Optimizing bus line platform assignment across bus stations in Utrecht

Abstract

Due to its increased popularity, public transport has grown considerably, which has resulted in more lines that are operated with higher frequencies. As a result, the current bus stations that are located in strategic places like city centers to serve as a hub are becoming too small. If there is no space to enlarge the station, then usually the best option is to create a second bus station close-by. This raises the problem of distributing the bus lines over the bus stations, such that good connections are offered to transfer passengers. We have considered this problem in the city of Utrecht, which is a middle-sized city in the center of the Netherlands. The central transit hub, which is located next to the central railway station, consists of several separate bus stations. The goal of the research is to minimize the total travel time for all passengers who want to transfer at the hub. Here we are not allowed to adjust the current timetable, and we have to take the capacity and vehicle limitations into account. To find out which journeys are made daily and by how many people we use data from a digital fare system. This results in passenger groups, and for each group we compute the relevant travel options given the current timetable. Thereto, the routes are split into an inbound itinerary, a transfer within the same bus station, and an outbound itinerary; the validity of a travel option depends on the assignment of the bus lines to the bus stations. We decompose the problem into first finding a distribution of the lines over the stations and then assigning them to a platform at the station of choice. In the first subproblem, we find the best set of transfers using Integer Linear Programming (ILP), resulting in a station assignment. In the second subproblem, for each station, we distribute the bus lines over the platforms. In this subproblem, there can be multiple lines assigned to a single platform, as long as there are never more vehicles at the platform simultaneously than physically fit. The goal is to maximize the comfort of the transfer passengers by assigning tight transfers to adjacent platforms; this problem is solved using ILP as well.