Research interests
Road traffic
  • There are some strong similarities between road traffic and fluid flows. As a consequence, the modeling methods are quite similar, making it quite natural for a physicist to get some interest in the subject. Until now, I concentrated mainly on highway traffic, on the point of vue of modeling or of data analysis (single car data obtained with magnetic loops)
  • Modeling
    • In collaboration with Ludger Santen (Lab. of Theoretical Physics, Univ. des Saarlandes), we have studied how the introduction of a reaction time in a simple transport model could introduce some metastability in the system. There are strong evidences that metastability also occurs in real traffic, and our model suggests one basic mechanism to explain this phenomenon.
    • In collaboration with H. Hilhorst and G. Schehr, we have proposed a kinetic model for a bidirectional road, with an explicit modeling of passing requirements, and shown that a spontaneous asymmetry could arise between the lanes, in a parameter range that should allow to observe it on national roads.
  • Analysis of single vehicle magnetic loop data :
    • I have proposed a simple scenario of rear end collisions in platoons, and used it as a criterium to evaluate the dangerosity of one-lane structures observed in real data. The impact of several security measures could also be discussed (collaboration with the Laboratoire des Ponts et Chaussées (Paris)).
    • The great amount of data I had allowed me to measure with high precision some velocity and time headway correlations, either on the same lane, or on adjacent lanes. Some platoons composed of cars very close to each other and moving almost as a solid block could be seen, involving up to 6-7 vehicles.
    • I have made the proposal that the response of a highway system to a local perturbation (on-ramp) could be used to test the validity of microscopic lane changing rules. Data analysis allowed to extract some basic properties of the relaxation length of the lane fractions after an on-ramp. The methodology was illustrated on a cellular automata model, but could be generalized to any simulator, including the commercial ones.