Discrete Transportation Network Design Under Stochastic Demand Among OD Pairs On Account Of Friendly Environment

Transportation network design in a broad sense is the configuration of network, which aims to solve urban traffic congestion problem by building new roads/lanes for the network or expanding link capacities on existing roads. Conventional transportation network designs mainly consider transportation system impedance and investment cost, while ignoring emission paradoxes and the origin-destination(O-D) demand uncertainty. Hence, it will pose a serious threat on future transportation network. Based on conventional transportation network design, this thesis organically combine environmental concerns and OD demand to improve the overall efficiency and robustness of the transportation planning.In this thesis, automobile exhaust pollution was analyzed quantitatively, Determine the emission prediction model vehicles based on the contents of this paper, and then the uncertainty of OD demand was also taken into account. With comparison of four type of optimization method, the mean model from stochastic programming theory has ultimately been chosen to be used as framework for modeling optimization in this thesis. And then a bi-level programming model of discrete transportation network design was proposed under stochastic demand among OD pairs on account of friendly environment. The objective of the upper-level programming model was constructed to minimize the summation of network impedance, investment costs and CO emission. The constraint conditions are set as land resources utilization and acoustic environment; the lower-level programming model was constructed using the user equilibrium assignment model. According to the characteristics of the bi-level programming level, Monte Carlo simulation and genetic algorithms were integrated to solve the problem effectively. In order to increase computing speed influenced by multiple loops in algorithm, uniform design test was applied and the optimal parameter combination was ultimately obtained, accelerating effective convergence of genetic algorithm.Finally, the model and algorithm in the thesis have been tested on the classic network.Referring to the thinking of solving slope in linear equations, the weight coefficients of the three goals has been calculated. With comparison of the conventional traffic network designmodel on this basis, the results of numerical examples indicate that the system impedance can be considerably reduced, and the traffic environment can be more effectively improved in the proposed network model.