Optimization methods for analysis of empirical rail dispatching relative to train plans

Abstract

A major factor in railroad operational efficiency and punctuality is the quality of train planning and dispatching. Schedules or dispatching plans may also not be actualized for a variety of reasons. This work proposes a methodological tool set, called the dispatch analysis problem, that can analyze recent, empirical train dispatching data against an optimal dispatching plan. A multitude of questions can be answered using the dispatch analysis methods and we address three: (1) At what times did dispatching actions reduce the optimality of future replanning? (2) What corrective actions could have mitigated the negative impacts of past dispatching actions? (3) Which trains introduced secondary effects to other train plans? We demonstrate the application of the dispatch analysis methods to these questions using illustrative case studies from a North American freight railroad and find: specific periods of time can be isolated that demonstrate a significant deterioration in replanning ability; small modifications to past actions are identified that could improve replanning outcomes; certain trains can exhibit small delays that lead to large secondary consequences for neighboring trains. The results of the case study demonstrate the types of actionable findings on real railroad data that are possible with the dispatch analysis methods.