such as accepting web cookies, and conditions bearing on the system and environment attributes, e.g., the time. Usage control is often coupled with information flow control to monitor how data are propagated. While usage control is well established and modeled in centralized systems, the literature has only partially addressed usage control for distributed systems, for instance by distributing the usage control system components. However, when it comes to assigning policy to certain data, it is always enforced by a central authority. This paper proposes an extended usage control model to integrate decentralized information flow control (DIFC), which enables users to decide collectively which policy to apply to their common data. Functions to handle connection status aspects are also considered, for dynamic Internet of Things (IoT) or peer-to-peer networks where parts of the distributed network can be disconnected. Architectural aspects and formal definitions to enable decentralized policies for shared data are proposed as a novelty, resulting from the integration of DIFC. We used the TLA+ formal specification language on the proposed model and its attached model checker TLC to detect potential issues. We detected potential deadlocks due to the new connection functions as well as temporal ordering issues then suggested mitigations accordingly. A privacy analysis is provided using a car-sharing scenario to highlight the benefits of usage control.