Computational Aspects and Applications in Large Scale Networks NET 2017 Nizhny Novgorod Russia June 2017 Valery A. Kalyagin

Computational Aspects and
Applications in Large Scale Networks

NET 2017 Nizhny Novgorod Russia
June 2017 Valery A. Kalyagin
Visit to download the full and correct content document:
https://textbookfull.com/product/computational-aspects-and-applications-in-large-scal
e-networks-net-2017-nizhny-novgorod-russia-june-2017-valery-a-kalyagin/
More products digital (pdf, epub, mobi) instant
download maybe you interests ...
Large Scale Scientific Computing 11th International

Conference LSSC 2017 Sozopol Bulgaria June 5 9 2017
Revised Selected Papers 1st Edition Ivan Lirkov
https://textbookfull.com/product/large-scale-scientific-
computing-11th-international-conference-lssc-2017-sozopol-
bulgaria-june-5-9-2017-revised-selected-papers-1st-edition-ivan-
lirkov/
Advances in Computational Intelligence 14th

International Work Conference on Artificial Neural
Networks IWANN 2017 Cadiz Spain June 14 16 2017
Proceedings Part I 1st Edition Ignacio Rojas
https://textbookfull.com/product/advances-in-computational-
intelligence-14th-international-work-conference-on-artificial-
neural-networks-iwann-2017-cadiz-spain-
june-14-16-2017-proceedings-part-i-1st-edition-ignacio-rojas/
Advances in Neural Networks ISNN 2017 14th

International Symposium ISNN 2017 Sapporo Hakodate and
Muroran Hokkaido Japan June 21 26 2017 Proceedings Part
I 1st Edition Fengyu Cong
https://textbookfull.com/product/advances-in-neural-networks-
isnn-2017-14th-international-symposium-isnn-2017-sapporo-
hakodate-and-muroran-hokkaido-japan-june-21-26-2017-proceedings-
part-i-1st-edition-fengyu-cong/
Advances in Neural Networks ISNN 2017 14th

International Symposium ISNN 2017 Sapporo Hakodate and
Muroran Hokkaido Japan June 21 26 2017 Proceedings Part
II 1st Edition Fengyu Cong
https://textbookfull.com/product/advances-in-neural-networks-
isnn-2017-14th-international-symposium-isnn-2017-sapporo-
hakodate-and-muroran-hokkaido-japan-june-21-26-2017-proceedings-
Analytical and Computational Methods in Probability
Theory First International Conference ACMPT 2017 Moscow
Russia October 23 27 2017 Proceedings 1st Edition
Vladimir V. Rykov
https://textbookfull.com/product/analytical-and-computational-
methods-in-probability-theory-first-international-conference-
acmpt-2017-moscow-russia-october-23-27-2017-proceedings-1st-
edition-vladimir-v-rykov/
Algorithms for Computational Biology 4th International

Conference AlCoB 2017 Aveiro Portugal June 5 6 2017
Proceedings 1st Edition Daniel Figueiredo
https://textbookfull.com/product/algorithms-for-computational-
biology-4th-international-conference-alcob-2017-aveiro-portugal-
june-5-6-2017-proceedings-1st-edition-daniel-figueiredo/
Information Science and Applications 2017: ICISA 2017

1st Edition Kuinam Kim
https://textbookfull.com/product/information-science-and-
applications-2017-icisa-2017-1st-edition-kuinam-kim/
Estimating Functional Connectivity and Topology in

Large-Scale Neuronal Assemblies: Statistical and
Computational Methods Vito Paolo Pastore
https://textbookfull.com/product/estimating-functional-
connectivity-and-topology-in-large-scale-neuronal-assemblies-
statistical-and-computational-methods-vito-paolo-pastore/
Analysis of Images Social Networks and Texts 6th

International Conference AIST 2017 Moscow Russia July
27 29 2017 Revised Selected Papers 1st Edition Wil M.P.
Van Der Aalst
https://textbookfull.com/product/analysis-of-images-social-
networks-and-texts-6th-international-conference-aist-2017-moscow-
russia-july-27-29-2017-revised-selected-papers-1st-edition-wil-m-
Springer Proceedings in Mathematics & Statistics
Valery A. Kalyagin · Panos M. Pardalos

Oleg Prokopyev · Irina Utkina Editors
Computational
Aspects and
Applications
in Large-Scale
Networks
NET 2017, Nizhny Novgorod, Russia,
June 2017
Volume 247
This book series features volumes composed of selected contributions from
workshops and conferences in all areas of current research in mathematics and
statistics, including operation research and optimization. In addition to an overall
evaluation of the interest, scientific quality, and timeliness of each proposal at the
hands of the publisher, individual contributions are all refereed to the high quality
standards of leading journals in the field. Thus, this series provides the research
community with well-edited, authoritative reports on developments in the most
exciting areas of mathematical and statistical research today.
More information about this series at http://www.springer.com/series/10533

Valery A. Kalyagin Panos M. Pardalos
•
Oleg Prokopyev Irina Utkina

•
Editors
Computational Aspects
and Applications
in Large-Scale Networks
NET 2017, Nizhny Novgorod, Russia,
June 2017
123
Editors
Valery A. Kalyagin Oleg Prokopyev
Higher School of Economics Department of Industrial Engineering
National Research University University of Pittsburgh
Nizhny Novgorod, Russia Pittsburgh, PA, USA
Panos M. Pardalos Irina Utkina

Department of Industrial Higher School of Economics
and Systems Engineering National Research University
University of Florida Nizhny Novgorod, Russia
Gainesville, FL, USA
ISSN 2194-1009 ISSN 2194-1017 (electronic)

ISBN 978-3-319-96246-7 ISBN 978-3-319-96247-4 (eBook)
https://doi.org/10.1007/978-3-319-96247-4
Library of Congress Control Number: 2018948592
Mathematics Subject Classification (2010): 05C82, 90B10, 90B15
© Springer International Publishing AG, part of Springer Nature 2018

This work is subject to copyright. All rights are reserved by the Publisher, whether the whole or part
of the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations,
recitation, broadcasting, reproduction on microfilms or in any other physical way, and transmission
or information storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar
methodology now known or hereafter developed.
The use of general descriptive names, registered names, trademarks, service marks, 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.
The publisher, the authors and the editors are safe to assume that the advice and information in this
book are believed to be true and accurate at the date of publication. Neither the publisher nor the
authors or the editors give a warranty, express or implied, with respect to the material contained herein or
for any errors or omissions that may have been made. The publisher remains neutral with regard to
jurisdictional claims in published maps and institutional affiliations.
This Springer imprint is published by the registered company Springer Nature Switzerland AG
The registered company address is: Gewerbestrasse 11, 6330 Cham, Switzerland
Preface
This volume is based on the papers presented at the 7th International Conference on
Network Analysis held in Nizhny Novgorod, Russia, June 22–24, 2017. The main
focus of the conference and this volume is centered on the development of new
computationally efficient algorithms as well as underlying analysis and optimization
in large-scale networks. Various applications in the context of social networks,
power transmission grids, stock market networks, and human brain networks are
also considered. The previous books based on the papers presented at the 1st–6th
Conferences International Conference on Network Analysis can be found in [1–6].
The current volume consists of three major parts, namely, network computational
algorithms, network models, and network applications.
The first part of the book is focused on algorithmic methods in large-scale
networks. In the Chapter “Tabu Search for Fleet Size and Mix Vehicle Routing
Problem with Hard and Soft Time Windows”, the authors present an efficient
algorithm for solving a computationally hard problem in a real-life network setting.
In the Chapter “FPT Algorithms for the Shortest Lattice Vector and Integer Linear
Programming Problems”, fixed-parameter tractable (FTP) algorithms are discussed.
The main parameter is the maximal absolute value of rank minors of matrices in the
problem formulation. The Chapter “The Video-Based Age and Gender Recognition
with Convolution Neural Networks” reviews the problem of age and gender
recognition methods for video data using modern deep convolutional neural net-
works. A new video-based recognition system is implemented with several
aggregation methods that improve the method identification accuracy. The Chapter
“On Forbidden Induced Subgraphs for the Class of Triangle-König Graphs” con-
siders a hereditary class of graphs with the following property: the maximum
cardinality of a triangle packing for each graph of the class is equal to the minimum
cardinality of a triangle vertex cover. Then, some minimal forbidden induced
subgraphs for this class are described. The Chapter “The Global Search Theory
Approach to the Bilevel Pricing Problem in Telecommunication Networks” deals
with the hierarchical problem of optimal pricing in telecommunication networks.
v
vi Preface
New methods of local and global search for finding optimistic solution are devel-
oped. In the Chapter “Graph Dichotomy Algorithm and Its Applications to Analysis
of Stocks Market”, a new approach to graph complexity measure is presented. This
new measure is used to study short-term predictions of crises at the stock market. In
the Chapter “Cluster Analysis of Facial Video Data in Video Surveillance Systems
Using Deep Learning”, a new approach is proposed for structuring information in
video surveillance systems. The first stage of the new approach consists of grouping
the videos, which contain identical faces. Based on this idea, a new efficient and fast
algorithm of video surveillance is developed. The Chapter “Using Modular
Decomposition Technique to Solve the Maximum Clique Problem” applies the
modular decomposition technique to solve the weighted maximum clique problem.
Developed technique is compared against the state-of-the-art algorithms in the area.
The second part of the book is focused on network models. In the Chapter
“Robust Statistical Procedures for Testing Dynamics in Market Network”, a sign
similarity market network is considered. The problem of testing dynamics in market
network is formulated as the problem of homogeneity hypotheses testing. Multiple
testing techniques to solve this problem are developed and applied for different stock
markets. The Chapter “Application of Market Models to Network Equilibrium
Problems” describes an extension of the network flow equilibrium problem with
elastic demands and develops a new equilibrium type model for resource allocation
problems in wireless communication networks. In the Chapter “Selective
Bi-coordinate Variations for Network Equilibrium Problems with Mixed Demand”,
a modification of the method of bi-coordinate variations for network equilibrium
problems with mixed demand is proposed. Some numerical results that confirm
efficiency of the method are presented. The Chapter “Developing a Model of
Topological Structure Formation for Power Transmission Grids Based on the
Analysis of the UNEG” studies the nodes degrees distribution in the United National
Electricity Grid (UNEG) of Russia. This study allows to develop a new Random
Growth Model (RGM) to simulate UNEG, and to identify the key principles for the
network formation. In the Chapter “Methods of Criteria Importance Theory and
Their Software Implementation”, a general approach for solution of the multicriteria
choice problem is developed on the basis of the criteria importance theory. These
methods are implemented in the computer system DASS. The Chapter “A Model of
Optimal Network Structure for Decentralized Nearest Neighbor Search” discusses
the problem of the best structure of network for the decentralized nearest neighbor
search algorithm. Mathematical programming model for the problem is proposed.
Optimal structures of small size networks are computed. A generalization for
large-scale networks is also discussed. In the Chapter “Computational Study of
Activation Dynamics on Networks of Arbitrary Structure”, new results on modeling
dynamical properties of collective systems are presented. A general technique for
reducing the problems to SAT problem is developed. This approach is then applied
to computer security networks. The Chapter “Rejection Graph for Multiple Testing
of Elliptical Model for Market Network” uses the symmetry condition of tail
Preface vii
distributions to test elliptical hypothesis for stock return distributions. Multiple

testing procedures are developed to solve this problem. These procedures are then
applied to the stock markets of different countries. Specific structure of the rejection
graph is observed and discussed. In the Chapter “Mapping Paradigms of Social
Sciences: Application of Network Analysis”, the network approach is applied to
studying the relationships between various elements that constitute any particular
research in social sciences. A set of relations between various elements from text-
books on methodology of social and political sciences is extracted and analyzed.
The third part of the book deals with network applications. The Chapter “Using
Geometry of the Set of Symmetric Positive Semidefinite Matrices to Classify
Structural Brain Networks” presents a method of Symmetric Positive Semidefinite
(SPSD) matrices classification and its application to analysis of structural brain
networks (connectomes). Existing Symmetric Positive Definite (SPD) matrix-based
algorithms are generalized to SPSD case. The performance of the proposed pipeline
is demonstrated on structural brain networks reconstructed from the Alzheimer‘s
Disease Neuroimaging Initiative (ADNI) data. In the Chapter “Comparison of
Statistical Procedures for Gaussian Graphical Model Selection”, the uncertainty of
statistical procedures for Gaussian graphical model selections is studied. Different
statistical procedures are compared using different uncertainty measures, such as
Type I and Type II errors, ROC and AUC. The Chapter “Sentiment Analysis Using
Deep Learning” analyzes advantages of the deep learning methods over other
baseline machine learning methods using sentiment analysis task in Twitter. In the
Chapter “Invariance Properties of Statistical Procedures for Network Structures
Identification”, optimality of some multiple decision procedures is discussed in the
class of scale/shift invariant procedures. In the Chapter “Topological Modules of
Human Brain Networks Are Anatomically Embedded: Evidence from Modularity
Analysis at Multiple Scales”, an MRI data set is used to demonstrate that modular
structure of brain networks is well reproducible in test–retest settings. These results
provide evidence of the theoretically well-motivated hypothesis that brain regions
neighboring the anatomical space also tend to belong to the same topological
modules. The Chapter “Commercial Astroturfing Detection in Social Networks” is
devoted to constructing a model capable of detecting astroturfing in customer
reviews based on network analysis. In the Chapter “Information Propagation
Strategies in Online Social Networks”, the authors discuss the problem of
predicting information propagation using social network interactions. They suggest
a new approach to construct the model of information propagation and test it on a
real data set from social networks. The Chapter “Analysis of Co-authorship
Networks and Scientific Citation Based on Google Scholar” investigates the
problem on how scientific collaboration represented by co- authorship is related to
citation indicators of a scientist. In the Chapter “Company Co-mention Network
Analysis”, a new company network is constructed on the basis of news mentioning
two companies together. Different types of social network analysis metrics
(degree centrality, closeness centrality, betweenness centrality, eigenvector
centrality, and frequency) are used to identify key companies in the network.
viii Preface
We would like to thank all the authors and referees for their efforts. This work is
supported by the Laboratory of Algorithms and Technologies for Network Analysis
(LATNA) of the National Research University Higher School of Economics and by
RSF grant 14-41-00039.
Nizhny Novgorod, Russia Valery A. Kalyagin

Gainesville, FL, USA Panos M. Pardalos
Pittsburgh, PA, USA Oleg Prokopyev
Nizhny Novgorod, Russia Irina Utkina
References
1. Goldengorin, B.I., Kalyagin, V.A., Pardalos, P.M. (eds.): Models, algorithms and technologies
for network analysis. In: Proceedings of the First International Conference on Network
Analysis. Springer Proceedings in Mathematics and Statistics, vol. 32. Springer, Cham (2013)
2. Goldengorin, B.I., Kalyagin, V.A., Pardalos, P.M. (eds.): Models, algorithms and technologies
for network analysis. In: Proceedings of the Second International Conference on Network
Analysis. Springer Proceedings in Mathematics and Statistics, vol. 59. Springer, Cham (2013)
3. Batsyn, M.V., Kalyagin, V.A., Pardalos, P.M. (eds.): Models, algorithms and technologies for
network analysis. In: Proceedings of Third International Conference on Network Analysis.
Springer Proceedings in Mathematics and Statistics, vol. 104. Springer, Cham (2014)
4. Kalyagin, V.A., Pardalos, P.M., Rassias, T.M. (eds.): Network models in economics and
finance. In: Springer Optimization and Its Applications, vol. 100. Springer, Cham (2014)
5. Kalyagin, V.A., Koldanov, P. A., Pardalos, P.M. (eds.): Models, algorithms and technologies
for network analysis. In: NET 2014, Nizhny Novgorod, Russia, May 2014. Springer
Proceedings in Mathematics and Statistics, vol. 156. Springer, Cham (2016)
6. Kalyagin, V.A., Nikolaev, A.I., Pardalos, P.M., Prokopyev, O.A. (eds.): Models, algorithms
and technologies for network analysis. In: NET 2016, Nizhny Novgorod, Russia, May 2016.
Springer Proceedings in Mathematics and Statistics, vol. 197. Springer, Cham (2017)
Contents
Part I Network Computational Algorithms

Tabu Search for Fleet Size and Mix Vehicle Routing Problem
with Hard and Soft Time Windows . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
Mikhail Batsyn, Ilya Bychkov, Larisa Komosko and Alexey Nikolaev
FPT Algorithms for the Shortest Lattice Vector and Integer Linear
Programming Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
D. V. Gribanov
The Video-Based Age and Gender Recognition with Convolution
Neural Networks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
Angelina S. Kharchevnikova and Andrey V. Savchenko
On Forbidden Induced Subgraphs for the Class of Triangle-König
Graphs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
Dmitry B. Mokeev
The Global Search Theory Approach to the Bilevel Pricing Problem
in Telecommunication Networks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
Andrei V. Orlov
Graph Dichotomy Algorithm and Its Applications to Analysis
of Stocks Market . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
Alexander Rubchinsky
Cluster Analysis of Facial Video Data in Video Surveillance
Systems Using Deep Learning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113
Anastasiia D. Sokolova and Andrey V. Savchenko
Using Modular Decomposition Technique to Solve the Maximum
Clique Problem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121
Irina Utkina
ix
x Contents
Part II Network Models

Robust Statistical Procedures for Testing Dynamics in Market
Network . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135
A. P. Koldanov and M. A. Voronina
Application of Market Models to Network Equilibrium Problems . . . . . 143
Igor Konnov
Selective Bi-coordinate Variations for Network Equilibrium
Problems with Mixed Demand . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 161
Igor Konnov and Olga Pinyagina
Developing a Model of Topological Structure Formation for Power
Transmission Grids Based on the Analysis of the UNEG . . . . . . . . . . . . 171
Sergey Makrushin
Methods of Criteria Importance Theory and Their Software
Implementation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 189
Andrey Pavlovich Nelyubin, Vladislav Vladimirovich Podinovski
and Mikhail Andreevich Potapov
A Model of Optimal Network Structure for Decentralized Nearest
Neighbor Search . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 197
Alexander Ponomarenko, Irina Utkina and Mikhail Batsyn
Computational Study of Activation Dynamics on Networks
of Arbitrary Structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 205
Alexander Semenov, Dmitry Gorbatenko and Stepan Kochemazov
Rejection Graph for Multiple Testing of Elliptical Model
for Market Network . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 221
D. P. Semenov and Petr A. Koldanov
Mapping Paradigms of Social Sciences: Application of Network
Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 235
Dmitry Zaytsev and Daria Drozdova
Part III Network Applications

Using Geometry of the Set of Symmetric Positive Semidefinite
Matrices to Classify Structural Brain Networks . . . . . . . . . . . . . . . . . . . 257
Mikhail Belyaev, Yulia Dodonova, Daria Belyaeva, Egor Krivov,
Boris Gutman, Joshua Faskowitz, Neda Jahanshad and Paul Thompson
Comparison of Statistical Procedures for Gaussian Graphical
Model Selection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 269
Ivan S. Grechikhin and Valery A. Kalyagin
Contents xi
Sentiment Analysis Using Deep Learning . . . . . . . . . . . . . . . . . . . . . . . . 281

Nikolay Karpov, Alexander Lyashuk and Arsenii Vizgunov
Invariance Properties of Statistical Procedures for Network
Structures Identification . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 289
Petr A. Koldanov
Topological Modules of Human Brain Networks Are Anatomically
Embedded: Evidence from Modularity Analysis at Multiple
Scales . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 299
Anvar Kurmukov, Yulia Dodonova, Margarita Burova,
Ayagoz Mussabayeva, Dmitry Petrov, Joshua Faskowitz
and Leonid E. Zhukov
Commercial Astroturfing Detection in Social Networks . . . . . . . . . . . . . 309
Nadezhda Kostyakova, Ilia Karpov, Ilya Makarov and Leonid E. Zhukov
Information Propagation Strategies in Online Social Networks . . . . . . . 319
Rodion Laptsuev, Marina Ananyeva, Dmitry Meinster, Ilia Karpov,
Ilya Makarov and Leonid E. Zhukov
Analysis of Co-authorship Networks and Scientific Citation
Based on Google Scholar . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 329
Nataliya Matveeva and Oleg Poldin
Company Co-mention Network Analysis . . . . . . . . . . . . . . . . . . . . . . . . 341
S. P. Sidorov, A. R. Faizliev, V. A. Balash, A. A. Gudkov,
A. Z. Chekmareva and P. K. Anikin
Contributors
Marina Ananyeva Higher School of Economics, National Research University,

Moscow, Russia
P. K. Anikin Saratov State University, Saratov, Russian Federation
V. A. Balash Saratov State University, Saratov, Russian Federation
Mikhail Batsyn Higher School of Economics, National Research University,
Moscow, Russia; Laboratory of Algorithms and Technologies for Network
Analysis, National Research University Higher School of Economics, Nizhny
Novgorod, Russia
Mikhail Belyaev Skolkovo Institute of Science and Technology, Moscow, Russia;
Kharkevich Institute for Information Transmission Problems, Moscow, Russia
Daria Belyaeva Skolkovo Institute of Science and Technology, Moscow, Russia;
Kharkevich Institute for Information Transmission Problems, Moscow, Russia
Margarita Burova Higher School of Economics, National Research University,
Moscow, Russia
Ilya Bychkov Laboratory of Algorithms and Technologies for Network Analysis,
National Research University Higher School of Economics, Nizhny Novgorod,
Russia
A. Z. Chekmareva Saratov State University, Saratov, Russian Federation
Yulia Dodonova Kharkevich Institute for Information Transmission Problems,
Moscow, Russia
Daria Drozdova Higher School of Economics, National Research University,
Moscow, Russia
A. R. Faizliev Saratov State University, Saratov, Russian Federation
xiii
xiv Contributors
Joshua Faskowitz Imaging Genetics Center, Stevens Neuroimaging and

Informatics Institute, University of Southern California, Marina del Rey, CA, USA;
Indiana University, Bloomington, USA
Dmitry Gorbatenko Matrosov Institute for System Dynamics and Control
Theory SB RAS, Irkutsk, Russia
Ivan S. Grechikhin Laboratory of Algorithms and Technologies for Network
Novgorod, Russia
D. V. Gribanov Lobachevsky State University of Nizhny Novgorod, Nizhny
Novgorod, Russian Federation; National Research University Higher School of
Economics, Nizhny Novgorod, Russian Federation
A. A. Gudkov Saratov State University, Saratov, Russian Federation
Boris Gutman Kharkevich Institute for Information Transmission Problems,
Moscow, Russia; Imaging Genetics Center, Stevens Neuroimaging and Informatics
Institute, University of Southern California, Marina del Rey, CA, USA
Neda Jahanshad Imaging Genetics Center, Stevens Neuroimaging and
Informatics Institute, University of Southern California, Marina del Rey, CA, USA
Valery A. Kalyagin Laboratory of Algorithms and Technologies for Network
Novgorod, Russia
Ilia Karpov Higher School of Economics, National Research University,
Moscow, Russia
Nikolay Karpov Higher School of Economics, National Research University,
Moscow, Russia
Angelina S. Kharchevnikova Higher School of Economics, National Research
University, Nizhny Novgorod, Russia
Stepan Kochemazov Matrosov Institute for System Dynamics and Control
A. P. Koldanov Laboratory of Algorithms and Technologies for Network
Novgorod, Russia
Petr A. Koldanov Laboratory of Algorithms and Technologies for Network
Novgorod, Russia
Larisa Komosko Laboratory of Algorithms and Technologies for Network
Novgorod, Russia
Contributors xv
Igor Konnov Kazan Federal University, Kazan, Russia; Kazan Federal University,
Institute of Computational Mathematics and Information Technologies, Kazan,
Russia
Nadezhda Kostyakova Higher School of Economics, National Research
University, Moscow, Russia
Egor Krivov Skolkovo Institute of Science and Technology, Moscow, Russia;
Kharkevich Institute for Information Transmission Problems, Moscow, Russia;
Moscow Institute of Physics and Technology, Moscow, Russia
Anvar Kurmukov Higher School of Economics, National Research University,
Moscow, Russia
Rodion Laptsuev Higher School of Economics, National Research University,
Moscow, Russia
Alexander Lyashuk Higher School of Economics, National Research University,
Moscow, Russia
Ilya Makarov Higher School of Economics, National Research University,
Moscow, Russia
Sergey Makrushin Financial University under the Government of the Russian
Federation, Moscow, Russia
Nataliya Matveeva Higher School of Economics, National Research University,
Moscow, Russia
Dmitry Meinster Higher School of Economics, National Research University,
Moscow, Russia
Dmitry B. Mokeev Department of Algebra, Geometry and Discrete Mathematics,
Lobachevsky State University of Nizhny Novgorod, N.Novgorod, Russia;
Laboratory of Algorithms and Technologies for Networks Analysis, Higher School
of Economics in Nizhny Novgorod, N.Novgorod, Russia
Ayagoz Mussabayeva Higher School of Economics, National Research
Andrey Pavlovich Nelyubin Mechanical Engineering Research, Institute of the
RAS, Moscow, Russia
Alexey Nikolaev Laboratory of Algorithms and Technologies for Network
Novgorod, Russia
Andrei V. Orlov Matrosov Institute for System Dynamics and Control Theory of
SB RAS, Irkutsk, Russia
xvi Contributors
Dmitry Petrov Kharkevich Institute for Information Transmission Problems,

Moscow, Russia; Imaging Genetics Center, University of Southern California, Los
Angeles, USA
Olga Pinyagina Kazan Federal University, Institute of Computational
Mathematics and Information Technologies, Kazan, Russia
Vladislav Vladimirovich Podinovski Higher School of Economics, National
Research University, Moscow, Russia
Oleg Poldin Higher School of Economics, National Research University,
Moscow, Russia
Alexander Ponomarenko Higher School of Economics, National Research
Mikhail Andreevich Potapov Institute of Computer Aided Design of the RAS,
Moscow, Russia
Alexander Rubchinsky Higher School of Economics, National Research
University, Moscow, Russia; National Research Technological University
“MISIS”, Moscow, Russia
Andrey V. Savchenko Laboratory of Algorithms and Technologies for Network
Novgorod, Russia
Alexander Semenov Matrosov Institute for System Dynamics and Control
D. P. Semenov Laboratory of Algorithms and Technologies for Network Analysis,
National Research University Higher School of Economics, Nizhny Novgorod,
Russia
S. P. Sidorov Saratov State University, Saratov, Russian Federation
Anastasiia D. Sokolova Laboratory of Algorithms and Technologies for Network
Novgorod, Russia
Paul Thompson Imaging Genetics Center, Stevens Neuroimaging and Informatics
Institute, University of Southern California, Marina del Rey, CA, USA
Irina Utkina Higher School of Economics, National Research University,
Moscow, Russia; Laboratory of Algorithms and Technologies for Network
Novgorod, Russia
Arsenii Vizgunov Higher School of Economics, National Research University,
Moscow, Russia
Contributors xvii
M. A. Voronina Laboratory of Algorithms and Technologies for Network

Novgorod, Russia
Dmitry Zaytsev Higher School of Economics, National Research University,
Moscow, Russia
Leonid E. Zhukov Higher School of Economics, National Research University,
Moscow, Russia
Part I
Network Computational Algorithms
Tabu Search for Fleet Size and Mix
Vehicle Routing Problem with Hard
and Soft Time Windows
Mikhail Batsyn, Ilya Bychkov, Larisa Komosko and Alexey Nikolaev
Abstract The paper presents a tabu search heuristic for the Fleet Size and Mix
Vehicle Routing Problem (FSMVRP) with hard and soft time windows. The objective
function minimizes the sum of travel costs, fixed vehicle costs, and penalties for soft
time window violations. The algorithm is based on the tabu search with several
neighborhoods. The main contribution of the paper is the efficient algorithm for a
real-life vehicle routing problem. To the best of our knowledge, there are no papers
devoted to the FSMVRP problem with soft time windows, while in real-life problems,
this is a usual case. We investigate the performance of the proposed heuristic on
the classical Solomon instances with additional constraints. We also compare our
approach without soft time windows and heterogeneous fleet of vehicles with the
recently published results on the VRP problem with hard time windows.
Keywords Soft time windows · Fleet size and mix · Vehicle routing · Variable
neighborhood search · Tabu search
1 Introduction
The Vehicle Routing Problem (VRP) is an NP-hard combinatorial optimization prob-

lem. The classical VRP consists in finding the optimal set of routes for a fleet of
vehicles located at the central depot to deliver goods to a given set of customers.
M. Batsyn (B) · I. Bychkov · L. Komosko · A. Nikolaev

Laboratory of Algorithms and Technologies for Network Analysis, National Research University
Higher School of Economics, 136 Rodionova, 603093 Nizhny Novgorod, Russia
e-mail: mbatsyn@hse.ru
I. Bychkov
e-mail: ibychkov@hse.ru
L. Komosko
e-mail: lkomosko@hse.ru
A. Nikolaev
e-mail: ainikolaev@hse.ru
© Springer International Publishing AG, part of Springer Nature 2018 3
V. A. Kalyagin et al. (eds.), Computational Aspects and Applications in Large-Scale
Networks, Springer Proceedings in Mathematics & Statistics 247,
https://doi.org/10.1007/978-3-319-96247-4_1
4 M. Batsyn et al.
The practical interest of this problem consists in its application to various spheres of
the real life: delivery of goods to supermarkets [26], waste collection [3], school bus
routing [20], mail delivery [25], transportation of people with disabilities [8], and
many others.
However, real-life vehicle routing problems differ considerably from classical
ones having various constraints such as hard and soft time windows, different vehi-
cles, capacity constraints, split deliveries, and others. All these additional features
make the process of finding the solution extremely difficult and time-consuming.
VRP problems with a heterogeneous fleet of vehicles are frequently encountered in
logistics. In real-world applications, fleets of vehicles with different capacities and
costs are more common than homogeneous ones [1]. There are two types of het-
erogeneous fleet VRP [16, 17]. The first one is Fleet Size and Mix Vehicle Routing
Problems (FSMVRP). In FSMVRP, there are no limits on the number of available
vehicles of each type. The second category is Heterogeneous Fleet Vehicle Routing
Problems (HFVRP), where the number of vehicles of each type is limited. In this
paper, the FSMVRP is considered [4, 23, 29, 30].
According to Laporte [18], time windows are also one of the most vital constraints
used in real-world applications. There are two types of them: hard and soft time
windows. In real life, it is rarely possible that a customer exposes only one type
of time windows. The most common situation is when there are certain working
hours of a customer—hard time window, and a desirable period of time for receiving
an order—soft time window. In the academic world, much attention is paid to the
Vehicle Routing Problem with hard Time Windows (VRPTW) [2, 5, 6, 10, 12, 13,
21, 25, 27, 28].
Fewer papers deal with the vehicle routing problem with soft time windows [9, 14,
22, 24]. And the papers considering both the heterogeneous fleet of vehicles and soft
time windows are not known to the authors. Note that the existing VRP models with
Soft Time Windows (VRPSTW) also have hard time window constraints. So hereafter
VRPSTW will denote a VRP problem with both soft and hard time windows.
In this paper, we suggest an efficient approach for the Fleet Size and Mix Vehicle
Routing Problem with Soft Time Windows (FSMVRPSTW). Our algorithm is based
on tabu search with variable neighborhoods. We use swap and relocate neighbor-
hoods for inter-route optimization and swap neighborhood for intra-route. A special
procedure is developed for soft time windows optimization.
The paper is organized as follows. In the next section, we provide an integer
linear programming model for the considered problem. In Sect. 3, the pseudocode
and description of our algorithm are presented. Section 4 contains computational
experiments and results, and Sect. 5 concludes the paper.
2 Mathematical Model
The fleet size and mix vehicle routing problem with soft time windows can be
described as follows. Let G (V0 , A) be a complete directed graph, where
Tabu Search for Fleet Size and Mix Vehicle Routing Problem ... 5
V0 {0, . . . , n} is a set of vertices representing customers V {1, . . . , n} with

the depot node 0, and A is a set of arcs. The nonnegative travel cost ci j and travel
time ti j are assigned to every arc (i, j) ∈ A, i j. Each customer i ∈ V has a non-
negative demand of quantity qi , service time si , hard time window—the earliest and
latest possible start of service—[e
i , li ], and soft time window—the earliest and latest
˜
preferred start of service— ẽi , li . K denotes the set of different types of vehicles
available at the depot. k is an index of the chosen type of vehicle. Parameters Q k and
f k correspond to the capacity and fixed cost of a vehicle of type k ∈ K .
The beginning of service of the customer i ∈ V by vehicle l of type k ∈ K in
a solution is denoted by variable bikl . Note that the number of available vehicles of
any type in this model is not limited, and therefore, index l can potentially take any
values from 1 to n. In case when a vehicle arrives to customer i ∈ V earlier than the
earliest possible start of service ei , it has to wait (without any restrictions) at least
until ei to start the service. If a vehicle starts service before
the earliest preferred
time of service ẽi and later than ei , then the penalty α ẽi − bikl is paid. Here, the
parameter α denotes the cost of one time unit of soft time window violation. Also
if a vehicle starts service later than the latest preferred time of service ˜
li and earlier
˜
than the latest possible time of service li , then the penalty α bi − li is paid.
kl
The objective of the FSMVRPSTW problem is to choose vehicles and their routes
to serve all the customers with minimum total cost including travel costs, fixed vehicle
costs, and penalties for soft time window violations. The following mixed integer
linear programming model for the fleet size and mix vehicle routing problem with
hard and soft time windows is introduced.
Decision variables
xiklj ∈ {0, 1}, xiklj 1, if vehicle l of type k travels by the arc (i, j) ∈ A in the
solution.
ykl ∈ {0, 1}, ykl 1, if vehicle l of type k is used in the solution for serving
customers.
z ikl ∈ {0, 1}, z ikl 1, if vehicle l of type k serves customer i in the solution.
bikl ≥ 0, beginning of service at customer i by vehicle l of type k.
vi−kl
≥ 0, vehicle l of type k violation of the earliest preferred time at customer i.
vi+kl
≥ 0, vehicle l of type k violation of the latest preferred time at customer i.
Objective function

n
n
n
kl
fk ykl + ci j xiklj + α kl
vi− + vi+ → min (1)
k∈K l1 k∈K l1 i∈V0 j∈V0 k∈K l1 i∈V
6 M. Batsyn et al.
Constraints:

n
z ikl 1, ∀i ∈ V (2)
k∈K l1
ykl ≥ z ikl , ∀k ∈
K , ∀l 1, n, ∀i ∈ V (3)

x0ikl kl
xi0 ykl , ∀k ∈ K , ∀l 1, n (4)
i∈V i∈V

xiklj x klji z ikl , ∀k ∈ K , ∀l 1, n, ∀i ∈ V (5)
j∈V0 j∈V0

qi z ikl ≤ Q k , ∀k ∈ K , ∀l 1, n (6)
i∈V

bklj ≥ bikl + si + ti j − M 1 − xiklj , ∀k ∈ K , ∀l ∈ 1, n, ∀i ∈ V0 , j ∈ V (7)
ei ≤ bikl ≤ li , ∀k ∈ K , ∀l 1, n, ∀i ∈ V (8)
kl
vi− ≥ ẽi − bikl ,∀k ∈ K , ∀l 1, n, ∀i ∈ V (9)
kl
vi+ ≥ bikl ˜
− li , ∀k ∈ K , ∀l 1, n, ∀i ∈ V (10)
The objective function (1) minimizes the total cost of all routes including travel
costs, fixed vehicle costs, and penalties for soft time window violations. Constraints
(2) require that each customer should be serviced exactly once. Inequalities (3)
guarantee that a vehicle is used in the solution and its fixed cost is taken into account
only if it serves at least one customer. Equalities (4, 5) specify that every used vehicle
has to leave the depot, arrive to all the appointed customers, leave them, and at the end
return to the depot. Constraints (6) require the total demand of all customers in one
route to satisfy the capacity of the vehicle serving it. Inequalities (7) guarantee that
the service of the customer is not started before the vehicle completes service at the
previous customer and travels to this one. Here, we use big-M notation, which makes
the inequality to be automatically satisfied for any values of variables, when xiklj 0.
Finally, constraints (8) require beginning times to lie inside hard time windows and
constraints (9, 10) determine how much the soft time windows are violated.
3 Algorithm Description
The developed algorithm is based on the tabu search and variable neighborhood
search metaheuristics. The main function in our algorithm is TABUSEARCH()
(Algorithm 1). The idea of the developed algorithm for the fleet size and mix vehicle
routing problem with soft time windows consists in application of tabu search with
several neighborhoods to an initial greedy solution. One of the advantages of the
tabu search is in tabu list which forbids some edges and thus diversifies the search.
This approach allows avoiding loops and going further from a local optimum out
from the neighborhood of the current best solution. New solutions are generated by
means of 2-opt, swap and relocate operators taking into consideration the tabu edges.
If a better value of the objective function is found and the new solution is feasible in
terms of hard time windows and vehicle loads, then it replaces the current optimal
solution. In such a case, the algorithm tries to optimize further the found solution by
means of intra-route improvement operators. If the solution is infeasible, then the
function value is penalized. All these steps are carried out until the current solution
remains infeasible during SC (Stopping Criterion) iterations.
Algorithm 1. Tabu search

function TABUSEARCH()
⊳ builds solution S* for the fleet size and mix vehicle routing problem with soft time windows
NS ← 1000 ⊳ the number of shakes – restarts from a new initial solution
σ ← 0,833 ⊳ diversification parameter, the best value was found empirically
TL ← ∅ ⊳ tabu list
S* ← ∅ ⊳ f (S*) = +∞
while NS > 0 do
S ← SOLOMONINSERTION(V)
β←1 ⊳ the penalty for violating a hard time window by 1 time unit
γ←1 ⊳ the penalty for exceeding a vehicle capacity by 1 unit
SC ← 1500 ⊳ stopping criterion – the number of infeasible steps in tabu search
while SС > 0 do
S' ← CHOOSENEIGHBORSOLUTION(S)
if f (S') ≥ f (S*) then
TL ← TL ∪ S
end if
S' ← S
if (w(S) = 0) & (q(S) = 0) then ⊳ w(S) – total violation of hard time windows in S
q(S) – total overload of vehicle capacities in S
INTRAOPTIMISATION(S)
SOFTWINDOWOPTIMISATION(S)
if f (S) < f (S*) then
S* ← S
end if
SC ← 1500
else
f (S) ← f (S) + β·w(S) + γ·q(S)
SC ← SC – 1
end if
UPDATEPARAMETER(β, w(S))
UPDATEPARAMETER(γ, q(S))
end while
TL ← ∅
NS ← NS – 1
end while
return S*
end function
8 M. Batsyn et al.
Algorithm 2. Updating objective function parameters

function UPDATEPARAMETER(parameter, criterion)
⊳ assigns a new value to parameter to control the direction of the search space exploration
if (criterion > 0) then
parameter ← parameter ⋅ (1 + σ)
else if (parameter > 1)
parameter ← parameter / (1 + σ)
end if
end function
3.1 Initial Solution
The algorithm of the initial solution construction for the stated problem is based on
Solomon [28] insertion heuristic. First, the route consisting of only one customer,
which is the farthest from the depot, is created. Afterward, new customers are grad-
ually added to the current route taking into account hard time windows and capacity
constraints. To evaluate the insertion cost for the new customer u between two adja-
cent customers i and j criterion c1 (i, u, j) is computed. And to choose the best
customer for insertion we compute criterion c2 (u).
Let (r1 , r2 , . . . , rm ) be the current route where the first and the last points are the
depots: r1 0, rm n + 1 (vertex n + 1 is a copy of vertex 0). For each unserved
customer, we determine the best insertion place taking into account the change in
travel cost, beginning times, and fixed vehicle cost. The evaluation of these factors
is taken into account in the c1 (i, u, j) criterion calculation. After application of this
procedure, the best insertion place with the least insertion cost is chosen. Next,
for every customer u the c2 (u) criterion is calculated. It takes into account how
expensive is to reach this customer from the depot—direct travel cost c0u , what is
the cost of the smallest vehicle to serve the route with this customer, and what is
the cheapest insertion cost c1∗ (u). The calculation of these criteria is implemented in
SolomonInsertion().
To better determine the best insertion place for the customer in terms of vehicle
costs, Dullaert et al.’s [7] approach AROS (Adapted Realistic Opportunity Savings)
10 M. Batsyn et al.
is used. The idea of Realistic Opportunity Savings (ROS) was suggested by Golden
et al. [11]. The principle of the method consists in taking into account the opportunity
to fill the unused capacity of a bigger vehicle needed after a new customer u insertion.
The new vehicle cost is taken into account only if the new total demand q q + qu
in the route is greater than the capacity Q of the current vehicle and requires at least
the vehicle of capacity Q . The difference F(q ) − F(q) shows how greater is the
fixed cost F(q ) of the smallest vehicle needed to serve the new total demand q than
the fixed cost F(q) of the current vehicle. The additional item F (Q − q ) is called
the opportunity savings. Here, Q − q is the unused capacity of the bigger vehicle
which gives an opportunity to add more customers with total demand not greater than
Q − q . So we can avoid using for such customers any vehicle having the capacity
not greater than Q − q . The fixed cost of the largest such vehicle is F (Q − q ).
3.2 Solution Improvement
As the tabu search implies, when the initial solution is known we should find the
neighbor solutions of it and choose one of them to be the current. Swap and relocate
operators are chosen as the neighborhood construction methods in the developed
algorithm. A notice should be made that while generating neighbor solutions it is
allowed to break restrictions regarding hard time windows and capacity constraints.
This is done to diversify the search because often the way to the best solution lies
through infeasible ones. After neighbor solutions are constructed, to avoid the sit-
uation of repeating solutions, the random one from the 3% of the best generated
solutions with the minimal function value is chosen as the current.
When a good feasible solution is found sometimes possible changes within each
route can be made which can lead to the decrease in the objective function value. Hard
time windows constraints set a certain sequence of customers in a route therefore
not all solution improvement methods are suitable for this operation. We considered
only neighbors’ swap. For details, see Braysy and Gendreau [2].
Besides taking into account soft time window violation penalty in the objective
function, the soft time window optimization procedure is developed. The principle
of it is in the following: for each route starting from its last customer, the method
tries to distribute soft time windows intervals with their smallest violations among
all the nodes in the route trying to use the latest preferred time where possible.
4 Computational Experiments
The experiments are carried out on Intel Xeon X5690 3,47 GHz with the following
parameter values determined empirically: Stopping Criterion SC 1500, Tabu List
size |TL| 24, diversification parameter σ 0.833, the penalty for one time unit of
soft time windows violation α = 4. As the main goal of the article is to develop the
algorithm for the fleet size and mix vehicle routing problem with soft time win-
dows—the problem regarding which the published papers are not found—therefore
the effectiveness of the proposed heuristic is tested on the VRPTW problems. Our
results are compared with the latest VRPTW results from the article of Jawarneh and
Abdullah [15].
4.1 VRPTW Results Comparison
As we compare the performance of our algorithm with the one from the article
Jawarneh and Abdullah [15], the computation times for every instance are set nearly
12 M. Batsyn et al.
the same: from 1 to 10 min. This is done by setting an appropriate Number of Shakes
(NS parameter) in the main algorithm.
Table 1 shows that in 34 cases of the 56 our approach is better than the approach
proposed by Jawarneh and Abdullah [15]. In eight cases, our algorithm shows the
same performance and for the remaining 14 cases, we have a slightly lower per-
formance. The best results we have on random instances (R101–R211), where we
obtain better solutions on 18 instances of the 23, and our improvement reaches 10%.
We have good results on random clustered instances (RC101–RC208), where our
algorithm gives lower cost on 10 instances of the 16, and the maximal improvement
is 6.4%. Finally, on clustered instances (C101–C208), our best improvement is 4.4%;
we are better on six instances of the 17 and have the same result on eight instances.
4.2 Results for FSMVRPSTW
As the fleet size and mix vehicle routing problem with hard and soft time windows
is not considered in literature we have generated instances for the computational
experiments. As a base, we take benchmarks of Liu and Shen [19] (Table 2) and add
soft time windows with different sizes of 10, 20, 30, and 40% of the hard time window
size. The position of the determined soft time window inside the hard one is chosen
randomly with uniform distribution. In each instance, there are several vehicle types
with different capacities and fixed costs denoted with letters A, B, C, D, E, and F. For
each VRPTW instance, we generate three instances with large, medium, and small
fixed vehicle costs marked with letters a, b, c correspondingly. In Tables 3, 4, 5 and 6,
the average results for each class of instances are presented. The generated instances
are available in the following link: https://nnov.hse.ru/en/latna/benchmarks.
The computational experiments show that the size of soft time windows influences
the value of the objective function: the larger is the soft time window interval the
better is the solution. However, the travel cost of routes in a solution can be even
smaller in case of narrow soft time windows, because only short routes can satisfy
such time windows. At the same time, such a solution will have more routes and a
greater fixed cost of used vehicles.
Table 1 Comparison of the results for the VRPTW with Jawarneh and Abdullah [15]
Instances Jawarneh and Proposed heuristic Improvement, %
Abdullah
C101 828,94 828,94 0,00
C102 828,94 828,94 0,00
C103 835,71 832,07 0,44
C104 885,06 847,45 4,44
C105 828,94 828,94 0,00
C106 828,94 828,94 0,00
C107 828,94 828,94 0,00
C108 831,73 828,94 0,34
C109 840,66 828,94 1,41
C201 591,56 591,56 0,00
C202 591,56 591,56 0,00
C203 593,21 636,08 −6,74
C204 606,9 648,07 −6,35
C205 588,88 588,88 0,00
C206 588,88 588,49 0,07
C207 590,59 602,99 −2,06
C208 593,15 588,49 0,79
R101 1643,18 1626,42 1,03
R102 1476,11 1452,20 1,65
R103 1245,86 1232,13 1,11
R104 1026,91 979,63 4,83
R105 1361,39 1362,88 −0,11
R106 1264,5 1246,08 1,48
R107 1108,11 1075,48 3,03
R108 994,68 955,35 4,12
R109 1168,91 1178,57 −0,82
R110 1108,22 1114,86 −0,60
R111 1080,84 1081,08 −0,02
R112 992,22 991,46 0,08
R201 1197,09 1152,34 3,88
R202 1092,22 1077,94 1,32
R203 983,06 893,26 10,05
R204 845,3 775,19 9,04
R205 999,54 983,99 1,58
R206 955,94 929,16 2,88
R207 903,59 858,53 5,25
R208 769,96 749,70 2,70
(continued)
Another random document with
no related content on Scribd:
The back is
made with one full
row of squares,
and two rows of
buttons at the
bottom, besides
the finishing
squares (see Fig.
50), then the swell
of the back is
carried up to
within 4 inches of Fig. 50.
the upper edge,
1⅛ inches being allowed for the swell. Then the top of the back is
finished with a large roll, of about 5 inches swell or girth, so that the
back has only one row of buttons in the upper sweep.
The arm-pieces, Fig. 51, are made in a
peculiar way, and the modus operandi is
rather difficult of explanation. In the place of
the usual arm-piece block, a piece of plank,
2¾ inches wide by ½ inch thick, is fitted in
with the usual sweep to it. Now fit four pieces
of single fly buckram, to form as it were a
Fig. 51.
funnel, the shape of the arm-piece desired;
then sew seaming lace to the two edges of
the funnel, which will show inside of the body; to the lower edge sew
a piece of cloth in smooth, so that it will cover the bottom of the
funnel or cylinder. To the same lace edge, blind sew in a piece of
cloth, for the purpose of forming a wrinkled roll on the inner face of
the funnel, wrinkled 1 inch in fulness for every 3 inches in length, and
as full the other way as desired, for it ought to be full enough to
come out with the bottom side quarter. Next, blind sew the outer
edge of this roll to the other or top seaming lace, and stuff lightly with
hair, thus forming one roll on the inner face, and having the lower
face covered with the smooth cloth.
Now blind sew another piece of cloth to the top lace, as in the
other case, to form another wrinkled roll on top of the funnel or
cylinder, but the outer edge of this roll is to be finished by nailing to
the outside of the piece of plank just mentioned. All this sewing is of
course to be done on the bench, one side of the funnel to be left
open for this purpose. Next, nail in the bottom side quarter made up
in squares, and then nail the side of the funnel, which is fitted against
the arm-board, to the board and over the quarter, thus finishing the
lower part. Then nail the fourth on top side to the top edge of arm-
board; next, stuff from the front the funnel, pretty solid, and finish the
top roll, which up to now has been left open, into the outside of the
arm-board, thus completing the arm-piece, which shows two
wrinkled rolls divided by two rows of seaming lace.
The door fall, Fig. 52, is made on three
fly buckram pasted together, but one fly is
cut off about an inch from the top to allow
the fall to hinge. The fall is made about 12
inches deep, the lower edge being
circular. The broad lace is bent to the
required shape, and the corners sewed
Fig. 52. and put on to the buckram, which is cut to
the shape intended for the fall. Mark
where the inside edge of the lace comes
all round the buckram, then mark 1¼ inches from the mark; paste a
piece of carpet into the buckram to come within ½ inch from this
mark, which will make the edge of the carpet 1⅜ inches from the
edge of the lace; cover this carpet with a piece of cloth, pasting on
the buckram; take a piece of seaming lace, long enough to reach
round the fall, and sew a piece of cloth to it, for the purpose of
forming a wrinkled roll round three sides of the fall, inside the broad
lace and outside the sewed carpet, between both, with 1 inch fulness
to each 3 inches of length; sew this seaming lace and roll to the
buckram at the mark, 1¼ inches from the broad lace, gather the
other edge with a running string, and sew down and stuff lightly,
finishing in such a manner that the broad lace shall cover this sewed
edge; next, paste on the broad lace and cover the wrong side with
silk or muslin. When dry, stitch both edges.
Fig. 53. Fig. 54.
Figs. 53 and 54 show two styles of trimming for a door. Fig. 53 is
made as follows:—Paste out three flies of buckram, and lay off for
block or biscuit pattern, leaving space enough all round for a broad
lace border, and at the top leave double the space. The top space is
formed into a plain cloth roll of the same goods as the job is trimmed
with.
In this case the trimming is brown cloth; the broad lace is silk and
worsted of a shade much lighter than the cloth. The diamond-shaped
and connecting figures are worsted and are raised.
The card-pocket is made of tin and covered with Turkey morocco
the colour of the trimming.
In Fig. 54 it will be noticed that the style is somewhat different from
the other; the surface of the door is trimmed plain, the fall alone
being stuffed. The fall is stuffed in diamond form and enclosed with a
lace border. This pocket runs up under the fall to the top, and is there
nailed.
The following remarks on lining and trimming are taken from
“Cassell’s Technical Educator:”—
“We may with advantage say a word to our carriage lace-makers,
who seem to have made but scant use of the various Schools of
Design for the improvement of their taste in producing new and
suitable patterns in the manufacture of their goods. For a long period
we had nothing but the old scroll or flower pattern, which was
handed down from father to son as if by a fixed law. At length, when
it was felt that some change was required, the absence of all taste in
design was shown in the production of entirely plain worked laces,
which deprived carriage linings of their chief element of lightness and
beauty. Thanks, however, to the taste and discernment of Messrs.
Whittingham & Walker, who, perhaps, have devoted more attention
to this branch of industry than any other house in London, the trade
was relieved from the necessity of either adhering to the old pattern
or of adopting the opposite scheme. They introduced small neat
designs in laces eminently adapted to the purpose, and in 1857 they
registered a pattern, now extensively known as the double diamond
pattern, which has not only become general in England, but is largely
patronised throughout Europe and America. This, and kindred
patterns, exactly fulfil modern requirements, and give us the
necessary relief without extreme.
“But with materials well and tastefully selected the trimmer has still
his work to do. The lining of a carriage is divided into many different
parts, all of which have to be designed. Canvas or paper patterns
have to be cut to these, and properly fitted before the material is
touched with which the carriage is to be lined. Wherever superiority
of workmanship is to be shown in this department, the French
method of trimming is adopted as being more elegant than the
English. We shall therefore confine our observations to this method.
“In adopting the French method silk is mostly used in the place of
morocco, and its peculiarities consist in the manner of quilting. The
different squabs are made up in horizontal pipes or flutes, which are
tufted in different ways. To proceed, cut a pattern in strong paper the
size and shape of the space to be trimmed, and draw on it with a
pencil the pipes, also mark the position of the tufts. On large pieces
only mark one-half, the other half being the same. The pipes of the
back are usually 12 inches high and from 3 to 5 inches wide.
“The position of the tufts is considerably varied. Next stretch a
piece of strong muslin in the stretching frame, lay the paper pattern
on it, and mark the position of the tufts with an awl. Mark the lines of
the pipes on the muslin with red chalk or pencil. In the same manner
the pattern must be transferred upon the inside of the material used
for covering, making of course due allowance for the depth of the
pipes; about 3 inches is a fair average for fulness at the top, 1 inch
for the height of the pipes, and 1½ inches for the width. For the last
pipe an extra allowance is made in a narrow strip sewed on to it.
“Next lay a quantity of hair on the frame and form it the swell
desired. Keep the hair in position with a few long stitches, and lay
the silk over it. Commence tufting in the middle of the lower row of
pipes, and continue equally to both sides. Silk cords stretched into
the channels between the pipes were at one time considered
elegant, but their main merit was that they aided materially in
preserving the original shape of the pipes. Backs are usually made
couch-shaped, with a roll all round on the top, which at the same
time form the elbow-pieces on the sides. In elegant carriages this roll
is often elaborately executed in a helical or screw-like shape, and
continued from the door-pillar down to the seat-frame, being made
by winding silk cords around the roll. These silk cords appear as a
single thread, but in reality there are three different cords which are
wound at even distances. A style of trimming much used of late both
in France and Germany for low backs, is a row of pipes at the lower
end, which are pinched to points at the top, and above these are
three rows of regular squares. Squares are preferred to diamonds as
they are softer.
“Usually the back is laid on spiral springs, which are fastened as
follows:—The back of the body is covered with coarse muslin, after
being slightly stuffed, and on this muslin four rows of seven small
springs each are set. For the lowest row, springs a little stronger may
be used than for the other rows. The highest row is set about 1½
inches below the edge of the back-board, and the lowest row at 6
inches above the seat-frame. The springs having been sewn on with
a bent needle, are tied first from right to left and then from top to
bottom. A thin cord will answer for this purpose.
“The cord is first cut in lengths, and when the tying begins about 6
inches are allowed at the ends. The cord is wound about the third
ring of the first and last spring in each row, and afterwards the first
ring is brought into the right position with the piece of cord allowed
over. This will make the spring stand upright, and it can be raised or
lowered on one side. The springs being thus all placed in position,
they are finally tied crossways.
“The squab, in this instance, is worked in coarse muslin or canvas,
stiffened with a little thin paste. It is set in the frame and marked as
we have described above. When the cushions press against the
back and side pieces, frequently no stuffing is made, but simply a
piece of fine linen is sewed reverse to the main piece, and this is
called the ‘false finish.’ In fine work the stuffing extends clear to the
seat-frame.
“Of course each of these variations requires a different calculation
for the muslin at the back as well as for the cover. For the latter an
allowance of 1½ inches is made for the pipes from the lower to the
upper end, and also for the points an addition of ¾ inch. For each
square in the height 1½ inch has to be calculated. The folds of
squares when laid over springs being diagonal, easily draw apart
when stretched out, while the folds of diamonds running up and
down may be drawn tighter to a certain degree of stretching.
“For the upper row of squares we have to allow for the backs at
least double what we have to allow for the other rows—namely, 3
inches. For the width of every pipe an addition of 1½ inches is
calculated.
“Both cover and muslin being thus marked, we commence to draw
in the tufts. Every point marked on the cover has to lie exactly on the
corresponding one of the muslin. The lowest tufts are first drawn in;
then turn the frame and commence on one side at top, every point of
the pipes being singly stuffed and the folds adjusted. This being
done, every fold of the squares can be tufted right through, stuffed,
and folded. Squares are easier to be worked than diamonds, but
pointed pipes give more trouble than the ordinary straight ones.
Fig. 55.—Double Brougham.

“The elbow-pieces of this finish consist of two rolls made of muslin;
they are thinner towards the front of the seats. After being stuffed, a
piece of muslin 8 inches wide is sewed on all the length to the
bottom of the roll, which serves, after the roll is tacked to the door-
pillar and back, to give it the required sweep in stretching and
tacking it to the sides of the body. Then mark on the roll the width of
the pipes, and cut the cover for it, allowing 1 inch of width; and as to
the height, the cover must go all round, the roll having to be sewed
back and front to the linen with which the roll is tacked to the body.
“After we have put in all the lining, we have to adjust the silk
curtains, the blinds and glasses to doors and front part, to cover the
iron dash-frame with the best patent leather, trim the coach-box seat,
put on all the mouldings and bearings to the body, arrange the
position of the lamps and fix them, and generally attend to all those
little finishing points which give the appearance of neatness and
finish to the whole.”
Fig. 55 shows a double brougham with a circular front.
CHAPTER XIV.
GENERAL REMARKS ON THE COACH-

BUILDING TRADE.
It is pleasing to be able to refer to the increased skill and ingenuity of
the coach workmen, especially among the rising generation of
operatives. This fact was elicited by the recent exhibition of coaches,
&c. Not only were there shown several excellent working drawings of
carriages—drawn to scale and difficult of execution, and showing
that there are forthcoming more highly educated and more
competent men well acquainted with the details of their crafts, and of
the proper and scientific manner of setting out their work, now that
frequent change of construction renders this knowledge so desirable
—but there were also shown many clever models of proposed
improvements, the work of the ordinary carriage artisans, showing
that their originators were men of thought and energy. And no doubt
if more prizes were offered, and these exhibitions more frequent,
greater competition would be aroused, which would be the means of
bringing out a great deal of talent which at present lies dormant for
want of some inducement to call it forth.
The art of the coachmaker being an intricate one, inasmuch as he
has to combine in one harmonious whole a number of most varied
products—wood, iron, steel, brass, paint, silver, cloth, leather, silk,
ivory, hair, carpet, glass, &c., each worked by a separate trade, but
generally in one manufactory, and each of which may be spoilt or
injured by careless or improper treatment in any process—it behoves
all engaged in the production of carriages to work in harmony, that
their united labours may approach perfection. It would add much to
this desirable end if in each manufactory, large or small, were issued
a series of printed “general directions” for conducting the work; not
rigid rules that would, if strictly enforced, reduce men to mere
machines instead of free and intelligent operatives, but such as
would so guide each worker in the execution of his work as not only
to give satisfaction to his employer by its excellent and honest
execution, but bring equal credit and satisfaction to himself. This
state of feeling would be a very desirable one to bring about; it would
be the means of bringing about mutual respect between employer
and employed, and lead the way to a more cordial appreciation of
each other’s wants and difficulties; at the same time it would lessen
the incessant watchfulness and anxiety necessary to insure the work
being executed in such a manner that it may be depended upon for
accuracy and excellence when completed.
It is not so generally known as it should be, that in France,
Belgium, Germany, and some other European States, the training of
workmen and apprentices receives a great deal of attention, the
Governments in these countries considering money and trouble
bestowed on such objects to be of national importance. Technical
schools in these countries furnish instruction in drawing, modelling,
the harmonious arrangement of colours, the application of chemistry
to manufactures, metallurgy, and the proper working of metals, the
principles and applications of mathematics and mechanics to
manufactures, together with much that is strictly technical. In some
parts of Germany, before an employer of labour can commence
business on his own account, he must prove to competent persons,
by the execution of some trial work, that he understands what he
undertakes; and, moreover, that he has travelled for three years in
foreign countries, working at his trade, to acquire a knowledge of its
processes in other countries besides his own. There is doubtless
much pedantry in many of the regulations that interfere with the free
exercise of trade, but culling the best points of the system there is
much good that results. The training of apprentices in most trades in
England is very unsatisfactory, and were public attention directed to
the matter, after discussing the subject in its different bearings, there
might be some good general recommendations circulated relating to
the subject.
The carriages of America are so different from our own and from
those of Europe, that they require special attention. It is quite
possible that in the future their style may greatly influence carriages
in all parts. The first noticeable trait in them is lightness, and English
coach-builders generally agree that they carry this lightness too far,
more especially in their larger carriages. We are supported in this
view by the fact, that for some years, these—such as landaus,
broughams, and coaches—have been materially modified by
European types. The Americans have adopted some of the shapes
of Europe, and the European mode of constructing the under-
carriages, retaining their own method of making the pole and
splinters, as giving greater freedom to the horses.
This principle of allowing the horses greater freedom for action is
well worthy of the attention of coach-builders. The manner in which
our horses are confined by tight, heavy strapping and traces, by tight
pole chains, by bearing reins, and the indiscriminate use of blinkers
to the bridles, has been much overdone in England. If a horse with a
heavy load and driven fast over slippery roads should stumble, it is
most difficult for him to recover himself. He generally falls, and is
pushed along by the impetus of the carriage, and is more or less
injured in his limbs or nerves by the accident, while it is a matter of
great difficulty, if not impossibility, for him to rise again till the harness
be unstrapped and the carriage is removed from above him. Our
horses are also harnessed too closely to their work in two-wheeled
carriages. We have thought only of the ease of turning and moving
the vehicle in crowded or narrow ways, without observing the
advantage of long shafts over short shafts. If the shafts are
considered as levers, by which the horse supports and moves the
weight behind him in a two-wheeled cart, it will at once be obvious
that although (whilst those levers are parallel with the road) it does
not so much signify whether they are long or short, yet the moment
they cease to be parallel with the road, when they point upwards, or
more particularly when they point downwards, the difference
between long and short levers is severely felt by the horse. We can
all of us lift a weight or support a weight more easily with a long lever
than with a short one, and it is the same with a horse.
Those who have travelled abroad must have noticed the great
weights placed upon two-wheeled carts in France and Belgium, and
the greater comparative distance the horse is placed from the
wheels, and yet he carries his load easily enough, because he does
not feel its weight upon his back. Many English drivers seem to have
observed this, and try and ease the horse and lessen his chance of
stumbling by tipping the shafts up in front; but in this way the horse is
made to feel a pressure on the under part of his body, which
certainly will not improve his health. It is very probable that in future
years public opinion will be in favour of longer shafts and poles. This
will also tend to preserve good carriages from the damages they at
present suffer from the heat of the horses and the quantity of mud
which is thrown by their heels upon the front of the vehicle. The reins
will of course have to be longer, but this cannot be of much
consequence; the driver of a brougham is farther from the horse than
the driver of a mail phaeton, but it is not by any means true that the
brougham is any more difficult to drive than the phaeton on that
account.
There is another fashion prevalent in this country which is certainly
a fallacy, viz. the supposed necessity for the driver to sit nearly
upright, which necessitates a deep boot and a clumsy, thick
coachman’s cushion. In America, Russia, and parts of Germany, the
driver sits low, but places his foot against a bar in front of the
footboard; this in their carriages is longer than in ours. Four horses
can be driven very well and easily in a low landau, and very
powerful-pulling and fast-trotting horses held in with apparent ease.
Our coachmen are often in danger of being pulled over by their
horses, and certainly when an accident happens in a collision they
are easily thrown from the boxes. They do not have the purchase
and security that the Russian drivers seem to possess.
One of the greatest novelties introduced by the Americans into the
United States is the “buggy,” a name first given in England a hundred
years ago to a light two-wheeled cart, carrying one person only, and
which we now call a “sulky.”
The Americans have lavished all their ingenuity upon these
buggies, and they have arrived at a marvellous perfection of
lightness. They are hung upon two elliptical springs. The axles and
carriage timber have been reduced to mere thin sticks. The four
wheels are made so slender as to resemble a spider’s web. Instead
of the circumference of the wheel being composed of a number of
felloes, they consist of only two of oak or hickory wood, bent to the
shape by steam. The ironwork is very slender and yet composed of
many pieces, and in order to reduce the cost these pieces are mostly
cast, not forged, of a sort of iron less brittle than our cast iron. The
bodies are of light work like what we call cabinet work. The weight of
the whole vehicle is so small that one man can easily lift it upon its
wheels again if it should be accidentally upset, and two persons of
ordinary strength can raise it easily from the ground. The four wheels
are of nearly the same height, and the body is suspended centrally
between them. There are no futchells; the pole or shafts are
attached to the front axletree bed, and the front of the pole is carried
by the horses in just the same way that they carry the shafts. The
splinter-bar and whipple-trees are attracted to the pole on swivels.
Some are made with hoods and some without. The hoods are made
so that the leather of the sides can be taken off and rolled up, and
the back leather rolled, removed, or fixed at the bottom, a few inches
away from the back, the roof remaining as a sun-shade. The leather-
work is very thin and of beautifully supple enamelled leather.
The perfection to which this vehicle has been carried is certainly
wonderful; and every part that is weak or likely to give way is
carefully strengthened. If well made they last a long time without
repair. The whole is so slender that it “gives” and recovers at any
obstacle. The defect in these carriages in English eyes consists of
the difficulty of getting in or out, by reason of the height of the front
wheel and its proximity to the hind wheel. It is often necessary to
partly lock round the wheel to allow of easy entrance. There is also a
tremulous motion on a hard road that is not always agreeable. It is
not surprising, however, that with the great advantages of extreme
lightness, ease, and durability, and with lofty wheels, these vehicles
travel with facility over very rough roads, as there is a great demand
for them in our colonies. It must be remembered that the price is
small, much less than the price of our gigs and four-wheeled dog-
carts. This cheapness is attained by making large numbers to the
same pattern, by the use of cast iron clips, couplings, and stays, and
by using machinery in sawing, shaping, grooving, and mortising the
timbers, and by the educated dexterity of the American workman,
always ready to adopt any improvement. An educated man will make
a nimble workman, just as an educated man learns his drill from the
military instructor more quickly than a clown; and an educated man
finds out the value of machinery and desires to use and improve it.
Instead of fearing its rivalry he welcomes it; he remembers that all
tools, even the saw and the hammer, are machines, and that the
hand that guides these tools is but a perfect machine obeying the
guidance of the brain more quickly and in a more varied manner than
any man-made machine. The American workman, therefore, uses
machines more and more.
In England machinery for wood shaping is used at Derby,
Newcastle, Nottingham, Worcester, and other towns, and in Paris
some very good machinery is at work in coach factories. In London it
is chiefly confined to patent wheel factories, a few steam-driven
saws, patent mills worked by hand, and drilling and punching
machines. But until the use of machinery is more generally adopted
in London, it is probable that the trade of building carriages for export
will drift more and more to the provinces and the continent. The
saving effected by machinery in cab and omnibus building would be
great, because the patterns vary so little, and all the other parts of a
carriage would correspond with another, and counter-change when
repairs were needed.
The coach-builders of the future will look to steam and hand
machinery as their great assistance in cheapening the cost of first-
rate carriages, in multiplying them for the probable increased
demand, and also to build carriages more speedily. It now takes from
two to three months to build a brougham, of which at least five
weeks are consumed simply in the wood and ironwork, a period
which by the use of machinery might easily be shortened.
There has been much controversy about difference in the length of
the front and hind axletree. It has been usual to make no greater
difference than will allow the higher wheel to follow in the same track
as the lower wheel. In France, however, it has been the practice
since the year 1846 to make the front axletree of broughams 6
inches shorter than the hind ones. The object has been to allow the
front wheel to be placed nearer to the body. As the front wheel of a
brougham must turn entirely in front of the body, the additional gain
of 3 inches was very desirable. Some English coach-builders have
followed the example of the French. There is a decided gain. The
eye is pleased with the proportions, the horse is eased, and upon
hard roads the difference of track is of no consequence. On the other
hand, in the country roads, the well-worn ruts make the running of
the carriage uneasy, whilst in town the driver often forgets that the
curbstones will strike his hind wheels sooner than his front ones, and
also more mud is thrown upon the panels. Under these
circumstances it is very probable that the French plan will not find
universal favour.
If carriages had always to move along perfectly smooth roads
such as a tramway of wood, stone, or iron, the use of wheels in
overcoming friction would be their sole utility, and their height would
be of small consequence. But as carriages are drawn along roads
with loose stones and uneven surfaces, wheels are further useful in
mounting these obstacles, and it is plain that a high wheel does this
more easily than a low wheel. To demonstrate this, let us suppose a
shallow ditch or gulley of a foot wide and 2 inches deep, a wheel 2
feet high would sink into this and touch the bottom, but a wheel 3
feet high would only sink an inch, and a wheel 4 feet 6 inches high
would only sink half an inch (the wheels are supposed to cross the
above-mentioned gulley at right angles), on account of their greater
diameters. Consequently, while the large wheel would have to be
lifted by a force sufficient to raise it half an inch, a force will have to
be applied to the smaller wheel to raise it 2 inches, and under more
disadvantageous circumstances, because the spokes are in this
case the levers, and we know that the longer the lever the more
easily is the load raised.
That the leverage power of a high wheel is very great is shown by
the advantages gained by a large wheel in locomotives and bicycles.
There is an idea deeply rooted among coach-builders and coach-
buyers too, that the draught of a vehicle is diminished by placing the
front part of the carriage as far back as possible. Intelligent men who
have given the subject great attention, and tested the actual working
of this idea, say that it is a fallacy; but other intelligent men, who also
say they have tested its working, say that it would effect a great
saving in the draught if it were successfully accomplished. The
general idea amongst practical men is, that it would not be an
advantage. We have already seen that the draught of a vehicle with
large wheels is less than that of a vehicle with small wheels. If,
therefore, a load has to be placed on a four-wheeled vehicle, it
should be so placed in relation to the front and hind wheels that the
greater part of the weight should rest on the higher wheels. To obtain
this result, it is sufficient to bring the hind-carriage part as far under
the body as it will work with comfort and safety, in order that as little
weight as possible rests on the fore-carriage part. English
coachmakers have been working at this for thirty years, but for the
most part blindly; they have copied well known builders in
construction as well as shape; they hear that these well known firms’
carriages run and follow very lightly, and if they could copy
accurately they would obtain the same reputation.
But while there is some doubt as to throwing the front wheels
backward for the purpose of lessening the draught, Mr. Offord of
Wells Street, Oxford Street, has been exercising his ingenuity for the
purpose of throwing the back wheels further forward, and has
produced a brougham that offers peculiar advantages in this respect
(Fig. 56). The hind wheel appears to be placed right across the door,
but the facilities for ingress and egress are quite equal to those given
in the ordinary brougham. This novel contrivance presents nothing
singular in appearance, while a very little reflection will satisfy the
practical thinker that the advantage sought after, of lessening the
draught of the carriage, must be obtained far more completely with
this arrangement than by throwing the front wheel backward.
Fig. 56.—Single Brougham.
The rattling so constantly complained of in carriages can in a great
degree be obviated by placing pieces of india-rubber so that the
doors shall press upon them when closed; it is a good thing also to
have india-rubber at the bottom of the doors for the windows to drop
upon when let down.
The difficulty of protecting carriages from the dirt has recently
been met by placing what is called a “mud scraper” just at the back
of the hind wheels. It is formed of a piece of india-rubber about 3
inches square and a ¼ inch thick, held in position by a short iron rod
attached to the end of the hind spring.
The elaborate dress carriages, hung with braces upon a C and
under-spring perch carriage, require so much skill and practice in
their manufacture that it is impossible to give ample directions for
their construction in a work like this. Some years ago it was thought
that no carriage could be made comfortable unless it was hung upon
the old-fashioned perch carriage with braces; but it was found that by
the introduction of india-rubber, especially in the ends of each spring,
that what are termed elliptic spring carriages (which are of course
much lighter in draught and less in cost) can be made extremely
pleasant in motion.
Fig. 57 shows an iron-framed or “skeleton boot” for a landau. It is
extremely light and strong.
Fig. 57.
It is desirable to direct attention to the proper horsing of carriages,
that the owners of carriages and horses may so adapt their plans as
to get the most satisfactory result from their arrangements. Not
unfrequently a carriage is ordered for one horse only; when it is
partly made, or perhaps finished, fittings are ordered for two horses;
and it sometimes happens that the two horses put to the light one-
horse carriage are coach-horses, between sixteen and seventeen
hands in height. Such horses, though well adapted to a family
carriage, are quite out of their place attached to a light one. Although
they can draw it at a good pace, and over almost any obstacle in the
road, and do their journey without fatigue, the carriage suffers
sooner or later. The lounging of such horses against a light pole, the
strain thrown on the pole in case of a horse tripping, the certain
breakage that must occur in case of a fall, and the risk of overturning
the carriage, should all be considered before putting a very light
carriage behind very large horses. It also sometimes happens that
miniature broughams and other very small carriages, built as light
and as slight as safety will allow, are afterwards used with a pair of
horses. In such cases, if accidents do not occur through the great
strain of a long pole acting as a lever on very light mechanism, the
parts become strained, do not work as they were intended to do, and
necessitate constant repair from not being adapted to the work put
upon them. Carriage owners should, in their own interest, have their
carriages and horses suited to what they ought and can undergo,
bearing in mind that there are advantages and disadvantages both
with heavy and light carriages. The former are easier and more
comfortable to ride in; they are safer for horses, drivers, and riders;
and the necessary repairs are less frequently required. The lighter
carriages follow the horses more easily, and can therefore do a
longer day’s journey; and, although the necessary repairs may come
more frequently, the saving of the horses may be an advantage that
many persons will consider of the utmost importance. Such light
carriages should, however, be made of the choicest materials and
workmanship, that they may do the work required of them.
A feature in the financial department of coach-making must not be
overlooked, as it has much influence on an important trade. In former
times a large proportion of the carriages were built to order for the
owners; the reverse is now the case; most persons select a finished
carriage which pleases their taste, or an advanced one, and get it
completed their favourite colour. This, of course, necessitates the
employment of a larger capital to meet the altered state of trade,
which now requires so large a stock of carriages to be kept ready for
use.
The excessive competition of recent years has so reduced the
profit on each carriage, that in order to carry on his business without
loss, the builder has to require a prompt payment from his customer
instead of giving a long credit.
The modern system enables the coach-builder to make his
purchases for ready money, and so buy not only better in quality but
at a less cost than for extended credit, in order that he might in his
turn give long credit to his customers, so that he is now obliged to
depend on small profits and quick returns by turning over his capital
more rapidly. He is not now, as much as in former times, the agent of
the persons who supply the materials that he and his workmen
convert into a carriage, but rather the designer, capitalist, and
director of those who seek his service or custom, whether to supply
labour or materials.
From the Government returns we find that carriages of all sorts
have increased from 60,000, in 1814, to 432,600 in 1874—a benefit
to the general population, it is clear, as well as to the workmen. In
1874, 125,000 carriages paid the Government duty.
The valuable library and fine series of photographs of state and
other carriages of the Coach and Coach Harness Makers’ Company
are open to coach artisans every Saturday afternoon. Tickets of
admission may be obtained at the principal coach-builders in
London.
In Calcutta there are several coach-builders of good reputation,
and who employ large numbers of native workmen. Messrs. Dyke
employ 600 hands; Messrs. Stewart and Co., 400; and Messrs.
Eastman, 300. The men are chiefly Hindoos, and are clever and
industrious, but have a singular habit of sitting down to their work.
Owing to the prejudices of the people in regard to the use of animal
fat, the labourers who have to use grease are chiefly
Mahommedans. The wages in the trade vary from sixpence to two
shillings per day.
“In Hindostan” (says Mr. Thrupp) “there are a large number of
vehicles of native build. It has been frequently remarked that there is
little change in Eastern fashions, that tools and workmen are
precisely as they were a thousand years ago, and the work they
produce is precisely the same. In examining, therefore, what is now
done by Indian coach-builders, we are probably noticing carriages of
a similar, if not identical, sort with those in use three thousand years
ago. The commonest cart in Hindostan is called ‘hackery’ by
Europeans; it is on two wheels, with a high axletree bed and a long
platform, frequently made of two bamboos which join in front and
form the pole, to which two oxen are yoked; the whole length is
united by smaller pieces of bamboo, tied together, not nailed. In
France two hundred years ago there was a similar cart, but the main

Computational Aspects and Applications in Large Scale Networks NET 2017 Nizhny Novgorod Russia June 2017 Valery A. Kalyagin

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Computational Aspects and Applications in Large Scale Networks NET 2017 Nizhny Novgorod Russia June 2017 Valery A. Kalyagin

Uploaded by

Copyright:

Available Formats

Computational Aspects and

Applications in Large Scale Networks

Large Scale Scientific Computing 11th International

Advances in Computational Intelligence 14th

Advances in Neural Networks ISNN 2017 14th

Advances in Neural Networks ISNN 2017 14th

Algorithms for Computational Biology 4th International

Information Science and Applications 2017: ICISA 2017

Estimating Functional Connectivity and Topology in

Analysis of Images Social Networks and Texts 6th

Valery A. Kalyagin · Panos M. Pardalos

More information about this series at http://www.springer.com/series/10533

Oleg Prokopyev Irina Utkina

Panos M. Pardalos Irina Utkina

ISSN 2194-1009 ISSN 2194-1017 (electronic)

Mathematics Subject Classiﬁcation (2010): 05C82, 90B10, 90B15

© Springer International Publishing AG, part of Springer Nature 2018

distributions to test elliptical hypothesis for stock return distributions. Multiple

Nizhny Novgorod, Russia Valery A. Kalyagin

Part I Network Computational Algorithms

Part II Network Models

Part III Network Applications

Sentiment Analysis Using Deep Learning . . . . . . . . . . . . . . . . . . . . . . . . 281

Marina Ananyeva Higher School of Economics, National Research University,

Joshua Faskowitz Imaging Genetics Center, Stevens Neuroimaging and

Dmitry Petrov Kharkevich Institute for Information Transmission Problems,

M. A. Voronina Laboratory of Algorithms and Technologies for Network

Mikhail Batsyn, Ilya Bychkov, Larisa Komosko and Alexey Nikolaev

The Vehicle Routing Problem (VRP) is an NP-hard combinatorial optimization prob-

M. Batsyn (B) · I. Bychkov · L. Komosko · A. Nikolaev

V0 {0, . . . , n} is a set of vertices representing customers V {1, . . . , n} with

Algorithm 1. Tabu search

Algorithm 2. Updating objective function parameters

3.1 Initial Solution

3.2 Solution Improvement

4.1 VRPTW Results Comparison

4.2 Results for FSMVRPSTW

Fig. 55.—Double Brougham.

GENERAL REMARKS ON THE COACH-

You might also like