Modélisation statistique et probabiliste du TEMPS INTER-VEHICULAIRE aux différents niveaux de trafic

Time Headway (TH) is a microscopic variable in traffic flow theories that has been studied since the 1930s. Distribution of this fundamental variable describes the arrival pattern of vehicles in traffic flow, so probabilistic modeling is the main approach to study TH and represents driving behavior. The applications of the variable in traffic engineering are varied ; include capacity calculation, microscopic simulation, traffic safety analysis, etc. This dissertation aims at modeling the TH distribution in different contexts. Firstly, the short-time sampling method and long-time sampling method are applied to obtain TH samples from the two data bases (the RN118 national roadway and the A6 motorway). Then, three probabilistic TH model types are analyzed and classified. An exhaustive comparison between the existing models and between the corresponding estimation methods leads to consider that the gamma-GQM is the best TH model in the literature. An estimation process is also proposed in order to obtain good and stable estimated results of the parameters. After that, the TH probabilistic modeling is developed by six new models. Except for the two ones which are worse, the four other models are statistically equivalent and/or better than the gamma-GQM. For practical reason, the Double Gamma model is selected, as a comparison model, with the gamma-GQM to calibrate all TH samples. Three traffic levels are considered : macroscopic, mesoscopic and microscopic. The effects of exogenous factors are also examined. Examining this factor in each macroscopic variable level leads to distinguish two following factor types : "impeding" factor and "propulsive" factor. Finally, different approaches for TH validation are tested. The proposed approach of "envelope of distributions" seems to be promising for future applications.