< Back to previous page

Publication

Transit Network Design: Narrowing the Gap between Theory and Practice

Book - Dissertation

Appropriate public transport systems are crucial in modern cities. Given the high costs that they represent and the impact they have on people's lives, effective tools are required to support their design. With this in mind, the Transit Network Design and Frequency Setting problem (TNDFSP) has been extensively studied in the domain of Operations Research. The TNDFSP aims to optimize the sequence of stops to visit and the frequency for each bus in the network. However, due to the complexity of the problem, previous research. Adtionall In this dissertation, several variants of the TNDFSP are addressed, incorporating assumptions that make it more realistic. The most relevant assumptions included are: considering a heterogeneous fleet, using a set of discrete frequencies, considering a limited set of nodes to be used as terminals, considering the effects of crowding in the passenger route choice, and including edge capacity constraints. A bi-objective approach is considered, in which at the same time the passenger travel time and the operatorĀ“s cost are minimized. Furthermore, in one of the variants of the problem, the minimization of the CO2 emissions is considered. A memetic algorithm is proposed to solve these problems. The good performance of the algorithm is demonstrated by comparing the results with the state-of-the-art. The proposed memetic algorithm obtains solutions as good or better than the best ones found before. Moreover, it obtains these solutions in less than one hour of computing time, even for the largest instances. The results also show the importance of considering the assumptions included in the extended versions of the problem in the tested instances. For example, better solutions are found when a heterogeneous fleet is considered. Moreover, the results show the relevance of considering the CO2 emissions explicitly as an optimization objective. Compared to the solution that only minimizes the travel time, the emissions can be reduced by 13% with an increase in the travel time of only 1.3%. Similarly, the results show that the algorithm can successfully address the problem considering the effects of crowding and the edge capacity constraints. The results show the importance of including these constraints already in the network design problem since, otherwise, the generated networks could turn out to be unfeasible later on. Additionally, a case study is performed to test the algorithm's potential of being useful in practice. A new realistic instance is generated using real data from the city of Utrecht, The Netherlands. The proposed memetic algorithm is used to generate alternative line plans, generating a set of non-dominated solutions that dominate the solution representing the current network. For example, there are solutions with the same fleet size as the current network but 6% lower passenger travel time. Important differences are identified compared to the current network, such as a smaller number of lines or leaving some demand unsatisfied. Further experiments demonstrate that the proposed solutions perform well in different demand scenarios, despite they are generated considering only the demand of morning peak hours. Moreover, it is shown that the computing time can be reduced by a 50% by ignoring OD pairs with very low demand, without affecting the quality of the solutions. After presenting the results to the bus operator in Utrecht, they confirm that the model could be useful to support the decision making in a future redesign of the bus network. Additionally, a new metric is proposed to easily compare two transit networks according to their topology. The results show that this metric could be useful as an additional input when selecting a compromise solution from a set of non-dominated solutions. Overall, the results demonstrate the capability and flexibility of the proposed algorithm to solve large and realistic instances of different variants of the TNDFSP. Moreover, the algorithm found good solutions in a very reasonable computing time. Considering the strategic nature of the problem, it seems reasonable to include new assumptions to make the modelling even more realistic, even if the computing time is increased. Indeed, despite the improvements proposed in this dissertation, there is still a gap between the simplified problem studied in theory and the extremely complex problem found in practice. However, the results of the case study show the great potential that the optimization methods have to support the design of more efficient and sustainable transit networks in practice.
Number of pages: 2
Publication year:2021
Accessibility:Open