Model And Algorithm Of Discrete Network Design Problem Under Demand Uncertainty

Network design is the most important part in comprehensive transportation planning. It is the crucial means to realize the planning goals. Demand forecasting is the data input of transport network design. Usually the forecasting results are treated as determinated value in traditional planning. While in practice there are many uncertain fators resulting in decision risk in transportation infrastructure investment. In this study demand is considered as an uncertaint input to transport network design. Discrete network design problem is thoroughly investigated through modeling, algorithm design and case study. The method proposed here can improve the whole efficiency and robustness of planning projects.Demand elements are supposed as stochastic variables. Then the framework of discrete network design under OD demand uncertainty is present based on stochastic bi-level programming and means-vairance model. Expected total system travel time and the standard variance of total system travel time are used to represent the planners'goals. The user equilibrium model under each demand realization is taken in the lower level program. For the discrete transport network design problem under stochastic demand is a combinatorial optimization, Monte Carlo simulation algorithm, genetic algorithm and traffic assignment algorithm are integrated to solve the problem. Numerical resulsts on Nguyen Dupuis network indicate that the final planning consequences are affected by the demand uncertainty level and the planner's risk preference.Based on the same modeling framework, four other aspects of transport network design including demand input, traveler's behaviour, government's planning goal and planning project scheduling are further researched in order to meet different purposes. When OD demand uncertainty is extended to trip production-attraction uncertainty, a discrete network design model based on combined trip distribution/traffic assignment model is formulated to reduce the data requirement. When travelers are divided into different groups by value of time, a novel multi-user discrete network design model based on demand uncertainty is established. Numerical resulsts show that the user heterogeneity has significant impact on planning outcome. Aiming to describe the multi-objective decision process of network design problem, a multi-objective stochastic bilevel programming model combined with personal travel choice under OD demand uncertainty is proposed with the conception of Pareto optimal solutions. Case study demonstrates that there are tradeoffs among construnction cost, expected total system travel time and the standard variance of total system travel time. A multi-stage discrete network design model under OD demand uncertainty is set up to determine the optimal project scheduling when planning period is divided into several phases. Numerical resulsts on Nguyen Dupuis network indicates that the distribution of funds has great impact on project scheduling and the system performance can be improved with more funds in the early days of the whole planning period.At last the process to employ the discrete network design theory and methods proposed in this study is summarized. Test on Sioux Falls network indicates the theory and methods can be used in medium scale network well.