Solving Transit Network Design Problem Using Many-Objective Evolutionary Approach

TLDR: This paper introduces the TNDP as a many-objective optimization problem that generates a diverse set of alternative solutions and finds $\theta $ -DEA to be the most robust among the employed algorithms.

Abstract: In many cities around the world, private vehicles are increasingly causing severe traffic congestion, pollution, and accidents. Public transports have been widely recognized as an effective way to improve urban life. To dissuade citizens from using private vehicles, it is necessary to design a practical, efficient, and economical public bus network. The transit network design problem (TNDP) determines the transit network (i.e., public bus network) for a city. It involves different stakeholders with diverse interests and values. To capture their conflicting expectations, numerous optimization objectives arise naturally. This paper introduces the TNDP as a many-objective optimization problem that generates a diverse set of alternative solutions. We apply several state-of-the-art many-objective evolutionary algorithms for the newly formulated TNDP. To efficiently explore the high-dimensional objective space of the TNDP, we develop problem-specific genetic operators for the evolutionary algorithm. We rigorously tested our approach on several benchmark datasets. The simulation results exhibit the effectiveness of the approach in addressing the challenges of a modern city. Based on the obtained results, we found $\theta $ -DEA to be the most robust among our employed algorithms. In addition, we observed the usefulness of the crossover operator, which randomly combines two solutions into one, and a simple mutation scheme, which is not biased to any objective function, to handle the many-objective nature of the TNDP.