Urban mobility is breaking with the past.
Due to pollution, accelerated urbanisation and network congestion, public authorities now have to deal with new ecological and public health issues.
All these challenges have impelled and accelerated the transformation of mobility through technology.
But before counting on innovation to meet new problems, we have to guide it by studying our mobility behaviour patterns.
Modelling mobility to better understand it
How can we perfectly understand and anticipate the mobility of a city, given that the variables are so numerous, unknown and interdependent? The only possibility is to use an intermediary between the real system and our preoccupations, to obtain answers to the questions that are asked. This intermediary is a model, i.e. a partial representation of reality, which is sufficiently simplified and yet detailed enough, to be able to study it and obtain relevant answers to the questions asked.
We can use such mobility models to better understand and anticipate the future according to different prospective scenarios (development of bicycle use, creation of new pedestrian zones, etc.).
Computers now enable us to make these models dynamic. We then enter the world of computer simulation, a tool that has become indispensable today for the study of complex systems. It is then important to understand that a model will produce relevant answers that are commensurate with the data it can be provided with, and with expert knowledge. SatNav systems, for example, are relevant because they have taken on board certain concepts of the highway code and receive data on traffic conditions in real time. Retrieving data is therefore an essential prerequisite.
Image recognition and Deep Learning algorithms also enable massive retrieval of information about mobility systems. There is no longer any need to install mechanical counting devices. Today, a simple video stream (from a security camera, for example) not only counts moving objects but also automatically classifies them (recognising whether they are cars, buses, pedestrians, bicycles, etc.) and records their characteristics and trajectories.
Data, another essential element
Other resources include the growing availability of Open Data made available by public authorities at both national and local levels. Thus, it is now possible to recover hundreds of geographical layers (GIS layers) capable of providing valuable information about an area in real time. This information is all the more valuable as it is possible to create new intermediate knowledge by cross-referencing and pre-processing it.
Finally, household-trip surveys (or travel surveys) are the ultimate source of mobility data on French cities. These large-scale surveys, standardised by the CERTU (Centre for the study of networks, transport, urbanism and public infrastructures) provide valuable indicators on citizens’ movements (mode of transport, home-to-work journeys, etc.). All these data improve our understanding of mobility and are the building blocks experts use to create models. A model is a simplification of reality that allows diagnoses and predictions to be made according to the data fed into it.
Historically, the field of mobility simulation is mainly based on the mathematical simulation of road traffic using laws inspired by physics and thermodynamics (e.g. pursuit laws, vehicle-mobility generation laws, etc.). However, these mathematical laws only offer global, generic, even disembodied answers because they do not take into account mobility phenomena which are very complex due to their richness (not to mention human behaviour which can sometimes prove irrational).
The use of Artificial Intelligence to better define human behaviour
More recently, to overcome the limitations of mathematical models, a new class of models resulting from collective artificial intelligence is rapidly expanding: multi-agent systems (or agent-based models). These can simulate individual behaviour and therefore mobility based on individualised entities. This means taking into account each individual inhabitant, his or her characteristics and motivations, and simulating their movements around an artificial city. Where classical models rely on systems of equations with no reference to individuals, agent-based models centre simulation around humans and take behavioural characteristics into account in prospective studies.
The example of HealthKer is a perfect illustration of the interest of this type of model for mobility. This programme – developed in partnership with the University of Rennes 1, the AirBreizh association and the Eagle company – is part of SCALIAN Group’s high-profile project to improve health in the city. As a valuable intermediary for understanding the city of today and reflecting on the city of tomorrow, HealthKer proposes an approach that is both systemic and transdisciplinary. By combining consideration of human behaviour, simulation and data from Open Data and sensors, the solution should make it possible to anticipate the complex links between mobility behaviour and its effects on air quality at the city scale. Based on the paradigm of multi-agent systems, HealthKer models the daily movements of the inhabitants of Rennes, in order to estimate greenhouse gas and fine particle emissions very locally. The reliability of these estimates is enhanced by the use of data measured by low-cost stations and prototype sensors.
All risk factors will be addressed: some of them are now fairly well known, such as air or water pollution, others are less obvious, such as the effect of urban travel on the spread of epidemics, or the consequences of weather conditions on accidents related to daily or exceptional activities.
In the same research field, SCALIAN is developing a tool capable of anticipating the impact of different scenarios for the establishment of “third places” (shared workspaces) in order to bring about a lasting change in the distribution of home-to-work journeys on a territorial scale (municipality, metropolis, département). Third places (coworking spaces, fablabs, makerspaces, etc.) are new hybrid spaces between home and work that significantly modify our daily mobility and promote sociability. The aim here is to provide decision-making support for development projects, both for saturated territories and for territories that need to be revitalized. This innovative solution will also enhance the value of open data and help identify unexpected places, which are often victims of territorial neglect.
Mobility, which is closely linked to our conception of society and territories, is a major aspect of our lives. AI can help anticipate the effect of new technologies (new transport solutions, autonomous vehicles, etc.) and public policies (pedestrianisation, bus lanes, urban development, etc.) in line with current ecological and economic challenges. Nevertheless, the results of the models must be interpreted by experts before being communicated to a wider public, strictly within the limits of the scope for which they were designed. And while the results of simulation studies are valuable input for informed decision-making, they should primarily be used to guide more in-depth field studies. The contribution of AI in anticipating the future of mobility is no less considerable and is transforming our way of conceiving cities of the future.