Junior Seminar

Link identifier archive #link-archive-thumb-soap-46249
Junior Seminar
Mercoledì 10 Maggio alle ore 14:00, Elia Onofri (dottorando, Università degli Studi Roma Tre)  terrà il seminario dal titolo "How machine learning can improve macroscopic models for traffic state estimation and forecast".

Abstract:
Within the last decades, many efforts were devoted to describing the dynamics of vehicular traffic flow on road networks,
either to analyse or to predict normal and abnormal situations (e.g. traffic jams, accidents, ...).
Mathematical models are used, in this sense, to describe many different aspects of traffic dynamics. 
Often, however, the increasing amount of data available through the omnipresent sensors scattered on the road networks
can not be used in their wholeness by mathematical models, hence leaving data full potential concealed.
The temporal-sequence nature of this kind of data makes Artificial Intelligence suitable to unleash its capabilities.

In this talk, we analyse the structure of a set of data provided by Autovie Venete S.p.A. and we present two AI approaches.
Autovie Venete is the company in charge of the management of the A4 Italian highway "Trieste-Venice" and its branches
and it provides us flux and velocity data gathered minute-wise from fixed sensors dispatched every 10--15 km.
We extensively analysed such data and we built two AI methodologies that enable us to:
(i) perform mid-term prediction of the boundary conditions for forecasting simulation,
(ii) make real-time classification of potentially critical situations, hence supporting the model clarifying the nature of the non-bijective fundamental diagram,
(iii) make short-term predictions of critical situations.


References:
 - Link identifier #identifier__112033-1https://doi.org/10.48550/arXiv.2303.12740



Il seminario avra' luogo in presenza presso il Dipartimento di Matematica e Fisica, 
Largo San Leonardo Murialdo, 1 - Aula 311
 
Link identifier #identifier__76342-1Link identifier #identifier__118190-2Link identifier #identifier__80741-3Link identifier #identifier__91516-4