Study On Key Technologies Of Bus Rapid Transit Network Optimal Design

As a new transit mode, which is supposed to provide better service at relatively low cost, BRT is expected to play a significant role in modern urban transit system. Although the problem of transit network design has been studied for decades, the BRT network optimal design still remains an important issue due to its specific characteristics different from that of the current transit modes, such as vehicle, operation, level of service, and road condition requirement.In this dissertation, we investigated into the multi-modal urban transit system, analyzed its architecture and collaboration among diverse transit modes, and constructed a hyper graph based on this architecture. We also further developed the conception of Trip Strategy mentioned in the past literatures and proposed a choice model to determine both the validity and the utility of trip strategy in the multi-modal transit network. In the choice model, we not only discussed the influence of trip distance and transfer fatigue over the travelers' preference of trips strategies, but also proposed a factor, Seat Expectation, in order to make the model closer to the travelers' strategy- choosing behavior in real world.To solve the BRT network optimal design problem, we applied a bi-level programming model, in which the upper objective is designed to achieve a goal such as to supplement current network service or to guide trip demands according to diversified purposes, and the lower-level problem is a User Equilibrium Assignment of multi-modal transit system, solved through a developed Frank-Wolfe algorithm. Several indexes are elaborately incorporated into the upper-objective to determine whether transit resources of different service levels are distributed equally around a certain area, which, meanwhile, help to perfect the evaluation system of transit network.Accordingly, we designed a hybrid heuristic algorithm, a combination of Genetic Algorithm (GA) and Particle Swarm Optimization (PSO), named Hybrid Particle Swarm Optimization (HPSO), to solve the upper-models, in which we compared traditional GA with HPSO on efficiency and performance. A numerical example follows to test the algorithm we designed, which indicates that the HPSO is better in performance though slower in convergence than the traditional GA. Finally, a simulation system was developed to test the models and the algorithms we proposed, and the system's architecture, functions, and general process are presented in the end.