A Survey on Container Processing in Railway Yards · PDF fileBoysen et al.: Survey on Container Processing in Railway Yards 314 Transportation Science 47(3), pp. 312–329, ©2013

This article was downloaded by: [137.250.43.101] On: 18 April 2016, At: 01:16Publisher: Institute for Operations Research and the Management Sciences (INFORMS)INFORMS is located in Maryland, USA

Transportation Science

Publication details, including instructions for authors and subscription information:http://pubsonline.informs.org

A Survey on Container Processing in Railway YardsNils Boysen, Malte Fliedner, Florian Jaehn, Erwin Pesch,

To cite this article:Nils Boysen, Malte Fliedner, Florian Jaehn, Erwin Pesch, (2013) A Survey on Container Processing in Railway Yards.Transportation Science 47(3):312-329. http://dx.doi.org/10.1287/trsc.1120.0415

Full terms and conditions of use: http://pubsonline.informs.org/page/terms-and-conditions

This article may be used only for the purposes of research, teaching, and/or private study. Commercial useor systematic downloading (by robots or other automatic processes) is prohibited without explicit Publisherapproval, unless otherwise noted. For more information, contact [email protected].

The Publisher does not warrant or guarantee the article’s accuracy, completeness, merchantability, fitnessfor a particular purpose, or non-infringement. Descriptions of, or references to, products or publications, orinclusion of an advertisement in this article, neither constitutes nor implies a guarantee, endorsement, orsupport of claims made of that product, publication, or service.

Copyright © 2013, INFORMS

Please scroll down for article—it is on subsequent pages

INFORMS is the largest professional society in the world for professionals in the fields of operations research, managementscience, and analytics.For more information on INFORMS, its publications, membership, or meetings visit http://www.informs.org

http://pubsonline.informs.org

http://dx.doi.org/10.1287/trsc.1120.0415

http://pubsonline.informs.org/page/terms-and-conditions

http://www.informs.org

Vol. 47, No. 3, August 2013, pp. 312–329ISSN 0041-1655 (print) � ISSN 1526-5447 (online) http://dx.doi.org/10.1287/trsc.1120.0415

© 2013 INFORMS

A Survey on Container Processing inRailway Yards

Nils BoysenLehrstuhl für Allgemeine Betriebswirtschaftslehre/Operations Management, Friedrich-Schiller-Universität Jena,

D-07743 Jena, Germany, [email protected]

Malte FliednerInstitut für Betriebswirtschaftslehre, Fachgebiet Operations Research, Technische Universität Darmstadt,

D-64289 Darmstadt, Germany, [email protected]

Florian Jaehn, Erwin PeschInstitut für Wirtschaftsinformatik, Universität Siegen, D-57068 Siegen, Germany

{[email protected], [email protected]}

In spite of extraordinary support programs initiated by the European Union and other national authorities,the percentage of overall freight traffic moved by train is in steady decline. This development has occurred

because the macroeconomic benefits of rail traffic, such as the relief of overloaded road networks and reducedenvironmental impacts, are counterbalanced by severe disadvantages from the perspective of the shipper, e.g.,low average delivery speed and general lack of reliability. Attracting a higher share of freight traffic on railrequires freight handling in railway yards that is more efficient, which includes technical innovations as well asthe development of suitable decision support systems. This paper reviews container processing in railway yardsfrom an operations research perspective and analyzes basic decision problems for the two most important yardtypes: conventional rail–road and modern rail–rail transshipment yards. Furthermore, we review the relevantliterature and identify open research challenges.

Key words : railway system; railway yard; container processing; decision support; surveyHistory : Received: January 2011; revisions received: July 2011, December 2011; accepted: January 2012.

Published online in Articles in Advance June 21, 2012.

1. IntroductionFrom a macroeconomic perspective, shifting freighttraffic from the road network to the railway systemis certainly desirable for several reasons. Increasedrail usage for mid- to long-distance freight can pro-vide an opportunity to relieve the often congestedroads of major trading countries, not only in cen-tral Europe, where freight traffic (and road freightin particular) is projected to increase considerablyover the next few decades (see, e.g., Progtrans 2007),but also in the United States (Frittelli 2003), Canada(TC 2004), and Australia (Meyrick and Associates2006). Moreover, rail traffic is typically preferred onthe basis of its reduced environmental impact, forinstance with regard to CO2 emissions, which areestimated to be more than four times higher perton-kilometer in the case of road traffic (Allianz-pro-Schiene 2008). In spite of extraordinary support pro-grams of the European Union (see, e.g., Tsamboulas,Vrenken, and Lekka 2007) and other national author-ities (e.g., U.S. DOT 1991, 1998), railway systemsstill seem to be considerably less attractive for ship-pers, especially when compared with freight traffic by

truck. Within the last 25 years, the fraction of over-all freight traffic moved by train declined from 20%(1970) to 10% (2005) (EU 2007a). This development ismainly because of the lack of investment in railwayinfrastructure over the last decade. As a consequence,the modest absolute increase in rail freight volumesled to an overutilization of critical resources and thusto severe competitive disadvantages, from a shipper’sperspective, compared with road traffic. According torecent studies, only 53% of freight trains reach theirdestination with fewer than 30 minutes’ delay (EU2007b), and the average delivery speed of a freighttrain is estimated to lie between 10 km/h (VDA 2006)and 20 km/h (EU 2001), predominantly because oflong waiting times in rail yards.

In addition to necessary investments in the networkinfrastructure, a second critical driver to increase themarket share of rail freight is therefore to establishmore efficient freight handling processes in existingrailway yards employing, for example, suitable opti-mization approaches and decision support systems.This paper surveys rail yard operations at conven-tional rail–road and modern rail–rail terminals from

312

Dow

nloa

ded

from

info

rms.

org

by [

137.

250.

43.1

01]

on 1

8 A

pril

2016

, at 0

1:16

. Fo

r pe

rson

al u

se o

nly,

all

righ

ts r

eser

ved.

Boysen et al.: Survey on Container Processing in Railway YardsTransportation Science 47(3), pp. 312–329, © 2013 INFORMS 313

(a) Tank container (left) and open topcontainer with canvas (right) on a flatcar

(b) Intermodel containeron flatcar

(c) Tank trailer on pocket wagon

(d) Hopper care for bulk cargo, e.g., coal

(e) Boxcar on covered

goods (f) Tank care for liquid

Figure 1 Railcars with Separable (a–c) and Non-Separable (d–f) Cargo CarriersaaEach of the six pictures is published under the Creative Commons Attribution ShareAlike License. The authors of the pictures are (a) Magnus Manske, (b)

Marwin Hawlisch, (c) Pantoine, (d) Steffen Mokosch, (e) Alwin Meschede, and (f) Harvey Henkelmann.

an operations research perspective by characterizingimportant decision problems and solution approachespublished in the scientific literature. On the basis ofthis analysis, future research challenges are identified.

The remainder of the paper is structured as fol-lows. Section 2 defines the scope of this review bydistinguishing different types of rail yards and brieflydescribing the associated decision problems. The twomost important yard types—conventional rail–roadterminals and modern rail–rail transshipment yards—are then studied individually in dedicated sectionswith regard to the core decision problems, exist-ing literature, and future research challenges. Finally,§5 concludes the paper.

2. Scope of ReviewA railway yard is a special transshipment nodein a rail network where loads for trains are pro-cessed, i.e., collected, rearranged, unloaded, interme-diately stored, loaded, and/or picked up. Our surveyexclusively treats freight railway yards where con-tainerized loads are processed. In spite of the largevariety of such load units, we simply use the term“container” throughout this paper for any cargo car-rier that is separable from its railcar and moveableby an alternative mode of transportation, for instanceby truck or ship. The variety of containers processedin rail yards is typically much higher than that ofseaports, where only a handful of different containertypes are transshipped. For instance, the German railnetwork distinguishes between 23 different containertypes (Kombiverkehr 2008) and a comparable numberhas been reported for North America (Muller 1999).A selection of these containers includes, for instance,

tank containers (a), standardized intermodal contain-ers (b), swap bodies, railroaders, and trailers (c) load-able on some flatcar (a and b) or pocket wagon (c)as depicted in Figure 1(a)–(c). Excluded from our sur-vey are therefore passenger railway systems (see, e.g.,Freling et al. 2005; Kroon, Lentink, and Schrijver 2008)or railway yards where nonseparable cargo carriersare processed. Examples of the latter yard type arethose handling bulk cargo (d), packaged goods (e),or liquids (f), which are loaded directly on a railcar(e.g., in the case of tree trunks) or in a nonseparableload carrier (see Figure 1(d)–(f)).

Usually a freight yard serves at least one of twomain purposes in a railway network:

(i) On the one hand, a terminal may serve as aninterface in intermodal transport so that shipmentscan be interchanged between the rail system and analternative mode of transportation, such as trucks orships. Typically in such a system, trucks serve cus-tomers on the last mile, whereas trains operate thelong-haul routes. Alternatively, an intermodal yardmight be located in (or nearby) a seaport for movingfreight to and from the hinterland.

(ii) On the other hand, a transshipment yard mightalso serve as a hub node in a hub-and-spoke net-work so that containers or even railcars themselvesare exchanged between different trains. This allows aconsolidation of trains (i.e., several short trains withloads for multiple destinations are reduced to fewerlong trains) so that economies in transportation aregenerated. Without hub nodes, rail freight is pre-dominantly executed as point-to-point traffic. Becausefixed costs for train traffic are high, point-to-pointshipment is only profitable if long trains travel overlong distances. Studies have estimated the break-even

Dow

nloa

ded

from

info

rms.

org

by [

137.

250.

43.1

01]

on 1

8 A

pril

2016

, at 0

1:16

. Fo

r pe

rson

al u

se o

nly,

all

righ

ts r

eser

ved.

Boysen et al.: Survey on Container Processing in Railway Yards314 Transportation Science 47(3), pp. 312–329, © 2013 INFORMS

Receiving tracks Hump Classification tracks Departure tracks

Figure 2 Outline of a Shunting Yard

point for a container being moved either by truck orby a full train over a range of 400 km (Williams andHoel 1998) or 500 km (van Klink and van den Berg1998). Thus, hub-and-spoke systems have been identi-fied as a promising starting point to attract rail freighttraffic for small freight flows over shorter distances(e.g., Trip and Bontekoning 2002).

To carry out these two tasks, different types ofyards have been established over the years, which,in accordance with the chronology of their appear-ance, can be generally grouped into three terminalgenerations (Boysen, Fliedner, and Kellner 2010):

• First generation: In traditional shunting (or clas-sification) yards, trains arrive at a set of receivingtracks, where railcars are decoupled and pushed overa hump or ramp. The cars are then redirected via asystem of track switches and classification tracks todeparture tracks, where they are assembled to out-bound trains (see Figure 2). Most of these yardsare dedicated to function (ii), of rearranging eitherwagons with nonseparable cargo carriers or railcarswhose container assignment is not altered to out-bound trains; however, loading operations (func-tion (i)), such as those of bulk cargo, might be carriedout. Shunting yards have a long history dating back tothe beginnings of rail transport, even though shuntingoperations have always been very time-consuming,especially when compared with the exchange of con-tainerized loads between trains without altering therailcar composition (see second- and third-generationterminals). It is estimated that in railway systems thatemploy shunting yards, 10% to 50% of a train’s totaltransit time is required for shunting (Bontekoning andPriemus 2004). Owing to the growth of container-ized transport, shunting yards have lost their formerimportance, and many have been put out of servicein recent decades (see, e.g., Rhodes 2003). Neverthe-less, there are still several operational shunting yardsin different railway networks; in some regions (in par-ticular in China), some have even been newly con-structed, mainly because of the comparatively lowinvestment cost of technical equipment.

• Second generation: At today’s conventional rail–road terminals, trains usually keep their railcars andonly containers are actually transshipped, typically bymeans of huge gantry cranes that span multiple par-allel railway tracks. Such yards often accommodate

additional elements, such as storage areas for theintermediate stacking of containers and adjacenttruck lanes for immediate transshipment from trainsto trucks and vice versa. Rail–road terminals havebecome one of the cornerstones of intermodal freight;their main purpose is to serve as an interface betweendifferent modes of transportation (function (i)). TheGerman railway network, for example, features 24rail–road terminals spread over the country (seeDUSS 2010). More recently, these yards have also beenapplied as part of hub-and-spoke networks (func-tion (ii)), for instance between Germany (with hubs inLudwigshafen, Munich-Riem, and the Port of Nurem-berg) and Italy (with hubs in Bologna, Busto Arsizio,Milan, and Verona; see Kombiverkehr 2009).

• The third generation of modern rail–rail trans-shipment is dedicated to a rapid consolidation oftrains (function (ii)). The layout of these yards issimilar to that of second-generation terminals. How-ever, for further acceleration of container transfers,a fully automated sorting system is employed insteadof conventional floor storage. Such a sorter consistsof shuttle cars that receive containers close to theirinitial positions on inbound trains and move themalongside the yard to their target positions. Only thendoes a gantry crane pick up the containers and trans-port them to their dedicated outbound trains. Mostof these terminals are still in the design phase; how-ever, some of these novel hub yards have alreadybeen constructed in the European Union (e.g., Port-Bou; see Martinez et al. 2004) and others are currentlyunder development. For instance, construction of theso-called German Mega Hub in Hannover-Lehrte isexpected to finally start in 2014 after a tedious designphase, which has been documented in more detail byAlicke (2002) and Rotter (2004).

This study exclusively focuses on second- andthird-generation terminals because first-generationshunting yards are fundamentally different in struc-ture and operation and traditionally dedicated toprocessing railcars with nonseparable cargo carri-ers. Therefore, shunting yards seem to be a subjectsuited to exclusive treatment. They are, moreover,rarely part of modern container-based rail networks,which are expected to be the main driver of futurerail freight; shunting yards are therefore of minorrelevance in the context of container transshipment.

Dow

nloa

ded

from

info

rms.

org

by [

137.

250.

43.1

01]

on 1

8 A

pril

2016

, at 0

1:16

. Fo

r pe

rson

al u

se o

nly,

all

righ

ts r

eser

ved.


For an introduction to shunting yard processes, seeGatto et al. (2009) or the valuable classification ofHansmann and Zimmermann (2008). Furthermore, allfreight terminals that are not explicitly designed totransship container units as well as innovative ter-minal concepts that have not yet passed throughthe purely conceptual phase are excluded from thescope of this survey. The former group includes, forinstance, special terminals dedicated to automobiletransshipment (see, e.g., Mattfeld and Kopfer 2003;Fischer and Gehring 2005) or company-owned rail-way sidings (e.g., Lübbecke and Zimmermann 2003),whereas the latter consists of concepts such as auto-mated shunting terminals (see, e.g., Hansen 2004) ormoving train techniques (see, e.g., Ballis and Golias2004). Instead, it is the aim of this study to reviewscientific approaches that tackle the long- to mid-termdecision problems of the design phase with regardto the layout and resource allocation of the termi-nal and short-term decision problems that are solvedas part of the daily operations of conventional rail–road and modern rail–rail terminals. The problemsare studied exclusively from the perspective of the ter-minal operator, and macroeconomic effects are thusnot considered.

Owing to the confinement of (isolated) terminaloperations, decision problems with regard to the com-plete railway network consisting of multiple termi-nals and the interconnecting track system are furtherexcluded. This exclusion comprises location plan-ning (e.g., see Klincewicz 1998; Arnold, Peeters, andThomas 2004), performance estimation of a rail-way network with respect to the capacity of nodesand connections (see, e.g., Ballis and Golias 2002,2004), distributing empty wagons within a network(see, e.g., Nozick and Morlok 1997), and train schedul-ing (e.g., see Newman and Yano 2000). Moreover,the smallest load unit considered in this survey isthe container, and the stowage planning of containers(see, e.g., Geng and Li 2001; Pisinger 2002) is thus notcovered. To conclude, Figure 3 schematically definesthe scope of this review.

Networkperspective

Long- to-mid-termlayout problems

Operationalproblems

Container loadingproblems

Scope of review:Layout planning and

terminal operations ofconventional rail–roadand modern rail–rail

terminals

Generation of freight rail terminalsOther rail terminals,e.g., for passengers

or automobiletransshipment

1st 2nd 3rd

Figure 3 Scope of the Review

Yard operations have already been discussed inreview papers with a wider scope, which cover railtransshipment yards as one part of a broader topic.For instance, surveys on general railway optimiza-tion (Assad 1980; Bussieck, Winter, and Zimmermann1997; Ferreira 1997; Cordeau, Toth, and Vigo 1998;Newman, Nozick, and Yano 2002), intermodal trans-port (Macharis and Bontekoning 2004; Bontekoning,Macharis, and Trip 2004; Crainic and Kim 2007; Caris,Macharis, and Janssens 2008), and seaport terminaloperations (Vis and de Koster 2003; Steenken, Voß,and Stahlbock 2004; Stahlbock and Voß 2008) alsobriefly elaborate on rail yards. However, the extendedscope of these surveys prevented an in-depth dis-cussion of decision problems, existing optimizationapproaches, and future research challenges of railyard operations.

3. Rail–Road Transshipment Yards3.1. Yard Layout, Transshipment Process, and

Decision ProblemsRail–road terminals mainly serve as interface nodesin intermodal transport, where gantry cranes trans-ship containers between trains and trucks and viceversa. A rail–road terminal is schematically depictedin Figure 4.

Freight trains are parked on parallel transshipmenttracks of the terminal. Typically, a terminal segmentconsists of between two and four parallel tracks, anda maximum gantry span of six tracks is possible (e.g.,Steenken, Voß, and Stahlbock 2004). Because freighttrains in Europe have a typical length of 600–750 m(Ballis and Golias 2002), the track area accessibleby cranes for container processing is typically ofabout the same length. Larger terminals, such asKöln-Eifeltor and Hamburg-Billwerder in Germany,consist of multiple parallel terminal segments. Trucks

Gantry crane

Storage a

rea

Driving la

ne

Parking

lane

Track

Track

Track

Track

Figure 4 Schematic Representation of a Rail–Road TransshipmentYard

Dow

nloa

ded

from

info

rms.

org

by [

137.

250.

43.1

01]

on 1

8 A

pril

2016

, at 0

1:16

. Fo

r pe

rson

al u

se o

nly,

all

righ

ts r

eser

ved.


arrive on parallel truck lanes, which are usually sep-arated into a driving lane and a parking lane. Fur-thermore, a floor storage area enables intermediatecontainer storage for whenever a delivered containercannot be immediately shipped to the respective out-bound truck or train. Typically, one or multiple gantrycrane(s) span all three elements (i.e., tracks, stor-age area, and truck lanes) so that containers can bedirectly moved to their destinations in a single step.Up to four of these gantry cranes serve a termi-nal segment in parallel. The most widespread typeof gantry crane is rail mounted and manually oper-ated, has a maximum load capacity of 41 tonnes, pro-cesses between 20 and 25 container moves per hour(see Rotter 2004), and is able to cross a maximumstacking height of three containers in the storage area(Ballis and Golias 2002). However, there exist yardswith alternative crane settings. Especially at smallerterminals, rubber-tired gantries or reach stackers areapplied (Ballis and Golias 2002) so that resources canbe more flexibly allocated to different segments of theyard. A yard terminal often features a holding lotfor trucks, a gate, and an office area for controllingand organizing access into and out of the terminal bytruck. Additional rail tracks may be provided, e.g., forthe holding and shunting of trains or storing railcars(Ballis and Golias 2002). All elements and their typ-ical layout inside a rail–road transshipment yard aredepicted in Figure 5.

During the design phase of a terminal, the elementsto be located and their general layout (as depictedin Figure 5) are already (more or less) fixed. How-ever, critical decisions with regard to the dimensionsof each terminal element are made. Specifically, thenumber of holding and transshipment tracks, thecapacity of the storage and parking area, as well asthe number and technology of gantry cranes need

Main track

Holding track

Gate

Transshipment tracks

Parking lane

Gan

try

cran

e

Driving lane

ParkingareaOffice

Storage area

Figure 5 Top View of a Rail–Road Transshipment Yard

to be determined because these interdependent lay-out factors heavily influence yard performance andare not easily reversible. Hence the job of the yardplanner is to carefully trade off investment cost ofa specific layout against estimated operational per-formance. Operations research methods are especiallysuited to quantifying the latter part of this tradeoff,and an accurate estimation of the performance ofa specific yard layout is thus an essential task forsupporting decisions of the design phase. Note that,although not in the scope of this review, the resultsfor yard performance need to be further evaluatedwith respect to the yard’s network integration becausean expansion of capacity in a non-bottleneck yarddoes not necessarily increase overall network perfor-mance. Existing approaches for performance estima-tion of rail–road terminals are reviewed in §3.2.

For a given yard layout, the operational processof container transshipment at a rail–road terminal isnow described in more detail. On the basis of a giventimetable of trains, container moves need to be pro-cessed periodically subject to arrival and departuretimes of trains. Typically, all trains arrive in the morn-ing, are processed over the course of a day, and leavethe terminal in the late evening. Owing to the generalright of way of passenger trains in many Europeancountries and, for example, Australia, freight trainsare often bound to travel during the night exclusively.Once a train arrives at the transshipment yard (after apotential interim stay on a holding track of the yard),the train is first assigned to a vertical and horizon-tal parking position of the yard. The vertical parkingposition relates to the actual track on which the trainenters the yard, but the notion of a horizontal parkingposition requires some additional explanation. Typ-ically, the yard area is subdivided horizontally intoslots of equal size measured in units of the length of a

Dow

nloa

ded

from

info

rms.

org

by [

137.

250.

43.1

01]

on 1

8 A

pril

2016

, at 0

1:16

. Fo

r pe

rson

al u

se o

nly,

all

righ

ts r

eser

ved.


standard railcar (or any other unit). The resulting gridis hence used to identify the coordinates of any givencontainer in the yard, and the horizontal parking posi-tion of a train refers to the slot in which the trac-tion vehicle is positioned. For rail–road terminals, theproblem of assigning parking positions to trains hasnot been studied in detail thus far. This is because itsimpact on yard performance is usually considered tobe rather small. Because trucks can be parked directlynext to the respective container of the train, cranesneed to move only vertically for the most part, andthe time required compared with the time-consumingpick and drop operations of cranes is often negligi-ble. The horizontal parking positions do determinethe accessibility of individual containers with respectto the different gantry cranes of the yard. Owing tothe immense fixed cost, however, freight transport bytrain is often only profitable if full trains (i.e., trainsused to capacity) are moved. Therefore, train and yardlengths are about the same and the degree of free-dom for varying horizontal parking positions is oftennot significant enough. In practice, parking positionsare therefore typically assigned according to a simplefirst-come-first-served policy (Kozan 1997).

As soon as a train is parked, the unloading of allinbound containers can commence. To avoid double-handling, a container is preferably transshippeddirectly from the train to its dedicated truck. In thefollowing, this form of container transfer is referred toas a direct move. Clearly, a direct move requires thesimultaneous presence of a respective truck and trainin the yard. The target truck is then called up fromthe holding area and is assigned a free parking posi-tion in the parking lane next to the respective railcar.If the target truck has not yet arrived and is there-fore not directly available, the container is moved tothe intermediate storage yard. This kind of double-handling is referred to as a split move (see Boysen,Jaehn, and Pesch 2011), where a storage location closeto the respective railcar is sought so that crane oper-ating times are reduced. However, because containersare usually stacked on top of each other, split movesto and from storage are subject to additional restric-tions, such as those relating to weight, stability, andestimated departure times of containers, to avoid sub-sequent reshuffling. Analogous to stacking logisticsin seaports (see Steenken, Voß, and Stahlbock 2004),three interrelated decision problems are associatedwith split moves. First, a suitable storage position isto be identified that minimizes the risk of containerblockages. Second, stored containers might need to bepre-marshaled on the basis of updated informationwith regard to arrival times of trucks, which is espe-cially reasonable during idle times of cranes. Finally,containers need to be efficiently retrieved from thestorage area as soon as the respective truck arrives,

which might in turn require additional handling ofany blocking containers. Typically, the frequencies ofdirect and split moves vary over time (see Bose 1983;Ballis and Golias 2002). Shortly after a train’s arrival,direct moves in particular are processed, i.e., movesfrom a train to trucks that are already waiting. In asecond phase, wagon-to-storage moves (i.e., from atrain to the storage area) predominate, and in the finalphase, storage-to-truck moves are processed.

At most terminals, outbound operations are exe-cuted only after inbound operations are completed.However, intermixed operations of inbound and out-bound containers are certainly possible. The process-ing of outbound operations is carried out analogouslyto that of inbound containers. Whenever trucks havedeployed containers prior to the train’s arrival, a splitmove occurs, and containers from the storage yard areloaded onto the train; deliveries that arrive during theloading process of the target train can be processedas direct moves. Therefore, prior to a train’s arrival,truck-to-storage moves prevail, which are then super-seded by direct moves processed after a train’s arrival.In the final phase, mainly storage-to-wagon movesare executed. In some yard settings, outbound con-tainers are moved by special container trailers ratherthan directly by customers’ trucks. The containers arethen carried to a separate storage area where cus-tomer trucks pick up and deliver containers. Clearly,this concept avoids double-handling of containers inthe yard at the price of an additional transshipmentin the storage area and higher investment cost of themany different container trailers required for the widerange of possible containers. Nonetheless, this prac-tice is often applied in North American yards (seeFerreira and Sigut 1993; Kozan 1997).

During the loading operations of an outboundtrain, there exist some degrees of freedom with regardto the exact position of each container on a train.Therefore, a load plan that determines the loadingpattern of containers on wagons is required. A typ-ical terminal faces a high variety of container typesand multiple different wagons, which can vary inlength between 40 and 104 feet (see Bruns and Knust2012). Given a specific setting of outbound contain-ers and railcars, the loading problem has to considerseveral hard constraints, i.e., wagon length, separa-tion of dangerous goods, weight restrictions, and trainheight. Furthermore, the quality of a load plan can bedetermined by different conflicting objectives, such asthe utilization of trains, setup time and/or cost forchanging a railcar’s pin configuration, or processingtimes for moving a container from its current positionto the respective wagon. From the overall networkperspective, load planning is also heavily interdepen-dent on the distribution of wagon types across yards.

Dow

nloa

ded

from

info

rms.

org

by [

137.

250.

43.1

01]

on 1

8 A

pril

2016

, at 0

1:16

. Fo

r pe

rson

al u

se o

nly,

all

righ

ts r

eser

ved.


This last aspect has been investigated by Powell andCarvalho (1998), for instance.

Once the load plan is determined, the set of con-tainer moves is finally fixed and the planning canfocus on determining transshipment schedules foreach crane. Because gantry cranes principally workin parallel, it seems especially desirable to split theoverall workload evenly among cranes so that trainprocessing is accelerated. However, in most yard set-tings, gantry cranes share a dedicated track for theirhorizontal movement along the yard, which preventsthem from passing by one another. Clearly, thesenoncrossing constraints of gantry cranes need to beintegrated into a suitable decision support tool ashard modeling constraints. Two distinct policies havebeen developed to avoid such crane interferences (seeBoysen and Fliedner 2010). On the one hand, theassignment of container moves to cranes can be static,which means that each crane receives a disjoint areaof operations, where all container moves falling intothe area are exclusively processed by the respectivecrane. On the other hand, containers can be assigneddynamically on the basis of the actual positions ofcranes and the set of moves that needs to be exe-cuted. Clearly, the latter policy offers more degrees offreedom for crane scheduling. However, the coordi-nation of cranes becomes more complex and requiresreal-time crane scheduling procedures to rule outany crane interference. In real-world yards, a staticcrane split with equally sized yard areas is the mostwidespread choice (see Boysen and Fliedner 2010).The sequence of moves falling into a crane’s area istypically not optimized with the help of a sophisti-cated decision support tool. Instead, the crane oper-ator simply chooses among those moves currentlybeing displayed on the monitor of the steeple cabon the basis of some nearest-neighbor decision rule(Boysen, Fliedner, and Kellner 2010).

To summarize, the operational process needs to ad-dress the following essential decision problems.

(I.i) Decide on storage positions of containers han-dled by split moves.

(I.ii) Assign each truck a parking position.(I.iii) Determine the positions of outbound contain-

ers on trains.(I.iv) Assign container moves to cranes.(I.v) Determine the sequence of container moves

per crane.In the following sections, the literature on layout

planning (§3.2) and operational container processing(§3.3) is summarized.

3.2. Literature on Layout PlanningExisting literature on layout planning exclusively con-sists of simulation studies. These simulations areapplied to anticipate yard performance for differentterminal layouts.

A discrete-event simulation study including botha macro (network) and micro (terminal) perspectivewas carried out by Rizzoli, Fornara, and Gambardella(2002). In that study, different technologies and oper-ational policies were compared with regard to theireffect on terminal and network performance. A simi-lar simulation model was described by Kondratowicz(1990). Lee et al. (2006) presented a simulation studythat was designed to support decisions on the numberand locations of rail terminals at a Korean containerport. Basic mathematical expressions were derived tocalculate the number of tracks and cranes requiredfor a specific number and locations of rail terminals.The authors simulated different train and truck arrivalpatterns as well as container move settings by apply-ing a simple crane scheduling rule; i.e., every craneprocesses containers successively while continuouslytraveling in a specified direction as long as a receivingtruck is available (if not, the crane changes directionfor the next container). The study was carried out fordifferent numbers and locations of terminals.

Ferreira and Sigut (1993, 1995) compared theresulting performance of container handling betweena conventional rail–road terminal and a roadrailer ter-minal. A roadrailer is a special trailer, which is pro-vided with a detachable bogie or a single rail axleso that they are capable of being hauled on roadand rail without requiring a wagon. The applicationof alternative wagon types was compared as part ofa simulation study for an Australian terminal. Theresults indicated the more efficient handling of con-tainers compared with roadrailers.

Another simulation tool dedicated to model a sin-gle terminal was introduced by Benna and Gronalt(2008). Terminal layout, arrival patterns of trains andtrucks, and container settings were specified as part ofthe input data. Simple priority rule-based approacheswere applied to determine crane schedules and inter-mediate storage positions of containers. As qualitymeasures, the study evaluated lifting performance,system capacity, and service level. A similar tool wasdescribed by Gronalt, Benna, and Posset (2007).

The results of a large European Union researchproject, which aimed to increase rail terminal perfor-mance, were presented by Ballis and Golias (2002,2004) and Abacoumkin and Ballis (2004). The authorsdeveloped an extensive expert system consisting ofa macro model that covers a complete railway net-work and a micro model simulating train processingin a single yard. A general overview of the macroand micro model was provided by Ballis and Golias(2004). They tested the micro model for 17 differ-ent terminal layouts with varying numbers of tracksand cranes as well as lifting technologies. Ideal ter-minal layouts for a given transshipment volume weredetermined by calculating the total cost per container.

Dow

nloa

ded

from

info

rms.

org

by [

137.

250.

43.1

01]

on 1

8 A

pril

2016

, at 0

1:16

. Fo

r pe

rson

al u

se o

nly,

all

righ

ts r

eser

ved.


The macro model was employed to anticipate themarket share of rail freight over a longer planninghorizon for a specific network structure and termi-nal configuration. They included a case study of therailway corridor from the large North Sea harbors toSwitzerland. A more detailed description of the micromodel was presented by Ballis and Golias (2002) andAbacoumkin and Ballis (2004).

Kozan (2006) used a simulation model for a ter-minal, in which gantry cranes can be supportedby additional lifting equipment (e.g., reach stackersand forklifts). For joint loading and unloading oper-ations over multiple days, the author compareddifferent crane settings to identify a suitable craneconfiguration that provides a reasonable tradeoff be-tween investment cost and operational performance.Sequencing of container moves for different arrivalpatterns of trucks and trains was guided by simplefirst-come-first-served policies. By means of simula-tion, Vis (2006) compared the use of manned strad-dle carriers with that of automated stacking cranes.The total travel time required to handle all containermoves was applied as a performance measure todetermine the yard layout for the landside of a sea-port terminal.

Although mainly dedicated to seaport containeroperations, a helpful paper for generating representa-tive simulation scenarios was provided by Hartmann(2004). The paper features a data generator for deriv-ing diverse transshipment scenarios. It can be directlyapplied to simulate rail–road terminals by generatingarrival patterns of trucks and the different containertypes to be (un)loaded.

3.3. Literature on Operational PlanningThus far, none of the literature has explicitly treatedthe determination of storage positions in a rail–roadyard (problem (I.i)). However, this decision problemis closely related to problems arising in the stackinglogistics of seaports. In this context, the problem hasattracted much research, for instance by de Castilloand Daganzo (1993); Kim (1997); and Kim, Park, andRyu (2000). The subproblem of re-sorting contain-ers during the idle time of cranes (so-called pre-marshaling) has been investigated, e.g., by Lee andHsu (2007), Choe et al. (2011), and Lee and Chao(2009). Finally, Kim and Hong (2006) and Caserta,Voß, and Sniedovich (2009), for instance, providedsolution procedures for determining a suitable pre-determined sequence of crane moves to remove con-tainers from intermediate storage. In addition to thesestatic problem settings, online stacking rules wereinvestigated by Dekker, Voogd, and van Asperen(2006) and Borgman, van Asperen, and Dekker (2010).A more detailed review of these approaches wasgiven by Steenken, Voß, and Stahlbock (2004) and

Stahlbock and Voß (2008). However, the extent towhich these approaches are directly applicable to rail–road terminals remains to be studied.

A first basic version of the train loading problem(I.iii), where load patterns of containers are deter-mined, was presented by Feo and González-Velarde(1995). Given a predetermined matrix defining whichcontainer can be assigned to any railcar on the basisof the pin configuration, the approach seeks to min-imize the number of wagons per train. However,the model and solution approaches are restricted toat most two containers per wagon. For the solutionof the basic train loading problem, a simple branch-and-bound approach based on LP-relaxation and aheuristic GRASP (see Feo and Resende 1995) proce-dure have been introduced, where initial solutionsare locally improved by a two-opt search. The pro-cedures are shown to be efficient in the case of real-world data for a North American terminal. Corry andKozan (2006) optimized load planning with respect tohandling times and the weight distribution within atrain. The authors considered only one type of con-tainer and leave out weight restrictions for wagons.Furthermore, they assumed that each container canbe loaded onto any wagon. The problem was formu-lated as an integer linear program and solved with anoff-the-shelf solver. In a subsequent paper, Corry andKozan (2008) aimed to derive a loading plan consider-ing multiple objectives. On one hand, the train lengthand on the other hand the total handling time of con-tainers is minimized. Multiple container types weremodeled. Load pattern restrictions were consideredfor the container-to-wagon assignment, but neitherweight restrictions with regard to the maximum loadper wagon nor that with regard to the whole trainwere integrated. The model was formulated as aninteger linear program, and solutions for real-worldproblem instances were generated by a local searchprocedure.

Recently, Bruns and Knust (2012) investigated an-other version of the loading problem (I.iii) of trainsthat are subject to weight and length restrictions ofwagons. In the objective function, three weightedobjectives are considered: maximizing the utilizationof trains, minimizing the setup cost for changing theexisting pin configuration, and minimizing the trans-portation cost from storage position to railcar. Twodifferent mixed-integer programs were introduced forthis problem setting and were shown to be solvableeven for real-world instances.

An additional aspect of the train loading prob-lem (I.iii) was first investigated by Lai and Barkan(2005). Intermodal trains often contain larger seg-ments of empty wagons (e.g., segments consisting offlatcars without a container), which lead to aerody-namic characteristics that are much worse than those

Dow

nloa

ded

from

info

rms.

org

by [

137.

250.

43.1

01]

on 1

8 A

pril

2016

, at 0

1:16

. Fo

r pe

rson

al u

se o

nly,

all

righ

ts r

eser

ved.


of full trains with close spacing, e.g., the case forhopper cars. Note that unloaded wagons are trans-ported whenever diverging wagon demands and sup-plies at the rail yards need to be balanced withina rail network (see, e.g., Nozick and Morlok 1997).Therefore, considerable savings in fuel cost can beachieved if train planning considers the additionalobjective of generating long chains of loaded railcarsalternating with long chains of unloaded railcars. Laiand Barkan (2005) quantified the aerodynamic andenergy penalties of specific load and car combinationsunder idealized conditions by assuming that contain-ers can be assigned to wagons without restrictions.Lai, Barkan, and Önal (2007) described a waysidemachine vision system that automatically monitorsthe gap lengths between intermodal loads on passingtrains, which allows automatic evaluation of the aero-dynamic efficiency of loading patterns. In a subse-quent paper, Lai, Barkan, and Önal (2008a) presenteda mixed-integer model for determining fuel-efficienttrain loads considering weight and length restrictionsof wagons, which was solved with an off-the-shelfsolver. It turned out that the estimated savings cal-culated with real-world data amount to a remark-able potential of annual fuel savings of US$28 million.The joint optimization of multiple trains’ load plans(with identical destination) and uncertain informationon future trains and incoming loads were incorpo-rated into the aforementioned mixed-integer model inanother paper by Lai, Ouyang, and Barkan (2008b).They iteratively solved the model on a rolling hori-zon scheme, where exponentially decreasing weightswere assigned to the objective functions with regardto the fuel efficiency of future trains.

Kozan (1997) provided a simple heuristic decisionrule for determining the crane split (I.iv) and a simpledispatching rule for the assignment of trains to rail-way tracks. He employed simple analytical measuresto anticipate the processing times of current trains andthus identify the track that would enable the earli-est expected departure of a current train. These sim-ple heuristic rules were then applied in a simulationstudy, where the resulting throughput times of con-tainers for different train arrival and loading patternswere compared for different yard layouts. Boysen andFliedner (2010) also investigated decision problems(I.iv) and introduced a polynomial dynamic program-ming approach to determine static and disjoint craneareas such that the workload is evenly shared amongcranes. In a simulation of real-world yard operations,they showed that their approximate surrogate objec-tive for determining the crane split is strongly pos-itively correlated with actual processing times, andsimple real-world policies are clearly outperformed.

Souffriau et al. (2009) proposed a holistic approach,which jointly determines load plans (I.iii) and crane

schedules (I.iv) and (I.v). They used a decomposi-tion approach to determine follow-up destinationsof trains, such that the number of resulting con-tainer moves is minimized. This problem was solvedas a linear assignment problem. The load plan washence determined by minimizing the transportationcost of container moves in a mathematical modelwith an off-the-shelf solver. Only three different con-tainer types as well as length restrictions for thewagons were considered. Finally, the crane sched-ule, which distributes container moves among cranesand sequences moves per crane, was modeled as asequential ordering problem and solved by a vari-able neighborhood search. Another holistic approachfor train processing at the landside of a seaport wasprovided by Froyland et al. (2008). They treated anintermodal terminal in Australia, where five succes-sive gantry cranes transship containers between trains(two tracks), trucks (60 slots), straddle cranes (serv-ing ships), and an intermediate storage area witha maximum capacity of 2,100 twenty-foot equiva-lent units. They jointly investigated decision prob-lems (I.i), (I.ii), and (I.v) and determined containerpositions in intermediate storage, parking positions oftrucks, and crane schedules, respectively. The problemwas decomposed into three stages, where problems(I.i) and (I.ii) were solved by mixed-integer program-ming and cranes (I.v) were scheduled on the basisof simple priority rules. Finally, Montemanni et al.(2009) modeled the sequencing of a given set of con-tainer moves per crane (I.v) as a sequential orderingproblem and provided local search and ant colonyoptimization (Dorigo and Stützle 2004) as solutionprocedures.

Moreover, dynamic crane scheduling as defined bydecision problems (I.iv) and (I.v) bears some similar-ities with quay crane scheduling at seaports, wherea given number of huge quay cranes is employed to(un)load container vessels at their berths. As at railterminals, quay cranes share a track and may notinterfere with nor cross each other during containeroperations. Typically, within quay crane scheduling,a vessel is separated into holds (or bays), whichare exclusively served by a dedicated crane, andnoncrossing constraints need to be considered when-ever cranes change holds. If, analogously, a rail yard isseparated into small horizontal areas, such as slots ofcontainer length comprising all containers of the par-allel tracks within the respective slot, then the solu-tion procedures developed for quay crane schedulingcould in principle be applied to solve crane schedul-ing in a rail yard. The first optimization approachesfor quay crane scheduling stem from Daganzo (1989)and Peterkofsky and Daganzo (1990). These stud-ies, however, did not consider noninterference con-straints of cranes. Kim and Park (2004) considered

Dow

nloa

ded

from

info

rms.

org

by [

137.

250.

43.1

01]

on 1

8 A

pril

2016

, at 0

1:16

. Fo

r pe

rson

al u

se o

nly,

all

righ

ts r

eser

ved.


noncrossing constraints and presented a model for-mulation along with exact and heuristic solutionprocedures. Alternative solution methods were pre-sented by Lee, Hui, and Miao (2008b), who alsoprovided an NP-hardness proof. Related contribu-tions stem from Lim et al. (2004); Zhu and Lim(2006); Lim, Rodrigues, and Xu (2007); and Lee, Hui,and Miao (2008a). A comprehensive review on quaycrane scheduling was provided by Bierwirth andMeisel (2010). Of particular interest with regard tocrane scheduling in rail–road yards are the resultsof Lim, Rodrigues, and Xu (2007), who showed thatunder given noncrossing constraints—and some addi-tional simplifying assumptions—optimal crane sched-ules are unidirectional, in the sense that each cranecan move from left to right while processing con-tainer moves without ever changing direction. Thisfinding is especially relevant to dynamic schedulesbecause it reduces the real-time effort for collisiondetection to a minimum and thus might make staticcrane bounds expendable. However, this propertyonly holds whenever containers can be processed inan arbitrary sequence, an assumption that many otherquay crane scheduling approaches equally make. Thisdoes not hold for the majority of rail–road termi-nals because dynamic arrival times of trucks needto be considered in the transshipment plans, andtherefore the aforementioned approaches cannot bedirectly applied and need to be adequately extended.

3.4. Future Research ChallengesAlthough there have been plenty of studies on con-ventional rail–road terminals, there are still manyopen questions for future research.

With regard to layout planning, it can be statedthat all existing simulation studies apply compara-tively simple priority rules to solve subordinate op-erational decision problems when estimating yardperformance. Notice that this might introduce thebias that yard layouts are systematically underratedwith regard to performance because sophisticatedscheduling procedures would allow for more effi-cient resource utilization than anticipated by simplerules of thumb. Consequently, existing studies mightrecommend more efficient layouts accompanied byhigher investment cost than actually required. It fol-lows that simulation studies incorporating sophisti-cated scheduling procedures could provide a valuablecontribution for promoting the success of intermodaltransport.

Furthermore, several potential yard layouts havenot yet been evaluated. For instance, crossover cranesoperating on different tracks could be employed,as in intermediate block storages at modern seaports,where even triple crossover gantry cranes are cur-rently in operation (Dorndorf and Schneider 2010).

Here, a pair of twin cranes running on the same tracksis supported by a large crossover crane on its ownrails. Evaluating the performance of these alternativecrane layouts in comparison to existing layout con-figurations would be valuable decision support forfuture terminal projects.

An important decision, which at the same timeheavily influences the yard layout, is the question ofwhether the European policy of allowing customertrucks to directly enter the transshipment area or theNorth American policy of applying container trailersthat transfer containers in the holding area is actuallythe better choice. Although the latter policy reducesthe load of gantry cranes by avoiding additional splitmoves and allows for better (deterministic) planningof crane schedules, it extends delivery times for cus-tomers. A detailed comparison of both policies andtheir impact on yard layout is a challenging subjectfor future research.

Furthermore, additional research in the operationalarea is required with regard to each of the decisionproblems defined in §3.1:

(I.i) Existing research on identifying appropriatestacking positions of containers in intermediate stor-age is mainly dedicated to operations in seaports.Although in principle the developed procedures canbe used in both fields of application, the dimensionsof the storage areas in rail terminals are much smaller.The mean stacking height, for instance, in many railyards is merely between 1 and 1.5 containers (Ballisand Golias 2002). Therefore, it would be a valuablecontribution to test whether applying the sophisti-cated procedures developed for seaports in fact accel-erates container processing sufficiently to justify theinvestment cost of the required information system.

(I.ii) The assignment of parking positions to trucksis a widely unexplored field of research. OnlyFroyland et al. (2008) have integrated this probleminto a holistic planning approach. Clearly, this prob-lem is of minor importance if only a few trucks enterthe yard simultaneously because each truck can beparked directly next to its respective container loca-tion. However, directly after a train’s arrival, typi-cally, many trucks are already waiting for containerprocessing. These trucks compete for scarce parkingpositions close to their respective containers. In sucha setting, the (dynamic) assignment of parking lotsto trucks such that congestion in the yard is reducedrequires sophisticated decision support and is thus achallenging problem for future research.

(I.iii) The train loading problem has attracted thelargest number of research contributions thus far.However, a versatile model integrating all real-world weight and loading constraints of wagons(as, e.g., defined by Bruns and Knust 2012) with aero-dynamic aspects (Lai, Barkan, and Önal 2008a) along

Dow

nloa

ded

from

info

rms.

org

by [

137.

250.

43.1

01]

on 1

8 A

pril

2016

, at 0

1:16

. Fo

r pe

rson

al u

se o

nly,

all

righ

ts r

eser

ved.


with suitable solution procedures is still missing. Fur-thermore, the degrees of freedom for train loading arediminished by the diversity of trailers and wagons tobe processed. A further standardization of containerspromises reduced effort in changing pin configura-tions of railcars and thus load plans that are moreefficient. Therefore, quantifying such standardizationeffects could help further encourage standardizationagreements for rail transport.

(I.iv) With regard to the assignment of containermoves to cranes, two basic policies are distinguishedin this survey: static assignment, where each craneoperates in a distinct yard area, and dynamic assign-ment, where the obstruction of cranes is to be avoidedin real time. Clearly, the dynamic approach promisescrane schedules that are more efficient but comes atthe price of a suitable information system. A thoroughquantitative comparison of both policies would offervaluable decision support for real-world yards.

(I.v) Once all container moves are specified andassigned to cranes, the sequence of container movesper crane resembles a sequential ordering problem(Montemanni et al. 2009). However, in reality, truckarrivals in particular are subject to uncertainties, andsequential ordering thus needs to be executed in anonline environment. It remains an open question as tohow to integrate sequential ordering in a rolling plan-ning horizon. Moreover, it should be tested whethersuch an approach is indeed able to considerably out-perform the common policy of experienced craneoperators making decentralized scheduling decisions.

Until now, freight traffic has only been profitableif full trains are moved over comparatively long dis-tances (see §2), and therefore, the degrees of freedomfor parking trains in the transshipment area are lim-ited. However, with the realization of hub-and-spokesystems, smaller trains might also become profitable,which in turn affects the operational planning envi-ronment. For instance, horizontal parking positionsmight be used to evenly balance the workload amongcranes if the lengths of trains vary sufficiently, whichgives rise to a parking problem (similar to the park-ing problem of rail–rail terminals; see §4.1). A carefulexamination of trends in intermodal transport mightyield interesting insights with regard to upcomingchallenges of yard planning.

Finally, in addition to an isolated investigationof the above decision problems, holistic approachesseem an especially promising field for future research.Currently, there are limited proposals for hierarchi-cal procedures, e.g., those presented by Froylandet al. (2008) and Souffriau et al. (2009). The highdegree of interdependence and relatedness of thediscussed decision problems makes determining theright sequence of decisions and hierarchical integra-tion of all or at least some decisions a challeng-ing task.

4. Rail–Rail Transshipment Yards4.1. Yard Layout, Transshipment Process, and

Decision ProblemsModern rail–rail terminals mainly serve as hub nodesin a hub-and-spoke rail network. Containers are trans-shipped among trains without the need to exchangerailcars, and inbound trains are thus consolidatedto a (reduced) set of full outbound trains. In someimplementations a terminal might additionally pro-cess rail–road operations, but these are usually ofminor relevance in comparison to the hub operations.A pure rail–rail terminal is schematically depicted inFigure 6.

The main technological innovation that rail–rail ter-minals introduce in comparison with conventionalrail–road terminals is that the simple floor storagearea is replaced by a fully (or partially) automatedsorting system. The sorter consists of moving andbuffer lanes, where automated guided vehicles (Bosteland Dejax 1998) or shuttle cars (Alicke 2002) receive acontainer from a gantry crane close to the container’sinitial position on the train and move it alongsidethe yard toward its dedicated container position onthe outbound train. A fully automated system can,for instance, employ rail-mounted shuttle cars, whichuse a rotation mechanism for changing tracks (Franke2002) and are propelled by contact-free linear syn-chronous motors with an electronic position detectionsystem that is able to direct shuttle cars with an accu-racy of ±3 mm (Bauer 1998). Simulation studies forthe Megahub in Hannover-Lehrte indicate that such asorting system increases container processing by up to45 container moves per hour and crane (Rotter 2004).

The impact of the high-performance sorting systemalso explains the critical importance of the systemduring the design phase. The choice of an appro-priate drive technology and the dimensioning of thesystem with regard to storage space and shuttle carsare two of the most important decisions in this con-text, in addition to the choice of general yard lay-out determined by the number of tracks and cranes.

FreighttrainTrack

Sorter

BufferingMoving

Gantry cranes

Figure 6 Schematic Representation of a Rail–Rail Transshipment Yard

Dow

nloa

ded

from

info

rms.

org

by [

137.

250.

43.1

01]

on 1

8 A

pril

2016

, at 0

1:16

. Fo

r pe

rson

al u

se o

nly,

all

righ

ts r

eser

ved.


In §4.2, operations research tools that support thedesign phase of a rail–rail terminal are reviewed.

The operational process of container consolidationat a rail–rail terminal is similar in principle to that ata rail–road terminal. Because the hub visit constitutesa time-consuming additional step in the distributionprocess, some organizational changes are necessaryto speed up transshipment and avoid a tedious stayfor an extra day. Generally, trains are required toexchange containers within a few hours so that theconsolidation process is rapid and trains can departto their next destinations the same day (or night).Therefore, a rail–rail terminal is operated in distinctso-called pulses (Bostel and Dejax 1998) or bundles(Alicke 2002; Rotter 2004) of trains. This means thatall tracks are occupied by trains, which are simulta-neously served and jointly leave the system only afterall container moves of the respective bundle are pro-cessed. Whenever the total number of incoming trainsexceeds the number of tracks, a first decision problemassigns each train to a bundle. This decision is sub-ject to release dates and departure times of trains asgiven by the train schedule and might consider sev-eral objectives that conflict in part, such as minimiz-ing the number of containers dedicated to a train ofan earlier bundle or maximizing the number of directmoves among trains of the same bundle. The formerobjective reduces the number of containers that do notarrive at their outbound trains (and thus need to beheld until the arrival of the next train that serves therespective destination), whereas the latter objectiveaccelerates train processing by reducing the amountof double-handling.

Once the pulses are determined, vertical and hori-zontal parking positions need to be assigned to eachtrain of a bundle. Trains that exchange a large num-ber of containers should be assigned to neighboringtracks to reduce the total distance that a loaded cranemoves. Because hub terminals also tend to be vis-ited by shorter trains, appropriate horizontal parkingpositions of trains can positively affect yard perfor-mance, for instance, by evenly spreading containermoves among cranes or by moving start and targetpositions of a container to the same area of operationso that a single crane can process the job instead ofrelying on the sorting system.

The problem of determining an appropriate loadpattern of containers on trains is similar to that aris-ing at rail–road terminals. A load plan, which min-imizes the overall distances that containers move,seems to have a somewhat smaller effect on the yardperformance if the sorting system is not a bottleneck,in which case containers can be prepositioned next totheir intended target positions. However, wheneverinbound and outbound loads are exchanged simulta-neously and the parking positions of trains and oper-ating areas of cranes are fixed, the load plan alone

determines the target position of the outbound con-tainer, and therefore the number of split moves canbe reduced by moving the target position closer to thestarting position.

The assignment of container moves to cranes canbe executed under a static or dynamic policy of dis-tinct or variable crane areas, respectively (see §3.1),and should generally aim to avoid split moves and toshare the workload evenly among cranes.

Finally, the schedule of container moves per craneis to be determined. The resulting problem consti-tutes an extension to crane scheduling at rail–roadterminals and is similar in structure to a sequen-tial ordering problem. One important aspect is thatcrane moves are asymmetric in distance; i.e., exe-cuting move A before B results in a distance differ-ent from that of the reverse order. This is becauseany two loaded moves need to be connected by anunloaded move of the crane and the distance betweenthe end position of move A and the start position ofmove B, which will typically differ from the distancebetween the end position of move B and the startposition of move A. Furthermore, container movesare subject to precedence constraints whenever a con-tainer is blocking another container’s target position.In addition, split moves via the sorting system needto be considered. This leads to heavily interdepen-dent crane schedules because the release date of acontainer transported by the sorter depends on whenanother crane has fed this container into the sortingsystem. This problem becomes even more complex ifthe sorter is a bottleneck and the availability of shuttlecars over time needs to be considered.

The basic decision problems of container processingin rail–rail yards can be summarized as follows.

(II.i) Schedule the service slots of trains by assign-ing them to bundles.

(II.ii) Assign each train a parking position.(II.iii) Determine the positions of containers on

trains.(II.iv) Assign container moves to cranes.(II.v) Schedule the shuttle cars in the sorter.(II.vi) Determine the sequence of container moves

per crane.Existing literature on decision support with regard

to these decision problems is reviewed in §4.3.

4.2. Literature on Layout PlanningThere have been very few studies investigating suit-able layouts of modern rail–rail terminals.

Meyer (1999) investigated the layout problem ofa rail–rail terminal that successively processes bun-dles of six trains. In addition, the terminal is assumedto handle a limited volume of rail–road containerexchanges. An animated computer simulation basedon Petri nets (see Peterson 1981) was developed to

Dow

nloa

ded

from

info

rms.

org

by [

137.

250.

43.1

01]

on 1

8 A

pril

2016

, at 0

1:16

. Fo

r pe

rson

al u

se o

nly,

all

righ

ts r

eser

ved.


determine the required capacity for cranes and inter-nal transport systems, and the most efficient arrivalpattern of trains was identified. Results were obtainedfrom simulation runs for a terminal planned to beconstructed in Germany.

Wiegmans et al. (2007) compared shunting yards,conventional rail–road terminals, and modern rail–rail yards with regard to their suitability of servingas a hub in a hub-and-spoke network for specificarrival patterns of trains and containers. The result-ing crane and shunting operations were simulatedfor all three yard types and several terminal layouts.Layout and operational costs are evaluated in a sep-arate cost module to derive suitable application sce-narios for all terminal types. It was established thata modern rail–rail terminal is beneficial only underhigh-capacity utilization, non-predefined load posi-tions of containers, synchronized train arrivals, andsynchronized crane operations. Suggesting a portfo-lio matrix, Wiegmans et al. (2007) concluded thata modern rail–rail yard is the appropriate choiceas a hub node whenever fast operations take pri-ority, whereas shunting yards and rail–road termi-nals are favorable for cost-efficient operations. Theresults presented were mainly based on the thesis ofBontekoning (2006).

4.3. Literature on Operational PlanningThe assignment of trains to bundles (II.i) was firstinvestigated by Boysen, Jaehn, and Pesch (2011), whoformulated a basic transshipment yard schedulingproblem that minimizes a weighted objective func-tion considering split moves between those trains thatare assigned to different bundles and the numberof revisits by trains that could not receive all con-tainers during their first visit to the yard. The prob-lem was shown to be NP-hard in the strong sense,and different heuristic and exact solution procedureswere presented. The study of Boysen, Jaehn, andPesch (2012) extended this research. The transship-ment yard scheduling problem was modified to con-sider an additional objective function that minimizedthe number of containers that could not be trans-shipped to their target trains in time. Additional com-plexity results were presented, and a more efficientexact branch-and-bound procedure and a very effi-cient heuristic ejection chain approach were provided.

Kellner, Boysen, and Fliedner (2012) presented asolution approach to solve decision problem (II.ii),i.e., the parking problem of trains. In their approach,the assignment of a given bundle of trains totracks (vertical position) and horizontal parking posi-tions along the spread of the yard aimed to mini-mize the makespan of train processing. Furthermore,split moves via the sorting system were consideredbetween cranes operating in disjoint crane areas.They presented a genetic algorithm for solving the

resulting problem and tested their approach in a sim-ulation study. An even simpler approach for deter-mining parking positions (II.ii) was presented byAlicke and Arnold (1998). They merely modeled thetrack assignment of trains (vertical position) in avery basic fashion without, for example, consideringhorizontal parking positions, sorter operations, andmultiple cranes. Instead, they developed a simple pri-ority value weighting the number of container moveswith their total horizontal distance to approximate theresulting workload between two trains. These weightswere then applied in a quadratic assignment problemto determine the track assignment.

Bostel and Dejax (1998) treated problem (II.iii) andaimed to jointly determine load plans for inboundand outbound containers. Start and target positionsof a container move on inbound and outbound trainswere to be brought as close together as possibleso that the travel distances of cranes and the result-ing costs of train processing were minimized. How-ever, with regard to load planning, the underlyingassumptions were rather limiting. It was assumedthat each container can be stored on any wagon, andweight restrictions were not considered. Four differ-ent models for this problem were derived by addition-ally considering container transfer by shuttle cars andstorage constraints in the sorting system. Each modelrequires different solution approaches, and therefore,several procedures based on the linear assignmentproblem, the minimum flow problem, and differentstart and improvement heuristics were developed.A computational study using real-world data for aFrench railway company showed a huge potentialfor accelerating train processing using simultaneouslyoptimized load plans.

Boysen, Fliedner, and Kellner (2010) assigned staticand disjoint crane areas (II.iv) to a bundle oftrains with given parking positions to minimize themakespan of train processing. The sorting system wasassumed to be activated whenever start and targetpositions of a container move fell into the differentcrane areas. They presented a polynomial dynamicprogramming procedure for solving the resultingproblem and tested the solutions against typical real-world policies in a simulation of yard operations.

Alicke (2002) jointly treated decision problems(II.iv), (II.v), and (II.vi). A given set of crane movesis assigned to cranes with overlapping areas of oper-ation, which are blocked whenever a crane enters anarea. Whenever a start or target position falls in anoverlapping area, the procedure dynamically decideswhich of two neighboring cranes processes the move.This decision also influences whether or not a con-tainer move uses the sorting system in a split move.The model takes the movement speed and availabilityof shuttle cars into account. The overall problemwas modeled as a constraint satisfaction problem

Dow

nloa

ded

from

info

rms.

org

by [

137.

250.

43.1

01]

on 1

8 A

pril

2016

, at 0

1:16

. Fo

r pe

rson

al u

se o

nly,

all

righ

ts r

eser

ved.


and tested on data sets of the German MegaHub inHannover-Lehrte. Different heuristic rules for fixingvariables of the constraint satisfaction problem werecompared.

At the border of two countries and railway systems,rail–rail terminals are also used to bridge differ-ent track gauges. This requires a special yard set-ting where complete train loads are transshipped bycranes onto a train with the gauge width of the desti-nation railway system. Martinez et al. (2004) investi-gated two simple rules for crane scheduling (II.vi) ata terminal on the border between France and Spain.Both rules were compared by means of a simula-tion study. The same terminal was investigated byGonzalez et al. (2008), who provided a mixed integermodel to jointly determine the load plan of outboundtrains (II.iii) and crane schedules (II.vi). Their objec-tive was to minimize crane distances while observ-ing the weight and length restrictions of wagons. Themodel was solved with an off-the-shelf solver thatwas shown to be suitable for real-world instances ofsmall size.

4.4. Future Research ChallengesBecause consolidating containers at a rail–rail termi-nal is still an emerging technology for railway sys-tems (Bontekoning, Macharis, and Trip 2004), thereremain many open fields for future research.

Similar to the situation for rail–road terminals, thefew existing simulation studies merely apply verysimple myopic decision rules when evaluating theperformance of terminal layouts. However, the threatof underestimating the performance of rail–rail termi-nals seems even more imminent because many termi-nals are currently in the conceptual evaluation phase.If poor scheduling rules lead to a poor forecast of yardperformance, it is to be expected that some projects(including new hub-and-spoke systems) will not beconstructed, which in turn might further deterioratethe market share of rail freight traffic.

The following open research challenges in opera-tional planning are identified.

(II.i) Assigning trains to bundles is currently per-formed in a deterministic setting. However, becausetrain timetables are bound to change, transshipmentyard scheduling might be more appropriately mod-eled on rolling horizons to account for unplannedtrain cancellations or arrivals.

(II.ii) Currently, there are only two approaches fordetermining parking positions of trains (Alicke andArnold 1998; Kellner, Boysen, and Fliedner 2012),both of which are heuristic in nature. Efficient exact-solution procedures might yield valuable insightsinto the solution structure of this specific assignmentproblem.

(II.iii) Bostel and Dejax (1998) introduced simulta-neous load planning for inbound and outbound trains

to minimize the distances for crane moves. However,they assumed that only a single container can beloaded onto a wagon and further that any containercan be assigned to any car. Real-world weight con-straints and length restrictions were not considered,and neither were split moves resulting from containermoves across different crane areas. Therefore, futureresearch should seek to integrate more of the diversereal-world constraints of train planning, which havealready been widely explored for conventional rail–road terminals (see §3.1).

(II.iv) Analogously to rail–road terminals, studieson the performance of static versus dynamic craneareas have not yet been undertaken (see §3.1).

(II.v) Existing research mostly assumes that thesorting system is not a bottleneck. However, when-ever the availability of shuttle cars is not guaran-teed, crane scheduling needs to be complementedby sophisticated scheduling of shuttle cars. An inter-esting yet unexplored problem in this context is tofurther the real-time control of sorting vehicles to gen-erate deadlock-free travel routes.

(II.vi) Analogously to rail–road terminals, thescheduling of crane moves can be modeled as a se-quential ordering problem as soon as all crane moveshave been specified and assigned to cranes. However,the problem becomes somewhat more complicated inrail–rail yards because split moves processed by thesorting system need to be considered. Some contain-ers become available only after they have been fedinto the sorter by another crane, and crane schedulesthus cannot be decomposed. Suitable crane schedul-ing procedures have not yet been developed.

A further field for future research is to integrate theabove decision problems into a holistic procedure. Forinstance, parking positions of trains and load plansof inbound and outbound trains both influence cranemoves and determine whether or not a containermove needs to use the sorting system. Integratingboth problems in a simultaneous planning procedureor developing a hierarchical framework that incorpo-rates the two decision problems would be a valuablecontribution.

5. ConclusionThis paper surveyed layout planning and operationaldecision problems arising in rail freight yards. Thecore decision problems of rail–road terminals andmodern rail–rail transshipment yards were character-ized and existing research was reviewed. To providea more concise overview of existing research, Table 1lists the literature of the field along with the prob-lem treated and the methodology employed. By con-trasting the structure of decision problems with thescope of existing research, several avenues for futureresearch are identified.

Dow

nloa

ded

from

info

rms.

org

by [

137.

250.

43.1

01]

on 1

8 A

pril

2016

, at 0

1:16

. Fo

r pe

rson

al u

se o

nly,

all

righ

ts r

eser

ved.


Table 1 Summary of the Literature

Source Terminal Decision problem Method

Abacoumkin and Ballis (2004) Rail–road Layout SimulationAlicke (2002) Rail–rail (II.iv), (II.v), (II.vi) Constraint satisfactionAlicke and Arnold (1998) Rail–rail (II.ii) Mixed-integer programmingBallis and Golias (2002) Rail–road Layout SimulationBallis and Golias (2004) Rail–road Layout SimulationBenna and Gronalt (2008) Rail–road Layout SimulationBontekoning (2006) Rail–road, rail–rail Layout SimulationBostel and Dejax (1998) Rail–rail (II.iii) Linear assignment, network flow, mixed-integer programmingBoysen and Fliedner (2010) Rail–road (I.iv) Dynamic programmingBoysen, Fliedner, and Kellner (2010) Rail–rail (II.iv) Dynamic programmingBoysen, Jaehn, and Pesch (2011) Rail–rail (II.i) Dynamic programming, beam searchBoysen, Jaehn, and Pesch (2012) Rail–rail (II.i) Branch and bound, ejection chainBruns and Knust (2012) Rail–road (I.iii) Mixed-integer programmingCorry and Kozan (2006) Rail–road (I.iii) Mixed-integer programmingCorry and Kozan (2008) Rail–road (I.iii) Mixed-integer programming, local searchFeo and González-Velarde (1995) Rail–road (I.iii) Branch and bound, GRASPFerreira and Sigut (1993) Rail–road Layout SimulationFerreira and Sigut (1995) Rail–road Layout SimulationFroyland et al. (2008) Rail–road (I.i), (I.ii), (I.v) Mixed-integer programming, priority rulesGonzalez et al. (2008) Rail–rail (II.iii), (II.vi) Mixed-integer programmingGronalt, Benna, and Posset (2007) Rail–road Layout SimulationKellner, Boysen, and Fliedner (2012) Rail–rail (II.ii) Genetic algorithmKondratowicz (1990) Rail–road Layout SimulationKozan (1997) Rail–road (I.iv) Priority rulesKozan (2006) Rail–road Layout SimulationLai and Barkan (2005) Rail–road (I.iii) Mixed-integer programmingLai, Barkan, and Önal (2008a) Rail–road (I.iii) Mixed-integer programmingLai, Ouyang, and Barkan (2008b) Rail–road (I.iii) Mixed-integer programmingLee et al. (2006) Rail–road Layout SimulationMartinez et al. (2004) Rail–rail (II.vi) SimulationMeyer (1999) Rail–rail Layout SimulationMontemanni et al. (2009) Rail–road (I.v) Local search, ant colony optimizationRizzoli, Fornara, and Gambardella (2002) Rail–road Layout SimulationSouffriau et al. (2009) Rail–road (I.iii), (I.iv), (I.v) Linear assignment, mixed-integer programming, variable

neighborhood searchVis (2006) Rail–road Layout SimulationWiegmans et al. (2007) Rail–road, rail–rail Layout Simulation

Clearly, there are future research challenges notonly in the context of each individual terminal typebut also with regard to their integration into an exist-ing railway network. Each terminal type varies ininvestment cost and operational performance, andchoosing the right terminal type with a proper lay-out and efficient operational transshipment processesis thus a challenging task. Furthermore, individualperformance assessment needs to be complementedby a network analysis that takes the relations to allother nodes of the rail network into account. It followsthat a concerted research effort is required to success-fully promote an efficient use of rail freight in thefuture.

AcknowledgmentsThe authors appreciate the careful reading and insightfulcomments of the referees and the editor. This work hasbeen supported by the German Science Foundation (DFG)with the grant “Optimierung der Containerabfertigung inUmschlagbahnhöfen” [BO 3148/1-1 and PE 514/16-1].

ReferencesAbacoumkin C, Ballis A (2004) Development of an expert sys-

tem for the evaluation of conventional and inovative tech-nologies in the intermodal transport area. Eur. J. Oper.Res. 152(2):410–419.

Alicke K (2002) Modeling and optimization of the intermodal ter-minal mega hub. OR Spectrum 24(1):1–17.

Alicke K, Arnold D (1998) Modellierung und Optimierungvon mehrstufigen Umschlagsystemen. Fördern und Heben8(10):769–772.

Allianz-pro-Schiene (2008) Umweltschonend mobil. Bahn, Auto,Flugzeug, Schiff im Umweltvergleich, press release, Berlin.

Arnold P, Peeters D, Thomas I (2004) Modelling a rail/roadintermodal transportation system. Transportation Res. Part E40(3):255–270.

Assad AA (1980) Modelling of rail networks: Toward a rout-ing/makeup model. Transportation Res. Part B 14(1–2):101–114.

Ballis A, Golias J (2002) Comparative evaluation of existing andinnovative rail-road freight transport terminals. TransportationRes. Part A 36(7):593–611.

Ballis A, Golias J (2004) Towards the improvement of a combinedtransport chain operation. Eur. J. Oper. Res. 152(2):420–436.

Bauer R (1998) Innovative linear motor-based transfer technologyallows intelligent container handling. Proc. EURNAV, Hannover.

Dow

nloa

ded

from

info

rms.

org

by [

137.

250.

43.1

01]

on 1

8 A

pril

2016

, at 0

1:16

. Fo

r pe

rson

al u

se o

nly,

all

righ

ts r

eser

ved.


Benna T, Gronalt M (2008) Generic simulation for rail–road con-tainer terminals. Mason S, Hill R, Mönch L, Rose O, Jefferson T,Fowler J, eds. Proc. 2008 Winter Simulation Conf., Miami,2656–2660.

Bierwirth C, Meisel F (2010) A survey of berth allocation and quaycrane scheduling problems in container terminals. Eur. J. Oper.Res. 202(3):615–627.

Bontekoning Y (2006) Hub Exchange Operations in Intermodal Hub-and-Spoke Networks (Delft University Press, The Netherlands).

Bontekoning Y, Priemus H (2004) Breakthrough innovations inintermodal freight transport. Transportation Planning Tech.27(5):335–345.

Bontekoning Y, Macharis C, Trip J (2004) Is a new applied trans-portation research field emerging? A review of intermodalrail-truck freight transport literature. Transportation Res. Part A38(1):1–24.

Borgman B, van Asperen E, Dekker R (2010) Online rules for con-tainer stacking. OR Spectrum 32(3):687–716.

Bose P (1983) Applications of computer model techniques for rail-road intermodal terminal configuration, equipment and oper-ational planning. Transportation Res. Record 907:45–52.

Bostel N, Dejax P (1998) Models and algorithms for container allo-cation problems on trains in a rapid transshipment shuntingyard. Transportation Sci. 32(4):370–379.

Boysen N, Fliedner M (2010) Determining crane areas in intermodaltransshipment yards: The yard partition problem. Eur. J. Oper.Res. 204(2):336–342.

Boysen N, Fliedner M, Kellner M (2010) Determining fixed craneareas in rail–rail transshipment yards. Transportation Res. Part B46(6):1005–1016.

Boysen N, Jaehn F, Pesch E (2011) Scheduling freight trains in rail–rail transshipment yards. Transportation Sci. 45(2):199–211.

Boysen N, Jaehn F, Pesch E (2012) New bounds and algorithmsfor the transshipment yard scheduling problem. J. Scheduling.Forthcoming.

Bruns F, Knust S (2012) Optimized load planning of trains in inter-modal transportation. OR Spectrum. Forthcoming.

Bussieck M, Winter T, Zimmermann U (1997) Discrete optimizationin public rail transport. Math. Programming 79(1–3):415–444.

Caris A, Macharis C, Janssens G (2008) Planning problems inintermodal freight transport: Accomplishments and prospects.Transportation Planning Tech. 31(3):277–302.

Caserta M, Voß S, Sniedovich M (2009) Applying the corri-dor method to a blocks relocation problem. OR Spectrum33(4):915–929.

Choe R, Park T, Oh M-S, Kang J, Ryu KR (2011) Generatinga rehandling-free intra-block remarshaling plan for an auto-mated container yard. Internat. J. Intelligent Manufacturing11(2):201–217.

Cordeau J-F, Toth P, Vigo D (1998) A survey of optimizationmodels for train routing and scheduling. Transportation Sci.32(4):380–404.

Corry P, Kozan E (2006) An assignment model for dynamic loadplanning of intermodal trains. Comput. Oper. Res. 33(1):1–17.

Corry P, Kozan E (2008) Optimised loading patterns for intermodaltrains. OR Spectrum 30(4):721–750.

Crainic TG, Kim KH (2007) Intermodal transport. Barnhart C,Laporte G, eds. Transportation, Handbooks in Operations Researchand Management Science, Vol. 14 (Elsevier, North-Holland),467–538.

Daganzo CF (1989) The crane scheduling problem. TransportationRes. Part B 23(3):159–175.

de Castillo B, Daganzo CF (1993) Handling strategies for importcontainers at marine terminals. Transportation Res. Part B27(2):151–166.

Dekker R, Voogd P, van Asperen E (2006) Advanced methods forcontainer stacking. OR Spectrum 28(4):563–586.

Dorigo M, Stützle T (2004) Ant Colony Optimization (MIT Press,Cambridge, MA).

Dorndorf U, Schneider F (2010) Scheduling automated triplecross-over stacking cranes in a container yard. OR Spectrum32(3):617–632.

DUSS (2010) DUSS—Daten and Fakten. Deutsche Umschlagge-sellschaft Schiene—Straße (DUSS) mbH, http://www1.deutschebahn.com/ecm2-duss/start/unternehmen/.

EU (2001) A Study of Single Wagonload Rail Traffic (Brüssel), http://ec.europa.eu/transport/rail/research/doc/wagon-fin.pdf.

EU (2007a) Mitteilung der Kommission an den Rat und daseuropäische Parlament: Aufbau eines vorrangig für den Güter-verkehr bestimmten Schienennetzes (Brüssel), http://ec.europa.eu/transport/logistics/freight_logistics_action_plan/doc/railway/2007_com_freight_de.pdf.

EU (2007b) Mitteilung der Kommission an den Rat und daseuropäische Parlament: Über die Überwachung der Entwick-lung des Schienenverkehrsmarkts (Brüssel), http://eur-lex.europa.eu/LexUriServ/LexUriServ.do?uri=COM:2007:0609:FIN:DE:HTML.

Feo T, González-Velarde JL (1995) The intermodal trailer assign-ment problem. Transportation Sci. 29(4):330–341.

Feo TA, Resende MGC (1995) Greedy randomized adaptive searchprocedures. J. Global Optim. 6(2):109–133.

Ferreira L (1997) Planning Australian freight rail operations: Anoverview. Transportation Res. Part A 31(4):335–348.

Ferreira L, Sigut J (1993) Measuring the performance of intermodalfreight terminals. Transportation Planning Tech. 17(3):268–279.

Ferreira L, Sigut J (1995) Modelling intermodal freight terminaloperations. Road and Transport Res. J. 4(4):4–16.

Fischer T, Gehring H (2005) Planning vehicle transhipment in aseaport automobile terminal using a multi-agent system. Eur.J. Oper. Res. 166(3):726–740.

Franke KP (2002) Transshipment of boxes between vessels andtrains—A dream may become reality. Proc. Internat. Congresson Freight Transport Automation and Multimodality, 23–24 May,Delft, The Netherlands.

Freling R, Lentink RM, Kroon LG, Huisman D (2005) Shunting ofpassenger train units in a railway station. Transportation Sci.39(2):261–272.

Frittelli J (2003) Intermodal connectors: A method for improvingtransportation efficiency. Report, Congressional Research Ser-vice, Washington, DC.

Froyland G, Koch T, Megow N, Duane E, Wren H (2008)Optimizing the landside operation of a container terminal. ORSpectrum 30(1):53–75.

Gatto M, Maue J, Mihalák M, Widmayer P (2009) Shunting for dum-mies: An introductory algorithmic survey. Ahuja R, MöhringR, Zaroliagis C, eds. Robust and Online Large-Scale Optimization,Lecture Notes in Computer Science, Vol. 5868. (Springer, Berlin),310–337.

Geng G, Li LX (2001) Scheduling railway freight cars. Knowledge-Based Systems 14(5–6):189–297.

Gonzalez JA, Ponce E, Mataix C, Carrasco J (2008) The automaticgeneration of transhipment plans for a train-train terminal:Application to the Spanish-French border. Transportation Plan-ning Tech. 31(5):545–567.

Gronalt M, Benna T, Posset M (2007) Strategic planning of hin-terland container terminals: A simulation based procedure.Koschke R, Herzog O, Rödiger K-H, Ronthaler M, eds. GIJahrestagung, Lecture Notes in Informatics (LNI), Vol. 1. 425–428.

Hansen IA (2004) Automated shunting of rail container wag-ons in ports and terminal areas. Transportation Planning Tech.27(5):385–401.

Hansmann RS, Zimmermann UT (2008) Optimal sorting of rollingstock at hump yards. Krebs H-J, Jäger W, eds. Mathematics—Key Technology for the Future (Springer, Berlin), 189–203.

Hartmann S (2004) Generating scenarios for simulation andoptimization of container terminal logistics. OR Spectrum26(2):171–192.

Kellner M, Boysen N, Fliedner M (2012) How to park freight trainson rail–rail transshipment yards: The train location problem.OR Spectrum 34(3):535–561.

Dow

nloa

ded

from

info

rms.

org

by [

137.

250.

43.1

01]

on 1

8 A

pril

2016

, at 0

1:16

. Fo

r pe

rson

al u

se o

nly,

all

righ

ts r

eser

ved.


Kim KH (1997) Evaluation of the number of rehandles in containeryards. Comput. Indust. Engrg. 32(4):701–711.

Kim KH, Hong G-P (2006) A heuristic rule for relocating blocks.Comput. Oper. Res. 33(4):940–954.

Kim KH, Park Y-M (2004) A crane scheduling method for port con-tainer terminals. Eur. J. Oper. Res. 156(3):752–768.

Kim KH, Park YM, Ryu K-R (2000) Deriving decision rules tolocate export containers in container yards. Eur. J. Oper. Res.124(1):89–101.

Klincewicz JG (1998) Hub location in backbone/tributary networkdesign: A review. Location Sci. 6(1–4):307–335.

Kombiverkehr (2008) Frachtbehälter Einteilung nach Längenklassen(UIC), http://www.kombiverkehr.de/neptun/neptun.php/oktopus/download/879.

Kombiverkehr (2009) Success Story—Southern Europe (UIC), http://www.kombiverkehr.de/neptun/neptun.php/oktopus/download/860.

Kondratowicz LJ (1990) Simulation methodology for intermodalfreight transportation terminals. Simulation 55(1):49–59.

Kozan E (1997) Increasing the operational efficiency of containerterminals in Australia. J. Oper. Res. Soc. 48(2):151–161.

Kozan E (2006) Optimum capacity for intermodal containerterminals. Transportation Planning Tech. 29(6):471–482.

Kroon LG, Lentink RM, Schrijver A (2008) Shunting of passen-ger train units: An integrated approach. Transportation Sci.42(4):436–449.

Lai Y-C, Barkan CPL (2005) Options for improving the energy effi-ciency of intermodal freight trains. Transportation Res. Record1916:47–55.

Lai Y-C, Barkan CPL, Drapa, J, Ahuja N, Hart JM, Narayanan PJ,Jawahar CV, Kumar A, Milhon LR, Stehly MP (2007) Machinevision analysis of the energy efficiency of intermodal freighttrains, Proc. Institution of Mechanical Engineers, Part F. J. Rail andRapid Transit. 221(3):353–364.

Lai Y-C, Barkan CPL, Önal H (2008a) Optimizing the aerodynamicefficiency of intermodal freight trains. Transportation Res. Part E44(5):820–834.

Lai Y-C, Ouyang Y, Barkan CPL (2008b) A rolling horizon modelto optimize aerodynamic efficiency of intermodal freight trainswith uncertainty. Transportation Sci. 42(4):466–477.

Lee Y, Chao S-L (2009) A neighborhood search heuristic for pre-marshalling export containers. Eur. J. Oper. Res. 196(2):468–475.

Lee Y, Hsu NY (2007) An optimization model for the container pre-marshalling problem. Comput. Oper. Res. 34(11):3295–3313.

Lee D-H, Hui QW, Miao L (2008a) Quay crane scheduling withhandling priority in port container terminals. Engrg. Optim.40(2):179–189.

Lee D-H, Hui QW, Miao L (2008b) Quay crane scheduling withnon-interference constraints in port container terminals. Trans-portation Res. Part E 44(1):124–135.

Lee BK, Jung BJ, Kim KH, Park SO, Seo JH (2006) A simulationstudy for designing a rail terminal in a container port. Proc.2006 Winter Simulation Conf., Monterey, CA, 1388–1397.

Lim A, Rodrigues B, Xiao F, Xu Z (2004) Crane scheduling withspatial constraints. Naval Res. Logist. 51(3):386–406.

Lim A, Rodrigues B, Xu Z (2007) An m-parallel crane schedul-ing problem with a non-crossing constraint. Naval Res. Logist.54(2):115–127.

Lübbecke ME, Zimmermann UT (2003) Engine routing andscheduling at industrial in-plant railroads. Transportation Sci.37(2):183–197.

Macharis C, Bontekoning YM (2004) Opportunities for OR in inter-modal freight transport research: A review. Eur. J. Oper. Res.153(2):400–416.

Martinez MF, Gutierrez IG, Oliveira AO, Arreche Bedia LM (2004)Gantry crane operations to transfer containers between trains:A simulation study of a Spanish terminal. Transportation Plan-ning Tech. 27(4):261–284.

Mattfeld DC, Kopfer H (2003) Terminal operations management invehicle transshipment. Transportation Res. Part A 37(5):435–452.

Meyer P (1999) Entwicklung Eines Simulationsprogramms fürUmschlagterminals des Kombinierten Verkehrs (Shaker, AachenVerlag, Germany).

Meyrick and Associates (2006) National intermodal terminal study:Final report. Department of Transport and Regional Services,Australian Government, Canberra, A.C.T., Dept. of Transportand Regional Services.

Montemanni R, Smith DH, Rizzoli AE, Gambardella LM (2009)Sequential ordering problems for crane scheduling in port ter-minals. Internat. J. Simulation and Process Modelling 5(4):348–361.

Muller G (1999) Intermodal freight transportation. Eno Transporta-tion Foundation.

Newman A, Nozick L, Yano C (2002) Optimization in the railindustry. Pardalos P, Resende M, eds. Handbook of AppliedOptimization (Oxford University Press, New York), 704–718.

Newman AM, Yano CA (2000) Scheduling direct and indirect trainsand containers in an intermodal setting. Transportation Sci.34(3):256–270.

Nozick L, Morlok K (1997) A model for medium-term operationsplanning in an intermodal rail-truck service. Transportation Res.Part A 31(1):91–107.

Peterkofsky RI, Daganzo CF (1990) A branch and bound solutionmethod for the crane scheduling problem. Transportation Res.Part B 24(3):159–172.

Peterson JL (1981) Petri Net Theory and the Modeling of Systems(Prentice Hall, Upper Saddle River, NJ).

Pisinger D (2002) Heuristics for the container loading problem. Eur.J. Oper. Res. 141(2):382–392.

Powell WB, Carvalho TA (1998) Real-time optimization of contain-ers and flatcars for intermodal operations. Transportation Sci.32(2):110–126.

Progtrans (2007) Abschätzung der langfristigen Entwicklung desGüterverkehrs in Deutschland bis 2050. Bundesministerium fürVerkehr, Bau und Stadtentwicklung, Bonn/Berlin, http://www.bmvbs.de/cae/servlet/contentblob/30886/publicationFile/10534/gueterverkehrs-prognose-2050.pdf.

Rhodes M (2003) North American RAILYARDS (Motorbooks Inter-national, MBI Publishing Company, Minneapolis).

Rizzoli AE, Fornara N, Gambardella LM (2002) A simulation toolfor combined rail/road transport in intermodal terminals.Math. Comput. Simulation 59(1–3):57–71.

Rotter H (2004) New operating concepts for intermodal transport:The mega hub in Hanover/Lehrte in Germany. TransportationPlanning Tech. 27(5):347–365.

Souffriau W, Vansteenwegen P, van den Berghe G, van OudheusdenD (2009) Variable neighbourhood descent for planning craneoperations in a train terminal. Sörensen K, Sevaux M,Habenicht W, Geiger MJ, eds. Metaheuristics in the Service Indus-try, Lecture Notes in Economics and Mathematical Systems,Vol. 624 (Springer, Berlin Heidelberg), 83–98.

Stahlbock R, Voß S (2008) Operations research at container termi-nals: A literature update. OR Spectrum 30(1):1–52.

Steenken D, Voß S, Stahlbock R (2004) Container terminal operationand operations research—A classification and literature review.OR Spectrum 26(1):3–49.

TC (2004) Literature review on intermodal freight transportation.Transports Canada, http://www.tc.gc.ca/media/documents/policy/tp14502e.pdf.

Trip JJ, Bontekoning YM (2002) Integration of small freight flowsin the intermodal transport system. J. Transport Geography10(3):221–229.

Tsamboulas D, Vrenken H, Lekka A-M (2007) Assessment of atransport policy potential for intermodal mode shift on aEuropean scale. Transportation Res. Part A 41(8):715–733.

U.S. DOT (1991) Intermodal surface transportation efficiency act of1991. Department of Transportation, Washington, DC.

U.S. DOT (1998) Transportation Equity Act for the 21st Cen-tury, Moving Americans into the 21st Century. Department ofTransportation, Washington, DC.

van Klink HA, van den Berg GC (1998) Gateways andintermodalism. J. Transport Geography 6(1):1–9.

Dow

nloa

ded

from

info

rms.

org

by [

137.

250.

43.1

01]

on 1

8 A

pril

2016

, at 0

1:16

. Fo

r pe

rson

al u

se o

nly,

all

righ

ts r

eser

ved.


VDA (2006) Auto Jahresbericht 2006. Verband der Automo-bilindustrie, Frankfurt am Main, http://www.vda.de/en/downloads/476/.

Vis IFA (2006) Survey of research in the design and control ofautomated guided vehicle systems. Eur. J. Oper. Res. 170(3):677–709.

Vis IFA, de Koster R (2003) Transshipment of containers at a con-tainer terminal: An overview. Eur. J. Oper. Res. 147(1):1–16.

Wiegmans BW, Stekelenburg DT, Versteegt C, Bontekoning YM(2007) Modeling rail–rail exchange operations: An analysis ofconventional and new-generation terminals. Transportation J.46(4):5–20.

Williams BM, Hoel LA (1998) Freight planning requirements forinterstate corridors. Transportation Quart. 52(2):39–48.

Zhu Y, Lim A (2006) Crane scheduling with non-crossing con-straints. J. Oper. Res. Soc. 57(12):1464–1471.

Dow

nloa

ded

from

info

rms.

org

by [

137.

250.

43.1

01]

on 1

8 A

pril

2016

, at 0

1:16

. Fo

r pe

rson

al u

se o

nly,

all

righ

ts r

eser

ved.

Documents

A Survey on Container Processing in Railway Yards · PDF fileBoysen et al.: Survey on Container Processing in Railway Yards 314 Transportation Science 47(3), pp. 312–329, ©2013