The 5th International Symposium on Dynamic Traffic Assignment (DTA 2014) edition:5 location:Solerno, Italy date:17-19 June 2014
In this paper an approximate gradient based dynamic origin-destination (O-D) estimation method is introduced which has the ability to incorporate congestion in a realistic way and is applicable to networks with route choice options. The method is an extension of the sensitivity based O-D estimation algorithm that exploits a Taylor approximation of the assignment procedure. It is based on a bi-level programming approach with on the upper level the O-D matrix estimation and on the lower level the dynamic traffic assignment. The dynamic traffic assignment itself is also a bi-level problem; with on the upper level the route choice matrix and on the lower level the dynamic network loading model that describes the physics of vehicles traversing over the network. A comparison is made with an estimator that is insensitive to route choice during the local search and it is concluded that our method is able to provide a better estimation of the O-D matrix in case of a network with route choice possibilities and congestion effects.