A potential demand model for a multi-circulation feeder network design

TL;DR

This paper introduces a modified demand model and genetic algorithm-based method for designing efficient multi-circulation feeder networks, demonstrated through a case study in Tehran that achieved high coverage and reasonable travel times.

Contribution

It presents a novel combined approach using a modified demand model, constructive heuristic, and genetic algorithm for designing multi-circulation feeder networks.

Findings

Four feeder routes with travel distances of 4.8 to 8.1 km were generated.

The proposed method covered 98% of the target area.

Feasible routes with reasonable travel times were obtained.

Abstract

Background: The development of public transit network can enhance the efficiency of the system as well as raise interest to use the system. Feeder bus service fills the area that is far from the railway system; therefore, designing a feeder network in a gap area causes the expansion of the main transit network. Purpose: To present a modified potential demand model for designing a multi-circulation feeder network. Materials and Methods: In this study, the potential demand is defined based on the traffic demand of each section related to each link, the average walking distance of passengers to specified links, the distance from the links to stations, and finally potential demand of accessing each main station for each link. A labelling method is used to create continuous circular and separated routes. After generating an initial solution using a constructive heuristic algorithm, a genetic algorithm was operated to solve the problem. In the second algorithm, each route is considered a gene, and each network is considered a chromosome. In fact, in this step, different routes from diverse solutions are combined as a unique feeder network. Tehran District 10 was selected as the case study and the model was tested on this area which is located near the central business districts. A machine learning approach has been applied to estimate the missing values of the related database. Results: By running the algorithm, four feeder routes were obtained with reasonable travel times and distances of 4.8, 5.5, 5.8, and 8.1 km, starting and ending at the same railway station. Moreover, 98% of the area was covered by resulted feeder network with a maximum access distance of 300m. Conclusion: What stands out from this method is that the combination of the modified feeder route design model and algorithms is feasible.