École Nationale Supérieure de
Techniques Avancées

Résumé

During this internship, we assess different algorithms designed to do the fourth and last step of a four step traffic model : trip assignment. This work specifically focuses on assignment algorithms for cyclist traffic. The need for such an assessment comes from the fact that cyclists have very different criteria choices than car users, for whom the traffic modeling methods have already been studied extensively. Numerous algorithms already in use in Goudappel are assessed, as well as a few more experimental ideas coming from scientific literature. The main outcome of this assessment is that the global load of the network is rather well transcribed regardless of the assignment method, even with the algorithms that are the easiest to implement and cheapest to run scoring high (AON Distance assignment, for instance). Using genetic algorithms in order to use the road types of the links proved to be a fruitful method to improve the accuracy, although we do not yet know how robust this calibration will prove to be on other networks. Experimental algorithms and cost functions were tried, with mitigated success due to the difficulty of data gathering and limitations of the framework.