Road pricing has long been recognized to be an efficient way to mitigate traffic congestion and to recoup construction and maintenance costs for transportation infrastructure. In view of the implementation difficulty and poor public acceptance of first-best pricing, the second-best link-based and cordon-based pricing schemes are investigated in the dissertation. The optimal selection of both toll levels and toll locations is formulated by Stackelberg games with mixed variables. The upper level objective is to maximize social welfare for elastic demand or to minimize system cost for fixed demand, and the lower level describes users’ route choice behavior. Genetic algorithms and simulated annealing approaches are used to solve the bi-level models. The concept of cutset in graph theory is...[ Read more ]

Road pricing has long been recognized to be an efficient way to mitigate traffic congestion and to recoup construction and maintenance costs for transportation infrastructure. In view of the implementation difficulty and poor public acceptance of first-best pricing, the second-best link-based and cordon-based pricing schemes are investigated in the dissertation. The optimal selection of both toll levels and toll locations is formulated by Stackelberg games with mixed variables. The upper level objective is to maximize social welfare for elastic demand or to minimize system cost for fixed demand, and the lower level describes users’ route choice behavior. Genetic algorithms and simulated annealing approaches are used to solve the bi-level models. The concept of cutset in graph theory is introduced to describe the mathematical properties of a toll cordon by examining the incidence matrix of the network. Furthermore, various equity issues in congestion pricing are examined.

Modeling private highway is another task of this dissertation. Toll charges for private highways depend on the entry and exit points, which are not link-additive. By proposing a network decomposition and transformation approach, the link non-additive traffic assignment problem is solved by the traditional link-additive method. The proposed method circumvents the difficulty of path enumeration or generation frequently involved in general non-additive traffic assignment problems. The optimal toll design problem is solved by a recently developed efficient marginal function approach.

In networks with mixed users following user equilibrium and Cournot-Nash principles, applying the traditional marginal-cost pricing for a system optimum requires that link tolls be differentiated across user classes. We then establish the existence of nonnegative uniform link tolls to support the mixed equilibrium as a system optimum resorting to the dual program approach. Finally, we extend the mixed equilibrium to consider multiclass multicriteria users, and prove that there still exist uniform link tolls for system optimum.