Network Science and Machine Learning for Resilience Characterisation and Optimisation of Large-scale

26/09/2021 – 27/09/2021 all-day

Offre en lien avec l’Action/le Réseau : – — –/Doctorants

Laboratoire/Entreprise : LICIT (Transport and Traffic Engineering Laborator
Durée : Three years
Contact :
Date limite de publication : 2021-09-26

Contexte :
In the last decade, the multi-modal mobility system of large cities has been profoundly jeopardized by a variety of sudden and extreme perturbations [1]. According to the World Economic Forum’s Global Risks Report 2019, extreme weather events are among the global risks of highest concern. Heavy precipitation, along with associated flooding in urban mega-regions, has been on the rise both in intensity and frequency under the dual forcings of climate change and rapid urbanization [2]. Similarly, in recent times, the COVID-19 pandemic has radically transformed human mobility habits, leading to globally unprecedented decline in transit ridership as well as drastic reduction of capacity of transit as a consequence of social distancing [3].

These factors of vulnerability related to transport are exacerbated by the fact that a transportation network is a complex entity composed of multiple interdependent subsystems (underground, train, tramway, bus transit, and road network), which are spatially constrained and that also rely on other urban infrastructure systems such as the power grid and communication networks. Thus, even limited disruptions in one component of this complex system, often triggered by exogenous and hardly predictable events, can lead to a severe loss of lifeline functions via cascading failures. Furthermore, as urban mobility systems are becoming increasingly connected and autonomous, one should also consider the growing threat of opportunistically targeted cyber-attacks designed to take advantage of natural hazard events [2].

This thesis focuses on the cutting-edge issue of characterizing, monitoring, and improving the resilience and robustness of complex systems, and, particularly, urban mobility systems of large metropolises. To advance the state-of-the-art in the field it is nowadays necessary to mobilize knowledge and skills from multiple research domains, including artificial intelligence and data science, mathematical optimization and graph theory.

In this context, it is expected to investigate hybrid approaches leveraging the combination of graph theory and data science, machine learning and, possibly, mathematical programming tools towards:
– advancing the understanding and characterisation of the resilience of temporal multi-modal urban mobility systems by means of graph modelling, network metrics and the analysis and learning of historical data available for a real-world multi-modal transport network (Lyon, France);
– defining original techniques and tools to design complex disruptive scenarios, that could couple targeted attacks, weather-related phenomena as well as sudden variations of the mobility demand and offer of the transport system induced by exogenous factors (floods, pandemic, etc.);
– evaluating their impacts on the performance of the existing transit system in terms of network science metrics and, possibly, by means of simulation.

The thesis may also explore solutions for resilience enhancement based on network redesign strategies, including:
– topological optimization based on network metrics computation, graph representation learning, and graph nodes/edge addition/removal;
– design and virtual deployment of on-demand mobility facilities (e.g., park-and-ride) that could help to support the dynamic adaptation of the system to such variations and rapid recovery from extreme perturbations with increased resilience.

Sujet :
Main activities

The thesis program will develop around the following scientific challenges:
– Modelling and coding of the multi-modal transport network of the Lyon urban area, by focusing on its transit system (bus, tramway, underground) and the city road network. An approach based on multi-layer networks [1, 4, 5] could be leveraged by relying on data from the National Institute of Geography (IGN) and from the local provider of the transit system of Lyon (Keolis-Sytral). The augmentation of the model with travel demand information and travel time data will be considered as an essential research direction, based on previous work from the team [6].
– Identification of complex networks metrics to describe the resilience and robustness of the multi-modal transport network. In particular, the size of the giant connected component (GCC), network efficiency, adapted to the context of multi-layer modelling and cascading failures [2], will be a potential candidate for robustness quantification in dynamic configurations. Additional metrics related to vulnerability, robustness, and resilience for characterizing the performance of transport systems under disturbance will be explored as well [7].
– Definition of a framework for the injection of multiple joint failures in the multi-modal transport system (disruptive scenario testing). More traditional strategies based on random failures as well as more complex approaches involving flood probability modelling and high centrality node failures will be investigated to simulate high-risk scenarios and evaluate their impact on the aforementioned robustness metrics. The expected solutions should allow modelling of compound disruptions, including flood scenarios combined with targeted attacks as well as global reduction of the transit capacity or travel demand.
– Analysis of optimization strategies for improvement of network robustness. Solutions based on optimal graph augmentation [8, 9], identification of the most critical sub-network, as well as the optimal allocation of on-demand mobility facilities (e.g., park-and-ride facility location [10]) for increased network robustness, will be investigated.

The PhD program is highly flexible and can be adapted based on the profile and skills of the candidate.

Profil du candidat :
Level of qualifications required:

Master-2 degree, 5-years engineer diploma or equivalent in Computer Science, Data Science, Machine Learning, Statistics, Mathematics or strictly-related field.


The ideal candidate should be highly motivated in carrying out research activity, possess high scientific curiosity and a high-degree of autonomy, and own a documented expertise in network science, data science, programming, and machine learning. A Master 2 or equivalent degree in one of the fields above is required.

Good knowledge of R and/or Python languages is also required.

Proven written and verbal communication skills with fluency in written and spoken English are a must.
Knowledge or previous experience with operations research tools, transport modelling, (traffic) simulation will be helpful and highly appreciated during the selection process.

The phd program is highly flexible and can be adapted based on the profile and skills of the candidate.

How to apply
Please send an email to:
including the following elements:

A curriculum vitae;
The complete record of master grades (relevé de notes M1 and M2 for French candidates)
A two-page-most motivation letter discussing how the candidate’s background and research interests relate to the proposed subject and bibliographic references.
Recommendation letter from supervisors, if any.
All applications without the elements above will be ignored.
Selection process
All received applications will be short-listed upon reception. The candidate will receive a reply only in case her/his application is selected for an interview. Multiple interviews could be necessary.

After the interviews with tutors, the candidate will be asked to prepare a presentation discussing the Ph.D. topic, personal research ideas, as well as a tentative schedule to carry out the research activities, according to her/his own perception of the subject.

A final interview with an external committee (expected around mid-September) will take place to validate the candidature.

Formation et compétences requises :
Research Objectives
The PhD student will have to advance current methodologies developed within the team on resilience analysis and optimization of complex transportation networks.
She/he will have to develop as well novel and efficient approaches based on network science, resilience metrics, and perturbation scenario design towards characterization of the resilience of the multimodal transport system.
It is expected that the successful candidate will contribute to top-tier network science, machine learning, data analytics and transportation conferences and journals (NetSci, Networks, Applied Network Science, IEEE Transactions on Network Science and Engineering, International Conference on Data Mining, KDD, Transportation Research Board, IEEE Intelligent Transportation Systems, Transportation Research Part B, C, E, etc.).

Adresse d’emploi :
The Transport and Traffic Engineering Laboratory (LICIT) is a Joint Research Unit under the dual administrative supervision of the French University Gustave Eiffel (UGE) and the National Post-Graduate School of Public Civil Engineering (ENTPE). It is recognized for its work on traffic modelling and engineering. The laboratory has already developed many successful applications for both traffic information and simulation tools.

Address: 25, avenue François Mitterrand, Cité des mobilités. F-69675 Bron (France)

More information on the lab are available here: