Traffic Assignment And Generalized Network Design In A Stochastic Road Network

With the rapid development of economics and society, the conflict of urban traffic supply and demand becomes more serious. In order to solve this problem effectively, the transportation infrastructure must be managed and utilized more scientifically, and the time and space distribution of traffic flow must be regulated more reasonably. Thus, the managers and researchers need to understand the theory of traffic assignment firstly. In the real road transportation network, there are many stochastic events (such as bad weather, traffic accidents, road maintenance, etc.) that disrupt the road conditions, and many of these stochastic events are unavoidable. As a result, it is significant both in theory and reality fields to study the road traffic assignment problem under stochastic circumstances and apply the research results to solve urban traffic congestion.Based on the existed research, the main contents contained in this dissertation are as follows.Firstly, a new traffic assignment model named stochastic mean-excess user equilibrium model is proposed. In the model, perceived mean-excess travel time is served as travelers’ route choice criterion. Mean-excess travel time is defined as the conditional expectation of route travel time beyond travel time budget, which reflects travelers’attention to both reliability and unreliability aspects of travel time variability in the route choice decision process. In addition, travelers’perception errors on travel time, the effects on traffic demand caused by the road network service level, and the road network’s topological relationship are taken into account. The calculation formula of mean-excess travel time is derived from conditions when the OD (Origin-Destination) demand follows a Log-normal distribution and the link capacity follows a Beta distribution, the user equilibrium model is built and formulated as an equivalent variational inequality problem, and a heuristic solution algorithm called self-adaptive projection and contraction algorithm is used to solve the proposed model.Secondly, a stochastic mean-excess user equilibrium model taking the differences in travel purposes into account is presented. In the road network, the travelers of the road network are divided into commuters and non-commuters, in which the traffic volume of commuters is deterministic while the traffic volume of non-commuters is stochastic. Both commuters and non-commuters take own perceived mean-excess travel time as separate route choice criteria, but they have different perception errors on travel time and different demand sensitivities. The calculation formula of mean-excess travel time is derived when the commuters’demand follows a Log-normal distribution and the link capacity follows a Beta distribution, and the user equilibrium model is set up and formulated as an equivalent variational inequality problem.Thirdly, considering travelers’heterogeneity in risk taking behaviors, a mixed user equilibrium model is founded. From the perspective of risk taking behaviors, the travelers are classified as risk-neutral travelers, risk-prone travelers, risk-averse travelers, and extremely risk-averse travelers, who take perceived expected travel time, perceived travel time budget (the reliability is less than0.5), perceived travel time budget (the reliability is more than0.5), and perceived mean-excess travel time as their route choice criteria, respectively. The calculation formulas of the travel cost for four types of travelers in the environment with stochastic supply and demand are derived, the user equilibrium model is founded and formulated as an equivalent variational inequality problem.Finally, the single-objective and multi-objective generalized network design models based on mixed user equilibrium are proposed respectively. In the single-objective bilevel programming model, the upper objective is to maximize the consumer surplus of the road network in which both link capacity enhancement and road congestion pricing are put into effect at the same time, while the lower model is a mixed user equilibrium model. In the multi-objective bilevel programming model, the upper objective includes two target functions, that is to maximize the consumer surplus of the road network and minimize the total construction cost in the situations where link capacity enhancement is implemented, while the lower model is a mixed user equilibrium model.