This thesis deals with so-called probit-type trip assignment models for uncongested situations, with special attention given to stochastic simulation approaches for complete networks. This class of assignment models is typified vis-a-vis other assignment approaches for uncongested problems. A detailed description and critical analysis is given of the behavioural foundations of this model type using the random-utility discrete-choice theory as a starting point. An efficient performance of a probit model based trip assignment on a network scale is only feasible using a simulation approach. In this case we use the term stochastic assignment. The various aspects of the simulation approach (random number generation, randomization sequence, number of iterations, etc) are critically reviewed and basis procedures of performing the simulations are compared with each other. This analysis leads to suggestions with respect to the specifications of the route-choice impedance function as well as about the best simulation procedure. It is recommended among other things to perform the stochastic assignment in a recursive fashion as a series of all-or-nothing assignments on a randomized network keeping the number of iterations dependent on the required output accuracy. Formulae are presented with which the convergence and the output accuracy can be examined. The problem of parameter calibration has been dealt with in a separate chapter. The thesis concludes with recommendations about to execution of stochastic trip assignments in planning practice.
Abstract