| In the United States, railways are the major means to trans-continentally move goods from ports to the various inland destinations. Due to mergers and abandonment of rail lines, there has been a reduction in the track capacity, concentrating rail traffic to fewer lines. In addition to this, the growth in the number of containers has already introduced congestion and threatened the capacity of the rail network system in many corridors. There is a need among U.S. freight railroads for better analytical tools to manage their capacity and scheduling. A challenging problem for railroad companies is to be able to plan the traffic and operating conditions over a network so that deadlocks are avoided and travel-times are below a threshold. This requires estimating travel-times and delays in a network, and determining the most efficient method of scheduling a set of trains.;The focus of this research is the development of a decision tool that can aid train planners in developing good quality routes and schedules, on a daily or weekly basis, within a short amount to time, to better manage the limited track capacity available for train movements. In daily operations, dispatchers use these plans to decide the movement of trains through a network. Due to extraneous factors such as train breakdowns, track maintenance etc., routes might need to be altered on a real-time basis. Dispatchers ensure that any such deviations from the planned routes and schedules are kept to a minimum, so as to minimize congestion and avoid deadlocks. In addition, dispatchers resolve conflicts between train schedules in real-time and determine priorities when trains meet at junctions, crossings and sidings. In this work, we concentrate ourselves to railroad routing and scheduling alone.;Towards achieving this goal, we develop an integer programming-based railway capacity management model that is capable of assigning trains to routes based on the statistical expectation of running times in order to balance the railroad traffic. This model is also capable of determining the best release times for trains to depart from origin stations and enter a network. We also present a simulation-based delay estimation methodology that can estimate the travel-time delay over any given single-track or double-track rail network. These estimates can be used to route and schedule trains to either ease or avoid congestion in a network or a section of it.;In the freight railroad companies in the United States, planners typically route and schedule trains, on a daily or weekly basis, for networks up to 50 miles long, depending on the prevailing traffic conditions. Due to this reason, efficient routes and schedules that improve capacity utilization need to be generated frequently and quickly. For this purpose, we develop solution techniques using approximation procedures and evolutionary methods to solve the aforementioned capacity management model in a reasonable amount of time. Our experimental results show that our recommended solution procedure is capable of lowering current real-world delays by up to 30%. As a whole, this research represents an original effort in developing a quantitative model to tactically plan the movement of trains through a complex network, with decisions based on an accurate representation of the delays these trains cause on the railroad and the possibility of real-time rerouting trains to alternative tracks. |
