TRICO DE LAS BASES DE DATOS
6.9 MEDIDORES DE CAUDAL
The application of Operations Research in the disruption management process of railway systems is still a relatively uncharted field. More research has been conducted, however, on the topic in other logistic settings. Especially the airline industry has attracted extensive research on the disruption management process. In this section we list a number of successful applications of Operations Research in disruption management in related fields.
For an extensive framework on disruption management, we refer to the work by Yu and Qi (2004). The authors introduce a number of modeling paradigms for disruption management and apply them to different settings. We will discuss the applicability of their methods to passenger railway operations in Chapter 4 where we develop a framework for disruption management in passenger railways.
2.8.1
Disruption management in the airline industry
There is a number of similarities between disruption management in passenger rail- ways and in the airline industry. In both cases the timetable is published in advance and thus has to be taken into account, and aircraft and rolling stock as well as crews have to be rescheduled in case of a disruption. However, disruptions are handled quite differently in the airline industry compared to the railway industry. This is due
to a number of key differences in the features of the two industries. First of all, trains are bound to the infrastructure network. Second, the number of operators that share both airports and airspace is usually much larger in the airline industry leading to different challenges than experienced in passenger railways. Third, aircraft usually only perform a few flights per day whereas a rolling stock unit may be assigned to dozens of train services during a day. Fourth, trains may consist of several coupled rolling stock units. Also the much stricter maintenance requirements for aircraft pose additional challenges. At the same time, more knowledge on passenger flows is avail- able in the airline industry due to reservation systems. This means that passengers can be routed individually in case of delays or cancellations.
Disruption management in the context of airlines is a well studied field. For an overview of the research on rescheduling in the airline industry we refer to Clausen et al. (2010) and Kohl et al. (2007).
Aircraft make up the resource that provides seat capacity in the airline setting where rolling stock is the corresponding resource in the railway setting. Yan and Lin (1997) and Thengvall et al. (2001) study the problem of recovering the circulation of aircraft after a disruption that involves the closure of an airport. The process is oriented to minimizing the cost related to delays and cancellations. Rosenberger et al. (2003) present a model for aircraft recovery that not only minimizes operational cost, but also implicitly takes crew and passenger recovery into account by penalizing crew disruptions and disruptions in the passenger flows. Aircraft recovery is also studied by among others Thengvall et al. (2000), Yan et al. (2005) and Ball et al. (2007).
2.8.2
Disruption management in vehicle scheduling applica-
tions
The vehicle scheduling problem is addressed in a vast number of publications and variants of the problem have applications in many logistic settings such as pick-up and delivery, dispatching of services and scheduling of shuttle services (see Bodin and Golden (1981) and Bunte and Kliewer (2009)). Many applications concern schedul- ing of buses or trucks from one or more depots with or without time-windows. A number of those applications address online variants of the problem where vehicles are dynamically dispatched at the arrival of new requests (see Ghiani et al. (2003)). Both online and disruption management in vehicle scheduling consider problems where existing schedules have to be adapted to a change in the environment. How- ever, although the online version considers an important and realistic operational
aspect, it differs from disruption management in the sense that disruption manage- ment concerns unwanted restrictions rather than new requests by customers.
Recently, disruption management in vehicle scheduling has received more atten- tion such as by Li et al. (2007), who consider the vehicle rescheduling problem for buses. There are several similarities between the rescheduling of buses and railway rolling stock. In both cases, vehicles are assigned to trips in a timetable, and vehicles are dispatched from one or more depots, or shunting yards in the railway case.
There is a number of differences between the two settings as well. Most notable is the difference in the underlying infrastructure, buses utilize the road network whereas trains of course run on rails. Rescheduling vehicles on the roads is often less involved as there may be many alternative routes to the destination, and buses are not subject to the complicated headway security requirements experienced in the one-dimensional railway setting. Another difference is that several railway rolling stock units are usually combined to make longer trains.
2.8.3
Robust planning and recoverability
Where disruption management is concerned with the actual reassignment of re- sources, robust planning is about incorporating disruption management measures already in the planning phase. As mentioned in Section 2.4, robust planning con- cerns pro-active features such as slack and reserve resources, and re-active features such as recoverability.
Several papers consider robust planning of railway resources. One robustness aspect that is considered in several publications is the built-in option of exchanging crew duties in case of a disruption. This is introduced as the concept of move-up crews by Shebalov and Klabjan (2006) in the airline industry. The concept is adapted to rolling stock schedules by Cacchiani et al. (2008b), who use move-up rolling stock units to increase the recoverability of a rolling stock schedule. They explore the trade-off between the efficiency and the recoverability of the resulting schedules.
Cacchiani et al. (2008a) further explore the recoverability of rolling stock schedules and quantify the trade-off as the price of recoverability. The authors define it as the loss in efficiency for the schedules to be inexpensively recovered in a number of disruption scenarios. A mathematical model is proposed for the robust version of the problem and a solution method based on Benders decomposition is applied. The approach is tested on a case from NS, and the authors observe that a robust solution with a slightly higher cost than an optimal non-robust solution performs dramatically better in a simulated disrupted scenario.
De Almeida et al. (2008) propose a model for robust rolling stock scheduling at the French railway company SNCF. Their model incorporates two aspects of robust- ness, limiting delay propagation and the possibility of swapping resources. They also observe that accounting for these aspects of robustness leads to schedules that propagate less delays.
The concept of robust crew planning in passenger railways is addressed by Flier et al. (2008) who study the complexity of the problem and develop algorithms for both creating robust crew schedules and for optimally performing the duty swaps in case of a disruption.
Potthoff (2010) introduces the concept of quasi robustness in the railway crew rescheduling context. It is an approach to rescheduling under uncertainty that con- siders the crew rescheduling problem under uncertainty as a multi-stage program. Rather than utilizing a stochastic programming approach, the quasi robust approach attempts to ensure that all crews have feasible alternative duties in case the disrup- tion lasts longer than expected.