Section: New Results

Transportation networks and vehicular systems

One of the regular performance metrics for qualifying the level of congestion in traffic networks is the travel time. Precision in the estimation or measurement of this variable is one of the most desired features for traffic management. The computation of the travel time is regularly performed based on instantaneous information so called instantaneous travel time (ITT), but regularly traffic changes on time and spaces and the computation depends dynamically on the speeds of the system and the notion of dynamic travel time (DTT) is required. Here the computation requires future information of speed so a short term forecast is required. First in [25] we have presented a framework for instantaneous travel time predictions for multiple origins and destinations in a highway. Secondly in [32], a detailed real time application to compute predictions of dynamic travel time (DTT) is presented. Speed measurements describing a spatio-temporal distribution are captured, from there the DTT is constructed. Definitions, computational details and properties in the construction of DTT are provided. DTT is dynamically clustered using a K-means algorithm and then information on the level and the trend of the centroid of the clusters is used to devise predictors computationally simple to be implemented. To take into account lack of information of cluster assignment of the data to be predicted, a fusion strategy based on the best linear unbiased estimator principle is proposed to combine the predictions of each model. The algorithm is deployed in a real time application and the performance is evaluated using real traffic data from the South Ring of the Grenoble city in France.

The work [19] contains Dominik Pisarski's major results which he obtained during the realization of his Ph.D. thesis at Inria-Rhone Alpes. In concerns the problem of optimal control for balancing traffic density in freeway traffic. The control is realized by ramp metering. The balancing of traffic was proposed as a new objective to improve the vehicular flow on freeways and ring-roads. It was demonstrated that the balancing may result in significantly shortened travel delays and reduced pollution. It may also be beneficial for safety and comfort during a travel. For the controller, a novel modular decentralized structure was proposed where each of the modules computes its optimal decision by using local traffic state and supplementary information arriving from the neighboring controllers. For such a structure, the optimal control problem was formulated as a Nash game, where each player (controller's module) optimizes its local subsystem with respect to decisions of the other players. In comparison to the existing solutions, this new approach significantly reduces the computational burden needed for optimal traffic control, allowing for on-line implementation over long freeway segments. In the paper, the proposed control method was tested via numerical examples with the use of Cell Transmission Model. Later, the performance of the designed method was validated by employing a micro-simulator and real traffic data collected from the south ring of Grenoble. The designed distributed controller resulted in 5% reduction of total time spent on the ring road, 18% reduction of total time spent in the on-ramp queues, 2reduction of the average fuel consumption, and 4% reduction of the traffic density.