An approach to Handling Irregular Oversaturation in Urban Subway Stations

Abstract

This Theses presents a data-based approach for a train scheduling that aims to minimize passenger waiting time by controlling train departure time and the number of skipped trains. In contrast to existing approaches that rely on a statistical model of passenger arrival, we develop a model based on real-world automated fare collection (AFC) data from a metro line in Daegu, a Korean city. The model consists of decomposing the travel time for each passenger into waiting, riding, and walking times, clustering of passengers by trains they ride and calculating the number of passengers in each train for any given time. Based on this, for a given train schedule, the passenger waiting time of each passenger for the entire AFC data period can be calculated. The problem is formulated using the model under realistic constraints such as headway, the number of available trains, and train capacity. To find the optimal solution, we employed a genetic algorithm (GA). The results demonstrate that the average waiting time is reduced up to 56% in the highly congested situation. Moreover, letting the trains directly go to the congested station by skipping previous stations further reduces the maximum waiting time by up to 19%. The effect of the optimization varies depending on the passenger arrival pattern of highly congested stations. This approach will improve the quality of the subway services by reducing passenger waiting time.