| It is recognized that traffic congestion usually comes out from the socially inefficient choices by individual travelers. To remove or at least to reduce such inefficiency in transportation system, road pricing has long been proposed and advocated by a bunch of economists. Now it further becomes one of the most favorable measures taken by transport policy makers throughout the world. Indeed, significant advance has been made towards better modeling and understanding of the essential mechanism of various road pricing schemes. There, however, still remain some questions around the actual design and implementation. One dilemma often being met is that the travel cost and demand functions are always referred to as the underlying elements in determination of optimal toll levels, but may not be fully available at ease. In practice, even with advanced transport modeling techniques, it is pretty hard to establish all of them.;To overcome this practical difficulty, a novel trial-and-error idea was put forward. It shows us a promising way, but few algorithms have been presented to truly realize it, except for a certain marginal cost pricing problem with unknown demand functions. Therefore, this thesis is focused on the development of practical implementation methods for several typical road pricing schemes including: the first-best pricing scheme, the second-best pricing scheme and the traffic restraint road pricing scheme. The fundamental idea behind the proposed sequential experimental approaches is to adjust toll charges at frequent intervals as traffic patterns changes brought about by price change. Given a set of trial link tolls, the revealed aggregate link flows can be observed easily, and based on the observed link flows, a new set of link tolls can be determined and used for the next trial. Through an iterative toll adjustment procedure, one can expect to identify the efficient toll rates without resort to the inaccessible demand functions. In the case of traffic restraint and pricing schemes, link travel cost functions are also not required. This does allow for traffic planners to estimate or update optimal toll charges easily.;In addition to describing those applicable methods with both theoretical convergence proof and numerical simulation tests, this thesis investigates the maximum efficiency gain of a first-best road pricing scheme as well. In other words, we are able to bound the maximum efficiency loss of any second-best toll patterns being tried in the iterative process. It thus helps us to determine when the sequential experiments could stop since the number of trials in real applications should be very limited. Besides of this, we also attempt to introduce some inexact criterion that allows the observed link flows not necessarily to be exactly in a deterministic user-equilibrium. This relaxed inexact manner reduces the requirement for operating time and resources and thereby makes the developed progressive methods more practical and cost-effective. |
