A new solution framework for the limited-stop bus service design problem. (November 2017)

• Record Type: Journal Article
• Authors: Soto, Guillermo; Larrain, Homero; Muñoz, Juan Carlos
• Is Part Of: Transportation research. Volume 105(2017)
• Page Start: 67
• Page End: 85
• Publication Date: 2017-11
• Subjects: Limited-stop service -- Transit network design -- Bus rapid transit -- Stochastic passenger assignment -- Bi-level optimization; Transportation -- Research -- Periodicals; Transportation -- Mathematical models -- Periodicals
• DOI: 10.1016/j.trb.2017.08.026
• Languages: English
• ISSNs: 0191-2615

Abstract

Highlights: We present a framework for addressing the limited-stop service design problem over a corridor, and formally introduce a family of subproblems involved in its solution. We introduce a bi-level optimization method to design these services while considering bus capacity, transfers, and even different types of behavioral models for the passengers. We drastically reduce the solution times of a benchmark algorithm on a deterministic setting. We show that, although it makes the problem much harder, working with stochastic assignment leads to more realistic and robust solutions. We show that neglecting transfers can lead to suboptimal solutions. Abstract: Limited-stop services are a key element to the successful operation of bus rapid transit corridors. In this study, we present a framework for addressing the limited-stop service design problem over a corridor, and formally introduce a family of subproblems involved in its solution. Using a bi-level optimization approach, we introduce a method of designing these services while considering bus capacity, transfers, and two behavioral models for passengers: deterministic and stochastic. The algorithm and its variants were tested on nine scenarios with up to 80 stops. Working with deterministic passenger assignment, our model solved the problem in a small fraction of the time required by a benchmark algorithm. We use this algorithm to show that neglecting transfers can lead to suboptimal solutions. We finally show that although it makes the problem much harder, working with stochastic assignment leads to more realistic and robust solutions.