Metodi e modelli matematici per sistemi complessi
Impact of the introduction of autonomous vehicles on vehicular traffic
Thesis in external company
External reference persons Brunella Caroleo (LINKS), Stefano Pensa (LINKS)
Research Groups Metodi e modelli matematici per sistemi complessi
Thesis type MASTER THESIS
Description The penetration of driver-assist/autonomous vehicles and their consequent impact on traffic is an emerging, thus still under-explored, research topic in the realm of Artificial Intelligence.
The modern mathematical theory of vehicular traffic dates back to approximately the beginning of the sixties, when in some pioneering works models of the flow of vehicles along a road were proposed at the three main representation scales: the microscopic one, which describes the vehicles as interacting particles; the macroscopic one, which assimilates the vehicles to a continuum with density; the mesoscopic one, which, inspired by the Boltzmann kinetic theory, studies the statistical distribution of the microscopic speeds of the vehicles. Subsequently, the microscopic and macroscopic models were intensely studied, giving rise to quite complete and refined mathematical theories. On the contrary, the kinetic models were rediscovered only towards the mid-nineties, parallelly to the initiation of the studies on multi-agent systems. Compared to the other two types of models, the strong point of the kinetic approach is the ability to link the elementary microscopic dynamics of an ensemble of interacting particles to the aggregate macroscopic manifestations of the whole system, based on rigorous mathematical-physical theoretical foundations. This makes kinetic models particularly suited to the study of the impact of small percentages of driver-assist/autonomous vehicles on the global traffic flow.
The thesis will be carried out in the UML area (Urban Mobility & Logistic Sytems) of the LINKS Foundation. There will be a strict interaction with the research group of Prof. Andrea Tosin at the Department of Mathematical Sciences “G. L. Lagrange” of Politecnico di Torino.
The thesis project is grounded on three main methodological aspects: modelling, model analysis and numerical simulation, which may be more or less developed according to the student’s taste. After deepening the state of the art on methods for traffic modelling, the student will be asked to develop a mesoscopic model for simulating the flow of vehicles and to analyse alternative penetration scenarios for driver-assist/autonomous vehicles. In particular, the model will be possibly tested on a case study of interest (such as e.g., the city of Turin) using the software MATSim and discussing critically the results. The latter may be used to outline recommendations and suggestions for the public decision maker and as an input to urban planning processes.
Required skills - Familiarity with differential models and partial differential equations
- Knowledge of the main programming languages (MATLAB/Python/C/C++)
- Interest in transport planning
- Ability to critically analyse and interpret the results
 L. Pareschi, G. Toscani. Interacting Multiagent Systems: Kinetic Equations and Monte Carlo Methods, Oxford University Press, Oxford, UK, 2013
 A. Tosin, M. Zanella. Control strategies for road risk mitigation in kinetic traffic modelling, IFAC-PapersOnLine, 51(9):67-72, 2018
 A. Tosin, M. Zanella. Kinetic-controlled hydrodynamics for traffic models with driver-assist vehicles, Multiscale Model. Simul., 17(2):716-749, 2019
 A. Tosin, M. Zanella. Uncertainty damping in kinetic traffic models by driver-assist controls, preprint, 2019 (available at: http://doi.org/10.13140/RG.2.2.35871.41124)
Deadline 16/09/2020 PROPONI LA TUA CANDIDATURA