Sustainable Deployment of Mobility on Demand via Spatial Machine Learning Methods

3 Nov 2023
Job Information

Organisation/Company

Institut Polytechnique de Paris

Research Field

Engineering
Computer science » Modelling tools

Researcher Profile

First Stage Researcher (R1)

Country

France

Application Deadline

29 Feb 2024 – 00:00 (Europe/Paris)

Type of Contract

Temporary

Job Status

Full-time

Hours Per Week

7

Is the job funded through the EU Research Framework Programme?

Not funded by an EU programme

Is the Job related to staff position within a Research Infrastructure?

No

Offer Description

This is a fully funded (3 years monthly salary) PhD. Starting date can be from now to January 2024. The PhD is co-supervised by:

• IPP, Institut Polytechnique de Paris (Palaiseau – Paris Region), excellence research and education institute (21st in Engineering in the QS World Ranking)[Cl23]  
• Padam Mobility (Siemens Mobility Group – Paris), enterprise that develops digital solutions for Mobility on Demand across Europe
• INRIA (Palaiseau – Paris Region) French National Research Institute on Informatics

 

You will develop geostatistical and Machine Learning methods for planning Mobility on Demand, in order to make human mobility more sustainable, from an environmental, economic and societal point of view.

Context

Mobility on Demand (MoD) services are offered via a fleet of vehicles that adapt their route on the fly to the observed user requests. Today, the efficiency of MoD is measured by basic metrics, such as the number of passengers served or average delay. In this PhD we want instead to construct MoD around accessibility,[Mi19,Lo19] an indicator that measures how many opportunities (jobs, schools, shops, etc.) one can reach within a limited time, starting from a certain location. In the areas where Public Transport (PT) offers good accessibility, people do not “depend upon” their (polluting) car to participate in society. Accessibility is thus a necessary condition for social,[Ma17] economic[Ca19,Ma17] and environmental[Sa22] sustainability. Unfortunately, PT provides insufficient accessibility in the suburbs. In such areas, MoD has been shown to be more efficient than traditional PT.[Ca23] We aim to exploit the potential of MoD with the objective to reduce the accessibility gap between city centres and suburbs.

Objective

The ambition of the PhD is to answer the following research question: given a territory, which improvement of accessibility (in terms of opportunities reached within 1 hour)  can we obtain via MoD? Currently, no method exists to estimate accessibility provided via MoD, based on real data. The main challenge is that accessibility is an indicator computed on a graph. While traditional PT can be modelled as a graph, MoD cannot, due to the dynamicity and stochasticity of vehicle routes.[Mo16] Our novel idea consists in modelling travel times in MoD as a random field,[Di23] which we can estimate via geostatistical methods and Machine Learning (ML), based on real observed trips and territory characteristics. Such estimations will allow us to build a virtual graph, on which we will compute accessibility. Our approach is multidisciplinary, as it combines Transport and Data Science, taking also into account urban planning concepts (concerning the distribution of opportunities across the territory), such as the 30 minutes city.[Ne16] This PhD will shed new light on Mobility on Demand, by proposing state-of-the-art quantitative methods to support human-centred deployment of territories.

Methodology

Accessibility computations will be based on open data[OSM, GTFS, Gr, INSEE, Sirene, Po] and historical real trips data from Padam Mobility. We will explore geostatistical methods such as Kriging,[De18,Ya85] uncertain networks,[Sa21] ML.[Liu20,He18,Is21]  To exploit the similarity between different territories, we will apply the concept of transfer learning, which has shown to be promising for spatial estimation.[Wu21,Od17] We will use open source software (e.g., CityChrone,[Lo19] qgis) as well as the development tools of Padam Mobility

Requirements to apply

Master 2 (or equivalent) in Engineering, Computer Science, Transport Science, Statistics or Applied Mathematics. Excellent analytical skills and programming skills. Knowledge of basic statistics. Experience with ML, Geostatistics, Geographical Information Systems is a plus. 

Interested candidates should send (i) a CV, (ii) an explanation of their previous relevant projects or skills (it can be included in the CV or as a separate document of no more than ½ page), (iii) all the marks of all the courses at BSc and MSc level, (iv) at least one recommendation letter, (v) a motivation letter. Such documents should be sent to Assoc. Prof. Andrea Araldo ([email protected] ) and Louis Zigrand ([email protected] ) and Dr. Aline Carneiro Viana ([email protected] ).

 

References

[Ca19] Camarero, L., Oliva, J. Thinking in rural gap: mobility and social inequalities. Palgrave Commun 5, 95 (2019). https://doi.org/10.1057/s41599-019-0306-x

[Ca23] Calabrò, G., Araldo, A.,… & Ben-Akiva (2023). Adaptive Transit Design: Optimizing Fixed and Demand Responsive Multi-Modal Transport via Continuous Approximation. In Transportation Research Part A.

[Cl23] https://www.ip-paris.fr/en/about/rankings

[Di23] Diepolder, S., Araldo, A., Chouaki, T., Horl, S., Maiti, S., Antoniou, C., On modelling accessibility of flexible mobility, 11th Symposium of the European Association for Research in Transportation (hEART) 2023

[GTFS] The Mobility Database. database.mobilitydata.org.

[Gr] Gridded population of the World v4. Socio Economic Data and Application Center – NASA. URL: https://sedac.ciesin.columbia.edu/data/collection/gpw-v4/ .

[Ha94] Handcock, M. S., & Wallis, J. R. (1994). An approach to statistical spatial-temporal modeling of meteorological fields. Journal of the American Statistical Association(426).

[He18] Hengl, T., Nussbaum, M., Wright, M. N., Heuvelink, G. B., & Gräler, B. (2018). Random forest as a generic framework for predictive modeling of spatial and spatio-temporal variables. PeerJ, 6, e5518.

[INSEE] Couples – Familles – Ménages en 2018, link

[Is21] Ismail, Hamza Y., et al. “Modelling of yields in torrefaction of olive stones using artificial intelligence coupled with kriging interpolation.” Journal of Cleaner Production 326 (2021): 129020.

[Liu20] Appleby, Gabriel, Linfeng Liu, and Li-Ping Liu. “Kriging convolutional networks.” Proceedings of the AAAI Conference on Artificial Intelligence. Vol. 34. No. 04. 2020.

[Lo19] Biazzo, I., Monechi, B., & Loreto, V. (2019). General scores for accessibility and inequality measures in urban areas. Royal Society open science.

[Ma17] G. Mattioli. “’Forced Car Ownership’ in the UK and Germany: Socio-Spatial Patterns and Potential Economic Stress Impacts”. In: Social Inclusion 5.4 (2017)

[Mi19] E. Miller. “Measuring Accessibility : Methods and Issues”. In: International Transport Forum Roundtable on Accessibility and Transport Appraisal. 2019.ù

[Mo16] Fortin, P., Morency, C., & Trépanier, M. (2016). Innovative gtfs data application for transit network analysis using a graph-oriented method. Journal of Public Transportation.

[Od17] Oda, H., Kiyohara, S., Tsuda, K., & Mizoguchi, T. (2017). Transfer learning to accelerate interface structure searches. Journal of the Physical Society of Japan, 86(12), 123601.

[OSM] Open Street Map. www.openstreetmap.org .

[Po] Population active occupée selon les catégories socio professionnelles des communes d’Île-de-France (données Insee), link

[Sa21] A. Saha, R. Brokkelkamp, Y. Velaj, A. Khan, and F. Bonchi, “Shortest paths and centrality in uncertain networks,” Proceedings of the VLDB Endowment, vol. 14, no. 7, pp. 1188–1201, 2021.

[Sa22] Saeidizand, P., Fransen, K., & Boussauw, K. (2022). Revisiting car dependency: A worldwide analysis of car travel in global metropolitan areas. Cities, 120, 103467.

[Sirene] Base Sirene, sirene.fr/ 

[Wu21] Wu, Y., Zhuang, D., Labbe, A., & Sun, L. (2021, May). Inductive graph neural networks for spatiotemporal kriging. In Proceedings of the AAAI Conference on Artificial Intelligence

[Ya85] Theorem 2.3 de “Yakowitz, S. J., & Szidarovsky, F. (1985). A comparison of kriging with nonparametric regression methods. Journal of Multivariate Analysis.”

Requirements

Research Field

Engineering

Education Level

Master Degree or equivalent

Additional Information
Work Location(s)

Number of offers available

1

Company/Institute

Institut Polytechnique de Paris

Country

France

City

Palaiseau

Postal Code

91120

Street

19 place Marguerite Perey

Geofield

Where to apply

E-mail

[email protected]

Contact

City

Palaiseau

Website

https://www.ip-paris.fr/en

Street

19 place Marguerite Perey

Postal Code

91120

E-Mail

[email protected]

STATUS: EXPIRED