Section: Partnerships and Cooperations

International Initiatives

• Title: COntrol and FOrecasting in Transportation networks

• International Partner (Institution - Laboratory - Researcher):

  • University of California Berkeley (ÉTATS-UNIS)

• Duration: 2014 - 2016

• See also: http://necs.inrialpes.fr/v2/pages/comfort/EA_homepage_COMFORT.html

• COMFORT is an Associate Team between Inria project-team NeCS and the Berkeley University project PATH. The joint team is in its 1st year of activity. COMFORT addresses open issues for Intelligent Transportation Systems (ITS). The goal of these systems is to use information technologies (sensing, signal processing, machine learning, communications, and control) to improve traffic flow, as well as enhance the safety and comfort of drivers. It has been established over the past several decades, through field studies and many scholarly publications, that the tools of ITS can significantly improve the flow of traffic on congested freeways and streets. Traffic operators can manage the system in a top-down fashion, for example, by changing the speed limit on a freeway, or by controlling the flow on the onramps (ramp metering). Individual drivers can also affect traffic conditions from the bottom up, by making decisions based on reliable predictions. These predictions must be provided by a centralized system that can evaluate the decisions based on global information and sophisticate modeling techniques. It is now crucial to develop efficient algorithms for control and prediction that are well adapted to current and emerging sensing and communication technologies. The areas of traffic modeling and calibration, state estimation, and traffic control remain central to this effort. Specifically, COMFORT will address issues related to model validation before developing new traffic forecasting and distributed control algorithms. In particular the crucial issue of robustness will be considered through a complementary approach based on both stochastic and deterministic methods. The efficiency of the derived methods will be assessed using large networks simulators and real data obtained from the Californian and the Grenoble's testbed. Three main objectives will be addressed in this collaboration: a) Model validation and robust modeling for traffic estimation, control and forecasting; b) New methods for traffic forecasting; c) New methods for distributed traffic control and estimation.