You are on page 1of 17

European Journal of Operational Research 153 (2004) 400–416

www.elsevier.com/locate/dsw

Opportunities for OR in intermodal freight transport


research: A review
a,* b,1
C. Macharis , Y.M. Bontekoning
a
Department of Business Economics and Strategic Management, Vrije Universiteit Brussel,
Pleinlaan 2, M 209, Brussels 1050, Belgium
b
Department of Infrastructure, Transport and Spatial Organisation, Research Institute OTB,
Delft University of Technology, Delft, The Netherlands

Abstract

Intermodal transport reflects the combination of at least two modes of transport in a single transport chain, without
a change of container for the goods, with most of the route traveled by rail, inland waterway or ocean-going vessel, and
with the shortest possible initial and final journeys by road. Operational Research has focused mostly on transport
problems of uni-modal transport modes. We argue that intermodal freight transportation research is emerging as a new
transportation research application field, that it still is in a pre-paradigmatic phase, and that it needs a different type of
models than those applicated to uni-modal transport. In this paper a review is given of the operational research models
that are currently used in this emerging field and the modelling problems, which need to be addressed.
 2003 Elsevier B.V. All rights reserved.

Keywords: Operational research; Intermodal transport; Transportation

1. Introduction cessive, various modes of transport (road, rail,


water) without any handling of the goods them-
Intermodal freight transport has developed into selves during transfers between modes (European
a significant sector of the transport industry in its Conference of Ministers of Transport, 1993). In
own right. This development has been followed by this paper we focus on inland intermodal freight
an increase in intermodal freight transportation transport, rail–truck and barge–truck transport.
research. Intermodal freight transport is the term Comparable research involving ocean shipping is
used to describe the movement of goods in one and not taken into account in this paper.
the same loading unit or vehicle which uses suc- Fig. 1 provides a simple depiction of road–rail
intermodal freight transport. A shipment that
needs to be transported from a shipper to a re-
*
ceiver is first transported by truck to a terminal.
Corresponding author. Tel.: +32-2-629-2286; fax: +32-2- There it is transhipped from truck to itÕs second
629-2060.
E-mail addresses: cathy.macharis@vub.ac.be (C. Macharis),
mode, in this instance a train. The train takes care
bontekoning@otb.tudelft.nl (Y.M. Bontekoning). of the terminal to terminal transport. At the
1
Tel.: +31-15-278-2174; fax: +31-15-278-3450. other end of the transport chain the shipment is

0377-2217/$ - see front matter  2003 Elsevier B.V. All rights reserved.
doi:10.1016/S0377-2217(03)00161-9
C. Macharis, Y.M. Bontekoning / European Journal of Operational Research 153 (2004) 400–416 401

Road haul Road haul – To what extent do the specific problems of inter-
modal transport contribute to the overall devel-
Rail haul opment of OR techniques?
Terminal Terminal
Section 2 describes our review procedure and the
set-up of a classification of OR techniques applied
Shippers or Receivers
in intermodal transport. In the subsequent sections
Fig. 1. A typical representation of road–rail intermodal freight (from 3 to 6) this classification is followed. For
transport. each actor in the intermodal transport system a
general problem description is given, followed by
the review of problems investigated and OR tech-
transhipped from train to truck and delivered by niques applied in intermodal research.
truck to the receiver. The trucking part of the
transport chain is called drayage, pre- and end-
haulage or pick-up and delivery. Instead of rail 2. Approach
transport between the terminals transport by barge
also is possible. The transport between two ter- We have performed a scientific literature re-
minals is called the long haul. For a comprehen- view. We chose a computerised search, due to itÕs
sive description of intermodal freight transport see speed and efficiency. However, we must note that
Muller (1999). electronic sources such as databases have limited
Intermodal freight transport is only just starting coverage. Their earliest date is 1988. Nevertheless,
to be researched seriously. Since 1990 a substantial this relatively short period of coverage is not really
number of analytical publications specifically ad- a significant bias in our review, as we presume that
dressing intermodal transport issues have appeared the majority of the intermodal literature has been
(see Bontekoning et al., in press). Various inter- published in the last ten years. We used a number
modal freight transport decision problems demand of channels when choosing our studies, in order to
models to help in the application of operation re- avoid bias in the coverage. As a preliminary step
search techniques. However, the use of OR in in- we searched the following databases: Transport,
termodal transport research is still limited. The Dissertation Abstracts, and the (Social) Sciences
intermodal transport system is more complex to Citation Index (SCI). However, the databases and
model than the mono-modal one and thus more SCI do not contain all journals; therefore, we
difficult to research. This gives very interesting and performed a separate search of the electronic
challenging tasks for the OR practitioners. journals concerning transportation which were not
Our contribution attempts to provide a com- covered by those channels. We also performed a
prehensive overview of the use of operations re- physical search of the collection in the Delft Uni-
search (OR) in intermodal freight research and to versity of Technology library, which holds the
elaborate a research agenda for the application largest collection of transportation literature in the
and development of OR techniques in intermodal Netherlands. In addition we included research we
freight transport research. Essentially, we want to already knew about from informal contacts with
answer the following questions: other researchers, as well as our own research.
Finally, we retrieved studies by tracking the re-
– What makes the problems in intermodal freight search cited in the literature that we had already
transport research interesting for OR scientists? obtained (ancestry approach). Most channels
– Which OR techniques have been applied, and cover the period from 1988 to 2002, with the ex-
for which problems? ception of the electronic journals, the Dissertation
– Have these techniques been applied appropri- Abstracts database, research of informal contacts
ately? Are there alternative approaches to the and our own research. These channels cover the
same problem? period from 1995 to 2002. Appendix A provides
402 C. Macharis, Y.M. Bontekoning / European Journal of Operational Research 153 (2004) 400–416

a detailed overview of the channels we investi- intermodal transport. The number of studies that
gated. require decisions from more than one operator
In the first round of the research it appeared and/or more time horizons are very limited. This is
that the studies reviewed could be categorised using already an important conclusion of the survey.
two criteria: (1) type of operator, and (2) time Intermodal transport, by definition, involves many
horizon of operations problem. Based on the four decision makers who all need to work in collabo-
main activities in intermodal transport we have ration in order for the system to run smoothly. If
distinguished the following operators: intermodal transport is to be developed it will re-
quire more decision-making support tools to assist
• drayage operators, who take care of the plan- the many actors and stakeholders involved in the
ning and scheduling of trucks between the ter- operation. A very good attempt at outlining these
minal and the shippers and receivers; tools can be found in the paper of Van Duin and
• terminal operators, who take care of the tran- Van Ham (2001) where a three-level modelling
shipment operations from road to rail or barge, approach is followed in order to take account of
or from rail to rail or barge to barge; the different goals of the different stakeholders.
• network operators, who take care of the infra- Table 1 is used to structure the review of the
structure planning and the organisation of rail papers in the following sections. Each section
or barge transport; starts with a general problem statement, which
• intermodal operators, who can be considered as provides a description of the intermodal opera-
users of the inter modal infrastructure and ser- tions related to the type of decision maker and
vices, and take care of the route selection for decision problems involved. In each section the
a shipment through the whole intermodal net- individual contribution to the general problem
work. statement of the paper reviewed is discussed.
Here, the categorisation along the time horizon is
These operators face operational problems with used.
different time horizons. Strategic, tactical and op-
erational problems can be distinguished. In the
strategic level, decisions are found on a very long 3. Drayage operator: OR problems and applications
term (10–20 years). The location of terminals, the
network configurations and the design and layout 3.1. General problem statement
of a terminal are typically decisions were a large
amount of capital is fixed for a long time and that Drayage operations involve the provision of an
are difficult to change. The tactical levels involves empty trailer or container to the shipper and the
a time period of months/weeks. Finally, at the subsequent transportation of a full trailer or con-
operational level, day-to-day or even real-time tainer to the terminal. The empty container may be
decisions are made. picked up either at the terminal, at an empty depot
The combination of these two categories pro- or at a receiver. Delivery operations involve the
vides a classification matrix with twelve categories distribution of a full container or trailer from the
of intermodal operations problems (see Table 1). terminal to a receiver, followed by the collection of
The classification is not exhaustive and some de- the empty container/trailer and its transportation
cision problems can be faced by several decision to the terminal, an empty depot, or a shipper. The
makers and can be relevant for the same decision possibility of separation of tractor and trailer al-
maker at the different time horizons. However, the lows two procedures: ‘‘stay-with’’ and ‘‘drop-and-
decision problems have been placed in the classi- pick’’. In the stay-with procedure, the tractor and
fication matrix of Table 1 were they were most driver stay with the trailer/container during load-
prominent. The studies reviewed have been as- ing and/or unloading. In the drop-and-pick proce-
signed to one of the categories, which provides a dure, a full or empty trailer/container is dropped off
structured overview of the type of OR problems in at the shipper/receiver. During loading/unloading,
C. Macharis, Y.M. Bontekoning / European Journal of Operational Research 153 (2004) 400–416 403

Table 1
Overview of publications reviewed––by research category
Decision maker Time horizon
Strategic Tactical Operational
Drayage operator Co-operation between drayage Allocation of shippers and re- Redistribution of trailer chassis
companies ceiver locations to a terminal and load units
Spasovic (1990) Taylor et al. (2002) Justice (1996)
Walker (1992)
Morlok and Spasovic (1994)
Morlok et al. (1995)

Truck and chassis fleet size Pricing strategies Scheduling of truck trips
– Spasovic and Morlok (1993) Wang and Regan (2002)

Terminal operator Terminal design Capacity levels of equipment Allocation of capacity to jobs
and labour
Ferreira and Sigut (1995) Kemper and Fischer (2000) –
Meyer (1998) Bostel and Dejax (1998)
Van Duin and Van Ham
(2001)
Redesign of operational rou- Scheduling of jobs
tines and layout structures
Voges et al. (1994) –

Network operator Infrastructure network configu- Configuration consolidation Load order of trains
ration network
Crainic et al. (1990) Jourquin et al. (1999) Feo and Gonzalez-Velarde
(1995)
Loureiro (1994) Janic et al. (1999) Powell and Carvalho (1998)
Jourquin et al. (1999) Newman and Yano (2000a)
Southworth and Peterson Newman and Yano (2000b)
(2000)
Location of terminals Type of production model Redistribution of railcars,
barges and load units
Rutten (1995) Nozick and Morlok (1997) Chih and van Dyke (1987)
Meinert et al. (1998) Chih et al. (1990)
Van Duin and Van Ham Bostel and Dejax (1998)
(2001)
Arnold and Thomas (1999) Pricing strategy
Groothedde and Tavasszy Yan et al. (1995)
(1999)
Macharis and Verbeke (1999) Tsai et al. (1994)

Intermodal operator n.a. n.a. Selection of routing and service


Barnhart and Ratliff (1993)
Boardman et al. (1997)
Source: own setup.

the tractor and driver are free to carry out other more). Each drayage company faces a trip sched-
activities. uling problem with trips between shippers, re-
An intermodal terminal may be used by a large ceivers and one or more terminals meeting several
number of drayage companies (a dozen or even requirements, such as customerÕs pre-specified
404 C. Macharis, Y.M. Bontekoning / European Journal of Operational Research 153 (2004) 400–416

pick-up and delivery times (time-windows), on- problems of this size are very difficult to solve.
road travel times, and realistic limits on the length Several redundant and definable constraints have
of the working day. Most shipments are known been introduced in order to push the model closer
about in advance. Only a fraction of loads in a to becoming a general network. The model is dy-
given are short notice. However, loads must namic as it considers the temporal variations in
sometimes be reassigned due to traffic, dock and demands for trailer load movements. The model
intermodal terminal delays. In addition, time must has been applied to a data set with 330 trailer
be allocated for local trailer chassis distribution. movements for an 8-day period, including trailer
Drayage companies carry out trips from and to origins and destinations and requested times for
several nearby terminals in an area. For example, pick-up and delivery. Due to imbalances in trade
in North-America trailer chassis are pooled over the set includes 215 movements from terminal to
several terminals in a region. At each terminal receiver and 115 from shipper to terminal.
sufficient chassis should be available. Due to im- In addition, the model is used to evaluate the
balances in flows, empty chassis need to be redis- efficiency of drayage rates charged by truckers in
tributed from terminals with abundant chassis to the current operation as well as rates used in a
terminals with a shortage. proposed operation with centralise planning of
The general problem of drayage operations is tractor and trailer movements (see Spasovic and
its cost effectiveness. Despite the relatively short Morlok, 1993). The marginal costs generated by
distance of the truck movement compared to the the integer linear program are used for this pur-
rail or barge haul, drayage accounts for a large pose. These are then evaluated by considering the
percentage (between 25% and 40%) of origin to change in total cost resulting from moving an
destination expenses. High drayage costs seriously additional load.
affect the profitability of an intermodal service, Walker (1992) addresses the similar problem of
and also limit the markets in which it can compete central planning as indicated above. His objective
with road transport (see Spasovic and Morlok, being to develop an efficient set of driver tours
1993; Morlok and Spasovic, 1994; Morlok et al., consistent with the shippersÕ pick-up and delivery
1995). Consequently, alternative, less costly oper- times, on-road travel times, and realistic limits
ations need to be designed. on the length of the working day, using a cost-
minimising vehicle-scheduling algorithm. This is
3.2. OR applications for strategic problems accomplished by constructing a computerised net-
work, containing arcs representing all trailer
Spasovic (1990), Morlok and Spasovic (1994) movements and all feasible bobtail connections
and Morlok et al. (1995) investigate whether cen- between trailer movements. The simulation is of a
tral planning of all pick-up and delivery trips of Monte-Carlo assignment and observes the trailer
several drayage companies in one terminal-service movements using a set of hypothetical trucking
area can reduce drayage costs. They examine the firms with a given size distribution. Trailers are
impact of a centralised planning on the total fleet randomly distributed to a given number of firms of
size and utilisation. The objective being to mini- equal size. The work hours and number of drivers
mise total cost of tractor and tractor-trailer required to service the schedule are then tabulated
activities subject to several time and service con- by company and for all trailer movements. The
straints. The model has been constructed as a algorithm is then applied to a data set with 300
large-scale integer linear program with time win- trailer movements distributed among 47 individual
dows and service constraints. The tractor-trailer drayage companies and 150 drivers. Factors such
delivery, repositioning and pick-up operation is as repositioning and redistribution of empty trail-
considered as a time-space network consisting of ers and the delivery of empty trailers to be loaded
nodes connected by links. The nodes represent are not taken into account in this example. Only
the terminals and the shippers/consignees, the links specific pick-up and delivery assignment of full
the tractor and tractor-trailer activities. Integer trailers from A to B are included.
C. Macharis, Y.M. Bontekoning / European Journal of Operational Research 153 (2004) 400–416 405

3.3. OR applications for tactical problems steps: (1) find planning horizon, (2) determine train
arrivals and departures, (3) obtain chassis supply
Taylor et al. (2002) deal with the assignment of and demand; (4) obtain unit costs with each sup-
shipper locations to terminal service areas. Their ply–demand pair; (5) optimise for minimum cost
objective is to develop two alternative heuristics solution through simplex based iterations. The
for intermodal terminal selection and determine model is applied to eight interconnected termi-
their robustness with respect to alternate termi- nals across the USA, three of them located in the
nal location assumptions and other pertinent para- same region. One scheduled double-stack train
meters, with respect to minimisation of total arrivals and departures per day at each terminal.
non-productive (empty) miles associated with cir- There is a known number of supplies and demand
cuitous (off-route miles) and empty travel (in for chassis at each terminal in a given period of
between two successive assignments). After the time.
mathematical formulation of the problem, soft- Wang and Regan (2002) are concerned with the
ware has been written to analyse 40 scenarios for problem of a single drayage company and how
the two alternatives: they can move within a local area with one or more
terminals as many loads as possible at the least
1. Circuit only: assignment of freight to termi- cost. The objective is to minimise the total cost of
nal pairs where load circuitry is the sole crite- providing service to loads within their time con-
rion. straints. Only pick-up time windows (and not de-
2. Total miles: assignment of freight to terminal livery time windows) are considered in order to
pairs based on the sum of total circuitry, empty simplify the process. The fleet size is fixed. New
miles associated with the geographical separa- assignments, trailer repositioning moves and re-
tion of pick-ups and deliveries, and the empty assigned moves due to traffic, dock and intermodal
miles associated with terminal imbalances. facility delays are added to the system as the day
progresses. The problem is solved several times
The 40 scenarios to account for all combinations within the day and more information becomes
of assignment heuristic (two alternatives), allow- available. Each loaded trip is treated as a node. In
able circuits per load (five possible values) and this way the problem may be viewed as an asym-
allowable dray length (four possible values). The metric multiple travelling salesman problem with
heuristics have been applied to 44,546 shipments time window constraints.
and 43 terminals.

3.4. OR applications for operational problems 4. Terminal operator: OR problems and applications

Justice (1996) deals with the problem of a 4.1. General problem statement
drayage company ensuring sufficient chassis avail-
able at terminals in order to meet demand. Trailer Transhipment is inherent to intermodal trans-
chassis are pooled over a set of terminals on both port. As Fig. 1 shows, load units are transhipped
sides of a long haul corridor. Reallocation can take at least twice between truck and train or barge;
place by truck within a region or by train between once at a beginning terminal and once at an end
regions. The objective is to determine when, where, terminal. This type of transhipment is called road–
how many and by what means (truck–train) chassis rail or road–barge exchange. A road–rail or road–
are redistributed and to develop a planning model barge terminal consists of:
with minimum cost solutions for daily decision
support. The problem is mathematically formu- • a road gate, where trucks enter and leave the
lated as a classic bi-directional time based (net- terminal,
work) transportation problem. Own software has • a rail or barge gate, where trains or barges enter
been developed to calculate solutions using five and leave the terminal,
406 C. Macharis, Y.M. Bontekoning / European Journal of Operational Research 153 (2004) 400–416

• a storage area for longer term storage of load them directly off onto another train, or onto the
units (24 hours or more), buffer or other transport system. In the latter two
• a buffer area, for temporary storage of load situations, the load unit is picked up by crane
units, which puts back onto train. When trains are not at
• lifting equipment to unload and load trains, the terminal at the same time but have an exchange
trucks and barges, correlation to each other, load units are sequen-
• storage and transport equipment. tially exchanged via the buffer or storage area (see
Bontekoning and Kreutzberger, 1999, 2001, for
Operations at such beginning or end terminals detailed descriptions of infrastructure and opera-
are globally as follows. When a loaded truck ar- tions of rail–rail terminals and barge–barge termi-
rives it joins the queue at the gate where admin- nals). Containers, swap bodies, trailers or complete
istrative handling takes place. Trucks arrive trucks need to be handled at a terminal. Equipment
randomly at the terminal. The truck enters the used at the terminal must suit these load units.
terminal and lines up either for the storage area or Terminals can apply different transhipment tech-
near the rail or barge siding for direct tranship- niques, layouts, operational strategies, dimensions,
ment to a train or barge. Next, the truck lines up etc. The features of an optimal functioning ter-
to deliver a second load unit or to pick-up one or minal depends on demand volume and type of
two new load units from the storage, buffer or exchange (road–rail, road–barge, rail–rail or barge–
directly from the train or barge. The latter trans- barge, or sometimes rail–barge). Shunting is a quite
action also applies to an empty truck after it has different operation, which consists of the receiving
passed the gate. Trains and barges arrive and de- of trains and an inbound inspection of rail wagons,
part according to a fixed timetable. They can enter classification (sorting) of rail wagons by pushing
the terminal when tracks or berths are empty re- wagons over a shunting hill into a yard with switch
spectively. Trains and barges are first unloaded and classification tracks before trains are assem-
and then reloaded. Load units are transhipped bled and permitted to depart.
from the train or barge either to the buffer, directly Exchange leads to an increase in chain lead time
to a truck, to the storage areas or onto an internal and total transport costs. Consequently, exchange
transport device that brings the load unit to a operations need to be efficient and fast. Terminal
remote storage area. operators have to make decisions on how to meet
Depending on the consolidation concept (see demand requirements. One factor specific to the
Bontekoning and Kreutzberger, 1999) applied in strategic level is the design of the terminal itself.
the rail or barge haul, additional intermediate Decisions regarding design include the type and
transhipment can take place. This is called rail–rail number of equipment used and type and capacity
or barge–barge exchange. The infrastructure of of load unit storage facilities, the way in which
rail–rail or barge–barge terminals is rather similar operations are carried out at the terminal and how
to that of road–rail and road–barge terminals, but the equipment is used, and the layout of the ter-
the layout of the terminals can differ. In addition, minal. At the tactical level a terminal operator
such terminals do not necessarily have facilities decides on the required capacity levels of equip-
to handle trucks. Rail–rail terminals are a new ment and labour, and on the redesign of opera-
concept and are still in the planning stage. Tradi- tional routines and layout structures. Finally, the
tionally, shunting of rail-wagons is applied to rail– terminal operator needs to make operational level
rail exchange. Barge–barge exchange also still decisions regarding capacity and scheduling. One
solely exists on the drawing board. may even add a real-time level, where relatively
Operations at a rail–rail terminal involve the detailed control decisions can be made regarding
exchange of load units between a group of related automatic equipment, for example, which route to
trains. When trains are in the terminal at the same take for an AGV, or to reschedule capacity to
time––this is called simultaneous exchange–– work, based on real-time data, such as equipment
cranes pick up load units from one train and drop brake-downs or delays.
C. Macharis, Y.M. Bontekoning / European Journal of Operational Research 153 (2004) 400–416 407

4.2. OR applications for strategic problems times of trucks (mean value and standard devia-
tion). A combination of Human Integrated Simu-
Ferreira and Sigut (1995) deal with the question lation (HIS) and pure computer simulation based
‘‘Which terminal design performs better, a con- on a Petri-net model has been applied. This com-
ventional road–rail terminal or the RoadRailer bined approach takes both objective influences
concept?’’ The RoadRailer technology uses trailers and human factors into account. It is a game ap-
with the capability of being hauled on road as well proach: real human beings play the role of oper-
as on a rail. Although normal road trailers are also ators at the terminal. The simulation has been
carried on rail there is a significant difference: the applied to a terminal with two cranes serving four
bi-modal trailers are not carried on rail wagons. tracks, two truck track and one storage lane.
Bi-modal trailers are connected by detachable Real data were recorded between 5 a.m. and 9
bogies, which roll on the tracks. The two concepts p.m. The simulation was constructed for a the time
are compared for one variable, the speed of oper- period between 5 a.m. and 2 p.m. All experiments
ation expressed as a mean loading finish time. For were done with the same distribution of arrival
both terminal concepts a discrete event simulation times and other characteristics of real data.
model has been constructed with Simview. Both Bostel and Dejax (1998) want to optimise the
types of operations have been divided into four initial loading allocation of containers on trains in
activities. For each terminal concept the service beginning terminals and their reloading after
time for each of the four activities has been de- transhipment at a rail–rail terminal. The objective
termined based on 120 observations. Experiments is to minimise transfers at rail–rail terminals as
were run for 30 RoadRailer trailers and 30 con- well as the use and volume of the handling
tainers respectively. equipment. The dimensions of the rail–rail termi-
Meyer (1998) faces the design problem of a rail– nal are: four tracks, seven storage lanes (42 places
rail terminal in a hub-and-spoke system for the per row), nine bridge cranes and one device to
exchange of a maximum of six trains at a time. In move load unit from one bridge crane to another.
addition, the terminal should be able to handle a The problem is considered as a minimum cost
limited volume of rail–road exchanges. Dynamic multi-commodity network flow problem with bi-
computer simulation (SIMPRO) with Petri-net nary variables, which are hard to solve optimally.
application and animation was developed to de- Therefore, the problem is broken down into four
termine required capacity for cranes and internal parts, applying heuristic methodology and linear
transport systems, and the most efficient arrival programming:
pattern of trains. Results were obtained from
simulation runs with the arrival of one group of 6 1. the problem of optimisation of container trans-
trains for the time period 2:30 to 3:42 and 62 load fers with imposed initial loading,
units per train. 2. the problem of joint optimisation of the initial
loading and the reloading area,
4.3. OR applications for tactical problems 3. both of the above problems with unlimited stor-
age capacity,
Voges et al. (1994) analyse operating proce- 4. both of the above problems with limited storage
dures for an existing terminal. They focus on three capacity.
questions. How should the dispatcher at the gate
and the crane drivers make their decisions on how This approach has been applied to four different
to continue the process? If a certain crane strategy data sets with each set of 31 trains divided into
would result in favourable waiting times for nine groups of three or four trains, comprising a
trucks, are the crane drivers able to follow it total of 1,000 containers.
without computer support? When would it be Kemper and Fischer (2000) model the transfer
useful to abandon the strategy and to work intu- of containers in a rail–road terminal with a sin-
itively? The value of measurement is the waiting gle crane. In this terminal a storage yard with a
408 C. Macharis, Y.M. Bontekoning / European Journal of Operational Research 153 (2004) 400–416

capacity of 40 containers is used to buffer con- optimal locations is often done by the private
tainers which need to be unloaded from a truck or sector although the government may also be in-
train, but cannot be directly reloaded for further terested to know which kind of investment to
transportation. The objective of the modelling is to support.
determine quality of service in terms of waiting At the tactical level the network operator has to
times and utilisation of resources, especially with determine which services he will offer. Unlike tra-
regard to the dimensions of the waiting areas for ditional rail and barge, fixed schedules are used in
incoming trucks. Other characteristics of the intermodal transport rail and barge systems. In
problem are: trucks carry two load units or none; traditional rail haul networks, trains run only
trains supply and pick up 20 containers in each when full and excessive classification at interme-
instance; and the arrival process of trucks is a diate nodes takes place. In traditional (bulk) barge
Poisson process. Stochastic Petri-nets are used as transport barges travel on demand. Hence, the
modelling language and results are obtained nu- network operator has to fix a service schedule for
merically by computation of the steady state dis- several months to a year in advance. This suggests
tribution of an associated Markov chain. two things.
Firstly, he has to decide which consolidation
network (point-to-point, line, hub-and-spoke or
5. OR problems facing network operators collection–distribution) to use. This must take into
account how to consolidate flows (Fig. 2), the
5.1. General problem statement routing of the trains or barges through the net-
work and which nodes to serve. Point-to-point
The network operator faces decision problems consolidation of flows is the easiest way to orga-
concerning infrastructure planning (strategic level), nise services and it has appeared to be feasible.
service schedules and pricing of services (tactical Therefore network operators most often apply this
level) and daily operations of the services (opera- system. In a point-to-point consolidation network,
tional level). The majority of the studies related trains or barges respectively travel between two
to intermodal infrastructure decisions deal with terminals, without intermediate stops. However,
the interconnectivity of modes in order to achieve this method of consolidation requires large vol-
intermodal transport chains and the location of umes in order to offer a daily service. Still little is
intermodal terminals. In order to achieve decision known about the feasibility of other consolidation
support models for these kinds of problems the
former uni-modal (road/rail/inland waterway)
network models have to be connected to each
other and a transfer of freight has to be made
possible at the nodes (terminals). We see a differ-
ence in the type of stakeholder that is involved in
these type of decisions between the European and
North American papers, especially with respect to
railway infrastructure. In Europe, it is often the
government dealing with railway infrastructure
and terminal location decisions, while in North
America it is the railway operatorÕs responsibility.
However, both European and North American
governments are mostly concerned with questions
regarding the interconnectivity of the modes, the
impact of a capacity increase and the effects of
price/cost increases/decreases on the use of the Fig. 2. Four basic consolidation networks (Source: Adjusted
different infrastructure networks. The search for from Bontekoning and Kreutzberger, 1999).
C. Macharis, Y.M. Bontekoning / European Journal of Operational Research 153 (2004) 400–416 409

forms such as line, hub-and-spoke and collection– complex as different types of flatcars exist, and
distribution models. A few researchers started to there are many types of trailers and containers.
investigate the relationships between consolidation
model, frequency, train length and costs. 5.2. OR applications for strategic problems
Secondly, the network operator has to decide
which production model to use, i.e. how to operate The network operator faces making decisions
the trains or barges. It involves decisions about about the development of the intermodal network,
frequency of service, train length, allocation of which includes decisions regarding investment in
equipment to routes and capacity planning of links and nodes (terminals). In the past, spatial
equipment. Publications show that only intermo- price equilibrium models and network models
dal rail transport has been investigated with re- have been developed for network infrastructure
spect to network operator decisions at the tactical planning. The network approach stems from
level. The intermodal rail system to be modelled is mathematical graph theory and practical experi-
a fairly complicated one, and is quite distinct from ence which was gained from analysing urban road
the traditional rail carload service which has been traffic and transit systems. The equilibrium ap-
the subject of much modelling in the last decade or proach stems from the study of spatially separated
so (see e.g. Assad, 1980). A substantial difference is markets and the concepts of spatial and economic
the interaction between a large variety of both equilibrium. Most models, however, have been
trailers/containers and railcars, and the multiple developed for one mode only, and cannot deal
level of service classes which must be considered. with intermodal flows. Crainic et al. (1990),
In addition, pricing strategy decisions have to be Loureiro (1994), Jourquin et al. (1999), and South-
considered at the tactical planning level. Pricing worth and Peterson (2000) have developed net-
the intermodal transport product is a complicated work models which are capable of dealing with
issue. The intermodal tariff is based on several intermodal flows, which implies that freight can be
tariffs determined by several actors related to the transferred from one mode to another in the model
various parts of the intermodal transport chain. via transfer points.
Intermodal agents, railroad, terminal and drayage In order to adapt the traditional network mod-
managers must each have their own tariff (pricing) els, more importance is given to the nodes, con-
strategy. Tariff determination itself can be very necting two or three networks. Crainic et al. (1990)
complicated. It requires an accurate cost calcula- extended uni-modal network models by adding
tion and insight in the market situation. To esti- links connecting the various modes in order to
mate his negotiating power, each actor must be derive an intermodal network model. The devel-
aware of his market position and the cost structure opment of geographic information system (GIS)
of other actors. The review shows a few studies technology yields new possibilities for the model-
that estimate costs for individual parts of the in- ling of large multi-modal freight networks as
termodal chain. In Section 5.3 we discuss pricing Jourquin et al. (1999) and Southworth and Peter-
of the rail haul part in the intermodal chain and in son (2000) show. GIS is in their work based on the
Section 3.3 pricing of drayage has been discussed. concept of the Ôvirtual networkÕ, that, in a system-
The operational level involves the day-to-day atic way, breaks down all the successive operations
management decisions about the load order of involved in multi-modal transport and includes a
trains and barges, redistribution of railcars or push detailed analysis of all costs. The generalised costs
barges, and load units (fleet management). A typ- are minimised according to the shortest path al-
ical management problem in intermodal rail/road gorithm. By simulation with different parameter
transport is the assignment of a set of trailers and values the software can provide indicators such as
containers to the available flatcars that can move tons per km, total distance, total cost, duration and
this equipment. In the European context the capacity utilisation of nodes and links.
problem focuses on assign containers and swap- Loureiro (1994) presents a multi-commodity
bodies to the different flatcars. The problem is quite multi-modal network model to be used as a
410 C. Macharis, Y.M. Bontekoning / European Journal of Operational Research 153 (2004) 400–416

planning tool for determining investment priorities network. Van Duin and Van Ham (2001) identify
for intercity freight networks. The model is de- optimal locations while incorporating the per-
signed to select the best set of investment options spectives and objectives of shippers, terminal op-
for a multi-modal regional network, given a lim- erators, agents, consignees and carriers. For each
ited investment budget. The main component of level, an appropriate model is developed. At the
the model incorporates the solution of a non-linear strategic level, a linear programming model sear-
bi-level multi-modal network design problem for- ches the optimal locations for intermodal termi-
mulated to choose investments. ItÕs main aim is nals. This model takes account of the existing
to minimise the transportation costs incurred by terminals in the Netherlands and can then be used
users (shippers) and the environmental impacts in order to find some new prospective sites (or
caused by the use of less efficient modes of trans- areas). In the next level a definite location in the
portation for moving freight. Investment options prospective area is found by means of a financial
to be considered by the model may involve the analysis. Here the location of large potential cus-
addition of new physical links to the network, the tomers is one of the most decisive factors. On the
improvement of existing links (i.e. an increase of lowest level or operational level a discrete event
capacity), and the location of intermodal transfer simulation model of the terminal gives the possi-
facilities at specified nodes of the network. bility to simulate the working of the terminal. This
Also for the location of terminals multi-modal model can be used to make decisions on the
networks were developed. These network models amount of cranes, amount of employees, etc. Ar-
are used in combination with origin/destination nold and Thomas (1999) minimise total transport
matrices, wherein the existing and/or the future costs in order to find the optimal location for in-
transportation flows are given. The traffic flows are termodal rail/road terminals in Belgium with the
assigned to the network by a generalised cost func- use of a linear programming model. Groothedde
tion. The possible routes/modes through which the and Tavasszy (1999) minimise generalised and
freight will pass are determined by the costs (in external costs in order to find the optimal location
terms of price, time, congestion phenomena or of intermodal rail/road terminals. They used the
other performance criteria) that are assigned to the simulated annealing technique. In order to find the
routes/modes. The intersections of the networks optimal locations of terminals, terminals are added
are seen as possible locations for the intermodal to the network randomly and, for each changed
terminals. Using this method the problem is re- network configuration, the total generalised (from
duced to a discrete (although sometimes large) set a user viewpoint) and external costs (from a sys-
of possible locations. A general problem when tems viewpoint) are calculated in order to find the
searching for an optimal location is the choice of optimal locations.
which objectives to optimise. Examples of these Meinert et al. (1998) investigate the location of
objectives are: minimising transportation costs on a new rail terminal in a specific region in which
the links, maximising terminal profitability, maxi- three rail terminals are already located. They spe-
mising modal shift from road to intermodal cifically consider the impact of the location of the
means, minimising total transport costs and mini- new terminal on drayage length and time. In order
mising drayage distance and costs. to accomplish this, they developed a discrete event
Network models for terminal location decisions simulation tool which provides the ability to ad-
have been applied by Rutten (1995), Arnold and dress individual rail terminal design considerations
Thomas (1999), Groothedde and Tavasszy (1999), such as handling capacity required, regional design
Van Duin and Van Ham (2001). RuttenÕs (1995) considerations related to terminal location and
objective was to find terminal locations that will trucking distances, and demand distribution over
attract sufficient freight to run daily trains to and time. A significant feature of this simulator is that,
from the terminal. He studies the effect of adding rather than modelling only the operation of the
terminals to the network on the performance of terminal, it also models the drayage to and from
existing terminals and of the overall intermodal regional destinations.
C. Macharis, Y.M. Bontekoning / European Journal of Operational Research 153 (2004) 400–416 411

Another kind of location analysis exists when a solutions. Thirty problems with one hub, three to
number of definite location sites are found but have six origins and destinations, and with different con-
to be evaluated in order to select the most optimal tainer demand patterns and cost structures have
site. Such location analysis allows more criteria, as been generated. Newman and Yano succeeded in
more specific information becomes available. Ma- obtaining optimal solutions for problems with
charis and Verbeke (1999) examined four potential three or four origins and destinations using their
sites for new barge terminals in Belgium. They decomposition procedure. However, larger prob-
followed a multi-criteria analysis approach, where- lems could not be solved.
fore a hierarchy of criteria was built. These criteria Nozick and Morlok (1997) developed a model
represent the aims of the actors who are involved, to plan different rail haul operations simulta-
namely the users of the terminal, the operators/in- neously: incorporating such variables as train
vestors and the community as a whole. The eval- length, motive power allocations, fleet allocations
uation of the terminal projects was executed by the to various traffic lanes, distribution of empty rail
PROMETHEE-method (Preference Ranking Or- cars and trailers, routing policies of traffic, work
ganization METHod for Enrichment Evaluations), allocation and estimation for terminals; but not
a multi-criteria analysis method developed by including train schedules and consolidation net-
Brans (1982) and extended for group decisions by works (point-to-point) as variables. In their study
Macharis et al. (1998). they use fixed examples of these factors. The model
aims at a planning period from one week to a
5.3. OR applications for tactical problems month. The complexity of the models is due to the
fact that it encompasses all elements of the rail
Janic et al. (1999), Jourquin et al. (1999), and haul operations. Integer linear program is used,
Newman and Yano (2000a,b) compare different which is computationally difficult to solve. There-
consolidation networks with each other. Janic et al. fore, a heuristic procedure was developed which
(1999) evaluates 23 state-of-the-art complex con- provides near-optimal solutions, to within 1% of
solidation networks in Europe by means of a the known optimal solution to the ÔrelaxedÕ (non-
multi-criteria analysis in order to assess the most integer) problem.
promising layouts. The simple additive weighting Yan et al. (1995) focus on cost calculations re-
(SAW) method is used here. A set of evaluation garding cost-related pricing strategies for the rail
criteria for network performances has been defined haul. They notice that the usual cost calculations
and quantified for the selected cases. Jourquin et al. are taking an average empty movement cost to
(1999) show how a network model and GIS can be determine the system incremental costs. In doing
used to model different consolidation networks this, they ignore the opportunity costs of using the
and their impact on the distribution of flows over conveyances on movements and on the availability
the available infrastructure and modalities (see of train capacity on a daily basis. A mathematical
also Section 5.2). Newman and Yano (2000a,b) program was written to address this problem in-
developed a model for determining a train sched- corporating an efficient algorithm for approxi-
ule with both direct and indirect (i.e. via a hub) mating reduced cost. The algorithm combines the
trains and the allocation of containers to any of use of Langrangian Relaxation with a minimum
these trains for a one to two week time horizon. cost flow algorithm and a shortest path algorithm.
Given container demands differentiated by origin, Tsai (1994) constructed two models to deter-
destination, arrival date at origin, and due date, mine optimal price and level of service for inter-
the objective is to determine a train schedule and modal transport in competition with truck
container shipment plan to minimise the total cost transport. The models consider the whole inter-
while meeting delivery punctuality requirements modal chain, contrary to Yan et al who only con-
and adhering to train capacity restrictions. The sidered the rail haul. This problem has a number of
problem is formulated as an integer program with a features which distinguishes it from typical pricing
new decomposition procedure to find near-optimal problems in other industries and services, because
412 C. Macharis, Y.M. Bontekoning / European Journal of Operational Research 153 (2004) 400–416

of the special technology and economies of trans- move flatcars is typically coupled with the problem
portation interconnection with headhauls and of planning the repositioning of empty boxes
backhauls (usually unbalanced flows). The models owned by the railroad (ROE). The management of
take into account not only carriersÕ pricing behav- ROE is a classic dynamic fleet management prob-
iour (supply side) but also shippersÕ mode choice lem. The object is to maximise returns. This is as-
behaviour (demand side). Solutions to find equi- sessed by calculating the proportion of fulfilled
librium are pursued by a mathematical program- requests (i.e. placing empty boxes where they are
ming approach. The objective of the models is to needed) against the costs of achieving this. For a
optimise intermodal profit within some constraints, review on dynamic models for fleet management
which include shippersÕ mode choice behaviour, see Powell et al. (1995) and Dejax and Crainic
non-negativity of carrier price and cargo amounts (1987) where special attention is given to the
and intermodal volume constraints. management of empty flows. In order to solve the
coupled problems, two dynamic assignment mod-
5.4. OR applications for operational problems els were developed: one assigns trailers to customer
requests and the other assigns flatcars to boxes. A
Feo and Gonz alez-Velarde (1995) are using an linear programming formulation is used. The LQN
optimisation procedure to tackle the problem of approach can be used in a real-time setting.
assigning highway trailers to railcar hitches (Ôpig- In Chih et al. (1990) a decision support system
gybackÕ transport). The problem is treated as a set- called RAILS is set up to optimally manage in-
covering problem. The idea is to ‘‘cover’’ all of the termodal double-stack trains. This assignment
trailers with a set of railcars at minimum cost. problem is even more complex than the mono-
With the use of an integer-linear programming problem as there are also height constraints and
model, solved by a branch and bound code, a so- choices between different modes have to be made.
lution can be found within a reasonable timescale. The system is to be used on a daily basis to ensure
Additionally, a heuristic for the Trailer Assign- the correct size of each train and to generate rail
ment Problem is developed. In the greedy rando- car repositioning instructions. The planning hori-
mised adaptive search procedure (GRASP) a zon is two weeks and takes the local and global
feasible assignment is contrived. This incorporates system needs into consideration. The problem is
a selection of the most difficult to use railcars formulated as a non-linear multi-commodity inte-
available together with the most difficult to assign ger network flow problem. As the problem is NP
trailers. In doing this, the least compatible and hard, a heuristic method had to be developed in
most problematic equipment is considered first. order to be able to solve the network optimisation
Powell and Carvalho (1998) take this a step problem within a reasonable time (13 minutes).
further. The previous model ignores the impor- The heuristic breaks the solution procedures into
tance the choice of destination has in the aim to several components and uses well developed traffic
fully utilise the equipment. For example, if the assignment and capacitated network transhipment
container is going to a destination that pools a optimisation algorithms to solve the problem. In
large number of trailers, flatcars are favoured that Chih and van Dyke (1987) a similar approach is
can carry trailers. Network information such as followed for the distribution of the fleetÕs empty
this can influence the decisions made by the local trailers and/or containers.
terminal. In the paper of Powell and Carvalho, the
model is set up to aid a local terminal manager to
determine how to assign several trailers and con- 6. OR problems facing intermodal operators
tainers to a flatcar, governed by complex assign-
ment rules. The problem, in this case, is formulated 6.1. General problem statement
as a logistics queuing problem, being a new for-
mulation of the dynamic fleet management prob- Intermodal operators organise the transporta-
lem. Furthermore, the problem of deciding how to tion of shipments on behalf of shippers. Intermodal
C. Macharis, Y.M. Bontekoning / European Journal of Operational Research 153 (2004) 400–416 413

operators buy the services offered by drayage, and the line rail haul costs) and the inventory costs
terminal and network operators. Decisions made (the sum of the in-transit inventory costs and the
by intermodal operators deal with route and ser- cost of additional stock resulting from the transit
vice choices in existing intermodal networks. This time). Two types of decision settings are identified
type of decision, by its nature, is an operational depending on who owns the equipment and who is
one, because it concerns the assignment of ship- providing the service. In the first group the rail-
ments to routes and carriers. Intermodal routing is road charges the shipper on a Ôper trailerÕ basis
rather more complex than the routing problems of and, in the second, the shipper is charged on a Ôper
road haulage. In road haulage, the minimum cost flatcarÕ basis. In the first case, the minimum cost
path algorithm was most commonly used in order routings are achieved with a shortest path finding
to find the route most suited to meeting the ob- procedure. In the second case, the optimal routings
jective of the company, for example, the least are determined with a matching or b-matching
costly or less time consuming route. A large vari- algorithm. These latter models are also able to
ety of combinations of transportation modes is incorporate non-monetary constraints such as
possible. In this case the routing decision is a mere schedule requirements and flatcar configuration.
modal choice problem for specific trajectories be- Indeed, many more modal choice variables have to
tween beginning and end points, taking into ac- be taken into account.
count specific freight volumes and, possibly, A very good example of this is given in the
specific time constraints. The three papers that will paper of Min (1991). This study focuses on the
be discussed are concerned with the route choice multi-objective nature of the modal choice deci-
for one shipment (container/flatcar/trailer). Any sion. A chance-constrained goal programming
cost reduction for groups of shipments has not (GP) model is constructed that best combines
been taken into account. The cost functions are different modes of transportation and best main-
quite comprehensive. Also time/service constraints tains a continuous flow of products during inter-
have been considered. modal transfer. The GP model is a multiple
objective technique for determining solutions. This
6.2. OR applications for operational problems technique satisfies multiple goals and their associ-
ated risks and uncertainties. The comparison be-
Boardman et al. (1997) built a decision support tween the transportation modes is based on the
system in order to assist the user in selecting the costs, market coverage, average length of haul,
least cost combination of transportation modes equipment capacity, speed, availability, reliability
(truck, rail, air, barge) between a given origin and and damage risk. The most service-cost-effective
a corresponding destination. GIS software assists transportation mode is sought for each segment in
in visualising the region to be analysed. Using a K- the international distribution channel.
shortest path algorithm model, both cost and time
are minimised. As an indicator of cost the average
transportation rates for each transportation mode 7. Conclusions
is used. This is a simplification of reality as there
would normally be a cost difference between long Intermodal transport research is an emerging
haul truck and short haul drayage costs. Also the research field. It is still in a pre-paradigmatic
cost of the inventory is not taken into account. phase, but is evolving and will soon be regarded
Barnhart and Ratliff (1993) pay special atten- as a legitimate branch of scientific research. For
tion to these types of costs. Their models are fo- several reasons modelling intermodal freight trans-
cused on the rail/road combinations compared to port is more complex than modelling uni-modal
uni-modal road transport. For a ÔpiggybackÕ ser- systems. Firstly, it involves at least two modes
vice the minimum cost routing for each shipment which have their own specific characteristics
is sought taking into consideration the total with respect to infrastructure and transport units.
transportation cost (the sum of the drayage costs Secondly, the control of the system has to be
414 C. Macharis, Y.M. Bontekoning / European Journal of Operational Research 153 (2004) 400–416

organised by a set of actors all of whom respon- Acknowledgements


sible for only a part of the whole. Thirdly, com-
plexity of assignment problems is increased due to We thank the anonymous referees for their very
the large variety of load units (type and size), rail useful suggestions.
wagons and trailer chassis.
A review has been executed in order to inves-
tigate how and which operational research tech- Appendix A. Specification of search strategy
niques have been applied to support the specific
decisions that have to be made by the different Sources
decision makers in the intermodal transport sys-
tem. From the review we can conclude the fol-
lowing: Channels Time period
covered
• Operations research has been used for various
Databases
strategic, tactical and operational problems of
–Transport1 1988–2002
the network operator, the terminal operator,
Social Sciences Citation Index2 1988–2002
the drayage operator and the intermodal opera-
Dissertation Abstracts3 1995–2002
tor. Almost all types of intermodal problems
Electronic Transport Journals
are covered, however, the number of studies in
–Journal of Transport Eco- 1997–2002
each category is still very limited.
nomics and Policy4
• A large variety of operations research tech-
–Transport Policy5 1995–2002
niques has been applied.
Library of Delft University of All years
• Due to the type of problems and/or size and
Technology6
complexity of problems, existing OR techniques
Research contacts 1995–2002
have been further developed and new heuristics
Own research 1995–2002
have emerged. However, further development
Cited references in retrieved Not applicable
of OR techniques and heuristics is needed in or-
literature
der to develop applications that provide good 1
solutions to intermodal problems. Transport database. Produced by: Organisa-
• Due to the limited number of studies, compari- tion for Economic Co-operation and Development
son of OR techniques and heuristics are still to (OECD), Transportation Research Board (TRB)
be carried out. These are needed in order to de- and European Conference of Ministers of Trans-
termine which techniques or heuristics best suits port (ECMT) (information provided by Interna-
which type of intermodal problem. tional Union of Railways).
2
Social Sciences Citation Index. Produced by:
• Some intermodal problems have yet to be tack-
Institute for Scientific Information (ISI), Phila-
led by OR techniques. To mention a few: the
delphia. See http://www.isinet.com/isi/.
optimal number of terminals in a network, loca- 3
Dissertation Abstracts. Produced by UMI/
tion decisions for hub-terminals, optimal con- Data Courier. See http://www.umi.com/ab-about.
solidation strategy, allocation of capacity to shtml.
jobs and scheduling of jobs in terminals, deter- 4
Journal of Transport Economics and Policy.
mining truck and chassis fleet size in drayage See http://www.swetsnet.nl/link/access_db?issn ¼
operations. 00225258.
5
Library of Delft University of Technology. See
To conclude, intermodal freight transport pro- http://delfi.library.tudelft.nl:4505/ALEPH/-/start/
vides OR scientists with an interesting and chal- tud01.
6
lenging field of study. There is still much work to Transport Policy. See http://www.sciencedi-
be done and much to discover on the subject. rect.com [publications] [transport policy].
C. Macharis, Y.M. Bontekoning / European Journal of Operational Research 153 (2004) 400–416 415

Search keys double-stack trains. Journal of Transportation Research


Intermodal AND freight Forum 31 (1), 50–62.
Crainic, T.D., Florian, M., Guelat, J., Spiess, H., 1990.
Intermodal service
Strategic planning of freight transportation: STAN, an
Combined transport interactive-graphic system. Transportation Research Re-
Drayage cord 1283, pp. 97–124.
Pre- and end-haulage Dejax, P.J., Crainic, T.G., 1987. Review of empty flows and
Rail–truck fleet management models in freight transportation. Trans-
portation Science 21, 227–247.
Rail–road
Feo, T.A., Gonzalez-Velarde, J.L., 1995. The intermodal trailer
Intermodal terminal assignment problem. Transportation Science 29 (4), 330–341.
Transhipment Ferreira, L., Sigut, J., 1995. Modelling intermodal freight
Hub–terminal terminal operations. Road and Transport Research 4 (4).
Hub and spoke Groothedde, B., Tavasszy, L.A., 1999. Optimalisatie van
terminallocaties in een multimodaal netwerk met simulated
Barge
annealing. In: Proceedings van de Vervoerslogistieke Werk-
dagen 1999, Connekt, Delft, 1999.
References Janic, M., Reggiani, A., Nijkamp, P., 1999. Sustainability of the
European freight transport system: Evaluation of innovative
Arnold, P., Thomas, I., 1999. Localisation des centres de bundling networks. Transportation Planning and Technol-
transbordement dans un systeme multi-reseaux: Essai de ogy 23, 129–156.
formalisation. LÕEspace Geographique 3, 193–204. Jourquin, B., Beuthe, M., Demilie, C.L., 1999. Freight bundling
Assad, A.A., 1980. Transportation Research A 14A, 205–220. network models: Methodology and application. Transpor-
Barnhart, C., Ratliff, H., 1993. Modelling intermodal routing. tation Planning and Technology 23, 157–177.
Journal of Business Logistics 14, 205–223. Justice, E.D., 1996. Optimization of chassis reallocation in
Boardman, B.S., Malstrom, E.M., Butler, D.P., Cole, M.H., doublestack container transportation systems (rail cars,
1997. Computer assisted routing of intermodal shipments. scheduling). Ph.D. thesis, University of Arkansas.
In: Proceedings of 21st International Conference on Com- Kemper, P., Fischer, M., 2000. Modelling and analysis of a
puters and Industry Engineering, vol. 33(1–2), pp. 311–314. freight terminal with stochastic Petri nets. IFAC Control in
Bontekoning, Y.M., Kreutzberger, E. (Eds.), 1999. Concepts of Transportation Systems, Braunschweig, Germany, pp. 267–
New-Generation Terminals and Terminal Nodes. Delft 272.
University Press, Delft. Loureiro, C.F.G., 1994. Modeling investment options for mul-
Bontekoning, Y.M., Kreutzberger, E., 2001. New-Generation timodal transportation networks. Dissertation at the Univer-
Terminals a Performance Evaluation Study. Delft Univer- sity of Tennessee. UMI Dissertation Service, Ann Arbor, MI.
sity Press, Delft. Macharis, C., Verbeke, A., 1999. Een multicriteria-analyse
Bontekoning, Y.M., Macharis, C., Trip, J.J., in press. Is a new methode voor de evaluatie van intermodale terminals.
applied transportation research field emerging?––A review Tijdschrift Vervoerswetenschap 4, 323–352.
of intermodal rail–truck freight transport literature. Trans- Macharis, C., Brans, J.P., Mareschal, B., 1998. The GDSS Prom-
portation Research A. ethee procedure. Journal of Decision Systems 7, 283–307.
Bostel, N., Dejax, P., 1998. Models and algorithms for Meinert, T.S., Youngblood, A.D., Taylor, G.D., Taha, H.A.,
container allocation problems on trains in a rapid tranship- 1998. Simulation of the railway component of intermodal
ment shunting yard. Transportation Science 32 (4), 370– transportation. Report. Arkansas University, Fayetteville, AK.
379. Meyer, P., 1998. Entwicklung eines Simulations programms f€ ur
Brans, J.P., 1982. LÕingenierie de la decision. Elaboration Umschlagterminals des Kombinierten Verkehrs. Disserta-
dÕinstruments dÕaide a la decision. Methode PROMETHEE. tion at University Hannover. Department of Mechanical
In: Nadeau, R., en Landry, M. (Eds.), LÕAide a la Decision: Engineering (Maschinenbau).
Nature, Instruments et Perspectives dÕAvenir. Presses de Min, H., 1991. International intermodal choices via chance-
lÕUniversite Laval, Quebec, Canada, pp. 183–214. constrainted goal programming. Transportation Research A
European Conference of Ministers of Transport, 1993. Termi- 25A (6), 351–362.
nology on combined transport. internetsite http:/www.cor- Morlok, E.K., Sammon, J.P., Spasovic, L.N., Nozick, L.K.,
dis.lu/transport. 1995. Improving productivity in intermodal rail-truck
Chih, C.K., van Dyke, C.D., 1987. The intermodal equipment transportation. In: Harker, P. (Ed.), The Service Produc-
distribution model. Transportation Research Forum XXVII tivity and Quality Challenge, pp. 407–434.
(1), 97–103. Morlok, E.K., Spasovic, L.N., 1994. Redesigning rail-truck
Chih, K.C.K., Bodden, M.P., Hornung, M.A., Kornhauser, intermodal drayage operations for enhanced service and cost
A.L., 1990. Routing and inventory logistics systems (Rails): performance. Journal of Transportation Research Forum 34
A heuristic model for optimally managing intermodal (1), 16–31.
416 C. Macharis, Y.M. Bontekoning / European Journal of Operational Research 153 (2004) 400–416

Muller, G., 1999. Intermodal Freight Transportation. Eno Taylor, G.D., Broadstreet, F., Meinert, T.S., Usher, J.S., 2002.
Transportation Foundation and IANA, Virginia. An analysis of intermodal ramp selection methods. Trans-
Newman, A.M., Yano, C.A., 2000a. Centralized and decen- portation Research Part E 38, 117–134.
tralized train scheduling for intermodal operations. IIE Tsai, J-F., 1994. Models for optimal price and level of service
Transactions 32, 743–754. positioning of intermodal service in competition with truck
Newman, A.M., Yano, C.A., 2000b. Scheduling direct and service. Dissertation at the University of Pennsylvania. UMI
indirect trains and containers in an intermodal setting. Dissertation Service, Ann Arbor, MI.
Transportation Science 34 (3), 256–270. Tsai, J.-F., Morlok, E.K., Smith, T.E., 1994. Optimal pricing
Nozick, L.K., Morlok, E.K., 1997. A model for medium- of rail intermodal frieght: models and tests, Department
termoperations plans in an intermodal rail-truck service. of Systems Engineering, School of Engineering and Ap-
Transportation Research A 31 (2), 91–107. plied Science, University of Pennsylvania, Philadelphia,
Powell, W.B., Carvalho, T.A., 1998. Real-time optimization of PA.
containers and flatcars for intermodal operations. Trans- Van Duin, R. Van Ham, H., 2001. A three-stage modeling
portation Science 32 (2), 110–126. approach for the design and organization of intermo-
Powell, W.B., Jaillet, P., Odoni, A., 1995. Stochastic and dal transportation services. In: Proceedings of the IEEE
dynamic networks and routing. In: Monma, C., Magnanti, International Conference on Systems, Man and Cyber-
T., Ball, M. (Eds.), Handbook in Operations Research and netics, Part 4, October 11–14, 1998, San Diego, CA,
Management Science, Volume on Network Routing. North- USA.
Holland, Amsterdam, pp. 141–295. Voges, J., Kesselmeier H., Beister, J., 1994. Simulation and
Rutten, B.J.C.M., 1995. On medium distance intermodal rail Performance Analysis of Combined Transport Terminals.
transport. Ph.D. thesis, Delft University of Technology, Delft. Intermodal 1994, Conference Paper.
Southworth, F., Peterson, B.E., 2000. Intermodal and interna- Walker, W.T., 1992. Network economics of scale in short haul
tional freight network modeling. Transportation Research C truckload operations. Journal of Transportation Economics
8, 147–166. and Policy XXVI (1), 3–17.
Spasovic, L.N., 1990. Planning intermodal drayage network Wang, X., Regan, A.C., 2002. Local truckload pickup and
operations. Ph.D. Dissertation University of Pennsylvania, delivery with hard time window constraints. Transportation
UMI Dissertation Service, Michigan. Research Part B 36, 97–112.
Spasovic, L.N., Morlok, E.K., 1993. Using Marginal Costs to Yan, S., Bernstein, D., Sheffi, Y., 1995. Intermodal pricing
evaluate drayage rates in rail-truck intermodal service. using network flow techniques. Transportation Research B
Transportation Research 1383, 8–16. 29 (3), 171–180.

You might also like