Monitoring mobility patterns: detection of punctual anomalies and long-lasting disruptions

Offer Description

Keywords

Anomaly detection, mobility monitoring, spatiotemporal analysis, short-term and long-term disruptions.

Context

In recent years, urban mobility patterns have undergone significant changes. Local authorities have implemented green policies to transform the transportation ecosystem, while COVID-19 has drastically altered behaviors. For instance, Paris saw its cyclist modal share increase significatively from 3% in 2010 [1] to 11.2% in 2023 [2]. Additionally, the city implemented a Low Emission Zone (LEZ) restricting certain vehicles from circulating in the territory. Current trends in mobility are thus highly dynamic and dependent on a combination of new constraints, both temporary and long-lasting. However, traditional methods for monitoring mobility rely on surveys only conducted every ten years. Therefore, local authorities must adopt more continuous observation techniques to effectively monitor and respond to local events.

Connected solutions and the growing digitization of transportation provide an enhanced capability to observe mobility patterns with fine detail. Counting stations and public transportation ticketing deliver precise traffic measurements at specific locations, capturing the flow and volume of vehicles and pedestrians. Free-floating vehicle information systems offer valuable insights into the origins and destinations of a significant portion of daily trips; for instance, in the Paris metropolitan area, an average of 171 288 Vélib’ trips were travelled each day in September 2023 [3]. For comparison, between January 2018 and March 2020, on a typical day, 610 000 bicycle trips were estimated [4]. GPS-based data, while representing only a small fraction of the total number of trips, provides widespread spatial coverage, offering a comprehensive view of travel behaviors across various areas. Together, these data sources enable a detailed and dynamic understanding of urban mobility, facilitating more effective planning and policy-making.

In this context, understanding mobility data on a fine spatial and temporal scale is a non-trivial task for municipalities. The complexity arises from many exogenous factors such as weather and calendar events [5,6], making the data noisy and difficult to interpret. The research will support the development of a mobility observatory for local authorities. As part of this, it aims to simplify mobility monitoring for stakeholders within their territories. This will help transition from a reactive approach, where data is used retrospectively to validate existing assumptions, to a proactive strategy in which data is employed to identify and reveal emerging events.

Objectives and approaches

The objective of the post-doctorate is to create innovative methods that identify significant events for local authorities, particularly those not explained by predefined factors such as weather conditions. The approach will focus on two key dimensions: spatial and temporal, and it will be designed to apply to various modes of transportation, including public transportation, bicycles, cars, or the total sum of flows, whose data will be provided. Additionally, the framework aims to detect long-term changes in mobility behavior.

Given the absence of a comprehensive database of all potential traffic disturbances, the methodology will be unsupervised. However, the approaches can be validated using a defined set of known cases. Various methodologies can be employed, including statistical analysis, similarity-based techniques, or pattern mining [8,9].

Overall, the work will be divided into the following tasks:

1. Data structuration and preprocessing. Given the diverse data sources, defining a consistent data format will be key for the generalizability of the forthcoming approach.
2. Definition of an initial approach using a comprehensive dataset, such as public transportation ticketing data, where information about the entire system is largely complete with minimal gaps.
3. Interpretation and categorization of anomalies, e.g., differentiation between sensor-related technical anomalies and genuine traffic anomalies. This requires contextualizing the anomalies and searching for recuring patterns.
4. Generalization of the approach to the broader transportation system and other transportation modes as well as other data sources. For instance, this might involve moving from an emphasis on traffic flows to a focus on Origin-Destination (OD) data.
5. Correlation analysis between anomalies in different transportation modes. This will help investigate if there are common structural causes of anomalies across transportation modes and how a disruption in one mode impacts others, as well as the associated modal shift.
6. Trend deviation phenomena analysis. Changes in trends indicate sustainable alterations in mobility patterns. These changes could result from effective mobility policies. This method will thus help to analyze long-term consequences of actions implemented by local authorities.

Team

The candidate will join an experienced research team in mobility analysis and traffic estimation [5,10-15] with close connections to local authorities. They will furthermore be part of the Mob Sci-Dat Factory project, in partnership with CEREMA, IGN-ENSG, INRIA, and Université Gustave Eiffel, which aims to improve methods for collecting, processing, and analyzing heterogeneous mobility data.

Application

For further information and to apply, please contact:

Alexandre Lanvin, Jean Charléty, Alexandre Chasse

Department of Control, Signal and System, IFP Energies Nouvelles, France

alexandre.lanvin@ifpen.fr, jean.charlety@ifpen.fr, alexandre.chasse@ifpen.fr

Bibliography

[1] Île-de-France Mobilités, OMNIL, DRIEAT (2010). Enquête Globale Transport 2010.

[2] Institut Paris Région (2024). Enquête Mobilité par GPS.

[3] Vélib’ en chiffres (2023, September 26). Vélib’ Métropole. https://blog.velib-metropole.fr/velib-en-chiffres/

[4] Île-de-France Mobilités, OMNIL, DRIEAT (2020). Enquête Globale Transport H2020.

[5] Lanvin, A., Michel, P., Charléty, J., & Chasse, A. (2024). Weathering heights: An updated analytical model of the nonlinear effects of weather on bicycle traffic. Journal of Cycling and Micromobility Research, 2, 100031.

[6] Petrović, D., Ivanović, I., Đorić, V., & Jović, J. (2020). Impact of weather conditions on travel demand–the most common research methods and applied models. Promet-Traffic&Transportation, 32(5), 711-725.

[7] Comptage vélo – Données compteurs (2024, July 17). Paris Data. https://opendata.paris.fr/explore/dataset/comptage-velo-donnees-compteurs/dataviz/?disjunctive.id_compteur&disjunctive.nom_compteur&disjunctive.id&disjunctive.name

[8] Djenouri, Y., Belhadi, A., Lin, J. C. W., Djenouri, D., & Cano, A. (2019). A survey on urban traffic anomalies detection algorithms. IEEE Access, 7, 12192-12205.

[9] Zhang, M., Li, T., Yu, Y., Li, Y., Hui, P., & Zheng, Y. (2020). Urban anomaly analytics: Description, detection, and prediction. IEEE Transactions on Big Data,8(3), 809-826.

[10] Feng, G., Diallo, A. O., & Chasse, A. Improving bike travel demand generation with dynamic data: an application to the Paris Metropolis.

[11] Feng, G., Jean, M., Chasse, A., & Hörl, S. (2021). Pre-calibration of a Discrete Choice Model and Evaluation of Cycling Mobility for Île-de-France. Procedia Computer Science, 184, 172-177.

[12] Michel, P., & Chasse, A. (2024). Aide à la décision au choix d’un véhicule. Les Techniques de l'Ingenieur, TRP904.

[13] Othman, B., De Nunzio, G., Laraki, M., & Sabiron, G. (2022, October). A novel approach to traffic flow estimation based on floating car data and road topography: Experimental validation in Lyon, France. In 2022 IEEE 25th International Conference on Intelligent Transportation Systems (ITSC) (pp. 2571-2576). IEEE.

[14] Laraki, M., De Nunzio, G., & Othman, B. (2022, October). A large-scale and data-based road traffic flow estimation method leveraging topography information and population statistics. In 2022 IEEE 25th International Conference on Intelligent Transportation Systems (ITSC) (pp. 2343-2349). IEEE.

[15] Carneiro, S. A., Chierchia, G., Charléty, J., Chataignon, A., & Najman, L. (2023, June). Swmlp: Shared weight multilayer perceptron for car trajectory speed prediction using road topographical features. In 2023 8th International Conference on Models and Technologies for Intelligent Transportation Systems (MT-ITS) (pp. 1-6). IEEE.

Where to applyE-mail : alexandre.chasse@ifpen.fr

Requirements

Research Field

Mathematics » Statistics

Education Level

PhD or equivalent

Skills/Qualifications

The candidate must hold a PhD in statistics, machine learning, transportation science or related discipline, with experience in anomaly detection. Proficiency in Python is required. The candidate must also demonstrate excellent written and verbal communication skills in English. Knowledge of French is a plus.

Specific Requirements

Level C1 in English. Willingness to learn French.

Languages: ENGLISH Level: Good

Work Location(s)

IFP Energies nouvelles, 1 et 4 avenue de Bois Préau, 92852 Rueil-Malmaison, France