Solving the trip based transport model using iterative optimization algorithms

This thesis proposes a more robust method for the estimation of lognormal cost function parameters within the trip-based gravity model for transport models. The parameters are currently calibrated using empirical trip length distribution, but the proposed method determines the parameters by using the mathematical relation between the parameters of the trip-based gravity model and the dual variables of the original optimization problem of finding the trip distribution with maximal entropy. First, the current trip-based gravity model together with its derivation from the entropy optimization problem, solving procedure and the currently implemented calibration method is described. Afterwards, the solving procedures for solving NLP of the entropy optimization problem are discussed, together with an extension that creates a hybrid between the lognormal cost function and a discrete cost function. These solving procedures are validated and tested on a realistic transport model for the Dutch city of Almere and its results are compared to those of the trip-based gravity model.