Solving transit network design problem using evolutionary many-objective approach

Abstract

Nowadays, the increasing private vehicles have caused severe tra c congestion, environmental pollution and road accidents in many cities around the world. As such, public transport has been widely recognized as an e ective way to improve urban life. To develop public bus service as a competitive alternative to private vehicles, we must design a practical, e cient and economic transit network. A transit network is composed of several connected routes for public buses. Transit Network Design Problem (TNDP) determines the transit network for a city while achieving some objectives and maintaining some constraints. In this modern age, TNDP involves di erent stakeholders with di erent interests and values. As a result, a large number of optimization objectives arise naturally. However, most of the researchers have ignored this issue by using single objective function to express solution quality. In this thesis, we proposed a new formulation for the many-objective TNDP which allows to generate a diverse set of alternative solutions. Then we developed problem speci c genetic operators for solving TNDP using evolutionary algorithms. For the rst time, we adapted several state-of-the-art many-objective evolutionary algorithms (MaOEAs) to explore the high-dimensional objective space of TNDP using our genetic operators. Our MaOEA based approach has been rigorously tested with several benchmark datasets. The experimental results show that, the proposed methodology is more e ective in addressing modern challenges than the existing approaches.