Metaheuristics For Resource Deployment Under Uncertainty in Complex Systems 1St Edition Shuxin Ding Online Ebook Texxtbook Full Chapter PDF

Metaheuristics for Resource
Deployment under Uncertainty in

Complex Systems 1st Edition Shuxin
Ding
Visit to download the full and correct content document:
https://ebookmeta.com/product/metaheuristics-for-resource-deployment-under-uncert
ainty-in-complex-systems-1st-edition-shuxin-ding/
More products digital (pdf, epub, mobi) instant
download maybe you interests ...
Optimal Operation of Integrated Multi-Energy Systems

Under Uncertainty 1st Edition Wu
https://ebookmeta.com/product/optimal-operation-of-integrated-
multi-energy-systems-under-uncertainty-1st-edition-wu/
Design of Deep Braced Excavation and Earth Retaining

Systems Under Complex Built Environment Theories and
Case Studies 1st Edition Zhang
https://ebookmeta.com/product/design-of-deep-braced-excavation-
and-earth-retaining-systems-under-complex-built-environment-
theories-and-case-studies-1st-edition-zhang/
The Science and Management of Uncertainty Dealing with

Doubt in Natural Resource Management 1st Edition Bruce
G. Marcot
https://ebookmeta.com/product/the-science-and-management-of-
uncertainty-dealing-with-doubt-in-natural-resource-
management-1st-edition-bruce-g-marcot/
Large Scale System Analysis Under Uncertainty With

Electric Power Applications 1st Edition Domínguez-
García
https://ebookmeta.com/product/large-scale-system-analysis-under-
uncertainty-with-electric-power-applications-1st-edition-
dominguez-garcia/
Methods of Mathematical Modelling and Computation for
Complex Systems Jagdev Singh
https://ebookmeta.com/product/methods-of-mathematical-modelling-
and-computation-for-complex-systems-jagdev-singh/
AI Metaheuristics for Information Security in Digital

Media 1st Edition Apoorva S. Shastri
https://ebookmeta.com/product/ai-metaheuristics-for-information-
security-in-digital-media-1st-edition-apoorva-s-shastri/
Metaheuristics for Finding Multiple Solutions 1st

Edition Mike Preuss
https://ebookmeta.com/product/metaheuristics-for-finding-
multiple-solutions-1st-edition-mike-preuss/
Defense Resource Planning under Uncertainty An

Application of Robust Decision Making to Munitions Mix
Planning 1st Edition Robert J Lempert Drake Warren Ryan
Henry Robert W Button Jonathan Klenk Kate Giglio
https://ebookmeta.com/product/defense-resource-planning-under-
uncertainty-an-application-of-robust-decision-making-to-
munitions-mix-planning-1st-edition-robert-j-lempert-drake-warren-
ryan-henry-robert-w-button-jonathan-klenk-kate/
Multiscale Dynamics Simulations Nano and Nano bio

Systems in Complex Environments 1st Edition Salahub
https://ebookmeta.com/product/multiscale-dynamics-simulations-
nano-and-nano-bio-systems-in-complex-environments-1st-edition-
salahub/
Metaheuristics for
Resource Deployment
under Uncertainty
in Complex Systems
Metaheuristics for
Resource Deployment
under Uncertainty
in Complex Systems
Shuxin Ding
Chen Chen
Qi Zhang
Bin Xin
Panos M. Pardalos
First edition published 2022
by CRC Press
6000 Broken Sound Parkway NW, Suite 300, Boca Raton, FL 33487-2742
and by CRC Press

2 Park Square, Milton Park, Abingdon, Oxon, OX14 4RN
© 2022 Shuxin Ding, Chen Chen, Qi Zhang, Bin Xin, Panos M. Pardalos
CRC Press is an imprint of Taylor & Francis Group, LLC
Reasonable efforts have been made to publish reliable data and information, but the author and
publisher cannot assume responsibility for the validity of all materials or the consequences of their use.
The authors and publishers have attempted to trace the copyright holders of all material reproduced
in this publication and apologize to copyright holders if permission to publish in this form has not
been obtained. If any copyright material has not been acknowledged please write and let us know so
we may rectify in any future reprint.
Except as permitted under U.S. Copyright Law, no part of this book may be reprinted, reproduced,
transmitted, or utilized in any form by any electronic, mechanical, or other means, now known or
hereafter invented, including photocopying, microfilming, and recording, or in any information
storage or retrieval system, without written permission from the publishers.
For permission to photocopy or use material electronically from this work, access www.copyright.com
or contact the Copyright Clearance Center, Inc. (CCC), 222 Rosewood Drive, Danvers, MA 01923,
978-750-8400. For works that are not available on CCC please contact mpkbookspermissions@tandf.
co.uk
Trademark notice: Product or corporate names may be trademarks or registered trademarks and are
used only for identification and explanation without intent to infringe.
Library of Congress Cataloging‑in‑Publication Data

Names: Ding, Shuxin, 1991- author. | Chen, Chen, 1982- author. | Zhang, Qi,
1968- author. | Xin, Bin, 1982- author. | Pardalos, Panos M., 1954-
author.
Title: Metaheuristics for resource deployment under uncertainty in complex
systems / Shuxin Ding, Chen Chen, Qi Zhang, Bin Xin, Panos M. Pardalos.
Description: First edition. | Boca Raton, FL : CRC Press, 2022. | Includes
bibliographical references and index. | Summary: “This book analyzes how
to set locations for the deployment of resources to incur the best
performance at the lowest cost. Resources can be static nodes and moving
nodes while services for a specific area or for customers can be
provided. Theories of modeling and solution techniques are used with
uncertainty taken into account and real-world applications used”--
Provided by publisher.
Identifiers: LCCN 2021010674 (print) | LCCN 2021010675 (ebook) | ISBN
9781032065205 (hardcover) | ISBN 9781032065243 (pbk.) | ISBN
9781003202653 (ebook)
Subjects: LCSH: Business logistics. | Metaheuristics.
Classification: LCC HD38.5 .D548 2022 (print) | LCC HD38.5 (ebook) | DDC
658.5/1015196--dc23
LC record available at https://lccn.loc.gov/2021010674
LC ebook record available at https://lccn.loc.gov/2021010675
ISBN: 978-1-032-06520-5 (hbk)

ISBN: 978-1-032-06524-3 (pbk)
ISBN: 978-1-003-20265-3 (ebk)
DOI: 10.1201/9781003202653
Typeset in LatinModern
by KnowledgeWorks Global Ltd.
Contents
Preface xi
Acknowledgments xv
Author Bios xvii
Chapter 1 Introduction 1
1.1 APPLICATIONS OF NODE DEPLOYMENT PROBLEM 1

1.1.1 Unmanned Systems 1
1.1.2 Wireless Sensor Networks 3
1.1.3 Healthcare 4
1.1.4 Public Sectors 6
1.1.5 Railway Network Design 6
1.1.6 Distributed Simulation Systems 8
1.2 FUNDAMENTAL ISSUES OF NODE DEPLOYMENT
PROBLEM 9
1.2.1 Task 10
1.2.2 Node 11
1.2.3 Environment 11
1.3 RESEARCH PROGRESS OF NODE DEPLOYMENT
MODELING 12
1.3.1 Deployment Space 12
1.3.1.1 Candidate Locations 12
1.3.1.2 Deployment Formation 14
1.3.2 Constraints 15
1.3.3 Objective Functions 16
v
vi Contents
1.3.3.1 Node Deployment in Wireless Sensor

Networks 16
1.3.3.2 Node Deployment in Air Defense 17
1.3.3.3 Other Types of Optimization Objective 20
1.4 RESEARCH PROGRESS OF NODE DEPLOYMENT
METHODS 20
1.4.1 Encoding 21
1.4.2 Constraints Handling 21
1.4.3 Multi-Objective Handling 21
1.4.4 Algorithms 22
1.4.4.1 Exact Algorithm 22
1.4.4.2 Metaheuristic Algorithm 23
1.5 MAIN ISSUES AND CHALLENGES 25
1.6 BOOK OUTLINE 27
Chapter 2 Stochastic Node Deployment for Area Coverage

Problem 29
2.1 INTRODUCTION 29
2.2 PROBLEM FORMULATION 31
2.2.1 Detection Models 31
2.2.1.1 Binary Detection Model 31
2.2.1.2 Probabilistic Detection Model 32
2.2.2 Network Model 32
2.2.3 Problem Statement 33
2.2.4 NP-Hardness Proof 34
2.3 SOLUTION ALGORITHMS 35
2.3.1 D-VFCPSO 35
2.3.2 Other PSO-Based Algorithm for Area Coverage
Problem 38
2.3.3 Complexity Analysis 39
2.4 EXPERIMENTS AND DISCUSSION 39
2.4.1 Test Instances 40
Contents vii
2.4.2 Parameter Setting 40

2.4.3 Analysis of Results 41
2.5 CONCLUSION 41
Chapter 3 Stochastic Dynamic Node Deployment for Target

Coverage Problem 45
3.1 INTRODUCTION 46
3.2.1 Mathematical Model 49
3.2.2 Scenario-Based Model Reformulation 49
3.3.1 NSGA-II 50
3.3.2 MOPSO 52
3.3.2.1 Personal Best Selection 52
3.3.2.2 Non-Dominated Solutions Maintaining
and Global Best Selection 53
3.3.2.3 Diversity Maintaining 53
3.3.3 Complexity Analysis 55
3.4.2 Performance Metrics 57
3.4.3 Parameter Turning 59
3.5 CONCLUSION 74
Chapter 4 Robust Node Deployment for Cooperative

Coverage Problem 75
4.1 INTRODUCTION 76
4.2.1 The Deterministic and Uncertain Two-Level
Cooperative Set Covering Problem 78
viii Contents
4.2.1.1 Two-Level Cooperative Set Covering

Problem 78
4.2.1.2 Generalized Uncertain Two-Level
4.2.2 Modeling the Robust Uncertain Two-Level
4.2.2.1 Compact Formulation of the
RUTLCSCP 87
4.3.1 Dealing with Subproblem 89
4.3.2 Rule-Based Heuristic for RUTLCSCP 92
4.3.2.1 Processing Procedure 94
4.3.2.2 Complexity Analysis of MRBCH-k 95
4.3.3 Proposed SaDE for RUTLCSCP 96
4.3.3.1 Encoding 97
4.3.3.2 Constraints Handling 97
4.3.3.3 Complexity Analysis of SaDE 100
4.4.2.1 Solving RUTLCSCP-LA-RC through
CPLEX 102
4.4.2.2 Comparisons of MRBCH-k with
Different k 105
4.4.2.3 Comparisons of SaDE and Its Variants 109
4.4.2.4 Comparisons on RUTLCSCP 109
4.5 CONCLUSION 113
Chapter 5 Fuzzy Node Deployment for Cooperative

Coverage Problem 115
5.1 INTRODUCTION 116

5.2.1 Fuzzy Conditional Value-at-Risk 118
Contents ix
5.2.2 Mathematical Model 118

5.2.3 Some Properties on CVaR-FTLCNDP 121
5.2.4 Linear Approximation of CVaR-FTLCNDP 122
5.3.1 Fuzzy Simulation 124
5.3.2 Improved Decomposition-Based Multi-Objective
Evolutionary Algorithms 126
5.3.2.1 Encoding 126
5.3.2.2 Updating of Individuals 127
5.3.2.3 Complexity Analysis 130
5.4.1 Performance Metrics 131
5.4.2.1 Case Study 1 131
5.4.2.2 Case Study 2 134
5.5 CONCLUSION 136
Chapter 6 Simulation-Based Evaluation Analysis of Node

Deployment under Risk Preference 139
6.1 INTRODUCTION 139

6.2 SIMULATION-BASED EVALUATION ANALYSIS OF
WORST-CASE CVAR NODE DEPLOYMENT 141
6.2.1 Uncertain Initial Position of Penetration Paths 142
6.2.2 Penetration Paths under Uncertainty 143
6.2.3 Scenario-Based Simulation 144
6.2.4 Evaluation Model with Decision Makers’ Risk
Preference 148
6.3.1 Case Study 1: Deployment of Sensor Nodes 148
6.3.2 Case Study 2: Deployment of Weapon Nodes 153
6.3.3 Case Study 3: Cooperative Deployment of
Sensor and Weapon Nodes 155
x Contents
6.4 CONCLUSION 164
Chapter 7 Overview and Future Directions 167
Bibliography 171
Index 189
Preface
Effective use of the resources at the best location is becoming more

important, bringing better services and lower costs. Optimizing the re-
source deployment problem is a critical issue in complex systems, which
involves how to substantially apply the limited resources to maximize
the systems’ ability. These resources can be unmanned aerial vehicles
(UAVs) and unmanned ground vehicles (UGVs) in unmanned systems
which provide continuous surveillance, reconnaissance, command, con-
trol, etc. They can also be static nodes, e.g., supermarkets, surveillance
cameras, etc. The resources are denoted as nodes in this book. In real-
world applications, the data obtained is uncertain. These data may be
stochastic, time-varying, unknown distribution, or fuzzy. These factors
should be considered when analyzing the resource deployment problem.
This book mainly focuses on the modeling and methods for solv-
ing the resource deployment problem under uncertainty. The models
are related to stochastic programming, robust optimization, fuzzy pro-
gramming, risk management, and single/multi-objective optimization.
Heuristic and metaheuristic algorithms will be applied to solve the pro-
posed problems. The resources are heterogeneous, which can be sensors
and actuators providing different tasks, e.g., sensing and attacking. As a
result, both separate coverage and cooperative coverage of the resources
are analyzed. The content of this book can be divided into six chapters
and summarized as follows:
Chapter 1: The applications, fundamental issues, literature review of
the modeling and methods, main issues, and challenges of the book are
presented.
Chapter 2: The stochastic node deployment problem for area cover-
age is considered. The detection probabilities of the sensors are under
uncertainty with known distribution. A perturbation mechanism is used
in the particle’s velocity updating equation, which prevents the particle
from local convergence. Comparative experiments show the advantage
of disturbance-virtual force co-evolutionary particle swarm optimization
(d-VFCPSO) in finding a good sensor deployment scheme.
xi
xii Preface
Chapter 3: The stochastic dynamic node deployment problem for

target coverage is investigated. The positions, the threat of the tar-
get, and the kill probabilities of the nodes are under uncertainty with
known distribution. It is a bi-objective problem, which minimizes the
threat of the targets and redeployment distance. Elitist learning strategy
(ELS) is used for mutation of the global best particles to keep diversity.
The Taguchi method with a novel response value is utilized to tune the
parameters of the proposed algorithms. Comparative experiments with
non-dominated sorting genetic algorithm with an elitist strategy (NSGA-
II) and other multi-objective particle swarm optimization (MOPSO) al-
gorithm show the effectiveness of the ELS.
Chapter 4: The robust node deployment problem for cooperative cov-
erage is discussed. The covering probabilities of the nodes are under un-
certainty with unknown distribution but a known range. Deployment
cost is used as the objective, while cooperative covering probability is
used as a constraint. It is assumed that there are two types of facilities
to be located for covering the demand node cooperatively. We proposed
a novel Two-Level Cooperative Set Covering Problem and its related
uncertain problems. A linear approximation method is proposed to deal
with the robust nonlinear constraints. The problem is solved by CPLEX
to obtain the linear approximation solution under relaxed constraints.
However, these solutions may violate the nonlinear constraints. There-
fore, a rule-based method is proposed to deal with the subproblem from
the robust nonlinear constraints. Besides, two marginal return values are
constructed by exploiting the objective and nonlinear constraints. Then,
a marginal-return-based constructive heuristic (MRBCH) method is de-
veloped by exploiting the two marginal return values. Meanwhile, an im-
proved self-adaptive differential evolution (SaDE) algorithm is also devel-
oped with four constraint-handling methods. Finally, a hybrid algorithm
is obtained by combining the constructive heuristic method MRBCH-1
with ordered repair operator SaDE-OR.
Chapter 5: We investigate the fuzzy node deployment problem for
cooperative coverage. To deal with the fuzzy uncertainty in the target
threat, we first proposed a method to calculate fuzzy conditional value-
at-risk (CVaR). What is more, we develop a fuzzy two-level cooperative
node deployment problem (CVaR-FTLCNDP), minimizing fuzzy CVaR
of target threat and deployment cost. Two decomposition-based multi-
objective evolutionary algorithms, MOEA/D and DMOEA-εC, are used.
Since the number of nodes is unknown a priori, the individual size varies
during the evolutionary process. Therefore, a variable individual size
Preface xiii
(VIS) encoding method is adopted, and MOEA/D-VIS and DMOEA-

εC-VIS are obtained. Experiments show the effectiveness of these two
algorithms, and MOEA/D-VIS shows better performance in diversity
and convergence. As a result, effective deployment schemes are obtained
at different costs.
Chapter 6: We study the simulation-based analysis of the node de-
ployment problem. The decision maker’s risk preference, the uncertainty
of the starting position, and the target’s penetration route are consid-
ered. To deal with the starting position’s uncertainty, a systematic sam-
pling method is adopted to obtain the scenarios of different starting
positions. As for the uncertainty of the target’s penetration route, the
Dijkstra algorithm is used for the target optimal penetration route plan-
ning to avoid detection by sensors or being shot by weapons. Worst-case
CVaR is used to describe the risk of the deployment scheme, and its
confidence level is used to describe the decision maker’s risk preference.
Experiments show that the proposed evaluation method has successfully
demonstrated the incoming target’s uncertainty and evaluated deploy-
ment schemes under different risk preferences.
Chapter 7: The overview and future directions are presented.
The models and approaches proposed in this book can be applied in
complex systems, e.g., unmanned systems, sensor networks, etc. Besides,
they can also be used in some facility location-related problems, e.g.,
railway network design, hospital layout optimization, locations problems
in emergency medical services (EMS), etc.
Shuxin Ding
Chen Chen
Qi Zhang
Bin Xin
Panos M. Pardalos
Acknowledgments
We wish to thank all the researchers of the School of Automation, Beijing

Institute of Technology, the State Key Laboratory of Intelligent Con-
trol and Decision of Complex Systems, and the Signal and Communica-
tion Research Institute, China Academy of Railway Sciences Corporation
Limited for reading and commenting on draft versions of the book. We
wish to acknowledge the support of the National Natural Science Foun-
dation of China under Grants U1834211, U1934220, 62022015, 62088101,
61773066, 61822304, 61673058, and 61790575 and the National Key R&D
Program of China (2018YFB1308000). Panos M. Pardalos was supported
by a Humboldt Research Award (Germany).
Most of all, Shuxin Ding would like to express his greatest gratitude
to his parents for their support, without which this book would not have
been finished.
We would appreciate any comments, questions, criticisms, or correc-
tions of this book. Readers may kindly provide to Dr. Shuxin Ding at
dingshuxin@rails.cn.
xv
Author Bios
Shuxin Ding received the B.E. degree in automation and the Ph.D. de-
gree in control science and engineering from the Beijing Institute of Tech-
nology, Beijing, China, in 2012 and 2019, respectively. He is currently
an assistant researcher with the Signal and Communication Research
Institute, China Academy of Railway Sciences Corporation Limited. His
current research interests include railway scheduling, evolutionary com-
putation, multi-objective optimization, and optimization under uncer-
tainty.
Chen Chen received the B.S. degree in automation and the Ph.D. de-
gree in control science and engineering from the Beijing Institute of
Technology, Beijing, China, in 2004 and 2009, respectively. She is cur-
rently a professor with the School of Automation, Beijing Institute of
Technology. Her current research interests include complicated systems,
multi-objective optimization, and distributed simulation.
Qi Zhang received the B.S. degree in electronic engineering from Bei-

jing University of Aeronautics and Astronautics, Beijing, China, in 1991,
the M.S degrees in computer applications, and the Ph.D. degree in traf-
fic information engineering and control from China Academy of Railway
Sciences, Beijing, in 1993 and 1998, respectively. He is currently a chief
researcher of China Academy of Railway Sciences Corporation Limited
and a leader in railway technical expertise. His research interests include
railway signal and communication, automatic train operation, train oper-
ation control, intelligent dispatching, and cooperative control of multiple
trains. Dr. Zhang was a recipient of the Youth Award, the Contributional
Award, and the Achievement Award of Zhan Tianyou Railway Science
and Technology.
Bin Xin received the B.S. degree in Information Engineering and Ph.D.
degree in Control Science and engineering, both from the Beijing Insti-
tute of Technology, Beijing, China, in 2004 and 2012, respectively. He
xvii
xviii Author Bios
is currently a professor with the School of Automation, Beijing Insti-

tute of Technology. His current research interests include search and
optimization, evolutionary computation, unmanned systems, and multi-
agent systems.
Panos M. Pardalos received the B.S. degree in mathematics from

Athens University, Athens, Greece, in 1977, the M.S. degree in mathe-
matics and computer science from Clarkson University, Potsdam, NY,
USA, in 1978, and the Ph.D. degree in computer and information sci-
ences from the University of Minnesota, Minneapolis, MN, USA, in 1985.
Dr. Pardalos is a Distinguished Professor in the Department of Industrial
and Systems Engineering at the University of Florida, Florida, USA, and
an affiliated faculty of the Biomedical Engineering and Computer Sci-
ence & Information & Engineering departments. In addition, he is the
director of the Center for Applied Optimization. Dr. Pardalos is a world-
renowned leader in Global Optimization, Mathematical Modeling, En-
ergy Systems, and Data Sciences. He is a Fellow of AAAS, AIMBE, and
INFORMS and was awarded the 2013 Constantin Caratheodory Prize of
the International Society of Global Optimization. In addition, Dr. Parda-
los has been awarded the 2013 EURO Gold Medal prize bestowed by the
Association for European Operational Research Societies. This medal
is the preeminent European award given to Operations Research (OR)
professionals for “scientific contributions that stand the test of time.”
Dr. Pardalos has been awarded a prestigious Humboldt Research Award
(2018–2019). The Humboldt Research Award is granted in recognition
of a researcher’s entire achievements to date—fundamental discoveries,
new theories, and insights that have had significant impact on their dis-
cipline. Dr. Pardalos is also a Member of several Academies of Sciences,
and he holds several honorary Ph.D. degrees and affiliations. He is the
Founding Editor of Optimization Letters and Energy Systems, and Co-
Founder of the International Journal of Global Optimization, Computa-
tional Management Science, and Springer Nature Operations Research
Forum. He has published over 500 journal papers and edited/authored
over 200 books. He is one of the most cited authors and has graduated
65 Ph.D. students so far.
CHAPTER 1
Introduction
Resource deployment problem analyzes how to set the locations for de-
ploying resources with the best performance and lowest cost. The re-
sources can be static nodes and moving nodes. These resources can
provide services for a specific area or some customers. This chapter in-
troduces the resource deployment problem. It provides some real-world
applications and fundamental issues of node deployment problem, and
research progress of node deployment modeling and methods. This chap-
ter provides a basic foundation for the whole work and gives the main
issue and challenges that lead to the following chapters.
1.1 APPLICATIONS OF NODE DEPLOYMENT PROBLEM

In this section, we outline a number of different applications related
to the node deployment problem. They can be complex systems, e.g.,
unmanned systems, sensor networks, etc. Besides, there are many ap-
plications in facility location problems, e.g., healthcare, public sectors,
railway network design, etc. There also exists resource deployment prob-
lems in distributed simulation systems.
1.1.1 Unmanned Systems

Unmanned systems are man-made and can be operated or managed
through advanced technologies. They are complex systems created by the
fusion of various technologies related to mechanics, control, computer,
communication, and materials [1]. They can be controlled by humans
or perform tasks autonomously. Various types of unmanned systems are
emerging include unmanned aerial (UAV), ground (UGV), and under-
water (UUV) vehicles, etc. They may be applied in applications from
DOI: 10.1201/9781003202653-1 1
2 Metaheuristics for Resource Deployment
:DWHU
Figure 1.1 UAVs, UGVs, and UUVs in an unmanned system.
the civil domain to the military domain, such as logistics, surveillance,

building and environment monitoring, search and rescue, intruder detec-
tion and attacking, etc. [2]. Thus, one objective might be to minimize
the number of vehicles for completing the tasks, and the other objective
might be to maximize the payoff of the tasks [3]. Figure 1.1 shows the
UAVs, UGVs, and UUVs in an unmanned system. A brief introduction
and example applications of unmanned systems are presented as follows.
• Video surveillance. Camera-mounted UAVs can provide coverage of
multiple-oriented targets. The positions of these UAVs are decided
by the ground control station with a master camera. Except for
surveillance [4] and crowd monitoring [5], they can be used for
infrastructure inspections [6], cinematography [7], etc.
• Networks. The UAVs can help formulate coverage when there are
disturbances and disruptions in the cellular networks caused by
concerts, natural disasters, etc. [8]. Problems such as minimizing
the number of UAVs required for continuous coverage, maximizing
the area coverage, and preserving network connectivity require an
optimized deployment strategy [2].
The vehicles may also perform different functional roles. Four main
functional roles are defined as sensors, actuators, decision makers (DMs),
and auxiliary facilities [9, 10]. For example, in UAV-UGV coordination
systems, the characteristics of UAVs and UGVs are strongly complemen-
tary. The combination of different functional roles makes cooperative
Introduction 3
UAV-UGV coordination systems promising. UGVs can act as actuators

limited by their speed and environmental occlusion, while UAVs can act
as sensors quickly deployed for finding targets. UAVs can also help in
formulating communication links for UGVs, which may be blocked by
obstacles. Besides, small-scale UAVs are restricted by their short voyage
due to the energy limitation, while UGVs can act as carriers providing
auxiliary facilities. This also suggests that we would like to make better
use of the systems by optimizing the deployment of the vehicles.
1.1.2 Wireless Sensor Networks

Wireless sensor networks (WSNs) are formed by small, inexpensive, low-
powered sensors. WSNs have recently become a popular research area for
many applications in military, environmental, industrial, home, medical,
etc. [11]. The objective is to monitor the environment and communi-
cate information with each sensor. Figure 1.2 shows a target detection
scenario in the three-dimensional (3D) space of WSNs. Decision makers
need to decide the numbers and locations for the WSNs. Some per-
formance metrics to be optimized in WSNs are introduced in Section
1.3.3.1. A brief introduction and example applications of wireless sensor
networks are presented as follows.
• Military applications. In command, control, communications, com-

puters, intelligence, surveillance, and reconnaissance (C4ISR) sys-
tems, WSNs can be rapidly scattered in critical terrains, routes
to provide battlefield intelligence. They can be used to detect and
track enemy targets.
• Environmental applications. Environmental problem is a critical

issue for human on the earth. WSNs can be used for wildlife mon-
itoring, fire detection, measuring CO2 level, flood detection, air
pollution detection, etc. [12].
• Industrial applications. WSNs can be used in manufacturing pro-

cess management, monitoring the gas, water, and electric, lighting
control, etc. They can also be used for monitoring the structural
health of buildings, bridges, roads, physical condition of water and
gas pipes, smart railway stations, etc.
• Home applications. WSNs in home environments connect every-

day objects and devices at home through networks and create an
Figure 1.2 Wireless sensor networks in 3D space for target detection.
Internet of Things (IoT)-based smart environment [13]. It is an en-

vironment that learns from our daily activity. People can conduct
remote control of the home devices through IoT-based smart home
systems [14].
• Medical applications. The Body Area Sensor Network (BASN) con-

sists of multiple interconnected nodes for sensing, data processing,
and wireless communication. These sensor nodes are placed on,
near, or within the human body. The BASN sensor nodes con-
stantly monitor and analyze different physiological signals, e.g.,
the electrical activities of the heart, muscles, and the brain; body
temperature, blood glucose, blood pressure, blood oxygen satura-
tion, etc. [15].
1.1.3 Healthcare
Healthcare is defined as the prevention and treatment for illness or injury
through professional medical services. The facility location problems re-
lated to healthcare covers from locating healthcare facilities to layout
problems in hospitals [16]. Figure 1.3 shows the locations of historical
cardiac arrests and candidate sites for AED deployment. A brief intro-
duction and example applications of healthcare are presented as follows.
Introduction 5
+LVWRULFDOFDUGLDFDUUHVWV
&DQGLGDWHVLWHV
Figure 1.3
Locations of historical cardiac arrests and candidate sites for
AED deployment.
• Healthcare facility location. The facility location problems for

healthcare are mainly related to healthcare facilities, e.g., com-
munity health clinics, public and private hospitals, etc. The opti-
mization criteria for healthcare facilities are minimizing access cost
for healthcare consumers, maximizing population with access, etc.
• AED location. Optimizing the deployment of public automated

external defibrillators (AEDs) can help to increase the probability
of survival when sudden cardiac arrest occurs [17, 18].
• Ambulance location. Ambulance location belongs to the emergency

vehicles sitting problems. The goal is to find the locations for the
ambulances (or ambulance bases) with a minimal number and pro-
vide a certain level of service. Meanwhile, relocation decisions for
ambulances should be periodically made to avoid areas unpro-
tected [19].
• Hospital layout planning. The layout planning problems for hospi-

tals aim at minimizing in-house travel distances or costs inside the
building. It is classified as a resource capacity planning problem,
which directly influences the quality and efficiency of healthcare,
and patient satisfaction.
1.1.4 Public Sectors

For public sectors, the main feature is the optimization criteria of the
decision makers compared with private sectors. Public sectors like gov-
ernment organizations aim at maximizing the provided services to people
with given resources, while private sectors aim at maximizing their profit
as well as minimizing their cost [20]. Possible applications include the
location of schools, bike sharing systems, hospitals, and bus stops [21].
Figure 1.4 shows the locations of sharing bakes in a bike sharing sys-
tem. A brief introduction and example applications of public sectors are
presented as follows.
• Locating schools. Schools are located to meet the students’ de-

mand. Some optimization criteria include student travel distance,
fixed and overhead school costs, the average number of students
enrolled per school, etc.
• Bike sharing systems. Bike sharing systems are designed to in-

crease the use of bicycles, to decrease congestion, and to provide a
service to people available to other means of transportation [22].
It consists of bike stations with several bike slots. A bike station
should have empty slots for arrivals and full slots for departures.
The bike sharing systems company would like to maximize the to-
tal profit related to the revenue from bike renting, cost of moving
empty bikes, cost of maintenance, etc. The number and location of
bicycle stations and the capacity of each station, and the number
of bikes in the system should be optimized.
• Locating electric vehicle charging stations. Electric vehicles (EVs)

are eco-friendly transport modes to reduce carbon emissions. The
need for EV charging or refueling stations increases with the devel-
opments in the EV industry. Some optimization criteria are max-
imizing the coverage, maximizing the traffic flow, and minimizing
the costs when optimizing the number, location of stations, and
the scale of each station.
1.1.5 Railway Network Design

Decisions involved in railway planning belong to different stages: strate-
gic, tactical, operational, and real-time. The first step in the railway plan-
ning process is the network design followed by line planning, timetabling,
Introduction 7
&KHFNRXW
DELNH
&KHFNLQ
DELNH
%LNHVKDULQJ VWDWLRQV
Figure 1.4 An illustration of a bike sharing system.
rolling-stock circulation and maintenance, crew scheduling, and real-time

management [23]. These problems can be analyzed separately or together
with two or more by iterative/integrated approach. Figure 1.5 shows a
railway network with lines and high-speed railway stations. A brief intro-
duction and example applications of the railway are presented as follows.
• Railway rapid transit network design. This problem consists of the

selection of nodes and links to formulate a potential or underly-
ing railway network with stations and links between them. Some
optimization criteria include maximizing the expected trip cover-
age and the total population covered of the planned network. The
existing network may also be extended by adding new lines or ex-
tending existing ones. The robustness of the network is also an
important factor in case of a disturbance or disruption. Complex
network theory may be applied in the assessment of transportation
networks [24].
• Location of Stations. Locating stations on a network from scratch
is influenced by the attraction to large volume of passengers. There
are also some problems dealing with locating new stations on ex-
isting railway lines. Besides, when extreme weather events cause
damage to the trains, proper locations of the railway emergency
rescue stations may help in saving rescue time [25].
• Line planning. This problem is the next task in the railway plan-
ning process after the network design. The origin, itinerary, and

+LJKVSHHGUDLOZD\VWDWLRQV

5DLOZD\OLQHV
Figure 1.5 An illustration of a high-speed railway network.
destination of each line, with the frequency of the train, are to be

decided with a given railway network. Cost related to the operat-
ing company and travel time related to the users are concerned for
line planning [23].
1.1.6 Distributed Simulation Systems

With the advantages of high utilization ratio, high scalability, and paral-
lelism, distributed systems can be run parallel to handle computationally
intensive problems. Besides, several difficult problems are derived from
the distributed structure, such as the communication problems among
the nodes, the mapping and assignment problems of system resources
and computing nodes. High level architecture (HLA) is widely used for
distributed simulation development. Although HLA supports large-scale
distributed simulation, the load balancing mechanism in the simulation
process is not standardized. When the simulation system runs under
non-optimized allocation, load imbalance problems will occur due to the
shortage of computing resources, which will seriously affect the simula-
tion efficiency and confidence.
The main purpose of load balancing is to even out the computation
load distribution among the nodes and minimize the communication
Introduction 9
Figure 1.6 Simulation process in a distributed simulation system.
load among them. The results of load balancing are shown in the fol-
lowing aspects: shortening the average response time of tasks effectively,
maximizing the use of simulation resources of the whole system, and
reducing the unnecessary waste of resources [26]. During the process of
simulation, additional federates or components will be added to complete
specific simulation tasks. As shown in Figure 1.6, when the simulation
process is advanced to step k, a certain number of federates or compo-
nents must be added. The key to solving this dynamic load balancing
problem is transferring dynamic states to static states, converting the
problems into static load balancing problems.
1.2 FUNDAMENTAL ISSUES OF NODE DEPLOYMENT PROB-

LEM
With the development of computer technology, communication tech-
nology, network technology, and other emerging technologies, there has
been a significant impact on optimizing resource deployment. Optimizing
the resource deployment problem is a critical issue in complex systems,
which involves applying the limited resource to maximize the systems’
ability. These resources can be UAVs and UGVs in unmanned systems,
which provide continuous surveillance, reconnaissance, command, con-

trol, etc. [9,10,27]. They can also be static nodes, e.g., surveillance cam-
eras, supermarkets, etc. The resources are denoted as nodes in this book.
Traditional decision-making methods for resource deployment are not
suitable for the large-scale, dynamic, and uncertain characteristics of the
system. The network connects the resources in an area as a united and
efficient system, which shares the condition of the resources and the
environment [28]. As a result, it decreases the total time consumed for
decision-making and increases the command speed.
Node deployment is affected and restricted by many factors in a com-
plex environment. Traditional deployment methods mainly rely on expe-
rience; thus, the nodes’ characteristics and environmental conditions are
not precisely described. As a result, the result is not ideal. The node de-
ployment problem in this book mainly studies how to determine the posi-
tion and number of nodes under the environment’s constraints, the total
number of nodes, etc., to obtain the deployment’s optimal performance.
Compared with the traditional experience deployment, it provides a rea-
sonable and effective deployment scheme. Optimizing the resource de-
ployment can be regarded as a resource-space matching problem.
In the following, the key elements of node deployment problem are
introduced, which are task, node, and environment.
1.2.1 Task
Optimizing the resource deployment problem is a critical issue in com-
plex systems. Heterogeneous resources, e.g., sensors and actuators, can
work cooperatively through the network, improving the system’s perfor-
mance. It has been applied in the network-centric systems with sensors
and weapon systems, which significantly improve the performance in
air defense systems [29]. In unmanned systems, both UAVs and UGVs
can act as sensor nodes or actuator nodes. Sensor nodes usually per-
form monitoring and surveillance tasks. Actuator nodes perform some
control tasks, e.g., weapon nodes perform attacking towards the enemy
targets. For example, by deploying UAVs and UGVs, surveillance tasks
on a specific area and attacking tasks on the targets within the area are
conducted. The sensor nodes’ environmental information and warning
information can be transmitted to the actuator (weapon) nodes through
the network. Under the guidance of sensor nodes, weapon nodes could
perform effective attacking.
Introduction 11
Many research works and reviews focused on deploying sensor nodes,

especially the wireless sensor networks (WSNs) [30–37]. Most of these
studies analyze the maximum sensor coverage satisfying the connectivity
and energy consumption. For weapon nodes, characteristics of the nodes
and the tasks should be considered [38]. Since there are fewer works
on actuator nodes, most of the studies will concentrate on deploying
actuator (weapon) nodes in this section.
Some research only focuses on deciding the number of nodes to be
deployed in some regions but neglects the specific position of the nodes
[39]. Therefore, this kind of problem can be regarded as an assignment
problem of nodes, assigning different numbers of nodes for regions with
different demands. Scholars have widely discussed assignment problems
in recent decades. For example, generalized assignment problem (GAP)
[40], resource assignment problem (RAP) [41], weapon target assignment
(WTA) [42], etc. Other research focuses on deciding the nodes’ positions,
which are more common and widely studied [43–47].
1.2.2 Node
In real-world applications, there are usually different kinds of nodes
deployed in a complex system. These nodes are deployed to comple-
ment each other’s advantages. However, these nodes are homogeneous
resources, which are belong to the same type, i.e., sensor node or actu-
ator node. Few studies consider the joint deployment of heterogeneous
resources, e.g., sensor and actuator nodes. Therefore, the cooperation
between heterogeneous resources and rational utilization of them needs
further investigation.
1.2.3 Environment
The node deployment problem can also be divided into static deploy-
ment and dynamic deployment. Most of the previous studies analyze
the static deployment problem. Nodes are deployed before any possible
related tasks. It is an off-line node deployment problem without any time
limit. However, when the states of the environment and nodes change
during the service, the deployment’s performance may not meet the fu-
ture requirements. Therefore, the nodes’ location should be adjusted by
redeployment to meet the future requirement.
Besides, most of the studies are deterministic, which means the pa-
rameters of the environment, resources, and targets are determined val-
ues. However, uncertainties always exist in real-world applications. For

example, some nodes are not reliable after deployment [44], the targets
may have uncertain route [48,49], etc. Therefore, it is necessary to model
and effectively deal with uncertainty.
As a result, previous research only deals with one resource type and
considers static and deterministic problems. Therefore, cooperative cov-
erage with heterogeneous resources and the resources’ uncertain and
dynamic properties should be considered. It is of great significance to
study this kind of problems both in theory and in practice.
1.3 RESEARCH PROGRESS OF NODE DEPLOYMENT MOD-

ELING
Node deployment problem brings a number of questions:
1. How many nodes should be deployed?
2. Where should the nodes be deployed?
3. How to select the type of the nodes?
The answers to this question depend on how the model is formulated
and the method is developed. To solve the node deployment problem,
we need to analyze the problem by mathematical modeling first. Three
factors are related to studying the model: deployment space, constraints,
and objective functions. The research progress on modeling is stated as
follows.
1.3.1 Deployment Space

Deployment space is the area for locating nodes with different amounts
and positions. Figure 1.7 shows the classification of deployment space
based on the candidate locations and deployment formation.
1.3.1.1 Candidate Locations

In order to obtain the candidate locations, the deployment space can
be classified into two types: discrete space-based and continuous space-
based. In discrete space-based deployment, different discrete points in
an area are regarded as candidate locations. A common way to obtain
the discrete points is by dividing the area by grids. As a result, grid
points are generated and regarded as both candidate locations and sam-
pling points for evaluating the coverage. The grid size should be selected
Introduction 13
Figure 1.7 Classification of deployment space.
(a) Square-shaped grid. (b) Ring-shaped grid. (c) Sector-shaped grid.
Figure 1.8 Methods for obtaining grid points.
appropriately based on the demand. More locations represent more cal-

culation, but the result will be more precise. There are several methods
to obtain grid points in the discrete space-based deployment. Most are
divided by squares and sectors shown in Figure 1.8. Except for grid
points, predefined positions can also be regarded as candidate locations.
Ref. [50] considers squared grids for node deployment in a square
area shown in Figure 1.8(a). The whole area is uniformly divided by
grids and covered by nodes. Refs. [39,47] adopt sectors to divide a circle
area shown in Figure 1.8(b). Ref. [51] considers a sector area for deploy-
ment shown in Figure 1.8(c). For the circle area and sector area, they are
divided by sectors from the original point. The original point is also re-
garded as a protected object. Therefore, more gird points are close to the
protected object for selecting locations. This non-uniform discretization
method can effectively reduce the unnecessary search space and make
the search more efficient. Besides squared grids and sectors, other dis-
cretization methods such as the triangular grid and hexagonal grid are
usually adopted in coverage problems in WSNs [36].
Since the discretization of the space will cause missing some deploy-
ment information inevitably, we should carefully select the size of the
grid. While in the continuous space-based deployment, nodes can be
deployed at any place in the predefined region, which ensures that the
optimal solution must be in the search space. However, the searching
cost of this method is higher than the discrete space-based deployment.
In Refs. [43, 44], the virtual force algorithm is adopted to deal with the
issues in the continuous space-based deployment. The total forces from
other nodes, protected objects, and boundaries are used to adjust the
locations of the nodes. It prevents the nodes from congesting in a local
area and keeps the nodes in a certain density with avoidance of blind
area.
1.3.1.2 Deployment Formation

Deployment formation refers to the form of the nodes in the deployment
space. Generally, the formations of the deployed nodes can be divided
into four types: ring deployment, sector deployment, line deployment,
and hybrid deployment [52].
In ring deployment, the nodes are deployed around some key points:
one ring, two rings, and multiple rings. So the targets from any direction
can be covered by nodes. Its advantage is to strengthen the coverage for
the main direction, but also has the omnidirectional capability. Therefore
the reliability and the efficiency are high. However, the number of nodes
required under the same conditions is greater. In Refs. [39, 47], grids
are used for ring deployment. Although it is continuous space-based de-
ployment in Refs. [43, 44], considering the targets are coming from all
directions, the final deployment schemes are still in ring deployment for-
mation.
In sector deployment, only the main directions are considered. The
deployed nodes will form like a sector. It is mainly used when the num-
ber of nodes is insufficient and terrain constraints cannot form a ring
deployment. The advantage is to save the number of nodes, but the re-
liability is not high. If the target changes the direction, the deployment
is easy to lose effectiveness.
In line deployment, the nodes are deployed along with the key points
Introduction 15
as a line, with a large width of interception front, but a small depth. It is

suitable for the situation where the target comes in many directions, and
the front is wide. This deployment is always seen in the barrier coverage
problem.
These three types of deployment formations are relatively typical,
but the actual deployment scenarios are complicated. Only one type of
formation cannot deal with the actual complex environment. The ring
deployment, sector deployment, and line deployment should be selected
according to the actual terrain conditions. Two or three formations are
combined to form a hybrid deployment formation, which is also called
group deployment [53].
1.3.2 Constraints
The resource deployment problem is a constrained optimization problem.
The constraints limit the range of solutions and ensure that the plan-
ning results are reasonable and effective. In the real-world deployment
problem, the feasible solution is limited by the deployment space, and
accurate geographic information is analyzed to determine the deploy-
able and non-deployable areas. For example, lakes, depressions, forests,
or locations where the slope is too large or the terrain is too low are not
suitable for deploying nodes. It is often difficult to mathematically de-
scribe non-deployable areas (especially irregular non-deployable areas).
Such constraints will be dealt with in a specific manner during the op-
timization process. Except for the constraints in the deployment space,
other features and demands of the nodes should be considered.
The number of the nodes is the first consideration when deploying
nodes. The number of available nodes in different areas is set as con-
straints. The nodes’ cost can also be a constraint since some of the
nodes may be very expensive. In Ref. [51], the coverage demands for
some important directions are set as constraints. Besides, the minimum
distance between nodes is set as a constraint to prevent electromagnetic
interference. Meanwhile, the distance between nodes and the protected
objects should satisfy some requirements. These are typical constraints
in the case of static node deployment.
In dynamic node deployment, the region for deployment changes ac-
cording to the nodes and targets. As a result, the corresponding de-
ployment space and constraints should be adjusted. In Ref. [44], some
deployed nodes are malfunctioning, and the corresponding redeployment
problem needs to be solved based on the updated constraints.
1.3.3 Objective Functions

The mathematical model of node deployment is presented as follows.
min [f1 (x), f2 (x), ..., fi (x)] (1.1)

s.t. gk (x) ≤ 0, k = 1, . . . , l (1.2)
hk (x) = 0, k = l + 1, . . . , p (1.3)
x∈X (1.4)
where X ⊆ Rn denotes the set of all feasible solutions. fi (x) denotes the
ith performance metrics for evaluating the deployment. hk (x) denotes
the kth constraint for deployment space requirement. gk (x) denotes the
kth other constraint. x denotes the decision variable, which is a vector
x = [x1 , x2 , ..., xn ]. In the discrete space-based deployment, the candi-
date location for deployment is denoted as p = [p1 , p2 , ..., pm ], where xi
denotes the ith node assigned with the jth location p. In the continuous
space-based deployment, xi denotes the position of the ith node.
The critical issue is how to describe the performance of the nodes
in different locations. Therefore, the objective function can be used to
evaluate the performance of the deployment. There are many different
types of node deployment problems, which involve different optimiza-
tion goals. We mainly consider the objective functions in WSNs and air
defense.
1.3.3.1 Node Deployment in Wireless Sensor Networks

In this subsection, an overview of the optimization objectives in WSNs is
provided. There are many performance metrics to be optimized in WSNs.
For example, coverage, connectivity, lifetime, and energy consumption
are important concerns for maintaining the quality-of-service (QoS) in
WSNs [37]. Meanwhile, these metrics usually conflict with each other.
For example, more energy consumed reduces the lifetime of the network.
• Coverage. In WSNs, coverage is the most widely used metric to

evaluate the QoS. WSN coverage can be classified into three types:
area coverage, target (point) coverage, and barrier coverage. In the
area coverage problem, the entire region needs to be covered. In
the target coverage problem, a finite set of discrete points are to
be covered. The barrier coverage problem deals with the detec-
tion of movements across a barrier of sensors [36]. The area cov-
erage problem is the most studied in WSNs. Two main problems
Introduction 17
related to area coverage are achieving satisfactory coverage and

maximum coverage, both NP-hard problems [36, 37]. Satisfactory
coverage means finding the minimum number of sensors that any
point in the covered region of interest (RoI) can be detected with
a probability that is equal or exceed a predefined threshold. An-
other method to describe satisfactory coverage is k-coverage, which
means any point in the RoI is within the sensing range of at least
k different sensor nodes [54]. Maximum coverage means to find the
optimal locations of a set of nodes to achieve maximum coverage.
• Network connectivity. Connectivity requires that the location of

any deployed node should within the communication range of one
or more active nodes, so that all active nodes can form a connected
communication network. Two sensor nodes are directly connected
if the distance of them is smaller than the communication range Rc .
The communication range Rc should be greater than the sensing
range Rs to achieve network connectivity. For a set of nodes that
cover a convex region, the network remains connected if Rc ≥ 2Rc
[37].
• Network lifetime. Lifetime is an important metric in WSNs. The

energy source of the sensor nodes is limited by the battery. It is also
crucial to maximize the lifetime of the sensor nodes with limited
resources. The networks’ lifetime is calculated by the time duration
from the activation of the network to the time when any sensor
node fails by depleted of energy.
• Energy consumption. Since the sensor nodes are equipped with

limited battery power, the energy consumption problem in WSN
should be taken into consideration. The total energy consumption
is calculated by the sum of the energy expended at the node along
the path for data acquisition, processing, and transmission phases.
1.3.3.2 Node Deployment in Air Defense

This subsection provides an overview of the optimization objectives in
air defense. The strength and depth of the coverage provided by the
deployed weapons show the ability in air defense. The strength refers to
multiple coverages, and the depth refers to the covered area. Multiple
factors should be considered for deployment evaluation. The main factors
are the deployed and protected resources and the targets for air defense.
The ability of the deployed resources and the condition of the protected
resources should be analyzed. Meanwhile, the moving direction, position,
and number of targets should also be considered. Similar to the area
coverage and target coverage in WSNs, two types of the node deployment
problem in air defense are usually studied, which are area air defense and
point air defense.
• Area air defense.
Area air defense represents that the entire area needs to be pro-
tected by the deployed nodes (e.g., weapon nodes) [55]. Since the
resources are limited compared with the protected area, it is nec-
essary to select several important sampling points for protected re-
sources to evaluate the deployment performance. In Ref. [55], some
important protected resources are chosen from the entire area and
are used for evaluation. The selection for the protected resources is
similar to the discretization method in candidate locations. Both
uniform and non-uniform sampling points are used for calculation.
To maximize the protection of these resources is belongs to the
coverage problem. The strength of the coverage is more suitable
for performance evaluation in area air defense.
Some studies consider the set covering problem (SCP) and maxi-
mal covering location problem (MCLP) to formulate this problem
and proposed a 0-1 programming problem [20, 56]. SCP minimizes
the deployment cost with satisfying coverage, while MCLP maxi-
mizes the coverage of the targets with limited resources. Ref. [57]
proposes firing range covering and firing angle covering based on
SCP. In Ref. [39], ship sector locations are optimized to provide
a robust air defense formation as the sector allocation problem
(SAP). The problem is formulated as a 0-1 integer linear program-
ming (ILP) problem, and the coverage provided to the ships is
maximized. This is a classic MCLP. Since there are always multiple
covering situations for covering the protected resources in the area
air defense, an aggregate signal function can be used to describe
the cooperative coverage1 [58]. The followings are two methods to
deal with it.
– Add up the coverage capabilities of different resources directly

[39].
1
The deployed nodes belong to the same type with different capabilities. It is
different with the idea in Section 1.5, where heterogeneous nodes are considered.
Introduction 19
– A certain probability is introduced to describe the resource’s

coverage capability. Multiple coverages consider calculating
the joint probability, which is the maximum expected cover-
age location problem (MECLP) [20].
Most of the studies consider joint probability to describe the co-

operative coverage rate of the deployment [43, 44, 50, 55, 59]. The
performances of the node deployment in WSNs and air defense
are evaluated by calculating the joint probability of the sam-
pling points. If a sampling point is within the sensing (attacking)
range of the sensor (weapon) nodes, the joint probability by sensor
(weapon) nodes can be calculated as follows.
Yn
P =1− (1 − pk ) (1.5)
k=1
where pk denotes the sensing (killing) probability of the kth node.

n denotes the total number of the nodes. P denotes the joint prob-
ability, which is the sensing (killing) probability from all n nodes.
In the area air defense, most studies of only consider the protected
resources and do not consider the threat target. Therefore, the
entire region for air defense is sampled by grids as protected re-
sources. However, this coverage evaluation method can only reflect
the coverage capability of deployment but cannot directly reflect
the effect of deployment in real combat. In Ref. [39], both the size
and the direction of the threat are considered and discretized as
sampling points as a MECLP.
Refs. [60, 61] describe the threat target by queuing theory. In de-
ploying weapon nodes in a multilayer defense, the incoming target
in the defense line is regarded as a random service system. The
target flow entering the defense area is regarded as Poisson flow.
The shooting efficiency of the defense is obtained by calculating
the killing probability of each layer of defense and the probabil-
ity of enemy penetration. Maximizing the defense capability of
the weapon nodes is the evaluation metric. This queuing theory
method does not consider the sampling of the entire defense area
and avoids the calculation difficulty caused by sampling.
• Point air defense.

In point air defense, there exist some critical resources for protec-
tion. The formation of weapon nodes should have enough strength
and depth for protection. The coverage strength and coverage

depth are two conflicting objectives. A deployment with too much
coverage strength will have less coverage depth. Point air defense
formations are usually circular or sector-shaped to ensure a certain
depth of coverage.
Ref. [62] proposes an objective function related to flight distance in
the defense area based on entropy. When the entropy is maximized,
the distance difference among all directions is the smallest. It en-
sures the uniform protection ability of air defense in all directions.
In Ref. [51], the coverage of some critical directions should be satis-
fied, and other directions should be balanced. The sampling point
can be determined by calculating the number and position of a sin-
gle target when it passes through the attacking range. We can max-
imize the killing probability (minimize the penetrating probability)
of the enemy target from different directions [43, 44]. Besides, the
queuing theory method can also describe the process of the enemy
target entering the defense area [47]. Besides, some other models
are proposed, e.g., minimum regret model [63], chance-constraint
programming model, dependent-chance programming model [64],
and mean-entropy model [65].
1.3.3.3 Other Types of Optimization Objective

Besides, some studies consider the multi-objective problem (MOP) and
dynamic optimization. In MOP, the coverage and the deployment cost
are two conflicting objective functions usually considered together. When
the environment changes, the previous deployment scheme may not be
optimal. Redeployment cost is another objective function that needs to
be considered. Since it is a real-time redeployment process, the comput-
ing time for solving the redeployment problem should meet the require-
ment.
1.4 RESEARCH PROGRESS OF NODE DEPLOYMENT METH-

ODS
As node deployment becomes complicated, the manual method to deter-
mine the locations may not achieve a satisfying result, thus using com-
puter algorithms in solving the node deployment is necessary. Different
methods are proposed to deal with the node deployment problem. The
Introduction 21
solution of deployment location involves encoding, constraint handling,

multi-objective optimization handling, and algorithms.
1.4.1 Encoding
Encoding is applied to describe the locations for the deployed nodes.
It is the decision variable and related to the deployment space. There
are differences between discrete space-based deployment and continuous
space-based deployment. In discrete space-based deployment, the deci-
sion variable is usually set as an integer representing whether a resource
has been deployed. Branch-and-bound [39,66] and genetic algorithm [55]
are applied to solve the problem. In continuous space-based deployment,
any position in the deployment space can be the location for deployment.
This encoding mechanism is usually adopted in the problem solving by
particle swarm optimization [32] and virtual force algorithm [43].
1.4.2 Constraints Handling

The complexity of the constraint increases the difficulty of the problem.
Therefore, constraints should be handled to improve the efficiency of the
algorithm. There are various techniques to deal with the constraints [67].
The constraints can be used for generating feasible solutions, e.g., repair
algorithms. It can replace an infeasible solution with a feasible one. An-
other popular technique is the penalty functions. This method aims to
transform a constrained optimization problem to an unconstrained one
by adding a penalty item based on the amount of constraint violation in
a particular solution [50]. However, the selection of the penalty function
may affect the efficiency of the algorithm since there are multiple con-
straints [44]. Therefore, an artificial potential field algorithm has been
applied to deal with the geographical and coverage constraints and ob-
tain deployment schemes [43, 44].
1.4.3 Multi-Objective Handling

The problem is mainly transformed into a single-objective optimization
problem or tackled by multi-objective evolutionary algorithms to deal
with the node deployment with multiple objective functions. Several
methods are presented as follows.
• Weighting method. The objective functions are combined as a single

objective function by weights.
• ε-constraint method. We optimize one of the objective functions,

and other objective functions are transformed into constraints. It
is better than the weighing method since different weights may
obtain the same solution. On the other hand, we may obtain a
different solution with the ε-constraint.
• Stratified sequencing method. Multiple objective functions are

ranked by different levels and solved according to the rank. In
[68], network coverage rate, energy consumption in redeployment,
and bi-connected network topology are considered objective func-
tions in a sensor redeployment problem. A distributed algorithm
based on the virtual forces is applied to solve the constraint of bi-
connected features. The network coverage is maximized first, and
then the energy consumption is minimized.
• Others. Multi-objective evolutionary algorithms (MOEAs) have

been applied to solve the MOP. In MOP, the objective functions
are conflicting, e.g., the coverage and the cost. The solution in the
obtained Pareto solutions set may be superior to others in one
objective function and inferior in another objective function. Be-
sides, analytic hierarchy process (AHP) [69], technique for order
preference by similarity to an ideal solution (TOPSIS) [70], expert
systems [69], and risk analysis [71] are applied to deal with the
multiple objective functions.
1.4.4 Algorithms
Exact algorithms are applied in node deployment, e.g., enumeration
method [72], branch-and-bound [39], etc. Besides, since node deploy-
ment is belongs to NP-hard problem, metaheuristic algorithms, e.g., ge-
netic algorithm [55,59], particle swarm optimization algorithm [50], sim-
ulated annealing algorithm [73], etc. are also applied. These algorithms
are nature-inspired and have a stochastic searching mechanism.
1.4.4.1 Exact Algorithm

The exact algorithm provides a precise and repeatable solution. In Ref.
[39], a branch-and-bound approach is proposed to deal with the sec-
tor allocation problem based on a linear MCLP. Tight lower and upper
bounds are proposed to speed up computation time better than CPLEX.
Ref. [66] proposes a new reformulation for the generalized partial cov-
Introduction 23
ering problem with lower bounding strategies. The problem is solved

by the branch-and-bound approach. In Ref. [46], the nonlinear objective
functions in MECLP are decomposed into linear summation by setting
the maximum covering times and then solved by using LINGO. The
number of nodes and candidate locations are small in these problems.
As the size of the problem increases, the problem will become hard to
solve with a larger searching space.
1.4.4.2 Metaheuristic Algorithm

In metaheuristic algorithms or intelligent optimization algorithms,
stochastic variables are introduced during the optimization process. The
obtained solutions are not optimal and not the same in each run. It is
a population-based algorithm. The best parameters of the algorithms
can be selected appropriately. The metaheuristic algorithms used in the
node deployment problem are as follows.
• Simulated annealing (SA). SA is based on the annealing in metal-

lurgy. It can tackle tough computational optimization problems
with nonlinear, discontinuous, and non-differentiable functions,
where exact algorithms fail. It can achieve an approximate so-
lution to the global optimum by controlling the cooling schedule.
Refs. [45, 73] adopt simulated annealing to deal with the node de-
ployment problem in air defense.
• Genetic algorithm (GA). GA simulates the natural selection of

biological evolution and the biological evolution process in genetics
for the optimal solution [74]. In Ref. [59], GA is proposed with a
new generation method of the initial population. Ref. [75] proposes
the parallel gene combination genetic algorithm to solve the air
defense deployment problem. The initial population is generated
by problem-specific knowledge and random selection. This method
overcomes the phenomenon of non-unique encoding and has a good
performance on the convergence time and solution quality. Refs.
[64, 65] use GA to solve the multilayer node deployment for air
defense in fuzzy environment.
• Particle swarm optimization (PSO). PSO is widely used in recent

decades [76]. The global best and personal best particles are used
for updating the position. It is easy to be applied in node de-
ployment problem [50]. However, the PSO algorithm will easily
converge to local minima and cause premature convergence. This

algorithm needs to be improved according to the corresponding
problem. Ding et al. [32] proposed a disturbance PSO (d-PSO)
with a Gaussian perturbation to update the velocity, which shows
fast convergence. Parallel particle swarm optimization (PPSO),
which divides the sensing area and the sensors equally into sev-
eral parts, and it is used when there are large numbers of sen-
sors to be deployed [77]. Thus, the searching space is partitioned,
and the computation time is saved. Soleimanzadeh et al. [78] pro-
posed a PSO-LA (learning automata) algorithm, and the velocity
is changed by using learning automata. Tang et al. [79] estab-
lished a three-dimensional sensor network model and provided an
improved PSO algorithm. Wang et al. [80] proposed a resampled
PSO (RPSO) to maximize the coverage and energy efficiency of the
WSN. In Ref. [81], the Voronoi diagram is combined with PSO to
obtain the best coverage. In this algorithm, PSO is used to find an
optimal position, while the Voronoi diagram is used for evaluating
the fitness of the solution.
• Artificial potential field (APF) algorithm. APF algorithm is also
called virtual force algorithm (VFA), which considers the attrac-
tive and repulsive force between nodes for optimal deployment [82].
The positions of the nodes will converge under the force from VFA.
Chen et al. [43] used virtual force between nodes, environment,
and obstacles to generate solutions and use GA to optimize the
parameters in APF under different deployment conditions. Zhou
et al. [83] proposed two adaptive distribution algorithms based on
virtual force to tackle the different obstacles and moving obstacles
in WSN. VFA has also been combined with other metaheuristics.
Wang et al. [30] presented an improved co-evolutionary PSO algo-
rithm that combines virtual force and PSO with a co-evolutionary
mechanism to solve the dynamic sensor deployment problem. Liang
et al. [84] used a virtual force-based coverage algorithm to achieve
area coverage. Four different forces caused by neighbor sensors
and uncovered regions are exerted on directional sensors. Binh et
al. [85] combined the heuristic initialization and modified virtual
force algorithm with both GA and PSO to deal with the maxi-
mization of obstacles constrained area coverage problem.
• Memetic algorithm (MA). MA is an algorithm framework com-
bining global search and local search. This hybrid algorithm can
Introduction 25
be a unique one with different search strategies [86]. Therefore,

it is easy to develop an improved version of it by this mechanism.
Chen et al. [51] used GA for global search and neighborhood search
for local search in sector-shaped point air defense problem. Liu et
al. [47] used GA for global search and hill-climbing algorithm for
local search in ring-shaped point air defense problem. Li et al. [87]
used PSO for global search and SA for local search in naval fleet
formation problem.
• Multi-objective evolutionary algorithm (MOEA). MOEA is a

population-based evaluation process to obtain Pareto solutions.
Wu et al. [88] used a non-dominated sorting genetic algorithm with
an elitist strategy (NSGA-II) to minimize the deployment cost and
penetrating probability. Liu et al. [89] proposed a group divided
dimensional reduction algorithm with strength Pareto evolution-
ary algorithm (SPEA2) in sector-shaped point air defense problem.
Zhang et al. [90] proposed a problem specific decomposition-based
MOEA for barrier coverage with wireless sensors considering the
power, reliability, and fairness of the deployment. A 2-tuple encod-
ing scheme is adopted, and a cover-shrink algorithm is proposed
to produce feasible and relatively optimal solutions. Besides, local
search is performed with problem-specific knowledge.
The above analysis shows that metaheuristics are commonly used

in node deployment problems. Many researchers propose improved and
hybrid intelligent optimization algorithm to tackle the problem. Besides,
most of the current algorithms are based on static and deterministic
problems. There are lacking algorithms for dynamic and uncertain node
deployment problems.
1.5 MAIN ISSUES AND CHALLENGES

Based on the above analysis, the following will be the main issues and
challenges in the node deployment problem.
• Node deployment with multi-objective functions. In order to deal

with the multi-objective node deployment problem, different meth-
ods are considered based on the decision maker’s (DM’s) prefer-
ences, which are listed as follows [91].
– A priori methods: The global preference information of the

problem is provided by the DM. Thus, weighted methods can
be applied.
– A posteriori methods: An approximation of the Pareto front is
obtained, and the most preferred one is selected. ε-constraint
method and MOEA can be adopted.
– Interactive methods: The DM specifies preferences progres-
sively during the solution process to guide the search towards
his/her preferred regions. This can be a promising research
direction that needs further investigation.
• Node deployment under uncertainty. In node deployment model-
ing, there are uncertainties with the number and state of the nodes
and targets. However, most current works are deterministic. Ac-
cording to the features of the parameters, it can be divided into
accurate information, information subject to a certain probabil-
ity distribution, information whose distribution is unknown but
belongs to a certain interval, fuzzy information, and real-time dy-
namic information. The latter four kinds of information types can
be formulated as follows.
– Stochastic node deployment: The parameters in this problem

are uncertain, but their statistic distribution can be obtained
by historical data. The expectation operator is usually used
to transfer the original problem into a deterministic one.
– Robust node deployment: The distributions of the parame-
ters in this problem are unknown but discrete or continuous.
The discrete events can be described by scenarios, while the
continuous ones can be described by intervals. The optimal
deployment scheme under different conditions has significant
limitations, so it is necessary to design a robust deployment
method. The deployment scheme provided by this method
has a good deployment effect for various complex situations.
Min-max, min-max regret operator, etc. are commonly used
to transfer the original problem into a deterministic one.
– Fuzzy node deployment: The parameters in this problem are
uncertain under certain fuzzy distributions, e.g., trapezoidal
fuzzy distribution, triangular fuzzy distribution, etc. Fuzzy
simulation [92] is usually used to transfer the original problem
into a deterministic one.
Another random document with
no related content on Scribd:
manufacture of textile machinery he added that of general machinery
and large tools for cutting, boring, rifling, planing and slotting. He had
a great reputation in his day, but his work seems to have been more
that of a builder of standard tools than an originator of new tools and
methods.
Charles Holtzapffel, another well-known engineer of that
generation, was the son of a German mechanic who came to
London in 1787. He received a good education, theoretical as well
as practical, and became a skilled mechanician and a tool builder of
wide influence. His principal book, “Turning and Mechanical
Manipulation,” published in 1843 in three volumes, is an admirable
piece of work. Covering a field much wider than its title indicates, it is
the fullest and best statement of the art at that time; and scattered
through it there is a large amount of very reliable mechanical history.
By 1840 the number of men engaged in tool building was
increasing rapidly, and it is impossible to consider many English tool
builders who were well known and who did valuable work, such as
Lewis of Manchester, B. Hick & Son of Bolton, and others. One
noteworthy man, however, ought to be mentioned—John George
Bodmer, who was neither an Englishman, nor, primarily, a tool
builder.[71] He was a Swiss who worked in Baden and Austria, as
well as in England, and his fertile ingenuity covered so many fields
that a list of the subjects covered by his patents occupy six pages in
the “Transactions of the Institution of Civil Engineers.”
[71] For a “Memoir” of Bodmer see “Transactions of the Institution of Civil
Engineers,” Vol. XXVIII, p. 573. London, 1868.
Bodmer was born at Zurich in 1786. After serving his

apprenticeship he opened a small shop for millwright work near that
city. A year or so later he formed a partnership with Baron d’Eichthal
and with workmen brought from St. Etienne, France, he started a
factory in an old convent at St. Blaise, in the Black Forest, first for
the manufacture of textile machinery and later, in 1806, of small
arms.
“Instead of confining himself to the ordinary process of gun-making
by manual labour, Mr. Bodmer invented and successfully applied a
series of special machines by which the various parts—more
especially those of the lock—were shaped and prepared for
immediate use, so as to insure perfect uniformity and to economise
labour. Amongst these machines there was also a planing machine
on a small scale; and Mr. Bodmer has been heard to observe how
strange it was that it should not have occurred to him to produce a
larger machine of the same kind, with a view to its use for general
purposes.”[72] He does not seem to have used the process of milling
until much later. Bodmer was thus among the first to discern and to
realize many of the possibilities of interchangeable manufacture, Eli
Whitney having begun the manufacture of firearms on the
interchangeable basis at New Haven, Conn., about 1800, only a few
years before. Why Bodmer’s attempt should have failed of the
influence which Whitney’s had is not quite clear. A possible
explanation may lie in the fact that the use of limit gauges does not
seem to have been a part of Bodmer’s plan. This use was
recognized by the American gun makers as an essential element in
the interchangeable system almost from the start.
[72] Ibid., p. 576. (The italics are ours.)
Bodmer was appointed, by the Grand Duke of Baden, director of

the iron works and military inspector with the rank of captain and for
a number of years much of his energy was given to the development
of small arms and field artillery. He invented and built a 12-pound
breech-loading cannon in 1814, which he had tested by the French
artillery officers. It failed to satisfy them, and was sent a few years
later to England, where it was decently buried by the Board of
Ordnance.
The following year he built a flour-mill at Zurich for his brother.
Instead of each set of stones being driven by a small waterwheel, all
the machinery connected with the mill was driven by a single large
wheel through mill gearing. The millstones were arranged in groups
of four. “Each set could be started and stopped separately, and was
besides furnished with a contrivance for accurately adjusting the
distance between the top and bottom stones. By means of a hoist of
simple construction, consisting in fact only of a large and broad-
flanged strap-pulley and a rope-drum, both mounted on the same
spindle (the latter being hinged at one end, so that it could be raised
and lowered by means of a rope), the sacks of grain or flour could be
made to ascend and to descend at pleasure, and the operatives
themselves could pass from one floor to any other by simply
tightening and releasing the rope.[73] The shafting of this mill was
made of wrought iron, and the wheels, pulleys, hangers, pedestals,
frames, &c., of cast iron, much in accordance with modern
practice.”[74] This was several years before Fairbairn and Lillie began
their improvements at Manchester.
[73] Apparently the modern belt conveyor.
[74] “Memoir,” p. 579.
Bodmer went to England for the first time in 1816 and visited all
the principal machine shops, textile mills and iron works. He returned
in 1824 and again in 1833, this time remaining many years. On his
second trip he established a small factory for the manufacture of
textile machinery at Bolton, in which was one of the first, if not the
first, traveling crane.[75] At the beginning of his last and long
residence in England, Bodmer appointed Sharp, Roberts &
Company makers of his improved cotton machinery, which they also
undertook to recommend and introduce. This arrangement was not
successful, and a few years later, in partnership with Mr. H. H. Birley,
Bodmer started a machine shop and foundry in Manchester for
building machinery.
[75] Ibid., p. 581.
Nearly all of the machinery for the Manchester plant was designed
and built by Bodmer himself and it forms the subject of two
remarkable patents, granted, one in 1839 and the other in 1841.[76]
The two patents cover in reality nearly forty distinct inventions in
machinery and tools “for cutting, planing, turning, drilling, and rolling
metal,” and “screwing stocks, taps and dies, and certain other tools.”
“Gradually, nearly the whole of these tools were actually constructed
and set to work. The small lathes, the large lathes, and the planing,
drilling, and slotting machines were systematically arranged in rows,
according to a carefully-prepared plan; the large lathes being
provided, overhead, with small traveling cranes, fitted with pulley-
blocks, for the purpose of enabling the workmen more economically
and conveniently to set the articles to be operated upon in the lathes,
and to remove them after being finished. Small cranes were also
erected in sufficient numbers within easy reach of the planing
machines, &c., besides which several lines of rails traversed the
shop from end to end for the easy conveyance on trucks of the parts
of machinery to be operated upon.”[77] There were, in addition to
these, however, “a large radial boring machine and a wheel-cutting
machine capable of taking in wheels of 15 feet in diameter, and of
splendid workmanship, especially in regard to the dividing wheel,
and a number of useful break or gap-lathes, were also constructed
and used with advantage. It is especially necessary to mention a
number of small, 6-inch, screwing lathes, which, by means of a
treadle acting upon the driving gear overhead, and a double slide-
rest—one of the tools moving into cut as the other was withdrawn,—
screw cutting could uninterruptedly proceed both in the forward and
in the backward motion of the toolslide, and therefore a given
amount of work accomplished in half the time which it would occupy
by the use of the ordinary means. Some of the slide-lathes were also
arranged for taking simultaneously a roughing and finishing cut.”[78]
[76] The first of these is described in the American Machinist of March
13, 1902, p. 369.
[77] “Memoir,” p. 588.
[78] Ibid., p. 597-598.
The latter part of Bodmer’s life was spent in and near Vienna,
working on engines and boilers, beet sugar machinery and
ordnance; and at Zurich, where he died in 1864, in his seventy-ninth
year.
Bodmer does not seem to have originated any new types of
machine tools, with the exception of the vertical boring-mill, which he
clearly describes, terming it a “circular planer.” It was little used in
England, and has been considered an American development.
It is hard now to determine how far Bodmer has influenced tool
design. It was much, anyway. Speaking of the patent just referred to,
John Richards, who has himself done so much for tool design, says,
“Here was the beginning of the practice that endured.” He has
described some of Bodmer’s tools in a series of articles which show
a standard of design greatly in advance of the practice of his time.[79]
Another writer says of Bodmer, “He seems always to have
thoroughly understood the problems he undertook to solve.” “One is
lost in admiration at the versatility of the inventive genius which could
at any one time—and that so early in the history of machine design
—evolve such excellent conceptions of what was needed in so many
branches of the mechanics’ art.”[80]
[79] American Machinist, Vol. XXII, pp. 352, 379, 402, 430, 457, 478,
507, 531, 559, 586, 607, 637.
[80] Ibid., Vol. XXV, p. 369.
Bodmer was elected a member of the Institution of Civil Engineers

in 1835, and his standing among his contemporaries is shown by the
fact that thirty-five pages in the “Transactions” of the Institution for
1868 are given to his memoir. For a foreigner to have won respect
and distinction in the fields of textile machinery, machine tools and
steam engines in England, where all three originated, was surely
“carrying coals to Newcastle.” Not only did he succeed in these
fields, but he invented the traveling crane, the chain grate for boilers,
the Meyer type of cut-off valve gear, the rolling of locomotive tires,
and introduced the system of diametral pitch, which was long known
as the “Manchester pitch,” from its having originated in his plant at
Manchester.
Though Bodmer was never regularly engaged in the building of
machine tools, his contribution to that field is far too great to be
forgotten.
CHAPTER VIII
JAMES NASMYTH
We know more of the life of Nasmyth than of any of the other tool
builders. Not only did Smiles give an account of him in “Industrial
Biography,”[81] but fortunately Nasmyth was induced in later life to
write his recollections, which were published in the form of an
autobiography, edited by Smiles.[82] With the exception of Sir William
Fairbairn, he is the only great engineer who has done this. His
intimate knowledge of the rise of tool building, the distinguished part
he himself had in it, and his keen and generous appreciation of
others, make his record valuable. We have already quoted him in
connection with Maudslay, and wherever possible will let him tell his
own story.
[81] “Industrial Biography,” Chap. XV. Boston, 1864.
[82] “James Nasmyth, Engineer, An Autobiography,” edited by Samuel
Smiles. London, 1883.
Unlike most of the early mechanics, James Nasmyth came from a

family of distinction dating from the thirteenth century. They lost their
property in the wars of the Covenanters and his direct ancestors took
refuge in Edinburgh, leaving their impress on the city as the
architects and builders of many of its most famous and beautiful
buildings. Alexander Nasmyth, the father of James, was a well-
known artist, the founder of the Scotch School of Landscape
Painting, and a friend of Burns, Raeburn and Sir Walter Scott. He
was a landscape architect and enough of an engineer to be included
in Walker’s engraving of “The Eminent Men of Science Living in
1807-1808,” reproduced in Fig. 8. He invented the “bow-string” truss
in 1794, the first one of which was erected over a deep ravine in the
island of St. Helena, and also the setting of rivets by pressure
instead of hammering. This last, by the way, was the result of trying
to do a surreptitious job on Sunday without outraging the fearsome
Scotch “Sawbath.” Alexander Nasmyth was one of the six men on
the first trip made on Dalswinton Loch, October 14, 1788, by the
steamboat built by Symington for Patrick Miller. This was the second
trip of a steam-propelled vessel, the first one being that of John Fitch
on the Delaware, August 22, 1787. It was an iron boat with double
hulls and made about five miles an hour. It barely escaped being the
first iron vessel, as Wilkinson’s iron boat on the Severn was
launched less than a year before. The picture of this trial trip which
has come down to us was made by Alexander Nasmyth at the
time.[83]
[83] Ibid., pp. 28-31.
James Nasmyth was born in 1808, the tenth in a family of eleven

children. Like all of his brothers and sisters, he inherited his father’s
artistic tastes. If he had not been an engineer he would probably
have become distinguished as an artist. He was ambidextrous, and
to the end of his life his skill with his pencil was a constant source of
pleasure and convenience. The notebook in which the later record of
his mathematical ideas is contained, is crowded with funny little
sketches, landscapes, little devils and whimsical figures running in
and out among the calculations. The leaf in this book on which he
made his first memorandum of the steam hammer is shown in Fig.
23. In 1817, Watt, then in his eighty-first year, visited Edinburgh and
was entertained at the Earl of Buchan’s, where Alexander Nasmyth
met him at dinner. Watt delighted all with his kindly talk, and
astonished them with the extent and profundity of his information.
The following day Watt visited Nasmyth to examine his artistic and
other works. James Nasmyth, a nine-year-old boy, returning from
school, met him at the doorstep as he was leaving, and never forgot
the tall, bent figure of “the Great Engineer.”
Figure 22. James Nasmyth
from an etching by paul rajon
Nasmyth’s father had a private workshop which was well equipped
for those days. Nasmyth played there from childhood and had
mastered the use of all the tools while still a schoolboy. “By means of
my father’s excellent foot lathe,” he says, “I turned out spinning tops
in capital style, so much so that I became quite noted amongst my
school companions. They would give any price for them. The peeries
were turned with perfect accuracy, and the steel shod, or spinning
pivot, was centered so as to correspond with the heaviest diameter
at the top. They could spin twice as long as the bought peeries.
When at full speed they would ‘sleep,’ that is, turn round without a
particle of waving. This was considered high art as regarded top-
spinning.”[84] He established a brisk business in these, in small brass
cannon, and especially in large cellar keys, which he converted into
a sort of hand cannon, with a small touch-hole bored into the barrel
and a sliding brass collar which allowed them to be loaded, primed,
and then carried around in the pocket.
[84] Ibid., p. 89.
He haunted all the shops and foundries in the neighborhood,

making friends with the skilled workmen and absorbing the mysteries
of foundry work, forging, hardening and tempering, and those arts
which were handed down from man to man. Speaking of Patterson’s
old shop, Nasmyth says: “To me it was the most instructive school of
practical mechanics. Although I was only about thirteen at the time, I
used to lend a hand, in which hearty zeal made up for want of
strength. I look back on these days, especially to the Saturday
afternoons spent in the workshops of this admirably conducted iron
foundry, as a most important part of my education as a mechanical
engineer. I did not read about such things; for words were of little
use. But I saw and handled, and thus all the ideas in connection with
them became permanently rooted in my mind....
“One of these excellent men, with whom I was frequently brought
into contact, was William Watson. He took special charge of all that
related to the construction and repairs of steam engines,
waterwheels, and millwork generally. He was a skillful designer and
draughtsman and an excellent pattern maker. His designs were
drawn in a bold and distinct style, on large deal boards, and were
passed into the hands of the mechanics to be translated by them into
actual work.”[85]
[85] Ibid., p. 92.
After telling of various workmen, Nasmyth says: “One of the most

original characters about the foundry, however, was Johnie Syme.
He took charge of the old Boulton & Watt steam engine, which gave
motion to the machinery of the works.... Johnie was a complete
incarnation of technical knowledge. He was the Jack-of-all-trades of
the establishment; and the standing counsel in every out-of-the-way
case of managing and overcoming mechanical difficulties. He was
the superintendent of the boring machines. In those days the boring
of a steam engine cylinder was considered high art in excelsis!
Patterson’s firm was celebrated for the accuracy of its boring.
“I owe Johnie Syme a special debt of gratitude, as it was he who
first initiated me into that most important of all technical processes in
practical mechanism—the art of hardening and tempering steel.”[86]
From another of his friends, Tom Smith, Nasmyth picked up the
rudiments of practical chemistry, as it was then understood.
[86] Ibid., p. 93.
Traveling with his father from time to time, he had good

opportunities for meeting many distinguished engineers and of
visiting the great iron works, the most famous of which was the
Carron Iron Works. “The Carron Iron Works,” he writes, “are classic
ground to engineers. They are associated with the memory of
Roebuck, Watt, and Miller of Dalswinton. For there, Roebuck and
Watt began the first working steam engine; Miller applied the steam
engine to the purposes of navigation, and invented the Carronade
gun. The works existed at an early period in the history of British iron
manufacture. Much of the machinery continued to be of wood.
Although effective in a general way it was monstrously cumbrous. It
gave the idea of vast power and capability of resistance, while it was
far from being so in reality. It was, however, truly imposing and
impressive in its effect upon strangers. When seen partially lit up by
the glowing masses of white-hot iron, with only the rays of bright
sunshine gleaming through the holes in the roof, and the dark, black,
smoky vaults in which the cumbrous machinery was heard rumbling
away in the distance—while the moving parts were dimly seen
through the murky atmosphere, mixed with the sounds of escaping
steam and rushes of water; with the half-naked men darting about
with masses of red-hot iron and ladles full of molten cast-iron—it
made a powerful impression upon the mind.”[87]
[87] Ibid., p. 109.
By the time he was seventeen Nasmyth had become a skilled

model maker. While he was still attending lectures in the Edinburgh
School of Arts and in the University, he had built up quite a brisk
business in engine models, for which he charged £10 each. He
made his brass castings in his own bedroom at night, arranging a
furnace in his grate. He had a secret box of moulding sand and
rammed his patterns gently so as not to awaken his father who slept
below. In the morning the room would be all clean and gave no
indication that it was serving for a foundry as well as a bedroom, and
by some miracle he managed to complete his practical education
without burning down the house. In 1827, when he was nineteen, he
built a steam road carriage which ran about the streets of Edinburgh
for many months, but the condition of the Scotch roads was such as
to make a machine of this kind almost useless. When he went to
London he broke it up, and sold the engine and boiler for £67.
From inspecting the engines constructed by different makers,
Nasmyth became impressed with the superiority of those turned out
by the Carmichaels of Dundee. “I afterwards found,” he writes, “that
the Carmichaels were among the first of the Scottish engine makers
who gave due attention to the employment of improved mechanical
tools, with the object of producing accurate work with greater ease,
rapidity, and economy, than could possibly be effected by the hand
labor of even the most skillful workmen. I was told that the cause of
the excellence of the Carmichaels’ work was not only in the ability of
the heads of the firm, but in their employment of the best engineers’
tools. Some of their leading men had worked at Maudslay’s machine
shop in London, the fame of which had already reached Dundee,
and Maudslay’s system of employing machine tools had been
imported into the northern steam factory.”[88] These reports built up
an ambition, which developed into a passion, to go to London and
work in Maudslay’s shop under “this greatest of mechanics.”
[88] Ibid., p. 123.
Consequently, in the spring of 1829, he went with his father to

London and made application to Maudslay to work with him as an
apprentice. Maudslay told them in the friendliest way, but
unmistakeably, that he had had no satisfaction from gentleman
apprentices and that he had definitely settled that he would never
employ one again. He showed them about his shop, however, and
began to melt when he saw the boy’s keen interest and intelligent
appreciation of everything about him. Nasmyth had brought with him
some of his drawings and one of his engine models. At the end of
the visit he mustered courage to ask Maudslay if he would look at
them. The next day Maudslay and his partner looked them over. “I
waited anxiously. Twenty long minutes passed. At last he entered the
room, and from a lively expression in his countenance I observed in
a moment that the great object of my long cherished ambition had
been attained! He expressed, in good round terms, his satisfaction at
my practical ability as a workman engineer and mechanical
draughtsman. Then, opening the door which led from his library into
his beautiful private workshop, he said, ‘This is where I wish you to
work, beside me, as my assistant workman. From what I have seen,
there is no need of an apprenticeship in your case.’[89]
[89] Ibid., p. 129.
“Mr. Maudslay seemed at once to take me into his confidence. He

treated me in the most kindly manner—not as a workman or an
apprentice, but as a friend. I was an anxious listener to everything
that he said; and it gave him pleasure to observe that I understood
and valued his conversation. The greatest treat of all was in store for
me. He showed me his exquisite collection of taps and dies and
screw-tackle, which he had made with the utmost care for his own
service. They rested in a succession of drawers near to the bench
where he worked....
“He proceeded to dilate upon the importance of the uniformity of
screws. Some may call it an improvement, but it might almost be
called a revolution in mechanical engineering which Mr. Maudslay
introduced. Before his time no system had been followed in
proportioning the number of threads of screws to their diameter.
Every bolt and nut was thus a specialty in itself, and neither
possessed nor admitted of any community with its neighbors. To
such an extent had this practice been carried that all bolts and their
corresponding nuts had to be specially marked as belonging to each
other....
“None but those who lived in the comparatively early days of
machine manufacture can form an adequate idea of the annoyance,
delay, and cost of this utter want of system, or can appreciate the
vast services rendered to mechanical engineering by Mr. Maudslay,
who was the first to introduce the practical measures necessary for
its remedy.”[90]
[90] Ibid., pp. 131-132.
There was no place in all England where Nasmyth could have

learned more. He was in close personal contact with one of the best
mechanics in the world. He had Maudslay’s warmest personal
interest and heard all the discussions of the engineers and famous
men who used to come to the workshop. “Among Mr. Maudslay’s
most frequent visitors was Gen. Sir Samuel Bentham, Mr. Barton,
director of the Royal Mint, Mr. Bryan Donkin, Mr. Faraday, and Mr.
Chantrey, the sculptor. As Mr. Maudslay wished me to be at hand to
give him any necessary assistance, I had the opportunity of listening
to the conversation between him and these distinguished visitors. Sir
Samuel Bentham called very often. He had been associated with
Maudslay during the contrivance and construction of the block
machinery. He was brother of the celebrated Jeremy Bentham, and
he applied the same clear common sense to mechanical subjects
which the other had done to legal, social and political questions.
“It was in the highest degree interesting and instructive to hear
these two great pioneers in the history and application of mechanics
discussing the events connected with the block-making machinery.
In fact, Maudslay’s connection with the subject had led to the
development of most of our modern engineering tools. They may
since have been somewhat altered in arrangement, but not in
principle. Scarcely a week passed without a visit from the General.
He sat in the beautiful workshop, where he always seemed so
happy. It was a great treat to hear him and Maudslay fight their
battles over again, in recounting the difficulties, both official and
mechanical, over which they had so gloriously triumphed.”[91]
[91] Ibid., pp. 151-152.
While with Maudslay, Nasmyth designed and built an index milling

machine for finishing the sides of hexagon nuts. After Maudslay’s
death in 1831, he remained a few months with Mr. Field to finish
some work in hand, and then left to start in business for himself.
Nasmyth speaks in the kindliest terms of Mr. Field, and doubtless
would have had more to say about him if his relationship with
Maudslay had not been so close.
Joshua Field was a man to be appreciated. He was a draftsman at
the Portsmouth dockyard when the block machinery was being built,
and showed so clear a grasp of the work in hand that Bentham had
him transferred to the Admiralty at Whitehall. In 1804 he left the
service and went to Maudslay’s, when he was at Margaret Street
and employed about eighty men. He rose steadily, was taken into
partnership in 1822, at the same time as Maudslay’s eldest son, and
was the senior partner after Maudslay’s death when the firm was at
the height of its long prosperity. He was one of those consulted in the
laying of the Atlantic cable and in the designing of machinery for
doing it.
“Mr. Field was one of the founders of the Institution of Civil
Engineers, the origin of which was very humble. About the year
1816, Mr. Henry Robinson Palmer, who was then a pupil of the late
Mr. Bryan Donkin, suggested to Mr. Field the idea of forming a
society of young engineers, for their mutual improvement in
mechanical and engineering science; and the earliest members were
Mr. Henry Robinson Palmer, Mr. William Nicholson Maudslay, and
Mr. Joshua Field. To these three were shortly added Mr. James
Jones, Mr. Charles Collinge, and Mr. James Ashwell. They met
occasionally in a room hired for the purpose, and to them were soon
attracted others having the same objects in view. Mr. Field was the
first chairman of the Institution, being elected to that post on the sixth
of January, 1818. Subsequently he became, in 1837, a vice-
president, an office he filled until he was elected president in 1848,
and in 1849, and he continued to the last to be an active member
and warm supporter of the Institution.”[92] Mr. Field did everything in
his power to give Nasmyth a start, allowing him to make the castings
for some machine tools which he proposed to finish later for use in
his own plant.
[92] Memoir, in “Transactions of the Institution of Civil Engineers,” Vol.
XXIII, p. 491. 1863.
Nasmyth returned to Edinburgh and took temporary quarters in a

little outbuilding 16 feet by 24 feet, within a few minutes’ walk of his
father’s home. He hired one mechanic, Archie Torry, who remained
with him the rest of his life and became one of his principal foremen.
His power plant consisted of one husky laborer who turned a crank.
Together they finished up the castings brought from Maudslay &
Field’s, making first a lathe, then a planer 20 inches by 36 inches,
and with these a few boring and drilling machines. He carried the
expense of this by doing some work for an enthusiastic inventor of a
wonderful rotary steam engine. Nasmyth honorably informed the
inventor that his machine would not work, but as the inventor was
bent on spending his money, Nasmyth executed the work for him,
and the proceeds enabled him to build his machinery.
In a few months he was ready to begin. He went to Liverpool and
Manchester looking for a location, and soon made many powerful
friends in both cities. In 1831 he rented a single floor in Manchester,
27 feet by 130 feet, with power, and ten days later Archie followed
with the tools. It was a particularly fortunate time and place for
starting such an enterprise. The success of the Liverpool &
Manchester Railway, just opened, created a great demand for
locomotives and for machine tools. Orders came in fast, and the
planer especially was busy all the time. If its profits were anything
like those of Clement’s planer, it must have been a very heavy
earner. As the business grew, Nasmyth added more tools, always
making them himself and steadily improving their design and
construction.
He soon outgrew his quarters; and in 1836 he secured land at
Patricroft, a mile or so outside of the city, admirably located between
the new railway and the Bridgewater Canal, and built a new plant
which he called the Bridgewater Foundry. In the new foundry he
used the first worm-geared tilting pouring-ladle. As it eliminated a
common and very dangerous source of accidents, he refrained from
patenting it and in a short time its use was universal. He formed a
partnership with Holbrook Gaskell, who took the business end of the
enterprise, and the firm of Nasmyth & Gaskell had a very prosperous
career until, sixteen years later, Mr. Gaskell was forced to retire on
account of ill health.
Nasmyth built machine tools of all kinds. In 1836 he invented the
shaper which was long known as “Nasmyth’s Steel Arm.”
Descriptions and illustrations of some of Nasmyth’s tools may be
found at the end of his autobiography,[93] in Buchanan’s “Mill
Work,”[94] and in the American Machinist.[95] He patented but few of
his inventions, relying for protection mainly upon the reputation
which he soon established. “In mechanical structures and
contrivances,” he says, “I have always endeavored to attain the
desired purpose by the employment of the fewest parts, casting
aside every detail not absolutely necessary, and guarding carefully
against the intrusion of mere traditional forms and arrangements.
The latter are apt to insinuate themselves, and to interfere with that
simplicity and directness of action which is in all cases so desirable a
quality in mechanical structures. Plain common sense should be
apparent in the general design, as in the form and arrangement of
the details; and a character of severe utility pervade the whole,
accompanied with as much attention to gracefulness of form as is
consistent with the nature and purpose of the structure.”[96] This was
written in later life. While his later work was in thorough conformity
with these principles, it was some time before he freed himself from
the tradition of Greek style in machine frames. He was one of those,
however, who led the way into the more correct practice indicated
above, though he was probably not so influential in this direction as
Whitworth.
[93] p. 400 et seq.
[94] Volume of Plates.
[95] Oct. 14, 1909, p. 654.
[96] Autobiography, p. 439.
His greatest invention unquestionably was that of the steam

hammer, which came about in an interesting way. He had built a
number of locomotives for the Great Western Railway. This railway
operated a line of steamers from Bristol to New York and was
planning a ship larger and faster than any then built, to be called
“The Great Britain.” It was to be a side-wheeler and the plans called
for a large and heavy paddle shaft, 30 inches in diameter. Mr.
Humphries, its designer, wrote to Nasmyth asking for help, saying so
large a shaft could not be forged with any of the hammers then in
use. Nasmyth saw at once the limitations of the prevailing tilt
hammer—which was simply a smith’s hand hammer, enlarged, with
a range so small that it “gagged” on large work,—and that the design
of large hammers must be approached in an entirely new way. “The
obvious remedy was to contrive some method by which a ponderous
block of iron should be lifted to a sufficient height above the object
on which it was desired to strike a blow, and then to let the block fall
down upon the forging, guiding it in its descent by such simple
means as should give the required precision in the percussive action
of the falling mass. Following up this idea,” he writes, “I got out my
‘Scheme Book,’ on the pages of which I generally thought out, with
the aid of pen and pencil, such mechanical adaptations as I had
conceived in my mind, and was thereby enabled to render them
visible. I then rapidly sketched out my steam hammer, having it all
clearly before me in mind’s eye. In little more than half an hour after
receiving Mr. Humphries’s letter narrating his unlooked-for difficulty, I
had the whole contrivance, in all its executant details, before me in a
page of my Scheme Book, a reduced photograph copy of which I
append to this description. (See Fig. 23.) The date of this first
drawing was the twenty-fourth of November, 1839....”[97]
[97] Ibid., p. 240.
Figure 23. First Sketch of the Steam Hammer Nov. 24, 1839
Figure 24. Model of the First Steam Hammer
In the South Kensington Museum, London
“Rude and rapidly sketched out as it was, this, my first delineation

of the steam hammer, will be found to comprise all the essential
elements of the invention.[98] Every detail of the drawing retains to
this day the form and arrangement which I gave to it forty-three
years ago. I believed that the steam hammer would prove practically
successful; and I looked forward to its general employment in the
forging of heavy masses of iron. It is no small gratification to me now,
when I look over my rude and hasty first sketch, to find that I hit the
mark so exactly, not only in the general structure but in the details;
and that the invention as I then conceived it and put it into shape, still
retains its form and arrangements intact in the thousands of steam
hammers that are now doing good service in the mechanical arts
throughout the civilized world.”[99]
[98] Compare Nasmyth’s sketch, Fig. 23, with Fig. 24, which was taken
from the model of his first hammer now in the South Kensington Museum
(Exhibit No. 1571). The description of it in the catalog is as follows:
“It consists of a base plate with a large central opening through which
projects the top of the anvil, so that a blow on the anvil is not transmitted to
the base plate. On the plate are secured two standards which form guides
for the hammer-head or tup, and also support an overhead cylinder, the
piston of which is connected with the tup by a piston rod passing through
the bottom of the cylinder. Steam is admitted to this cylinder by a stop valve
in the form of a slide, and then by a slide valve on the front of the cylinder,
which by a hand lever can be moved so as to let steam in below the piston
and so raise the heavy tup. When it is lifted to a height proportionate to the
energy of the blow required, the steam is by the slide valve permitted to
escape and the hammer falls upon the forging placed on the anvil. The
cylinder is therefore only single-acting, but the top is closed, and a ring of
holes communicating with the exhaust pipe is provided at a little distance
down inside. In this way an air cushion is formed which helps to start the
piston downwards when a long stroke is being taken, and also the steam
below the piston is permitted to escape when the tup has been lifted as high
as it can safely go. Soon after its invention the steam hammer was greatly
increased in power by accelerating the fall of the tup by admitting steam
above the piston in the downstroke and so changing it into the usual
double-acting steam hammer.” Cat. Machinery Collection, Part II, p. 255.
[99] Autobiography, p. 242.

Metaheuristics For Resource Deployment Under Uncertainty in Complex Systems 1St Edition Shuxin Ding Online Ebook Texxtbook Full Chapter PDF

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

Metaheuristics For Resource Deployment Under Uncertainty in Complex Systems 1St Edition Shuxin Ding Online Ebook Texxtbook Full Chapter PDF

Uploaded by

Copyright:

Available Formats

Metaheuristics for Resource

Deployment under Uncertainty in

Optimal Operation of Integrated Multi-Energy Systems

Design of Deep Braced Excavation and Earth Retaining

The Science and Management of Uncertainty Dealing with

Large Scale System Analysis Under Uncertainty With

AI Metaheuristics for Information Security in Digital

Metaheuristics for Finding Multiple Solutions 1st

Defense Resource Planning under Uncertainty An

Multiscale Dynamics Simulations Nano and Nano bio

and by CRC Press

CRC Press is an imprint of Taylor & Francis Group, LLC

Library of Congress Cataloging‑in‑Publication Data

ISBN: 978-1-032-06520-5 (hbk)

Author Bios xvii

1.1 APPLICATIONS OF NODE DEPLOYMENT PROBLEM 1

1.3.3.1 Node Deployment in Wireless Sensor

Chapter 2 Stochastic Node Deployment for Area Coverage

2.4.2 Parameter Setting 40

Chapter 3 Stochastic Dynamic Node Deployment for Target

Chapter 4 Robust Node Deployment for Cooperative

4.2.1.1 Two-Level Cooperative Set Covering

Chapter 5 Fuzzy Node Deployment for Cooperative

5.1 INTRODUCTION 116

5.2.2 Mathematical Model 118

Chapter 6 Simulation-Based Evaluation Analysis of Node

6.1 INTRODUCTION 139

6.4 CONCLUSION 164

Chapter 7 Overview and Future Directions 167

Effective use of the resources at the best location is becoming more

Chapter 3: The stochastic dynamic node deployment problem for

(VIS) encoding method is adopted, and MOEA/D-VIS and DMOEA-

We wish to thank all the researchers of the School of Automation, Beijing

Qi Zhang received the B.S. degree in electronic engineering from Bei-

is currently a professor with the School of Automation, Beijing Insti-

Panos M. Pardalos received the B.S. degree in mathematics from

1.1 APPLICATIONS OF NODE DEPLOYMENT PROBLEM

1.1.1 Unmanned Systems

Figure 1.1 UAVs, UGVs, and UUVs in an unmanned system.

the civil domain to the military domain, such as logistics, surveillance,

UAV-UGV coordination systems promising. UGVs can act as actuators

1.1.2 Wireless Sensor Networks

• Military applications. In command, control, communications, com-

• Environmental applications. Environmental problem is a critical

• Industrial applications. WSNs can be used in manufacturing pro-

• Home applications. WSNs in home environments connect every-

Figure 1.2 Wireless sensor networks in 3D space for target detection.

Internet of Things (IoT)-based smart environment [13]. It is an en-

• Medical applications. The Body Area Sensor Network (BASN) con-

• Healthcare facility location. The facility location problems for

• AED location. Optimizing the deployment of public automated

• Ambulance location. Ambulance location belongs to the emergency

• Hospital layout planning. The layout planning problems for hospi-

1.1.4 Public Sectors

• Locating schools. Schools are located to meet the students’ de-

• Bike sharing systems. Bike sharing systems are designed to in-

• Locating electric vehicle charging stations. Electric vehicles (EVs)

1.1.5 Railway Network Design

Figure 1.4 An illustration of a bike sharing system.

rolling-stock circulation and maintenance, crew scheduling, and real-time

• Railway rapid transit network design. This problem consists of the

Figure 1.5 An illustration of a high-speed railway network.

destination of each line, with the frequency of the train, are to be