Springer operational freight carrier planning basic concepts optimization models and advanced memetic algorithms 2005 ISBN3540253181

GOR ■ Publications
Managing Editor
Kolisch, Rainer

Editors
Burkard, Rainer E.
Fleischmann, Bernhard
Inderfurth, Karl
Möhring, Rolf H.
Voss, Stefan


Titles in the Series
H.-O. Günther and P. v. Beek (Eds.)
Advanced Planning and Scheduling
Solutions in Process Industry
VI, 426 pages. 2003. ISBN 3-540-00222-7


Jörn Schönberger

Operational
Freight Carrier Planning
Basic Concepts, Optimization Models
and Advanced Memetic Algorithms

With 43 Figures
and 24 Tables

123

Dr. Jörn Schönberger
University of Bremen
Lehrstuhl für Logistik
Fachbereich 07
Wilhelm-Herbst-Straße 5
28359 Bremen
Germany
E-mail:

Library of Congress Control Number: 2005922933

ISBN 3-540-25318-1 Springer Berlin Heidelberg New York
This work is subject to copyright.All rights are reserved, whether the whole or part of the material
is concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation,
broadcasting, reproduction on microfilm or in any other way, and storage in data banks. Duplication of this publication or parts thereof is permitted only under the provisions of the German
Copyright Law of September 9, 1965, in its current version, and permission for use must always
be obtained from Springer-Verlag.Violations are liable for prosecution under the German Copyright Law.
Springer is a part of Springer Science+Business Media
springeronline.com
© Springer-Verlag Berlin Heidelberg 2005
Printed in Germany
The use of general descriptive names, registered names, trademarks, etc. in this publication does
not imply, even in the absence of a specific statement, that such names are exempt from the
relevant protective laws and regulations and therefore free for general use.
Cover design: Erich Kirchner
Production: Helmut Petri
Printing: Strauss Offsetdruck
SPIN 11407584

Printed on acid-free paper – 42/3153 – 5 4 3 2 1 0


Preface

This book represents the compilation of several research approaches on operational freight carrier planning carried out at the Chair of Logistics, University
of Bremen. It took nearly three years from the first ideas to the final version,
now in your hands. During this time, several persons helped me all the time
to keep on going and to re-start when I got stuck in a dead end or when I
could not see the wood for the trees. I am deeply indebted to them for their
encouragement and comments.
Prof. Dr. Herbert Kopfer, holder of the Chair of Logistics, introduced me
into the field of operational transport planning. He motivated and supervised
me. Furthermore, he supported me constantly and allowed me to be as free
as possible in my research and encouraged me to be as creative as necessary.
In addition, I have to thank Prof. Dr. Hans-Dietrich Haasis, Prof. Dr. Martin
G. Mohrle and Prof. Dr. Thorsten Poddig.
On behalf of all my colleagues, who supported me in numerous ways, I have
to say thank you to Prof. Dr. Dirk C. Mattfeld, Prof. Dr. Christian Bierwirth,
Henner Gratz, Prof. Dr. Elmar Erkens, Nadja Shigo and Katrin Dorow. They
all helped me even with my most obscure and dubious problems.
My family supported me all the time. They always showed me their trust
and encouraged me continuously. Special thanks are dedicated to my parents
Monika and Heinz-Jiirgen.
However, there is somebody who helped and supported me much more
than any other person. It's my beloved wife Ilka. She believes in me more
often than I beheve in myself. But more importantly, she periodically rescues
me from the jungle of science and guides my attention to other wonderful
aspects of life. Thank you very much.

Bremen,
January 2005

Jorn Schonberger


Contents

1

Transport in F'reight Carrier Networks . . . . . . . . . . . . . . . . . . . . . I
1.1 Recent Trends in Freight Transportation . . . . . . . . . . . . . . . . . . . 1
1.2 Carrier Tkansport Networks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.3 Network Design, Configuration and Deployment . . . . . . . . . . . . . 9
1.4 Distribution and Collection Planning . . . . . . . . . . . . . . . . . . . . . . . 11
1.5 Aims of this Book and Used Methods . . . . . . . . . . . . . . . . . . . . . . 13

2

Operational Freight Transport Planning . . . . . . . . . . . . . . . . . . .
2.1 Decision Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.1.1 Request Acceptance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.1.2 Mode Selection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.1.3 Routing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.1.4 Freight Optimization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.2 Hierarchical and Simultaneous Planning . . . . . . . . . . . . . . . . . . . .
2.2.1 Hierarchical Approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.2.2 Simultaneous Routing and Freight Optimization . . . . . . .
2.3 Generic Models for Simultaneous Problems . . . . . . . . . . . . . . . . .

2.3.1 Maximal-Profit Selection . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.3.2 Bottleneck Selection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.3.3 Selection with Compulsory Requests . . . . . . . . . . . . . . . . .
2.3.4 Selection with Postponement . . . . . . . . . . . . . . . . . . . . . . .
2.4 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

15
16
16
17
19
20
22
22
23
24
25
25
26
27
29

3

Pickup and Delivery Selection Problems . . . . . . . . . . . . . . . . . . .
3.1 Problems with Pickup and Delivery Requests . . . . . . . . . . . . . . .
3.1.1 Problems with Depot-Connected Requests . . . . . . . . . . . .
3.1.2 Problems with Direct Delivery Requests . . . . . . . . . . . . . .
3.1.3 Simultaneous Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.2 Pickup and Delivery Paths and Schedules . . . . . . . . . . . . . . . . . . .

3.3 Optimization Problem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.4 Problem Variants . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

31
31
33
33
34
34
36
37


VIII

Contents

3.4.1 The PDSP with LSP Incorporation . . . . . . . . . . . . . . . . . .
3.4.2 The Capacitated PDSP . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.4.3 The PDSP with Compulsory Requests . . . . . . . . . . . . . . .
3.4.4 The PDSP with Postponement . . . . . . . . . . . . . . . . . . . . . .
3.5 Test Case Generation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.5.1 Generation of Pickup and Delivery Requests . . . . . . . . . .
3.5.2 Freight Tariff . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.5.3 Benchmark Suites . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.6 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

38
39
39

40
42
42
45
46
48

4

Memetic Algorithms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
4.1 Algorithmic Solving of Problems with PD-Requests . . . . . . . . . . 49
4.2 Evolutionary Algorithms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
4.3 Genetic Algorithms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
4.3.1 Terminus Technici . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
4.3.2 General Framework . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
4.3.3 Applicability of Genetic Search . . . . . . . . . . . . . . . . . . . . . . 57
4.3.4 Limits of the Genetic Search . . . . . . . . . . . . . . . . . . . . . . . . 58
4.4 Repairing and Improving the Genetic Code . . . . . . . . . . . . . . . . . 60
4.5 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64

5

Memetic Algorithm Vehicle Routing . . . . . . . . . . . . . . . . . . . . . . .
5.1 Genetic Sequencing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.2 Genetic Clustering . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.3 Combined Genetic Sequencing and Clustering . . . . . . . . . . . . . . .
5.4 Advanced MA-Approaches: The State-of-the-Art . . . . . . . . . . . .
5.4.1 Multi-Chromosome Memetic Algorithms . . . . . . . . . . . . .
5.4.2 Co-Evolution with Specialization . . . . . . . . . . . . . . . . . . . .
5.4.3 Co-Evolution of Partial Solutions . . . . . . . . . . . . . . . . . . . .

5.5 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

6

Memetic Search for Optimal PD-Schedules . . . . . . . . . . . . . . . . 77
6.1 Permutation-Controlled Schedule Construction . . . . . . . . . . . . . . 78
6.1.1 Construction of Routes for more than one Vehicle . . . . . 78
6.1.2 Parallel Time-Window-Based Routing . . . . . . . . . . . . . . . . 78
6.1.3 Algorithm Steps . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
6.1.4 Determination of the Request Instantiation Order . . . . . 84
6.2 Representation of a PD-Schedule . . . . . . . . . . . . . . . . . . . . . . . . . . 84
6.3 Configuration of the Memetic Algorithm . . . . . . . . . . . . . . . . . . . 85
6.3.1 Initial Population . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
6.3.2 Recombination . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
6.3.3 Mutation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90
6.3.4 Population Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92
6.4 Computational Experiments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
6.4.1 Parameterization of the MA . . . . . . . . . . . . . . . . . . . . . . . . 94

65
65
68
71
71
72
74
75
76

Contents

IX

6.4.2 Impacts of Spatial Distribution and Time Window
Tightness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
6.4.3 Identification of Profit-Maximum Request Selections . . . 100
6.4.4 Consideration of Capacity Limitations . . . . . . . . . . . . . . . 102
6.4.5 Identification of Deferrable Requests . . . . . . . . . . . . . . . . . 109
6.5 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113
7

Coping with Compulsory Requests . . . . . . . . . . . . . . . . . . . . . . . . 115
7.1 Limits of Fitness Penalization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116
7.1.1 Static Penalties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116
7.1.2 Dynamically Determined Penalties . . . . . . . . . . . . . . . . . . . 118
7.1.3 Adaptive Penalization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119
7.2 A Double-Ranking Approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 120
7.3 Converging-Constraint Approach . . . . . . . . . . . . . . . . . . . . . . . . . . 121
7.3.1 Alternating and Converging Constraints . . . . . . . . . . . . . . 121
7.3.2 ACC-Algorithm Control . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124
7.4 Assessing QC-MA and ACC-MA: Numerical Results . . . . . . . . . 125
7.4.1 Experimental Setup . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125
7.4.2 Numerical Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125
7.4.3 Impacts of Intermediate Cost Reductions: An Example . 130
.
7.5 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .133

8

Request Selection and Collaborative Planning . . . . . . . . . . . . . 135
8.1 The Portfolio Re-composition Problem . . . . . . . . . . . . . . . . . . . . .136
8.1.1 Literature Review . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 136
8.1.2 Formal Problem Statement . . . . . . . . . . . . . . . . . . . . . . . . . 137
8.2 Configuration of the Groupage System . . . . . . . . . . . . . . . . . . . . . 139
8.2.1 Bundle Specification by the Carriers . . . . . . . . . . . . . . . . . 140
8.2.2 Bundle Assignment by the Mediator . . . . . . . . . . . . . . . . . 140
8.3 Computational Experiments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 141
8.3.1 Test Cases . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 142
8.3.2 Collaborative Planning Approach . . . . . . . . . . . . . . . . . . . . 142
8.3.3 Reference Approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143
8.3.4 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143
8.4 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 147

9

Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 149
9.1 Understanding Freight Carrier Decision Problems . . . . . . . . . . . . 149
9.2 Model Building . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 150
9.3 Methodological Enhancements . . . . . . . . . . . . . . . . . . . . . . . . . . . . 151

References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 153
Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 161


Transport in Freight Carrier Networks

The division of labor among the continents, countries or regions over the
world enables the production of goods in the most efficient manner. Goods
are produced at different locations so that the overall costs are minimized. The

manufacture of a certain product often concentrates on few places in a region,
a country, a continent or even in the world. However, the demand for the
products manufactured at certain locations in an economic zone is typically
scattered over the complete zone. In order to satisfy this demand with the
centrally produced goods, extensive transport is needed. Transport describes
the spatial transformation of goods or persons with the goal of balancing
supply and demand. The increase of goods transport is accompanied by a
significant extension of passenger transport. The movement of manpower to
the centralized production facilities becomes necessary and additionally, the
enlarged incomes are used for private travel.
In Sect. 1.1of this chapter, the economic importance of freight transport
is explored. Some current trends, from which the demand for a reinforced
planning arises, are shown by means of the examples European Union (EU)
and United States (US). In Sect. 1.2 the structure of a freight carrier network and the transport processes in such a network are analyzed. Planning
problems regarding the design, configuration and deployment of the transport
system are discussed in Sect. 1.3. The distribution and collection of freight
from providers or suppliers to a consolidation facility, and in the reverse direction, is identified as a very critical phase of the transport and the need
for additional planning support is emphasized in Sect. 1.4. The goals and the
organization of this thesis are given in Sect. 1.5.

1.1 Recent Trends in Freight Transportation
The commonly used indicator for the performance of the goods transport sector is the amount of realized ton-kilometers (tkm) expressing the product of


2

1 Transport in Freight Carrier Networks

the quantities moved and the sum of traveled distances. For passenger transport the number of passenger-kilometers (pkm) gives an adequate measure for
the quantity of passengers moved and the bridged distances.

From 1990 to 2000, the performance of the transport sector has grown
significantly in the US as well as in the EU. For the United States a growth of
more than 20% (goods) and 24% (passengers) is reported (Fenn, 2004). The
European rates show an increase of 29% for goods and 17% more passengers
(Eurostat, 2002).
At the same time, the Gross Domestic Product (GDP) of the US has
expanded by 39% (Fenn, 2004) and the EU-GDP has improved by 21% in
the observed decade(Eurostat, 2002). The absolute values of the performance
indicators have increased from 5.76 billion tkm t o 6.93 billion tkm, from 6.23
billion pkm to 7.72 billion pkm (US), from 2.33 billion tkm to 3.08 billion tkm
and from 4.041 billion pkm to 4.839 billion pkm (EU).
Different studies forecast a further significant increase in required transport (Eurostat, 2002; Arendt and Achermann, 2002; ICF, 2002). An annual
growth of around 3.4% (US) and 3.0% (EU) is expected in the transport
of goods. Relative to 2000, a growth of 40% (US) and 34% (EU) in goods
transport will be achieved by 2010.
This thesis is about problems in planning the transport of goods, often
called freight transport (Crainic and Laporte, 1997). Freight transport is performed by different means of transport: road transport by trucks and vans,
rail transport by trains, waterway transport by inland navigation (barges)
and vessels and pipeline transport of fluid products. The contributions of
each mode (the modal split) have changed during the last decade. In both
economies, the contribution of pipeline transport remains on a approximately
unchanged level. Waterway transport's part has lost significantly in the US as
well as in the EU. Different directions are observed for rail transport: in the
US it has grown but in the EU it has declined.
In both economic zones, the main share of the internal freight transport
is performed on the roads. Trucks and vans make up 32% of the US-domestic
freight transport (Moore, 2002) and 44% of the intra-EU transport of goods
(Eurostat, 2002).
The domain of road transport mostly takes place in the short or medium
size quantity field with shipments of less than 45 tons and distances of typically

not more than 500 km (Eurostat, 2002; Moore, 2002).
Two modes of road transport are performed. In the own account (AC)
mode, the owner of the vehicles and of the moved goods are identical. Typically, such a company requires transport in order to manage the flow of their
goods along their supply chain from the suppliers, through production stages
and warehouses to the customers. The contribution of AC transports to the
performance of the transport sector in the US has decreased slightly from
28% to 24% (Moore, 2004) in the nineties. A significant performance loss for
AC has been observed in the EU-zone. For example, in Germany it has fallen
from 42% down to 30%, and in France the value has dropped from 28% down


1.1 Recent Trends in Freight Transportation

3

to 18% (Oberhausen, 2003). An increasing number of enterprises outsource
their transport departments, basically to reduce their overhead costs. They
hire independent logistics service providers to execute the necessary transport. Such a transport company is called a (freight) carrier: it operates in the
so-called hire or reward (HR) mode. This sector's contribution to the overall
performance varies from 50% (in Portugal) to over 76% (in the US) and nearly
90% (in Spain) with an average value of nearly 70%. In both modes, AC and
HR, short-distance (max. 50km) and medium-distance transport operations
(max. 150km) are the most often demanded services (Oberhausen, 2003) in
the EU. In the United States the average haul distance is 730km, all other
modes operate at longer average haul lengths (Moore, 2002).
The environmental impacts of road transport are significant: 19% (US) and
24% (EU) of the annual carbon dioxide emission is produced by this mode of
transport (Eurostat, 2002; Fenn, 2004). An increase of nearly 16% (US) and
20% (EU) during the nineties of the last century has been observed. Besides
the emissions, several other negative impacts to the environment are observed:

intensified traffic congestion lowers the performance of the road transport,
large surfaces are sealed by roads, parking lots and transshipment areas and
noise emissions lower the quality of residential areas. Regulations and laws
have been announced in order to alleviate the pollution and to maintain several
standards of life quality (European Commission, 2001).
For the EU some additional current developments are observed which are
expected to have significant impacts on road transport and the companies
involved.
The transport enterprises in the current EU member countries expect the
complete integration of the accession countries, e.g. Poland or the Czech Republic. These new competitors can offer lower transport costs since their labor costs are low and the regulations are not as restrictive as the laws in the
current full member countries of the EU. The low cost structure in the preaccession countries will lead to a new wave of competitors intruding onto the
EU transport market. Therefore, the existing transport sector in the current
EU now has to improve the efficiency of their business.
Roads are free of charge in most EU countries. The construction and maintenance of the expressways, highways and other roads is under the responsibility of public authorities. For some special connections (typically including
large bridges or tunnels) the payment of a toll is required. Recently, several
member countries like Germany, Austria and the Netherlands announced their
intention to establish a toll system for using their national roads, and in most
of the pre-accession countries road pricing has been established for years. The
consideration of this additional kind of costs cannot be avoided in the future.
The public authorities are not able anymore to provide the necessary funding
in inter-urban transport infrastructure (roads, bridges, tunnels). Private investors are sought for funding, constructing and maintaining these facilities.
They are allowed to collect a toll from the users.


4

1 Transport in Freight Carrier Networks

Several urban governments (e.g. such as the municipality of the City of
London, UK) have recently either introduced or plan to introduce a toll for

driving on the urban roads. The urban-toll has been introduced mainly for
the reason of preventing traffic congestion caused by private road transport
with the goal of improving the speed of business transport in the urban areas
(Nash and Niskanen, 2003).
As mentioned above, road transport is an important player in the US and
European economy with a significant contribution to the overall performance
of the economic zone. However, it is faced with several challenges in the near
future, which make it unconditionally necessary to strive for an improvement
in the efficiency of the height transport operations on the roads. Carrier companies, which only operate in HR-mode, will be most affected by the challenges
due to the increasing importance of road transport in the United States and
in the EU, and their unconditional need of external customers.

1.2 Carrier Transport Networks
A customer request describes a single transport demand. The location of the
pickup and the location of the delivery are specified as well as the quantity to
be moved. Typically, additional requirements like time limits for the loading
or unloading operation or special handling requirements are stipulated. A
transport department or a carrier company derives internal processes from
the requests in order to satisfy the customer demands using the available
resources in an efficient manner. Several independent requests with coinciding
or adjacent origins and coinciding or adjacent destinations are bundled to
shipments or truckloads of larger quantity. A significant decrease of the relative
price for the movement of one single request is achieved, because overhead
costs are split into small amounts that are assigned to each request. The
realization of maximum economies of scale requires consolidation processes,
which both support as well as depend upon the operation mode of a transport
company.
Firstly, the AC-mode is analyzed. This mode is typically selected for the
distribution of finished goods in industrial productions. From a small number of factories, large quantities of different goods are moved over relatively
long distances to regional distribution centers (DCs) in order to replenish the

DCs with goods. From a DC, the goods are distributed to customers, who
are situated relatively near to the DC. Typically, the distance from a DC to
a customer is less than 100 km, whereas the distance between a production
facility and a DC is often more than several 100 km. The links between the
production facilities, DCs and the customers defines a distribution network.
In a distribution network, the flow of goods is uni-directional from few sources
(the production facilities) to many sinks (the customers). Since the quantities moved for replenishment are large, this network topology supports the
realization of economies of scale. In Fig. 1.1, a distribution network with two


1.2 Carrier Ti-ansport Networks

5

production facilities Fl and Fz is shown. These two factories produce goods
that are used to replenish the three distribution centers DC1, DCz and DC3.
Each customer is assigned to one of the DC from which he is supplied. The
traveled distance and the moved quantities in the distribution are different
from those observed in the replenishment. In distribution tasks, the distances
to travel are significantly reduced and the quantities moved along a link are
significantly smaller.

production
facilities
replenishment

distribution
centers
distribution
customers (

Fig. 1.1. Structure of a distribution network

The strict uni-directional flow of goods from the factories to the DCs or
customers leads to an inefficient use of the deployed vehicles. They drive fully
loaded from the production facilities to the DCs, from a DC to customers or
directly from the factory to a customer (if enough load is available). However,
they have to travel back to the DC or production facilities. If no back-freight
is available, which is carried on the way back to the DC or to the production
facility, half of the traveled distance consists of empty miles.
Often, the transport of finished goods cannot be combined with backfreights flowing in the reverse direction. National laws in several countries
(e.g. Germany) do not allow vehicles operating in the AC-mode to transport goods that are owned by another company. In order to use the existing
network and equipment also for moving goods of third parties, several manufacturing enterprises decided to outsource their distribution systems and let
them operate (at own responsibility) as freight carriers in the HR-mode.
A freight carrier transports goods for different customer companies. Particular batches of different customers are combined in one vehicle if the origins
and the destinations match. A vehicle of a carrier picks up goods of one customer at a certain place and moves it either directly to the final destination
or with intermediate loading, unloading or transshipment stops belonging to
other requests. Afterwards, it continues to a new pickup location situated


6

1 Transport in Freight Carrier Networks

near the destination of the former transport task and loads new goods that
are then moved towards the specified destination.
Typically, a carrier company operates for several customers. The flow of
the goods is not limited to a small number of relations and it is not unidirectional, but rather bi-directional (Fleischmann, 1998). There are many
sources and many sinks of flow spread over the whole operational area. The
quantity of goods associated to a certain pair of origin and destination is small,
and the particular flows are numerous. Long-distance transport demands and

medium-distance demands must be satisfied as well as short-distance bridging. Exclusive origin-destination transport is typically not achievable (Trip
and Bontekoning, 2002). Due to the small quantities of particular requests,
an efficient consolidation strategy is necessary in order to reduce the part
of the overhead costs that have to be assigned to each single request. The
consolidation of small flows from a huge number of locations into large longdistance flows has to be supported as well as the deconsolidation into flows to
the particular customer destinations.
Therefore, the operations area of a carrier is hierarchically organized and
the origin-to-destination transport process is partitioned in sub-processes according to the partition of the area. Few large regions form the operations area,
and each region is divided into several small zones. A hierarchically organized
network of transshipment facilities and connections between these facilities is
maintained. The network structure permits a successive aggregation of flows
from customer origins into high quantity flows and a subsequent resolution
from bundled flows into deliveries to the several destinations. Whenever a single transport demand requires the crossing of an organizational border, the
goods of a request are combined with goods of other requests or extracted
from a large volume flow.
Transport between different regions takes place only between hubs (H).
A hub is a transshipment facility where all inbound flows from other regions
are received and where all outbound flows are released. In each region, the
hub receives the goods from different transshipment points (TP) situated in
the zones and forwards incoming goods into the right destination zones. In
each zone, the goods are distributed from the TP to the customer locations by
vehicles within several tours. The same vehicles are used to collect goods from
customer locations. These goods are delivered to the destinations in the same
tour if the location is situated in the same zone. Otherwise, they are brought
to the TP, where they are merged with other collected goods and forwarded
to the regional hub. A typical layout of such a hierarchically structured freight
transport network is given in Fig. 1.2. The overall operations area is partitioned into four regions. The thin continuous lines mark the borders. In each
region one hub (HI,Hz, H3 and H4) is available. All extra-regional flows of
goods out of a region are realized through this transshipment facility as well
as all inbound flows. Each region is separated into several zones. Their borders

are given by the thin dotted lines. In each zone one T P is maintained where
the goods flowing out of the zone are bundled and forwarded and where the


1.2 Carrier Transport Networks

s

Hi= hub in region i

Ti= transshipment point

Fig. 1.2. Hierarchical network structure of a carrier network

flow of goods destined for this particular zone is received. The distance from
the customer location to a T P in a zone is typically less than the distance
between the T P and the corresponding regional hub. However, the distance
between the hubs often causes the main part of the distance necessary for
moving a packet from its origin to its destination.
The hierarchically organization is typical for a freight transport network
operated by a carrier. It permits the economically reasonable service of geographically scattered locations with averagely low flow between particular
origins and destinations (Fleischmann, 1998). However, the described original
structure is often modified adapted to meet the special requirements (Wlcek,
1998). A hub serves as the T P for a zone or a complete region, if the quantity of the flow of goods does not require the strong tree structure in a region.
Direct origin-to-destination shipments among different regions or zones are offered in the event that the quantity of goods to be moved and the associated
revenues are sufficiently large.
In the following, the transport process for carrying the less-than-truckload
(LTL) packet p, of the request r is analyzed in detail. Figure 1.3 shows the
five phases of the process from an origin C3 to the destination D.
Initially, the packet p, is collected. This phase is called collection. Typically, a vehicle of small or medium capacity is deployed for the fulfilment of

1 Transport in Freight Carrier Networks

this task. This vehicle collects packets from several requests with origins in
one selected zone and carries them to the transshipment point (TP) of the
zone. A T P is a special facility, in which the packets of a zone, collected by
several vehicles, are consolidated into larger quantities (truckloads). A truckload consists of all quantities belonging to requests originating out of a zone
with destinations in zones embedded in different regions. Requests in which
both the origin and the destination are included in the same zone are served
without involving a TP.
In the forward feeding phase, the truckload is carried from the T P of
the origin zone to the regional hub. All truckloads of a region arrive synchronized at this large and high-performance transshipment facility (Fleischmann,
1998). In contrast to a TP, incoming and outgoing goods are merged while
passing a hub. Bundled truckloads from different TPs are resolved before the
packets are re-consolidated into shipments so that all packets in a shipment
have to be carried to customer locations in the same destination region.

hub

\

I
line haul

I
origin O

TP

hub

hub

hub

TP destination 'l/

Fig. 1.3. Process-chain of an origin-destination carrier transport

The complete shipment containing p, is now carried to the hub of the destination region, which includes the final customer location. If the hubs are fully
connected, then no intermediate stop at any other hub is necessary because
there is a direct connection between each two hubs. Otherwise the shipment
is moved to one or several intermediate hubs, reconsolidated if necessary, and
then finally transferred to the hub of the destination region. In this line haul
phase (Daganzo, 1999), large distances are traveled. The means of transport is
often different to those in the previous phases. Since the flow of goods is continuous and of balanced quantity, the inter-hub connections are often served
in a regular way following a fixed schedule (Crainic, 2000). For this reason,
it is necessary that the feeder transport schedule be synchronized with the
departures from the hub. All feeder truckloads should arrive in time so that
they can be considered for the inter-hub transport departures. A synchronized
arrival enables the most effective re-consolidation of the incoming truckloads
from other hubs and from the TPs.


1.3 Network Design, Configuration and Deployment

9

At the destination hub, the shipment containing p, is resolved and merged

with other incoming shipments into truckloads, so that each truckload comprises packets for different customers who are situated in the same distribution
zone of the considered region. Each truckload is transported to the TP of the
corresponding zone. This phase is denoted as backward feeding.
At the T P of the zone, which contains the destination of r, the truckload is
broken into the packets and the packets are distributed to the customers that
are situated scattered over the destination zone. Delivering p, to the customer
specified delivery location completes the request r .
The first two phases of the transport process are subsumed under the name
pickup and the last two phases are referred to as delivery. The pickup in an
origin region is typically combined with the delivery of back-freight destined
for this region. Thereby, the flows of goods in both directions are combined
in an effective and efficient way.
The correspondence of the five-phase transport process and the hierarchically organized network is shown by means of a transport of a packet from a
customer situated at the small black point in the north-western zone to the
location marked by the small black point in the south-eastern zone in Fig. 1.2.
A vehicle following the route that visits all customer locations in the origin
zone, shaded in grey, picks up the considered packet. At TI, the TP of the
origin zone, the packet is consolidated with all other packets originating from
this zone into a truckload and it is fed to the regional hub H I . With an intermediate stop a t H3,the packet is carried to H4, the hub in the destination
region. There, it is transshipped and fed to T4, the T P in the destination
zone. F'rom T4 it is delivered on the route visiting all grey shaded locations,
including the particular delivery site.

1.3 Network Design, Configuration and Deployment
The construction, the management and the usage of an effective and efficient
freight carrier transport network require the solution of numerous often interdependent decision problems.
Logistics System Design. The design of a freight transportation network affects several problems related to the location and the layout of the network components such as the TPs, hubs or traffic routes. Three main classes
of design problems are distinguished (Crainic and Laporte, 1997):
0

0

How many hubs and TPs are needed? Where should they be installed?
How large should they be? (location and layout)
How should the hubs and TPs be linked? Which means of transport should
be used for the connections? How should the flow of goods be distributed
over the connections between the hubs? (network design)
How can the network be protected against disadvantageous external influences and evolutions arising from infrastructure modifications, the evolu-


10

1 Transport in Freight Carrier Networks

tion of demand and from new governmental or industrial policies? (regional
multi modal planning)
Network design problems are strategic. The necessary funding and the necessary time for the construction or modification of a n existing infrastructure
do not allow short or medium-time changes. Design problems are solved using
estimated data expressing the expected flows of goods. The decision for a certain network architecture is based upon the costs for installing, maintaining
and using the facilities and the traffic links between them.
Logistics System Configuration comprises three main mid-term planning categories (Crainic, 2000): service selection, traffic distribution and terminal policies. A service describes a repeated transport operation connecting
hubs or hubs and TPs. In a service selection problem, the offered services in a
network are determined. Typically, the repetition of a service follows a regular
schedule. The departure and the arrival times a t the first, the last and the
intermediate stops are defined and announced. The necessary work power and
transport capacity to offer the intended services is procured.
Several services are compiled into closed routes (itineraries). For each
itinerary, a vehicle is allocated and the corresponding necessary terminal operations are fixed. Services are determined only for connections of hubs with
T P s or other hubs. The derived schedule is valid for up to several months in
the future, but exact long-term planning data are not available. The quantities of goods have to be estimated, e.g. based on observations from the past.

However, reliable estimates need reliable input data, which is typically available only for the feeder or line haul connections. The consolidation of packets
from the customers ensures a predictable and balanced flow of goods between
the hubs.
A terminal policy describes the offered activities a t a given hub or TP. The
type of performed consolidation tasks a t a certain hub or T P is specified and
defines the available throughput that can be handled. Additional resources
have to be maintained in order to compensate for peaks in the demanded
services.
Logistics System Configuration aims at establishing services that allow efficient operations t o answer customer demands and t o ensure the profitability
of the operations (Crainic, 2003). Efficiency is typically measured in terms of
costs for fulfilling the customer demands a t a predicted quality that allows the
customers t o maintain complex and reliable production systems (Rodrigue,
1999).
Logistics System Deployment comprises short-term planning problems
in a freight carrier network. In contrast to the design and configuration of
the network, deployment decisions are mainly based on known problem data.
These data are derived from the known or declared flows of goods extracted
from the customer demands. The goal is to allocate labor and capacities in
order to support the efficient fulfilment of known customer demands with
respect to the policies and services determined in the configuration step. These


1.4 Distribution and Collection Planning

11

planning problems are solved following the rolling horizon planning paradigm
in order to handle the continuously updated information about additional,
cancelled or modified customer requests. The necessary operations for the next
period are definitively determined together with a tentative determination of

the operations planned for the subsequent planning periods. The following
short-term planning problems occur (Crainic and Laporte, 1997):

0

0

Assignment of crews, reserve crews or maintenance teams t o vehicles or
transshipment facilities in order t o support the planned operations (crew
scheduling)
Preparation of the operations for the next planning period (empty balancing)
Scheduling of the services for the pickup and the delivery phases (vehicle
routing and scheduling)

Logistics System Deployment mainly impacts the short-term planning of
pickup and delivery operations. The operations during the long haul phase are
determined by the valid regular schedules. For two reasons the determination
of a long-term schedule for the pickup and the delivery operations is not
achievable:

1. The locations that have to be visited are typically not known in advance.
2. There is no balanced flow of goods that permits the prediction of necessary
services and/or necessary capacities in a zone.
The costs for the operations in the first and in the last phase of the carrier
transport process are, expressed in terms of money units per tkm, the most
expensive part in the complete transport from the pickup location to the
final delivery location. Herry (2001) states that the costs per tkm in shortdistance transport are a t least three times larger than the costs per tkm in
the middle or long distance case. The reason for this extreme increase of the
costs can be seen in the relatively small quantities that are delivered or picked
up a t a customer site stop, and in the lack of consolidation options due t o

the scattered locations of customer sites that requires a visit. Furthermore,
the unconditional need for the consideration of tight time windows for the
pickup and the delivery visits confines the realization of economies of scale that
are otherwise achieved by the bundling of several requests (Punakivi et al.,
2001). The consideration of time windows is necessary in order t o synchronize
the transport processes performed by the carrier company with the internal
processes of the customers.

1.4 Distribution and Collection Planning
Each customer request is split into five internal requests according t o the
subprocesses described in Sect. 1.2. A collection task is necessary t o carry the


12

1 Transport in Freight Carrier Networks

demanded quantity from the customer specified origin t o the next TP. Within
the forward feeding task, this quantity has to be carried t o the regional hub,
typically together with quantities from other customer requests. The line haul
tasks expresses the necessity to transport the quantity to the destination hub,
the backward feeding task requires the movement of the quantity t o the TP
in the destination region and the distribution task describes the final carriage
to the destination. In the remainder of this thesis, the term request is used as
a synonym for task or internal request.
Each line haul task is assigned t o one of the regular services, which ensures
the execution of the task. The remaining tasks cannot be assigned t o such a
regular service, because there are no regular services for tasks within one
region. It is necessary t o allocate resources for the remaining four tasks in
each planning period. To fulfil the required tasks associated with different

requests, company-owned vehicles can be used or, for the payment of a fee,
other carriers can be instructed to complete selected tasks.
Pickup and delivery planning problems comprise the allocation of the resources for fulfilling the tasks within a region for a given planning period.
A solution of a pickup and delivery planning problem is the transportation
plan (Crainic and Laporte, 1997) and the necessary costs are called fuZjilment
costs. The transportation plan is determined only for the next planning period,
which comprises just a day. or often even only several hours. For subsequent
periods, the transportation plan is renewed considering the recently released
information of additional, cancelled or modified requests and the currently
available transport resources.
Altogether, the following issues characterize pickup and delivery planning
problems.
The request portfolio cannot be modified. In some cases a postponement
of some of these requests is allowed.
The consideration of a large number of low-quantity packets instead of
bundled shipments requires a large number of stops a t customer sites.
The composition of several requests into routes is necessary in order t o
achieve profitability (Trip and Bontekoning, 2002).
Compared t o the line haul tasks, the costs for fulfilling a (regional) pickup
task or delivery task are tripled (Herry, 2001).
No regular services are available. For each planning period, a new transport
plan has t o be determined (logistics system deployment problems).
The flow in both directions t o and from the TP/hub is typically not synchronized. A time gap between freight and back freight has t o be managed.
Several time constraints have to be taken into account: earliest departure
times from the TP/hub (availability), latest arrival times a t the TP/hub
and customer site time windows.
The routes are determined following a rolling horizon planning, the period
length depends upon the frequency of incoming feeder or long haul services

1.5 Aims of this Book and Used Methods

0

0

13

a t the TP/hub and upon the portfolio of released, but so far unconsidered,
customer requests.
The demanded transport capacity is not balanced: it varies from planning
period to planning period.
Other carrier companies are allowed to be ordered to fulfil tasks (subcontractor incorporating, externalization).

Pickup and delivery planning aims a t finding a reasonable trade-off between necessary costs (resulting in reasonable offered prices) and service quality. Mathematical optimization models are proposed in which the costs are
expressed in a cost function that is minimized. In order to ensure customer
satisfaction, several constraints have to be considered, especially time windows
that restrict the delivery or collection time.
Existing models for pickup and delivery problems typically refrain from
involving external carriers. However, their consideration can be profitable if
the charges are below the costs for using an own vehicle. Additionally, the
occasional involvement of external carriers for a fixed charge allows a quick
and temporary capacity expansion when the available resources do not suffice
to serve all tasks.

1.5 Aims of this Book and Used Methods
Systems for freight transport require an exact and efficient coordination of
the different processes in order to offer a reliable service to customers and to
keep the necessary costs within an acceptable budget.
The setup of regular services in a particular zone or region is generally

not targeted because the local flow of goods is unpredictable. Neither the
demanded quantities nor the locations to be visited can be estimated in a
sufficient manner. Since the costs for each performed tkm are significantly
enlarged compared to the long-haul operations the determination of efficient
long valid schedules is hardly possible. The determination of the necessary
tasks to fulfill the demand for transport in a particular region is furthermore
compromised by a significant reduction of available and maintained transport
resources due to the current poor market conditions and due to the intrusion
of low-cost carriers from the EU candidate countries into the market.
The involvement of carrier companies to fulfill requests for a fixed charge
becomes more and more attractive for freight carrier companies who offer and
manage wide-area transport networks. Instead of maintaining own equipment
in all regions (with often severe costs for employment, depreciation, maintenance and insurance), subcontractors are involved for a previously known
charge whenever it is possible (due to lower costs) or necessary (due to a
short-term capacity bottleneck).
The planning of the regional collection and distribution traffic of a freight
carrier company is a very sophisticated challenge, since it is impossible to


14

1 Transport in Freight Carrier Networks

establish regular services between customer locations and transshipment facilities. Planning support for the incorporation of external carriers has received
only minor attention so far, although it is extensively required in practice. Adequate planning models are rarely proposed but they are becoming more and
more necessary t o support freight carrier planning. The intention of this thesis
is to contribute new ideas to close this gap between real world requirements
and available planning support.
The first goal of this thesis is to derive general modeling approaches for
incorporating external logistics service providers in the fulfillment planning of

a freight carrier network. Chapter 2 provides a n introduction t o operational
carrier planning problems. The corresponding scientific literature is surveyed
and four main modeling approaches with special consideration of the external
request fulfillment are derived.
As the second goal, the extension of existing pure routing models by the
incorporation of a carrier incorporation feature is performed. One of the most
general routing problems which represents a variety of different planning situations, the pickup and delivery problem with time windows, is generalized.
Therefore, the possibility of the usage of a n external carrier for unprofitable
requests is added. Four different problem variants representing several special
planning environments are set up. For each single variant, adequate test problems are generated. Chapter 3 comprises all these investigations of so-called
pickup and delivery selection problems. A pickup and delivery selection problem can be modeled as a multi-constraint mathematical optimization problem.
The solving of instances of such a problem requires massive computational effort even for small number of incorporated requests and vehicles. Optimality
guaranteeing algorithms are not available.
The third goal of this thesis is the derivation of adequate algorithms t o
solve instances of pickup and delivery selection problems. The configuration
of Memetic Algorithms is proposed. Memetic Algorithms combine the proven
exploration capability of Genetic Algorithms for restricted combinatorial optimization problems with the exploitation capabilities of problem-specific algorithms. General concepts of the memetic search are presented in Chap. 4.
This class of algorithms has been successively applied to routing problems for
vehicles. The general configurations of a memetic search algorithm for routing
problems are surveyed in Chap. 5 . The configuration of a Memetic Algorithm
for solving pickup and delivery selection problem instances is presented in
the Chapters 6 and 7. Its applicability is assessed within extensive computational experiments. It is shown, how it might possibly be incorporated into
and exploited by a carrier service. The presented framework can be used for
problems based on any of the four carrier incorporation approaches.
The applicability of the derived models and algorithms is investigated for
a cooperative planning scenario described in Chap. 8.
This thesis concludes with a summary of the core findings of the investigations and formulates topics for future research in the field of planning models
for freight transportation with the possibility of incorporating a paid carrier.

Operational Freight Transport Planning

This chapter is about operational (short-time) planning problems that must be
resolved in order to determine the transportation plan of a carrier company for
the next planning period. The focus of the investigations is on the coordination
of the operations to realize the outer legs of the freight carrier transport
process described in the first chapter. The aim is to determine the necessary
operations to fulfill the demand for transport between customer locations and
the next transshipment facility in a particular region.
A carrier has to solve different decision problems in order to setup the
transportation plan for a certain planning period. The first decision concerns
the question as to whether the responsibility for a request fulfilment is taken
on. Each accepted customer request is resolved into several internal tasks
(internal requests) distinguishing between pickup, line-haul and delivery requests. Line-haul tasks are assigned to the designated regular and scheduled
services.
The second decision problem comprises the selection of the fulfilment mode
of the pickup and the delivery requests. Mode selection means to decide
whether a request is fulfilled with own vehicles (carrier controlled vehicles)
or whether it is externalized and fulfilled by another carrier for a charge (external order processing).
To utilize the own fleet in a most efficient manner it is necessary to solve
a routing problem in which the requests are assigned to different vehicles and
execution orders are setup (third decision problem). In order to reduce the
overall sum of charges to be paid to other carriers, a freight charge minimization problem has to be solved (fourth decision problem). These decision
problems are analyzed in Sect. 2.1.
The four decision problems are interdependent. In a hierarchical solving
approach, the problems are solved successively: Initially, the mode of execution is selected for each accepted request and afterwards the resulting routing
and the resulting freight minimization problem are solved. In a simultaneous
approach, the decisions upon the selected mode, the routing and the freight
charge minimization are derived simultaneously within one closed optimiza-

16

2 Operational Freight Transport Planning

tion problem. Hierarchical and simultaneous approaches are described in Sect.
2.2.
The simultaneous approach has received only minor attention so far although it is most promising and required in many real-world applications.
Adequate decision models or problem representations are required. In Sect.
2.3, four generic frameworks for combined mode selection, routing and freight
charge minimization problems are introduced. Each one represents a simultaneous problem for a special environment.

2.1 Decision Problems
The determination of the transportation plan for one period requires the solving of different sophisticated decision problems. First of all, the carrier has
to decide whether a certain request r is accepted for completion, then the
mode of completion (with own equipment or by another carrier who receives
a charge) is selected. Routes have to be established for the own vehicles and
the sum of charges to be paid for all externalized requests has to be minimized.
In both cases, the consolidation of several requests helps to reduce the costs
for each moved capacity unit so that the solving of these problems requires
the simultaneous treatment of more than one request.

2.1.1 Request Acceptance
If a carrier accepts a customer request then it takes over the responsibility
for the reliable completion of the customer specified transport demand. The
carrier company receives a certain revenue for each satisfactorily completed
request. If a request is not completed to the customer's satisfaction, the carrier
is sanctioned. The agreed revenues and the agreed penalties are typically
codified in a contract between the carrier and the customer.
Solving the request acceptance problem for a request r means to decide

whether r can be fulfilled in a profitable manner or not. Therefore, the request
acceptance problem is a binary decision problem.
A carrier company is faced with different request acceptance problems.
Tactical request acceptance problems require a general decision about
the future acceptance of different request types. Such a type comprises typically all the requests of a certain customer. This acceptance problem is a
management problem, because the general acceptance of all requests of a customer requires medium- or long-term investments for additional transport
or transshipment resources. The general acceptance is fixed in medium- or
long-term contracts. In such a contract, the expected quality of the request
fulfilment is codified. Furthermore, the revenues achieved for the request fulfilment are determined. The general acceptance is recommended if, and only
if, the agreed revenues cover the sum of necessary investment and operation
costs. A reasonable contribution to the profit has to be achieved.