Optimizing Military Tactical MANETs efficiently using PSO  
 
 
by 
 
Younchol Cho 
 
 
 
 
A dissertation submitted to the Graduate Faculty of 
Auburn University 
in partial fulfillment of the 
requirements for the Degree of 
Doctor of Philosophy 
 
Auburn, Alabama 
December 18, 2009 
 
 
 
 
Keywords: mobile ad-hoc network, hop-count, shortest path,  
particle swarm optimization, network centric warfare 
 
Copyright 2009 by Younchol Cho 
 
 
Approved by 
 
Alice E. Smith, Co-chair, Professor of Industrial & Systems Engineering 
Jeffrey S. Smith, Co-chair, Professor of Industrial & Systems Engineering 
Abdullah Konak, Associate Professor of Information Science & Technology, Penn State Berks 
 
 
 
 
 
 
 
Abstract 
 
 
 A mobile ad-hoc network (MANET) is a self-configuring network of autonomous agents 
designed to continuously support users who change the topology of the network dynamically and 
independently. The dynamic mobility of user nodes can cause communication links to disconnect 
if the network is not managed correctly. The autonomous agents are controlled to maximize the 
connectivity of user nodes considering military aspects. Military operations require the efficient 
use of limited resources and sometimes sacrifice network performance to accomplish given 
missions. Especially, military MANETs are exposed to a dangerous environment due to the 
enemy activities, so these hostile effects should be considered. Under these requirements and 
circumstances, the primary objective of the agents is to maximize the connection and quality of 
communication between user nodes and a control node. This is formulated as a shortest path 
problem using a hop-count metric. A population based heuristic algorithm, Particle Swarm 
Optimization (PSO), is developed to solve the military MANET problem. To best accommodate 
the aspects of military MANETs several crucial constructs are devised and tested. These are the 
Pre-deployed Agent Level (PAL), the Messenger Agent and the Priority node. Testing involves 
the common random waypoint model and two specific military scenarios - search and rescue and 
patrol. The proposed model in this thesis tries better represents military MANETs by considering 
enemy obstacles and military operation characteristics. It can be used to evaluate network 
performance before deploying an actual network in the battlefield and to properly size the 
 ii
 iii
number of agent nodes. Also, the proposed approach could be used for commercial applications 
by slight modification of the objective function. 
 
 
 
 
 
 
 
Acknowledgments 
 
 
 I want to thank God who allowed me to experience many valuable things through the 
difficulties I faced while studying for my doctoral degree at Auburn University. I will try to keep 
the teachings of the Lord in my heart forever. I would like to thank my country and Navy for 
giving me this great opportunity. I will dedicate my life to the Republic of Korea as a Korean 
navy officer to repay the benefits given by my country and Navy.  
I would like to thank my wife Yokyung Kim, and my children, Eunju and Mansu for their 
love and support in the pursuit of this degree. This work would not have been possible without 
their understanding and support. I would also like to thank my parents, Younghwan Cho and 
Jonghwa Lim for their help and encouragement throughout my life. 
Finally, I would like to thank Dr. Alice Smith and Jeffrey Smith, my thesis advisors and 
Dr. Abdullah Konak, a committee member for their great help in my research throughout my 
years at Auburn.  
 iv
 
 
 
 
 
Table of Contents 
 
 
Abstract ......................................................................................................................................... ii 
Acknowledgments ....................................................................................................................... iv 
List of Tables ............................................................................................................................. viii 
List of Figures .............................................................................................................................. ix 
Chapter 1 Introduction  ................................................................................................................. 1 
 1.1 Background   ............................................................................................................... 1 
 1.2 Research Objectives   .................................................................................................. 8 
 1.3 Dissertation Overview   ............................................................................................ 11 
Chapter 2  Literature review ....................................................................................................... 13 
 2.1 MANET performance   ............................................................................................. 13 
 2.2 Network connection   ................................................................................................ 16 
 2.3 MANET performance metrics   ................................................................................ 17 
 2.4 MANET routing   ...................................................................................................... 22 
 2.5 Mobility  ................................................................................................................... 25 
 2.6 Particle Swarm Optimization (PSO)   ....................................................................... 30 
 2.7 Optimization in dynamic environments   .................................................................. 39 
Chapter 3 Understanding Network-Centric Warfare (NCW)  .................................................... 42 
 3.1 What is NCW ?   ....................................................................................................... 43 
 3.2 Tenets and governing principles for NCW   ............................................................. 47 
 v
 3.3 Domains of conflict in NCW   .................................................................................. 48 
 3.4 Information warfare  ................................................................................................. 51 
 3.5 Decision making and security in NCW environment   ............................................. 52 
 3.6 Wireless networks   ................................................................................................... 55 
 3.7 Tactical battlefield network   .................................................................................... 58 
Chapter 4 Military MANET description and mathematical model  ........................................... 62 
 4.1 Proposed military MANET operation system   ......................................................... 62 
 4.2 Military MANET nodes   .......................................................................................... 64 
 4.3 Network states   ......................................................................................................... 66 
 4.4 Military MANET optimizer  ..................................................................................... 67 
 4.5 Mathematical formulation   ....................................................................................... 74 
Chapter 5 Military MANET system and simulation environment .............................................. 81 
 5.1 Routing by network state   ........................................................................................ 82 
 5.2 User node mobility model   ....................................................................................... 83 
 5.3 Isolated agent behavior   ........................................................................................... 86 
 5.4 Enemy behavior  ....................................................................................................... 86 
 5.5 Prediction of user future location   ............................................................................ 90 
 5.6 Messenger agent behavior   ....................................................................................... 92 
 5.7 Network node behavior in an enemy effect ranges   ................................................. 94 
 5.8 PSO parameters   ....................................................................................................... 97 
Chapter 6 Simulation experiments and analysis ......................................................................... 98 
 6.1 Simulation environment   .......................................................................................... 99 
 6.2 Test mobility models  ............................................................................................. 100 
 vi
 vii
 6.3 Performance measures   .......................................................................................... 106 
 6.4 Effect of metrics under different mobility  ............................................................. 109 
 6.5 Effect of metrics for medium and large sized problem with RW  .......................... 128 
 6.6 Cost benefit analysis  .............................................................................................. 137 
Chapter 7 Conclusions .............................................................................................................. 155 
References   ............................................................................................................................... 158 
 
 
 
 
 
 
 
 
 
 
List of Tables 
 
 
Table 4-1 Path Loss vs. Data rate ............................................................................................... 73 
Table 5-1 Distance vs. Jamming effect level  ............................................................................. 90 
Table 5-2 Combat power of resources and Index probability  .................................................... 95 
Table 6-1 Test problems  ............................................................................................................ 99 
Table 6-2 PAL effect with the hop-count based algorithm  ...................................................... 110 
Table 6-3 PAL effect with the shortest distance based algorithm  ........................................... 113 
Table 6-4 Efficiency of algorithms as measured by number of hops per connected user ........ 126 
Table 6-5 PAL effect with medium and large problem ............................................................ 129 
Table 6-6 Efficiency (#hops/connected user) of algorithms with medium and large problems 
................................................................................................................................................... 131 
 
 
 viii
 
 
 
 
 
 
List of Figures 
 
 
Figure 2-1 PSO basic mechanism ............................................................................................... 33 
Figure 2-2 Common topologies in PSO  ..................................................................................... 39 
Figure 3-1 Domain of conflict  ................................................................................................... 50 
Figure 3-2 Wireless networks and their coverage  ...................................................................... 56 
Figure 3-3 MANET interoperability ........................................................................................... 60 
Figure 5-1 Main optimization framework  .................................................................................. 81 
Figure 5-2 Optimization by network states ................................................................................. 83 
Figure 5-3 Enemy effect zones  .................................................................................................. 88 
Figure 5-4 Pseudo code for a messenger  ................................................................................... 93 
Figure 5-5 Kill probability of resources in a kill effect zone  ..................................................... 96 
Figure 5-6 Agent behavior in the kill zone ................................................................................. 96 
Figure 6-1 Initial formation with RW  ...................................................................................... 101 
Figure 6-2 Operation formation with RW ................................................................................ 102 
Figure 6-3 Initial formation with CD ........................................................................................ 103 
Figure 6-4 Operation formation with RW ................................................................................ 104 
Figure 6-5 Initial formation with SR ........................................................................................ 105 
Figure 6-6 Operation formation with SR .................................................................................. 106 
Figure 6-7 Comparison of PAL and No PAL (RW, 1E)  .......................................................... 111 
Figure 6-8 Comparison of PAL and No PAL (CD, 1E) ........................................................... 112 
 ix
Figure 6-9 Comparison of PAL and No PAL (SR, 1E) ............................................................ 113 
Figure 6-10 Messenger effect (RW)  ........................................................................................ 115 
Figure 6-11 Comparison of Messenger and No messenger (RW)  ........................................... 115 
Figure 6-12 Messenger effect (CD)  ......................................................................................... 116 
Figure 6-13 Comparison of Messenger and No messenger (CD, 1E)  ..................................... 117 
Figure 6-14 Messenger effect (SR)  .......................................................................................... 118 
Figure 6-15 Comparison of Messenger and No messenger (SR, 1E)  ...................................... 119 
Figure 6-16 MCR changes by priority node weights (RW)  ..................................................... 120 
Figure 6-17 Priority node effect (RW, 1E)   ............................................................................. 121 
Figure 6-18 Priority node effect (RW, 2E)   ............................................................................. 121 
Figure 6-19 Priority node effect (RW, 3E)   ............................................................................. 122 
Figure 6-20 MCR changes by priority node weights (CD)  ...................................................... 123 
Figure 6-21 NCU changes by priority node weights (CD)   ..................................................... 124 
Figure 6-22 MCR changes by priority node weights (SR)  ...................................................... 124 
Figure 6-23 NCU changes by priority node weights (SR)   ...................................................... 125 
Figure 6-24 Comparison of number of hops used for user a user connection (RW, 2E)  ........ 127 
Figure 6-25 Comparison of number of hops used for user a user connection (CD, 2E)   ........ 127 
Figure 6-26 Comparison of number of hops used for user a user connection (SR, 2E)   ......... 128 
Figure 6-27 Comparison of PAL and No PAL (Medium, no enemy)   .................................... 130 
Figure 6-28 Comparison of PAL and No PAL (Large, no enemy)   ........................................ 130 
Figure 6-29 Comparison of number of hops used for a user connection (Medium)   ............... 131 
Figure 6-30 Comparison of number of hops used for a user connection (Large)   ................... 132 
Figure 6-31 Messenger effect (Medium, 2E)   .......................................................................... 133 
 x
Figure 6-32 Messenger effect (Medium, 4E)   .......................................................................... 133 
Figure 6-33 Messenger effect (Large, 2E)   .............................................................................. 134 
Figure 6-34 Messenger effect (Large, 2E)   .............................................................................. 134 
Figure 6-35 Priority node effect (Medium, 2E)   ...................................................................... 135 
Figure 6-36 Priority node effect (Medium, 4E)   ...................................................................... 135 
Figure 6-37 Priority node effect (Large, 2E)   .......................................................................... 136 
Figure 6-38 Priority node effect (Large, 4E)   .......................................................................... 136 
Figure 6-39 Messenger effect (S1)  .......................................................................................... 139 
Figure 6-40 Comparison of network performance by messenger (S1)  .................................... 139 
Figure 6-41 Efficient number of agents in the no enemy case (S1)  ......................................... 140 
Figure 6-42 Efficient number of agents in the enemy case (S1)   ............................................. 141 
Figure 6-43 Network performance and number of killed resources (S1)   ............................... 141 
Figure 6-44 Messenger effect (S2)   ......................................................................................... 142 
Figure 6-45 Comparison of network performance by messenger (S2)   ................................... 143 
Figure 6-46 Efficient number of agents in the no enemy case (S2)   ........................................ 143 
Figure 6-47 Efficient number of agents in the enemy case (S2)   ............................................. 144 
Figure 6-48 Network performance and number of killed resources (S2)   ............................... 145 
Figure 6-49 Animation picture of combat scenario 2 (Red: user, Green: agent, Blue: control,           
Black: enemy) ........................................................................................................................... 146 
Figure 6-50 Efficient number of agents in the no enemy case (S3)   ........................................ 147 
Figure 6-51 Messenger effect (S3)   ......................................................................................... 148 
Figure 6-52 Comparison of network performance by messenger (S3)  .................................... 148 
Figure 6-53 Network performance and number of killed resources (S2)   ............................... 149 
 xi
 xii
Figure 6-54 Efficient number of agents in the enemy case (S3)   ............................................. 150 
Figure 6-55 Messenger effect in the combat scenario with CD   .............................................. 151 
Figure 6-56 Comparison of network performance by messenger in the combat scenario with CD 
................................................................................................................................................... 151 
Figure 6-57 Efficient number of agents in the combat scenario with CD ................................ 152 
Figure 6-58 Network performance and number of killed resources in the combat scenario with 
CD  ............................................................................................................................................ 153 
Figure 6-59 Efficient number of agents in the combat scenario with SR   ............................... 154 
 
 
 
Chapter 1 
Introduction 
 
1.1 Background 
Today, most people carry at least one portable information device, such as laptops, 
mobile phones and PDAs, for use in their professional and private lives. The purpose of these 
devices is to exchange and acquire valuable information through well established communication 
technology. Let us imagine what would happen to the world if we could not use those devices for 
a day. It would seriously affect our normal life, and the world would immediately fall into 
disorder due to cessation of normal functioning of the technological infrastructure that supports 
many important parts of our world. The dependence of modern human life on information 
technology (IT) based on telecommunication networks is so prevalent that we cannot imagine a 
life without them [34]. 
The rapid development of information technology has integrated most parts of the world 
into one well organized system. Networking is a complex part of computing that makes up most 
of the IT industries. Without networks, almost all communication in the world would cease. 
A telecommunications network is a network of links and nodes arranged in a way that 
messages may be passed from one part of the network to another through those links. These 
telecommunication networks can be split into two categories: wireline and wireless networks 
[105]. 
Wireline networks have provided users with very fast and reliable communications, and a 
large portion of the world economy now relies completely on these telecommunication networks. 
1 
Although a wireline network has the benefits of speed and security, it requires a great deal of 
time and cost for installation and maintenance. Moreover, it cannot respond to a dynamic 
situation or environment because of its requirement for infrastructure [26]. 
Therefore, wireless technologies are becoming quite popular and mobile internet service 
is a big trend. According to the survey conducted by the Pew Internet & American Life Project, 
54% of those who have internet-ready phones have used their phone to go online. Also, 56% of 
PDA owners have used their portable device to connect to the internet [78]. Wireless networks 
include infrastructure-based networks and ad-hoc networks. Infrastructure-based wireless 
networks need some infrastructure, such as access points or base stations. So, they have a similar 
limitation to wireline networks, although not as severe. Most wireless infrastructure-based 
networks are established by a one hop radio connection, which is an intermediate connection 
within a string of connections linking two network devices in a network, to a wired network. On 
the other hand, mobile ad-hoc networks are decentralized networks that develop through self-
organization, in which multi-hop communication is normal [76]. 
A wireless network is basically the same as a Local Area Network (LAN) or a Wide Area 
Network (WAN) of a wireline network, but there are no wires between hosts and servers. A 
Wireless LAN (WLAN) is a wireless local area network that links two or more computers or 
network devices without using wires. This gives users the mobility to move around within a 
broad coverage area and still be connected to the network. The most common WLAN, IEEE 
802.11, covers ranges from hundreds of meters to a few kilometers, depending on antennas [86]. 
A Wireless Wide Area Network (WWAN) is different from a WLAN because it 
uses cellular network technologies.  These cellular technologies are offered regionally, 
2 
nationwide, or even globally and are provided by many different wireless service companies. 
Cellular modems and mobile phones are good examples of this technology. 
Wireless cellular systems have been in use since the 1980s. Wireless systems operate with 
the aid of a centralized supporting structure such as an access point. These access points assist 
the wireless users to keep connected with the wireless system when they roam from one place to 
the other [6]. Mobile Ad-hoc Networks (MANET) have been developed from the wireless 
cellular system to overcome the limitation for use in an environment that does not have readily 
available infrastructure [18]. 
There are different types of agents which can be classified depending on the agent?s 
abilities into static or mobile, reactionary or not; work alone or with other agents, autonomous or 
not,. A MANET is an autonomous collection of mobile nodes, users and agents forming a 
dynamic wireless network. Autonomous agents can independently respond to events, network 
state changes, and move to other locations and adjust their behavior accordingly to accomplish 
goals [8]. So, agents can work more intelligently when they are well informed about a network 
state and its users. Autonomous mobile agents in a military MANET placed in an unknown 
terrain is challenging under the following conditions of military applications [85]:  
1) Dramatic change of geographical area (mission area) over time since dynamic nature 
of tasks,  
2) Network nodes may be decreased by hostile attacks or malfunctions,  
3) Isolation of network nodes, and 
4) Intermittent communication due to a hostile environment 
 
3 
Mobile ad-hoc networks operate without any fixed infrastructure and offer quick and easy 
network deployment in situations where infrastructure is not possible [6]. In these networks, 
nodes typically cooperate with each other by forwarding packets for nodes which are not within 
the direct communication range of the source node. A MANET provides a practical way to 
rapidly build a decentralized communication network in an area where there is no existing 
infrastructure or where temporary connectivity is needed, e.g. emergency situations, disaster 
relief scenarios, and military applications. The distribution/routing technique used depends on 
multi-hop protocols in order to deliver messages between nodes connected to the network [65, 84, 
96]. 
Nodes in a mobile ad-hoc network are free to move and organize themselves in an 
arbitrary fashion. The path between each pair of users may have multiple links and the radio 
between them can be heterogeneous if the network devices have different configurations. This 
allows an association of various links to be a part of the same network.  
Mobile ad-hoc networks can operate in a standalone fashion or be connected to a larger 
network such as the Internet. Mobile ad-hoc networks can turn the dream of getting connected 
"anywhere and at any time" into reality [6]. Historically, mobile ad-hoc networks have primarily 
been used for tactical network related applications to improve battlefield 
communications/survivability and rescue operation in disaster sites. 
In cases where group members have specific tasks to accomplish, group communication 
and synchronization are required. In military situations, for example, the members of a group of 
users have different targets. Thus, communication among group elements is an important issue 
and must be maintained for as long as possible [65]. MANET is a wireless network that 
continually re-organizes itself in response to environmental changes without using any 
4 
infrastructure. In order to accomplish assigned missions for a group like a military force or a 
rescue team, MANETs continuously communicate, collaborate, and interact among members in 
the network.  
MANET is a key enabler for achieving the goals of Network-Centric Warfare (NCW) 
[69] that represent modern warfare trends. Advanced technology in combat can provide the right 
information at the right place and time, and shorten the ?kill chain? (targeting cycle consisting of 
detecting a target, attacking it and assessing the result of the attack) by combining the several 
tactical levels of MANETs into a large strategic level of MANET. An example of this would be 
the U.S. armed forces? Global Information Grid (GIG) [27]. The ability to make more informed 
decisions faster is a central theme in the network centric warfare concept, since the key element 
for victory in modern warfare depends on the capability to deal with information related to the 
situation at hand [93]. The major challenges of MANET are to provide wireless, high-capability, 
secure, and networked connectivity. Compared with wired links and infrastructure based 
networks providing stable service, MANET capability is restrictive and has the potential for 
intermittent connectivity failure. So, participants must communicate using limited bandwidth 
through wireless links [82].  
Military operations are usually performed in highly deteriorated and varied conditions 
due to natural and artificial obstacles. The effort for maintaining robust communication in a 
network or among networks in these military environments is an essential element. In fact, a 
network is the most important weapon for modern military operations. The information 
superiority based on a network will enable agile deployment of a lighter, leaner, more lethal 
combat enterprise that overwhelms any potential adversary before it responds [69]. In order to 
5 
meet the above practical needs, the performance of MANETs should be evaluated accurately 
under more realistic environments for before deployment.  
We will focus on developing a more realistic and robust military MANET model in this 
study. The needs and problems encountered in developing a realistic military MANET model are 
related to simulating its operation environment as follows: 
 First, the representation of enemy effects in a military MANET operation is an 
indispensible factor for accurate evaluation before commissioning it into an actual tactical 
operation. However, hostile effects of enemies in the military MANET operation have not been 
fully studied in the literature, as far as we know. The enemies in the operation area attack the 
MANET nodes electrically or physically. Most research has focused only on electrical attack 
limiting the communication capability of MANET nodes [15, 50]. The transmission range of 
MANET nodes in a tactical battlefield fluctuate by an electrical attack (such as jamming), and 
transmission of radio signals by enemy forces to disrupt communications among MANET nodes, 
or the nodes could be destroyed by a physical attack. Also, the quality of wireless channels in the 
real world is variable due to a variety of propagation phenomenon based on the surrounding 
conditions, such as multipath resulting in radio signals reaching the destination node by two or 
more paths, atmospheric effects like fading and reflection, and obstacles [40]. As a result, the 
transmission range of network nodes cannot be constant, but varies depending on the 
surrounding conditions. So, implementing these characteristics to military MANET modeling is 
necessary to represent a realistic operational environment, which will help evaluate network 
performance more accurately.  
Second, military MANETs have different operation objectives and functional 
characteristics from commercial ones. Maximizing connectivity between nodes in the mobile 
6 
wireless network is an important and common performance measure for MANETs. However, 
military MANETs, unlike commercial ones, may often pursue other objectives. The 
accomplishment of a given mission could be more important than other performance measures. 
Also, military MANET nodes need to be specified in terms of functions within the military unit 
structure. User (client node) and agent (service node) is a common classification for commercial 
networks. The military MANET may not be represented by this simple classification. 
Third, the military MANET is required to be more efficient for longer operation since it 
tends to be operated with limited resources. Situation awareness established by robust 
networking is an effective weapon for the forces and commanders in waging war in a tactical 
operation area. This dramatically increases combat power [46]. However, because of the 
dynamic and unpredictable nature of military combat operations, it may be impossible to 
maintain full connection continuously. Especially, inefficient use of the network may shorten the 
operation time since battery capacity is limited. Sometimes, this situation forces military 
commanders to think about how to operate limited resources to maximize their networking 
capability. 
Finally, MANET nodes? mobility in wireless networks should also be examined as a 
major consideration since it plays a key role in the performance evaluation of MANET [68]. The 
main role of a mobility model is to emulate the movement behavior of actual users in a network. 
There are many mobility models proposed in the literature such as random walk, random 
waypoint, group mobility, and reference point group mobility [5, 13, 58, 83, 87]. These models 
vary widely in their movement characteristics. Among those, the random waypoint model has 
generally been used as the default mobility model in many network simulations regardless of 
application since it describes the movement pattern of independent nodes by simple terms. 
7 
However, some military operations show different movement patterns from the random waypoint, 
in which the users? movement may be directed or coherent by operation purposes instead of 
moving randomly and independently. Zhou et al. (2004) suggest that users? mobility patterns in 
military scenarios may not be independent, but are related to one another. In other words, the 
mobility of nodes in a military network is much more coherent and directed than in civilian 
applications. One typical example is group mobility. In battlefields, nodes with the same mission 
usually move in groups such as swarms or tank battalions [109]. Consequently, the mobility 
pattern in military operation depends on the mission. Therefore, the mobility of a military 
MANET should represent the movement characteristics of a given mission. 
The issues above are necessary for realistic representation of military MANET to test the 
effective operation of military MANETs under varied challenging conditions. In this thesis, a 
military MANET model which can deal with these issues is devised. 
 
1.2 Research objectives 
The primary objective of this dissertation is to develop a solvable realistic military 
MANET with autonomous agents under more realistic environments including employed enemy 
obstacles, and to develop a heuristic algorithm using Particle Swarm Optimization (PSO) to 
solve the model. The developed heuristic optimizer is expected to deploy the autonomous agents 
to the best locations at each time step based on given information such as MANET nodes? and 
obstacles? locations. MANET nodes have limited velocity and transmission ranges and user 
nodes are especially free to move around the tactical operation area. The decision variables 
which define the optimal solution are the agents? directions and magnitudes of motion. In order 
8 
to accomplish these objectives, the focus of the research has centered on the following detailed 
objectives: 
To begin with, for a realistic military MANET model, we will implement three realistic 
considerations as follows:  
First, the military network nodes are categorized to describe the different units under a 
command node in the military operation. Most MANET routing studies in the literature use two 
types of network node: user and agent. But in this study these are four types of network node: 
user, priority, agent, and control. Each node has a different responsibility and characteristic in the 
network. The control node is responsible for controlling the network nodes and also performs as 
an agent. An agent node is only responsible for supporting network connections. The priority 
node is basically the same type of node as the user. The difference between a user node and a 
priority node is the preference to be connected to the control node due to any urgent or crucial 
operation situation. That is, if a user node faces a situational event such as positioning in a 
dangerous location, fighting against enemies, and obtaining important operational information 
needed to be reported or broadcasted through the network immediately, it would be designated as 
a priority node. The priority node is converted to a user node when it gets out of the situational 
event.  
Second, enemy obstacles constraining network nodes? capability electrically or physically 
are included. The network nodes? communication capabilities are degraded or destroyed by the 
enemy?s hostile activities, such as electrical or physical attacks. These effects are modeled using 
the distance between a network node and enemies and the combat power of each node. The 
insertion of enemies to an operation environment will help increase the realism of the MANET 
model. 
9 
Third, a messenger option is implemented. The messenger is not a new resource, but an 
agent which has a temporary mission to search for disconnected users. During operation, users 
could be disconnected from the network. The agent node closest to the lost contact point is 
designated as a messenger searching for the disconnected user for an allowed time. This allowed 
time depends on the network state. That is, if there is extra capability of agents for the search, 
messenger operation time would be increased, otherwise it would be shorten by the capacity 
available. The main purpose of messenger implementation is to increase the performance of 
military MANETs. Sometimes, messenger agents are expected to improve network performance 
by reconnecting the disconnected users from the network. 
Fourth, based on the realistic environment constructed above for a better military 
MANET evaluation, a heuristic algorithm is developed to find the efficient network paths in 
terms of military perspective by using proper network performance metrics such as hop-count, 
bandwidth, and a newly defined metric, Pre-deployed Agent Level (PAL). Routing by the 
minimum hop route has been used for a long time in wireless networks. The strong points of this 
method are simplicity and lower consumption of network resources than the shortest-distance 
path [80]. Minimum hop based routing is one of the popular methods in mobile ad-hoc networks 
and is suitable for the military MANETs. However, minimum hop routing does not take into 
account the link quality and stability [24]. So, it sometimes lacks response to network changes. 
We employ an important metric, PAL, to make up for this weakness. The hop-count based 
algorithm may not respond quickly to network changes as mentioned above since it uses the hop 
as its primary network performance metric. It does not consider the pre-deployment of agent 
nodes to the proper locations to prepare for abrupt needs. That is, a link between network nodes 
may be broken if it is not properly supported by agent nodes. The disconnection of the link 
10 
between users is more frequent quicker than other links due to users? free movement. Also, an 
agent node cannot immediately be deployed anywhere it is needed because of its velocity 
limitation. Consequently, hop-count alone may worsen network performance. So, PAL is 
developed and used in this study to deal with this problem.  
Finally, in addition to the mechanisms proposed above to represent actual movement of 
specific military operation scenarios, two other mobility models are used. This expands our study 
beyond the random waypoint model which has been used as the default user mobility model in 
MANET research.  
The experiments in this study are divided into two parts, the verification of proposed 
mechanisms effect and a cost benefit analysis. The three important mechanisms, PAL, messenger, 
and priority node, developed and for better network performance and representation of military 
MANETs are first assessed using different mobility models and scenarios without the physical 
destruction of network resources by combat. Then, a cost benefit analysis for computing the most 
efficient number of agents for a given operational scenario is performed, in which the physical 
destruction effect is included. 
 
1.3 Dissertation overview 
Chapter 2 reviews previous studies and knowledge related to military mobile ad-hoc 
networks and particle swarm optimization.  
Chapter 3 discusses network-centric warfare (NCW) to increase understanding of new 
war paradigm. The key concept of modern warfare is NCW, and the mobile ad-hoc network as a 
tactical battlefield network is a basic component to enable NCW. 
Chapters 4 and 5 present the military MANET developed for this study.  Chapter 4 
11 
12 
describes problems in a military operation environment and the mathematical models 
representing them and the detailed model description is shown in Chapter 5.  
In Chapter 6, experiments are described and simulation results are provided with detailed 
analysis.  
Chapter 7 presents a brief conclusion along with some suggestions for further study. 
 
Chapter 2 
Literature review 
  
2.1 MANET performance 
A wireless ad-hoc network allows network nodes to communicate with each other over a 
common wireless channel without support from a fixed infrastructure. The basic parameter for 
the MANET performance measure is the link, also called the arc, between nodes in a network. 
Regardless of what type of routing protocol is used, the first and most important requirement for 
communication between the nodes in a network is to have at least one path linking them.  
Network nodes can directly talk to other network nodes if they are within their 
communication range, but in many cases in order to talk with a specific node they are required to 
communicate through other network nodes within transmission range. 
The transmission range of network devices in a network is affected by factors such as 
transmission power of the devices and environmental conditions. In particular, military MANETs 
may operate under vulnerable environmental conditions, in which network nodes? capabilities 
may be limited. Especially, the enemy?s hostile activities in a battlefield can seriously affect 
network communications. So, under these dynamic environments, the connectivity may not be 
guaranteed at all times for network nodes that are moving around the operation area continuously. 
In order to maintain the connectivity between the nodes exceeding their communication 
capability we can add more relay nodes into the network. However, by simply doing this we may 
not improve the MANET performance as expected, since additional nodes may cause a traffic 
burden to the network. As an alternative, we can extend the transmission range by increasing the 
13 
signal strength but this may cause increased interference among the nodes. Consequently, the 
capacity of MANET may be decreased by the mutual interference of concurrent transmissions 
among network nodes. 
Grossglauser and Tse [36] propose a communication model to improve the capacity of an 
ad-hoc network under various conditions affecting the MANET performance, with which a 
number of relay nodes are used to relay data and the data is relayed only when the relay node is 
closer to the destination, within a two-hop path. The drawback of this proposal is that the 
applications need to be delay tolerant because the communication is delayed until the mobile 
relay nodes are close to the destination nodes. This causes large delays when the size of the 
system is increased. Therefore, it is unsuitable for real time applications such as voice 
communications or remote control. 
A fundamental characteristic of mobile wireless networks is the time variation of the 
channel strength of the underlying communication links. Such time variation is due to multipath 
fading, path loss via distance attenuation, shadowing by obstacles, and interference from other 
users. By considering such variations, a variety of properties have been proposed to evaluate a 
routing protocol. The metrics are usually divided into two groups: qualitative and quantitative. 
These property metrics used for evaluation of protocols should be independent from the routing 
protocols and represent the properties of given protocols [60]. Distributed operation, demand-
based/proactive operation, security, bidirectional/unidirectional routing and sleep period 
operation are categorized into the qualitative property group. In particular, sleep period operation 
is required to avoid detection/jamming by an enemy. During the sleeping period, transmitting 
and/or receiving are stopped. On the other hand, for quantitative property, data throughput and 
delay, routing acquisition time and efficiency are considered. Data throughput and delay are 
14 
usually used for measuring network performance. Routing acquisition time indicates the time 
required to establish a route in a network. 
MANET?s performance should be measured using proper performance metrics based on 
properties due to the dynamic nature of mobile ad-hoc networks. In this research, the number of 
connected user nodes to the control node and the network bandwidth are the primary metrics are 
used to evaluate network performance. This research will focus on developing a method to model 
and optimize military MANETs under vulnerable situations, such as enemies in the operation 
area.   
Some different routing algorithms which evaluate network performance based on the 
network connectivity have been described in the literature [1, 3, 28, 41, 47, 55, 75, 76, 79, 101]. 
The MANET performance of a military MANET in this study is evaluated in a vulnerable 
environment added by hostile forces? jamming and physical attacks. It is not well known how 
well MANET can perform in a vulnerable environment. 
Jamming can be as simple as sending out a strong noise signal in order to prevent data 
packets in the network from being received. Hostile effects in military MANET operations are 
important considerations for planning and efficient use of limited resources since hostile action 
may seriously affect MANET performance. Providing efficient networking services in military 
MANETs is also very challenging in presence of mobility of users, unpredictable radio channel 
due to surrounding conditions and interference among network nodes.  
Karhima et al. (1996) address the vulnerability of 802.11b (a set of IEEE standards that 
govern wireless networking transmission methods) based MANET to intentional jamming by 
Unattended Jammer (UAJ) and Unmanned Aerial Vehicle (UAV). These jammers use higher 
transmit powers than MANET nodes to obstruct the connectivity. In this research, the jammers 
15 
are fixed at locations through the simulation time span and have abilities to attack the MANET 
nodes based on the predefined combat power. 
 
2.2 Network connection 
A graph G = G (V, E) consists of a set of nodes (vertices) and a set of edges (links, arcs). 
Graph theory is the study of graphs of mathematical structures that are used to model pair wise 
relations between objects from a certain collection in mathematics and computer science [34]. 
The set of nodes, denoted by V = {v
1
, ..., v
n
}, represents the ad-hoc network devices; and the set 
of edges, denoted by E = {e
1
, ..., e
m
}, represents the wireless communication links between nodes. 
Wireless multi-hop networks are generally modeled as communication graphs. In a 
wireless multi-hop network, each node has a certain transmission range, and is able to send 
messages to other nodes within its own transmission range. A wireless network can be viewed as 
a communication graph, where the existence of each edge is decided by the transmission range of 
related nodes. Therefore, the connectivity of a network depends on the transmission ranges of all 
nodes. So, two nodes are able to communicate directly via a wireless link if they are within the 
range of each other. In terms of communication networks, all nodes of a connected network can 
communicate with each other over one hop or multiple hops, whereas in a disconnected network 
we may have several isolated sub networks which cannot communicate with one another. 
Connectivity is a function of the number and locations of the nodes and wireless transmission 
range [6]. 
 
 
 
16 
2.3 MANET performance metrics  
Mobile networks have employed a variety of routing strategies to improve their 
performance under given environmental and operational conditions. The process of selecting a 
path among possible ones between two end users in a network depends on the variable link state 
over time. The link state can be expressed by costs or quality parameters that are computed based 
on many common types of network performance metrics such as hop-count, bandwidth, 
propagation delay and jitter [61, 72]. Furthermore, in many cases, the combined use of the 
metrics can improve the network performance as well as reduce the computation complexity. 
Various combinations of metrics are possible according to the application. The combination of 
hop-count and bandwidth is one of the most popular ones in the literature. However, the metrics 
should be able to represent the network properties of the system under consideration and be 
practical. Some metrics, such as delay and jitter, are not preferable due to their difficult 
computation. So, the metrics for a network must be combined together carefully by considering 
the complexity of computing paths based on them and the quality requirements of network flows 
[80].  
In the long run, network routing can be chosen using either a single metric or a 
combination of metrics. To help the computation of combined metrics, the sequential filtering 
method has been utilized in many studies [80, 48, 49]. With the sequential filtering method, the 
metrics are divided into primary and secondary ones. Paths are computed first based on the 
primary metrics. Many routes are eliminated from the candidate set by the primary metrics. The 
candidate set of possible routes will be narrowed down using secondary metrics until a single 
path is found. The number of steps of the sequential filtering could be different with each 
application.  
17 
2.3.1 Hop-count 
Mobile nodes in a wireless network are operating on a battery supply. So, mobile nodes 
naturally have strong power constraints and network life depends on the efficient use of this 
resource. In the real world, the number of relay nodes tends to be small. Therefore wireless 
communications share this limited number of relay nodes. Thus, each additional transmission 
causes the relay device to be busy for the length of the transmission and use beyond capacity 
prevents other nodes from transmitting. Therefore, unnecessary transmission should be 
minimized since additional hops or communications increase the delay of transporting the data 
packet due to the additional buffering, contention for resources, and transmission time required. 
Furthermore, efficient resource management can result in additional benefits other than reducing 
power consumption. For example, under a combat situation the least amount of transmission 
power minimizes the probability of detection or interception by enemy forces. As a result, a 
MANET routing protocol should be power efficient [39]. 
The hop-count is the number of hops separating a source node from its destination along 
the minimum path [97]. That is to say, it is the number of links passed by a packet between a 
source and a destination node. The hop-count based routing approach is one of the popular 
routing protocols for efficient use of the network. Most MANET protocols try to minimize the 
hop-count of the selected route. This is important in multi-hop wireless networks. For example, if 
a link between two nodes in a network requires at least 6 Mb/s bandwidth for delivering a data 
packet the more number of links on a selected shortest path, the more additional cost in terms of 
total amount of network resources consumed [37]. 
Under the situation of limited resources and battery efficiency issues, the hop-count based 
approach could be more effective than alternatives since it minimizes resource consumption. 
18 
However, the major drawback of this approach is potentially uneven network traffic. In addition, 
link quality may vary during the lifetime of a network because of the distance between the 
network nodes, transmitting power, antenna shape and orientation, radio interference, and 
environmental conditions. The links may be asymmetric by such variation. For instance, 
connectivity may change even under nodes at fixed locations and configured with the same 
transmission power. Connectivity could be affected by surrounding conditions other than 
location and transmission power factors. Due to these drawbacks, link quality based routing 
methods (e.g. QoS) have been considered and used [96]. 
The nodes in a MANET change location over time because of the dynamic nature. This 
mobility of MANET nodes may break the connection between nodes and lead to failure of the 
communication between them. Therefore, the consideration of link quality and stability in the 
context of link connection is as important as the selection of the shortest path. In the following 
section we will discuss this issue. 
 
2.3.2 Link quality (Bandwidth) 
As a network performance parameter, quality represents the state of a link by using 
metrics such as bandwidth, delay, etc. In this respect quality based routing is a method for traffic 
flow to meet quality requirements. The main purposes of this approach are to select routes 
satisfying a particular quality requirement and to provide efficient utilization of the network. 
There are three popular quality based routing algorithms based on the hop-count and bandwidth, 
which are Shortest-Widest Path (SWP), Widest-Shortest Path (WSP) and Shortest-Distance Path 
(SDP). WSP and SWP are relatively simple routing approaches [74]. 
19 
WSP first considers the minimum hop-count as a primary metric, but if there is more than 
one such path it uses bandwidth to break the tie. That is, the shortest path which has maximum 
bandwidth will be chosen. Conversely, SWP chooses a path with maximum bandwidth as a 
primary metric, and then consider the minimum hop-count as a secondary metric. Preferring the 
shorter path (such as WSP) will minimize the consumption of network resources, while 
preferring the widest path (such as SWP) will balance network load as well as maximize the 
chance to meet the required bandwidth in case of inaccurate network state information [72].  
Here, the bandwidth of a path indicates the minimum available bandwidth along the path. 
For example, if there is a path consisting of three links and each link?s bandwidth is 10, 8 and 5 
Mb/s, this path?s bandwidth will be 5 Mb/s. SDP can dynamically balance the effect of hop-count 
and path load using a distance function. However, resource consumption and load distribution 
conflict with each other and both objectives in network operation cannot be accomplished at the 
same time. So, we should pick the one most suitable for a given situation, condition or 
operational objective [37]. 
 
2.3.3 Link stability 
Link stability refers to the ability of a link to survive for a certain time period. In this 
respect link stability based routing is unique to wireless networks. The stability of a link has a 
direct relationship to the distance and signal strength between two nodes. That is, it depends on 
how long the two nodes remain within each other?s communication range or signal strength is 
above a threshold.   
Sridhar et al. (2005) propose an algorithm called Stability and Hop-count based algorithm 
for Route Computing (SHARC) that uses hop-count and residual lifetime of the link as 
20 
performance metrics. SHARC uses the shortest path algorithm by hop-count as the initial filter to 
narrow down route selections and then uses path stability by residual lifetime, a less robust 
indication, to choose the best route from among the available routes. Here, link stability is 
represented by the residual lifetime computed based on link age. The residual lifetime can be 
computed by the following equation [96]: 
 
 
alilR
aiai
iia ??=
??
>>
)/)((  
(2?1) 
       Where: 
Ra: the average residual link lifetime when the current link age is a 
i: the link duration in seconds 
li: the number of links with link duration in seconds 
 
Link stability can be much more important in military MANETs in combat situations than 
in commercial networks since combat operations require a robust network connection to respond 
quickly and to maintain operational continuity. Because of these military aspects, some user 
nodes need to be supported with a higher priority than others even though it may sacrifice 
network performance. So, we designate priority nodes in this study. We will discuss this in more 
detail in Chapter 4. 
In order to make a more efficient algorithm for maximizing network performance it is 
important to decide on proper metrics for network characteristics which are correspond to the 
surrounding conditions and operational properties. This will enable the routing algorithm to find 
the most available path efficiently and in a less expensive manner. Although a great variety of 
routing approaches have been employed, many previous studies have shown that algorithms with 
21 
a strong preference for minimum-hop routes almost always outperform algorithms that do not 
consider path length. The WSP is a good example. The shortest-path routing is particularly 
attractive in a large, distributed network, since path length is a relatively stable metric, compared 
with dynamic measurements of link delay or loss rate. 
The completeness of the specified missions is usually the primary goal in military 
operations. So, network performance could be sacrificed under certain circumstances to 
accomplish a given mission. Successful mission performance under the modern warfare 
paradigm could be enhanced by maintaining a tight communication network for real time 
information exchange. That is, link connections in the networks should be able to be guaranteed. 
Hop-count, quality and stability are the most common considerations for network algorithms. For 
a better solution we have to combine those factors properly using the strong point of each factor 
[96]. 
 
2.4 MANET routing 
Routing is the mechanism of directing data packet flow from the source to the destination 
[26] and is very crucial in the MANET, because changes in network topology and other states 
occur frequently and continuously.  
One common and traditional way of achieving routes in mobile Ad-hoc routing is to 
consider each host as a router [12]. Ad-hoc mobile routing protocols can usually be categorized 
into three types: Table driven proactive, On-demand-driven reactive/Source initiated, and Hybrid 
protocols. 
 
 
22 
2.4.1 Table-driven routing (Proactive) 
Table-driven routing is a proactive protocol. The most distinguishing characteristic of this 
routing is to continuously search for routing information within a network to keep routing tables 
available anytime they are needed, so it is called table-driven routing [28]. These routing 
protocols react to any change in the topology even if no traffic is affected by the change, and 
they require periodic control messages to maintain routes to every node in the network. The rate 
at which these control messages are sent must reflect the dynamics of the network in order to 
maintain valid routes. Thus, the maintenance of the routing tables requires significant bandwidth 
[17].  Representatives of this protocol category are briefly listed below. 
? Destination-Sequenced Distance Vector Routing (DSDV): each node maintains a list of 
all destinations and number of hops to each destination [74]. 
? Clustered Gateway Switch Routing Protocol (CGSR): each node maintains a cluster 
member and a routing table [16]. 
? Wireless Routing Protocol (WRP): each node maintains four tables; distance, routing, 
link cost and message retransmission list [62]. 
 
2.4.2 On Demand-driven routing (Reactive) 
As an alternative to proactive routing, On Demand-driven routing was introduced, which 
constructs paths when they are explicitly needed to route packets. This prevents the nodes from 
updating every possible route in the network, and instead allows them to focus either on routes 
that are being used, or on routes that are in the process of being set up.  
The best routing can be found based on the available information about link state at the 
moment. In other words the routing decision depends much on the current information. Link 
23 
state information can be propagated by two ways, periodic and responding to a significant 
change in the link state metric. For example, the amount of change required for triggering an 
update could be decided by the user and any link advertises its available bandwidth metric 
whenever it changes by more than the set amount since the previous update message. 
However, network performance could be affected by the frequency of propagating 
updated information. That is, frequent messaging of the information may cause too much 
network overhead, which may make an accurate routing decision not worth the expensive costs. 
So, a careful understanding of the trade-off between network overheads and an accurate routing 
decision should be required for adjusting the frequency of the link state update message [56, 90]. 
The proactive approach depletes too many resources by updating. If the update interval is 
too long, the network will simply contain a large amount of stale routes in the nodes, which 
results in a significant loss of packets. In every test case these reactive routing algorithms 
outperformed the proactive algorithms in terms of throughput and delay [17, 25]. Moreover, the 
reactive protocol is more realistic than the proactive way, considering limited resources available 
in the real world.  
? Ad-hoc-On Demand Distance Vector (AODV): This routing protocol builds on the DSDV 
algorithm but improves it by minimizing the number of required broadcasts by creating 
routes on an on-demand basis, as opposed to maintaining a complete list of routes [76]. 
? Dynamic Source Routing (DSR): DSR is a routing protocol for wireless mesh networks 
and is similar to AODV. However, DSR uses source routing instead of relying on the 
routing table at each intermediate device [47].  
24 
? Signal Stability Routing (SSR): SSR selects the route based on signal strength and 
stability. A path that has stronger power and longer duration is chosen as the best route [1, 
12]. 
 
2.4.3 Hybrid 
Hybrid protocols try to combine proactive and reactive protocols. Zone Routing Protocol 
(ZRP) is a good example of a hybrid protocol, which divides the topology into zones and seeks 
to utilize different routing protocols within and between the zones based on the weaknesses and 
strengths of these protocols. Any routing protocol can be used within and between zones since 
ZRP is modular [70, 71]. 
As discussed before, common classification of routing protocols is to divide them into 
proactive, reactive and hybrid routing. However, in [96] the authors classify routing protocols 
into hop-count based, QoS (e.g. bandwidth) based, and stability (or availability) based. Each of 
these classes of routing protocols can be either proactive or reactive [7].  
 
2.5 Mobility 
The introduction of relatively low-cost mobile computing devices is having a profound 
impact on everyday life. Commerce and family life are both affected by the ability to remain 
connected to other people and to exchange an ever increasing amount of information [88, 93]. 
The continuous improvement of mobile computing devices will extend the service range of 
mobile wireless networks and enable users? more dynamic movements. In a MANET we assume 
that nodes are free to move. The network topology and wireless link status are changed due to 
the mobility of nodes. 
25 
Many different measures of mobility for evaluating mobile ad-hoc network performance 
have been proposed, since a mobility model is imperative to evaluate a routing protocol of 
MANET using simulation [38, 93]. The most important characteristic of a mobility model is the 
degree of realism with respect to the movement of users, because more realistic models enable 
more accurate simulation and evaluation of network parameters. There are two broad categories 
of mobility models, trace-driven models and synthetic models. Trace-driven models use recorded 
histories of real users? traces. However, movement traces are unavailable in ad-hoc networks due 
to their decentralized nature. Therefore, we have to employ mobility models that determine the 
movement patterns of mobility nodes synthetically. These models supply the movement 
behaviors of mobile users using particular constraints or mathematical equations [68].   
Other researchers [54] classify mobility models into stochastic and event-based group. 
They state that regardless of the selection of a mobility model, being able to measure the amount 
of mobility is as important as the realism of the model itself. 
To achieve the greatest realism, mobility modeling must take into consideration three 
essential factors [97], which are spatial environments, user travel decisions and user   movement 
dynamics. Moreover, a mobility model must address both regular and random components of a 
user?s movement. 
Signaling cost for maintenance of routes for a MANET is proportional to the rate of link 
changes and can be expressed as a function of the mobility of the nodes. Therefore, the 
performance of a MANET is closely related with the efficiency of the routing protocol in 
adapting to changes to the network topology and the link status. For assessing different routing 
protocols for MANETs, it is important to use mobility model with some index or quantitative 
measure that is relevant to the performance of the network [11, 76, 107]. 
26 
Many authors have used different measures of mobility in their research. In [13, 93] the 
average speed of the nodes is used to represent their mobility, while the maximum   speed is used 
in [39, 107]. The problem with using average or maximum speed as a measure of mobility is that 
the relative motion between the nodes is not reflected. Also, the same average or maximum 
speed in different mobility models or in networks with different physical dimensions often 
results from different rates of route changes [54]. 
Some researchers [39, 56] propose a remoteness function as a mobility measure for 
MANET. The remoteness is generally defined by the distance between two nodes, i and j simply. 
 
               ))(()( tdFtR ijij =
(2?2) 
 
However, a more sophisticated definition is more useful for MANETs. For example, if a 
wireless node has R communication range and is located at a 3R distance, it can be considered as 
remote as a node located at a 10R distance. Similarly, if a node is well within communication 
range R, the node would not seem very remote even if the distance were doubled. The 
remoteness is changeable as the movement of the node may change the wireless link status with 
the node. Based on these observations, they present requirements that F satisfies: 
 
.0   
)()(
    )(
0
)(
lim    )(
0
)(
    )(
0   0
)(
    )(
1)(lim ,0)0(    )(
0
??
=
=
??
==
=
??
=
??
xallfor
dx
xdF
dx
xdF
e
dx
xdF
d
dx
xdF
c
xallfor
dx
xdF
b
xFFa
Rx
x
x
x
  
(2?3) 
27 
Requirement (a) normalizes F to have unity maximum value and Requirement (b) 
guarantees that the remoteness is a monotonically increasing function of distance, and as a result 
0 ? F ? 1 from (a). Requirements (c) and (d) describe the boundary condition of F, which 
guarantee that the remoteness of a node at extreme locations does not change with the movement 
of the node. Finally, the remoteness is the most sensitive to the movement of the node by 
Requirement (e). The signal strength of a node in this study is described by the distance it can 
communicate with other node in the network. So, the remoteness of a node can be measured by 
its Euclidian distance to other nodes. Using this, our heuristic optimizer responds to links at the 
edge of a communication range.  
Ishibashi and Boutaba [39] represent the relationship between mobility and MANET 
topology, where average link lifetimes exponentially decrease with increasing maximum velocity. 
Chu and Nikolaidis [19] analyze the relationship between mobility and connectivity, where the 
higher the velocities, the better the connectivity. Although it seems contradictory at first glance, 
basically it represents the same thing from a different perspective. The lifetime of a link has been 
described with three different definitions. First, the lifetime of a link is the time from when the 
nodes first move into each other?s range so the link can be formed until the link is broken when 
they move out of communication range. However, it does not mean actual time that the link is 
available for use since it has to be detected by a node first to be used. This is true for breakage of 
the link. The second definition is the perceived link lifetime, which represents the elapsed time 
from the first detection to the link breakage detection. This usually extends beyond the end of the 
existence of a usable link. As a final definition they propose the time the link is first included in a 
path by the routing protocol. A failure not due to a network device but the nodes moving out of 
transmission range has the expected time to failure equal to half of the perceived link lifetime. 
28 
Military MANET is usually much more coherent and directed when it is compared with 
commercial ad-hoc networks and nodes will, in general, be more concentrated, rather than 
dispersed. Typical mobility models (e.g, random waypoint) used in most MANET analyses may 
not be sufficient for military MANETs [40]. The random waypoint model is an extension of a 
random walk, in which a user randomly selects a direction and a speed from a distribution within 
the problem space, and then randomly moves to the destination with the selected speed. When it 
reaches the destination, it pauses for some time and then this process is repeated again. The 
movement of one user node in this mobility model is modeled independently from all others [43].  
The mobility of a military tactical battlefield network depends on the nature of the 
assigned mission. Perisa et al. (2007) analyze a military war game exercise and assert that many 
assumptions of random mobility models from the literature do not hold in military MANETs [73]. 
Although this is true, a recent survey of MANET simulation studies found that 65.8% of studies 
used the random waypoint model for mobility [68].  
Chlamtac et al. (2005), in [18], present three essential factors for mobility modeling to 
achieve the greatest realism for military MANETs. There are spatial environments, user travel 
decisions and user movement dynamics. Moreover, a mobility model must address both the 
regular and the random components of a user?s movement. For an accurate evaluation of the 
performance of a protocol, the mobility model must supply a stable movement pattern during the 
simulation time and attain its steady state for most of the simulation time [68].   
In this thesis, two military operation based mobility models and the random waypoint 
model are used to represent military MANETs.  
 
 
 
29 
2.6 Particle Swarm Optimization (PSO) 
Particle Swarm Optimization (PSO) is a population based stochastic optimization tool 
and this optimization technique is generally used for continuous nonlinear functions. PSO was 
first introduced by James Kennedy and Russ Eberhart in 1995 [28, 30, 53, 92].  
PSO is based on social-psychological principles and includes two important concepts: 
experience and knowledge [29]. It has roots in artificial life in general, and particularly bird 
flocking, fish schooling, and swarming theory. It is also related to evolutionary computation, and 
has ties to both genetic algorithms and evolutionary programming [30, 52]. However, there is no 
survival of the fittest concept in PSO.  
In [52] the authors state that "In theory at least, individual members of the school can 
profit from the discoveries and previous experience of all other members of the school during the 
search for food. This advantage can become decisive, outweighing the disadvantages of 
competition for food items, whenever the resource is unpredictably distributed in patches". This 
statement suggests that social sharing of information among conspeciates offers an evolutionary 
advantage: this hypothesis was fundamental to the development of particle swarm optimization. 
 
2.6.1 Basic concept 
PSO is initialized with a group of random particles which are candidate solutions chosen 
randomly within the problem space. Each particle in the swarm is represented by three 
component vectors: X, P and V and two fitness values. The vector X represents the current 
location of the particle in the search space, P records the particle?s best location found so far by 
the particle and V is the gradient vector that defines the direction and magnitude the particle will 
travel if not disturbed. V is used to update X every iteration [26]. On the other hand, the two 
30 
fitness values are the x-fitness and p-fitness. The x-fitness records the fitness of the vector X and 
the p-fitness records the fitness of the vector P.  
These particles interact with one another through the communication structure or defined 
social network. Each particle is evaluated by a fitness function and updated by the best fitness 
obtained by itself and by the swarm at each time step. Each particle keeps track of its coordinates 
in the problem space which are associated with its best solution so far. There are three aspects to 
evaluate solutions as follows: 
? A global best that is known to all and immediately updated when a new best position is 
found by any particle in the swarm.  
? A neighborhood best that the particle obtains by communicating with a subset of the 
swarm (in case of neighbor topology).  
? The local best (or particle best), which is the best solution that the particle itself has been 
forced.  
 
After each iteration, the particle updates its velocity and positions as shown in the 
following equations. As seen below the velocity of particles is updated by one of three different 
equations in equation (2-4) and then the position is updated based on the velocity. The desire to 
better control the scope of the problem space motivated researchers to implement an inertia 
weight (?) or a constriction coefficient (K).  The first equation among those represents the 
standard case that does not use either parameter, whereas the second and third equations are 
modified by using inertia weight (?) or constriction coefficient (K), respectively. We will discuss 
these parameters in more detail in the next section. 
 
31 
 
 
 
)()( )1(22)1(11)1( ?????? ?+??+= tttt XGrandXPrandVV ??  
)()( )1(22)1(11)1( ?????? ?+??+?= tttt XGrandXPrandVV ???  
[])()( )1(22)1(11)1( ?????? ?+??+= tttt XGrandXPrandVKV ??  
 
(2?4) 
   
 VXX tt += ? )1(   (2?5) 
   
        Where,   
? Vt: velocity of a agent at time t 
? Xt: current location of a agent at time t 
? ?: inertia weight 
? K: constriction coefficient 
? ?1: control parameter representing experience information by itself 
? ?2: control parameter representing knowledge information about other agents 
? P: local best of an particle 
? G: global best of the swarm 
 
Each particle modifies its position using its current position, current velocity (Vx, Vy), 
distance between the current position and particle personal best (P) and distance between the 
current position and global best (G). Two main steps in the PSO are initialization of swarm 
particles and fitness evaluation. The pseudo code in the Figure 2-1 shows the basic mechanism of 
PSO.  
32 
 
Initialize swarm particles 
{ 
 X = U (Xmin, Xmax) 
 V = U (Vmin, Vmax) 
 P = X 
} 
Do  
{ 
 )()( 2211 XGrandXPrandVV ?+??+= ??? ??  
  VXX +=
 If (f(X) ? f(P)) then P = X 
 If (f(X) ? f(G)) then G = X 
} While (Stopping criteria not met) 
 
Figure 2-1 PSO basic mechanism 
 
Particles have two essential reasoning capabilities: memory of their own best position 
and knowledge of the global or their neighborhood's best. Members of a swarm communicate 
good positions to each other and adjust their own velocity and position based on these good 
positions [103]. As shown in the pseudo code the fitness evaluation of solutions is performed at 
every time step. The two best locations, P and G, are tracked by the particles while flying 
through the problem space.  
33 
Maintaining the G vector relies on a communication scheme within the swarm. As 
mentioned above, G is the best found in its neighborhood of particles or G is the global best if 
the swarm employs a global neighborhood. Although different neighborhood topologies have 
been studied in the literature, global neighborhoods seem to perform better in terms of 
computational costs [14]. For this reason, in this research, we use a global neighborhood. 
There are two methods to determine the global best, synchronous and asynchronous. 
With the synchronous method, the global best location is updated after updating all particles. On 
the other hand, the asynchronous method determines the global best after each particle has been 
updated. In terms of number of iterations the asynchronous method finds solutions quicker 
because it uses a newly found global best location in subsequent particle updates immediately. 
This advantage usually makes asynchronous updates a better choice for a standard PSO. 
However, the synchronous method is easier to implement than the asynchronous method since it 
does not require evaluation of particles one by one. We use the synchronous method in this 
research. 
PSO has been successfully applied in many research and application areas and often got 
better results more quickly than other methods. There are few parameters to adjust and it can be 
implemented easily and works well in a wide variety of applications with only slight parameter 
variation. The use of real numbers for decision variables leads it to be one of the most popular 
global optimizers [14, 52, 110]. Dengiz [26] develop a strategy optimizing the location of 
autonomous mobile agents in MANETs to maintain network connectivity using PSO. These 
mobile agents try to maximize network data flow, which is formulated as an all-pair maximum 
flow problem.  
34 
Baburaj and Vasudevan [4] propose a new PSO using the On Demand Multicast Routing 
Protocol (PSO-ODMRP) to improve the performance in routing messages in mobile ad-hoc 
networks. ODMRP is a mesh?based demand driven multicast protocol, where a mesh consists of 
a set of nodes, called forwarding nodes, responsible for forwarding data packets between a 
source and a receiver. PSO-ODMRP is well suited for MANETs that have a constrained power 
and frequently change topology. 
Ji et al. (2004) describe how PSO algorithms can be applied to clustering techniques in 
MANETs. Here, the Weighted Clustering Algorithm (WCA), in which cluster heads are selected 
based on the weight of each node, is revised to be suitable for dense mobile nodes. A Divided 
Range PSO, in which particles are divided into groups and each group has four neighborhood 
nodes, is applied for the revised WCA above. It can assign different weights to and consider a 
combined effect of the transmission power, mobility and battery power of network nodes. This 
approach is efficient and effective when the distribution of mobile nodes is dense according to 
the simulation study [44]. 
 
2.6.2 PSO Parameters 
As discussed above, particles update themselves with the internal velocity while going 
through problem space seeking best solutions. Thus, the velocity is the most important factor 
deciding the search process. Particles? velocities are limited by a maximum velocity. So in case 
where velocity exceeds the limit, it is reset to the specified maximum velocity [54]. Velocity 
changes in PSO are due to three parts, social, cognitive and momentum. The balance among 
these parts determines the balance of the global and local search ability. If all velocities are at the 
35 
maximum, particles will tend to search the periphery of the problem space. To control this 
problem PSO has introduced several parameters.  
Shi and Eberhart [91] have shown that PSO searches wide areas effectively, but usually 
lacks precision for local search. Their suggestion for solving this issue is to implement an inertia 
weight (?) in the standard equation that adjusts the velocity correction over time, gradually 
concentrating the algorithm into a local search, as shown in the following equation. 
 
 
iter
iter
current
?
?
?=
max
minmax
max
??
?
?   
(2?6) 
Where,  
? ?max : initial weight 
? ?min : final weight 
? iter
max
 : maximum iteration number 
? iter
current
: current iteration number 
 
The inertia weight is used to balance global and local search abilities. A large inertia 
weight facilitates global searches while a small inertia weight facilitates local searches. The 
introduction of the inertia weight also eliminates the requirement of setting the maximum 
velocity each time the PSO algorithm is used. 
There are other strategies to adjust the inertia weight. Eberhart and Shi (2000) describe 
that a fuzzy inertia weight improved PSO performance. Also, Eberhart and Shi (2001) suggest 
the adaptation of inertia weight with a random component, rather than time-decreasing [33]. In 
[77, 108], the authors assert that an increasing inertia weight obtained good PSO performance.  
36 
Another parameter, called the constriction coefficient, was introduced with the hope that 
it could ensure a PSO to converge to the optimal solution. Maurice Clerc [22] introduced the 
constriction factor (K) that improves the control of the velocities, as given by the following 
equation: 
 
 
??? 42
2
2
???
=K , where ? (=?
1
 + ?
2
) > 4      
(2?7) 
 
The constriction factor approach controls system behavior, and results in convergence of 
the PSO over time. Unlike other methods, the constriction factor approach ensures the 
convergence of the search procedures based on mathematical theory. The amplitude of each 
particle?s oscillation decreases as it focuses on a previous best point. The constriction factor 
approach can generate higher quality solutions than the conventional PSO approach [32, 67, 76, 
91].  
The PSO algorithm with a constriction factor can be considered as a special case of the 
PSO algorithm with inertia weight [22, 23, 31]. With use of an inertia weight or a constriction 
coefficient, Vmax is no longer necessary for damping the swarm?s dynamic. A PSO with a 
constriction coefficient is algebraically equivalent to a PSO with an inertia weight. 
Eberhart and Shi (2000) describe the best approach to use with PSO as a ?rule of thumb? is to 
implement the constriction factor (K) approach while limiting Vmax (Maximum velocity) to Xmax 
(Maximum distance), or implement the inertia factor approach while selecting ?, ?1, and ?2. 
 
 
37 
2.6.3 PSO Topology 
There are static and dynamic versions of topology. The neighbors and neighborhood in a 
static topology are not changed during a run. However, the neighborhood changes in the dynamic 
version topology. The most common topologies applied in PSO are global (star) and local (ring) 
topology [81].  
In star topology, each particle flies through the search space with a velocity that is 
dynamically adjusted according to the particle?s personal best and by the global best by all 
particles achieved so far. In the ring topology, each particle?s velocity is adjusted according to its 
personal best and the best performance achieved within its neighborhood.  
The neighborhood of each particle in the local topology is generally defined as 
topologically near particles to the particle. The global topology also can be considered as a local 
topology with each particle?s neighborhood the whole population. The number of particles in the 
neighborhood set can be expanded or contracted by adding particles that are two positions away, 
three positions away, etc. If the neighborhood set includes every single particle in the swarm, 
then the neighborhood is fully-connected, or global. 
Local topology has the benefit of allowing parallel search, which results in a more 
thorough search strategy. Thus, it might have improved chances to find better solutions but 
converges more slowly than the global topology. Figure 2-2 shows an example of both topologies. 
 
 
 
 
 
38 
 
 
 
 
 
 
 
                    Local topology (Ring)                                       Global Topology (Star) 
Figure 2-2 Common topologies in PSO 
 
2.7 Optimization in dynamic environments 
 The military mobile ad-hoc network is very dynamic due to its movement and variables 
concerning the natural or artificial factors existing in the operation space. Network performance 
is varied by changes in environment or condition, which should be managed by an optimizer. 
Change in environment requires re-optimizing the MANET system and a continuous tracking of 
the changing optimum [9]. Consequently, the optimal solutions are changed over time in the 
dynamic MANET environment. 
Uncertainties in optimization are categorized into four classes by Jin and Branke [45]. 
1) Time-varying fitness functions: The evolutionary algorithm should be able to 
continuously track the changing optimum rather than requiring a repeated restart of the 
optimization process because the fitness function is deterministic at any time point but is 
dependent on time. The most difficult part here is to reuse information from previous 
environments to speed up the optimization process after a change. 
I1 
I6 
I3 
I4 
I0 
I7 
I1 
I7 
I6 
I2 
I0 
I2 
I5 
I3 
I5 
I4 
39 
2) Noise: Sensory measurement errors or randomized simulations in the fitness evaluation 
bring about noise. 
3) Robustness: A common requirement in evolutionary optimization is that a solution should 
still work satisfactorily when the design variables change slightly since the design 
variables are subject to perturbations or changes after the optimal solution has been 
determined. 
4) Fitness approximation: The fitness function can be approximated when the function is 
very expensive to evaluate, or an analytical fitness function is not available. 
 
Military MANET problems are very dynamic. The optimal solution should be 
recomputed to respond to the dynamic changes. Therefore, our military MANET optimization 
problem falls into the first class of the category.  
A MANET continues to face changes of the network topology and state over time due to 
dynamic factors in the environment such as mobility, various variables and obstacles, etc. The 
most distinctive and simplest way to respond to the changes is to consider each change as a new 
optimization problem that has to be solved from scratch. However, it may be inefficient under a 
limited time frame and not even applicable. So, using information transferred from previous time 
step helps increase the search speed after a change.  
There are many different ways to transfer information from the previous search step. A 
common way is to keep the individual particles in the final population of the previous time step. 
However, in order to have populations respond properly and find the new optimum easily some 
special attention should be put on the changes in the environment at each time [10, 26, 45]. 
40 
Similarly, in this study the best swarm particles? information at current time step is transferred to 
next time step.  
The algorithm to solve the problem of optimizing the location for mobile agents should 
be flexible enough to react to network changes even though useful knowledge can be transferred 
from previous search step. Some meta-heuristics have been applied to dynamic optimization 
problems. PSO is one of those. Most meta-heuristics lose their adaptability to changes because of 
convergence during the run. Thus, a successful meta-heuristic should be able to maintain 
adaptability by introducing other metrics besides transferring knowledge [45]. If it is not 
supplemented, the hop-count based approach used in this research may lack adaptability when it 
is not required to maintain network connectivity at the moment. In other words, agent nodes may 
be placed at locations distant from the optimal ones near from future. That is, the mobile agents 
need to be within the proximity of the new optimal locations which need agents for network 
connection. If not, network cannot respond quickly to changes and network performance will 
degrade.  
The maximum available velocity of the agents and other possible constraints limit the 
search in a MANET. Therefore, the mobile agent optimization problem is appropriate to dynamic 
environment solution methods because of these intrinsic characteristics.  
In this research, a PSO with dynamic objective functions is developed to dynamically 
manage the motion of the mobile agents operated under various environments including some 
obstacles. This is expected to improve network performance. We will discuss this in more detail 
in Chapter 4 
41 
 
Chapter 3 
Understanding Network-Centric Warfare (NCW) 
 
Without distinction of field, information has played a key role as a means to improve 
current status or performance in each field throughout history. The 21
st
 century is called the 
?Information Age? since its usage and importance in all fields, particularly military operations, 
has greatly increased. It is clear that that we must consider information as the most important 
means to survive in the competitive world and to assure victory in military combat. There are 
many limitations for military forces to combat effectively in the past when the communication 
and information systems were not developed sufficiently. All communications for information 
sharing or support were done by conventional methods such as express messenger or signal fire, 
etc. As a result, the geographically dispersed force was weak since it is almost impossible to 
respond to any operational changes immediately. In [15], the author states that ?one purpose of 
network-centric warfare is to eliminate the geo-locational constraints by networking the forces 
using the most advanced technologies available.? With the rapid development of information, 
communication and other associated technologies, these limitations are being reduced. 
Today, we can easily see and experience many remarkable changes in our daily life due to 
technological innovations. Particularly, combat performed by U.S. forces in Iraq and Afghanistan 
demonstrated the changes in military operations and the trends in modern warfighting. It is time 
consuming work for a force to defeat an enemy using a poor network system providing only low 
level and delayed information. In this Chapter, we will figure out the importance of maintaining 
the individual network for real time information flow in modern military operation and combat.   
42 
3.1 What is NCW? 
NCW is a new military theory of war pioneered by the U.S. DoD and is the term used by 
military personnel to define information-based warfighting. The Office of Force Transformation 
(OFT) was established in 2001 by the Office of Secretary of Defense to transform U.S. military 
capabilities. OFT describes that ?NCW represents a powerful set of warfighting concepts and 
associated military capabilities that allow warfighters to take full advantage of all available 
information and bring all available assets to bear in a rapid and flexible manner?[64]. 
McKenna [99] addresses that network-centric warfare is a theory which proposes that the 
application of the ?Information Age? concepts speed, communications and increased situational 
awareness through networking improves both the efficiency and effectiveness of military 
operations. 
Alberts et al. (2000) describe that NCW is an information superiority-enabled concept of 
operations that generates increased combat power by networking sensors, platforms, shooters 
and other military forces within a battle space. The increased combat power is obtained by 
shared awareness, increased speed of command, higher tempo of operations, greater lethality, 
increased survivability, and self synchronization [2].  
In [66], authors describe the common picture of the combat situation enabled by NCW, 
which is available for all forces in a tactical network. This common picture can play the critical 
role of reducing uncertainty. In addition, military forces wherever they are, in air, ground and sea, 
once they are connected to the network can self-synchronize, or rapidly and effectively respond 
to combat situations. 
The ultimate objective of NCW is to combine all forces in a network in order to obtain 
information superiority and increased combat power over enemy forces. This can be 
43 
accomplished by employing well developed information and related technologies. Most potential 
power comes from the proper combination of those available technologies. Walter (1997) 
describes three concurrent technical revolutions for the technological combination: information, 
sensor, and weapons technology [100]. The first, information technology, is the most basic one to 
multiply the military forces? power. Information can be produced using information technologies 
such as communication devices and various sensors to collect data from a battlefield. Under well 
developed and combined information technologies, information distribution to forces anywhere 
in the operation area is possible in real-time. The second technological revolution required for 
the network centric operations is in sensors. The technology of sensors pursues smaller, cheaper 
and numerous sensors for real-time surveillance over wide operation areas, so that much good 
quality data can be collected and provided. The third one is in weapons technology. The weapons 
revolution is a matter of increasing numbers of precise munitions by reducing costs. These 
technical revolutions will interact and multiply each other?s impacts and maximize the effects 
that will change the character of war as we know it. These revolutions and changes in how we 
think about war have come to be embodied in the idea of network-centric operations. 
Information was also used in traditional military operations in the past. The application 
level was comparatively lower than in modern military operations. In [95], the author addresses 
the difference between traditional military operations and network-centric operations. The 
traditional operation is performed by a set of steps: assign, plan, generate, and operation, whereas, 
network-centric operation is conducted based on the continuation concept. That is, between each 
step in a set of stepped cycle of traditional operations is a period of relative inaction for 
coordination. Thus, conventional operation cannot keep the continuation between steps. However, 
44 
the network-centric operation is not required to pause before deciding on further action thanks to 
the real-time situational awareness developed by networked forces in the tactical space.  
NCW translates information advantages by a networking force in battle into combat 
power. It also enables more rapid, effective decision making and more precise power deployment 
at all levels of warfare and military operations. Consequently, new forms of organizational 
behavior may be expected under the NCW concept [59, 64]. 
The tactical environment of battlefields has been getting more complicated and requires 
fast reaction to environment changes. Under this circumstance, information superiority over an 
adversary can be achieved by NCW based on well developed information and communication 
technologies. Information through the network system can effectively be collected and 
distributed over large geographical areas in near real-time. As a result of effective use of 
information in military operations, the warfighting paradigm has shifted from mass and platform 
centricity to information and effective network-centricity. Platform-centric forces lack the ability 
to leverage the synergies created through a network, while network-centric forces are more 
adaptive and ready to respond to uncertainty in a very dynamic environment at all levels of 
warfare and across the range of military operations [15, 64]. In short, the collaboration among 
networked forces by enabling the free flow of information and quickly providing it to combat 
units in the battle space increases the forces? power in waging war. 
NCW is also commonly called network-centric operations (NCO). However, we can 
specify the relation between them as NCW is the theory, while NCO is the theory put into action. 
In other words, NCO represents the implementation of NCW.  
The goal of NCO is to maximize the effectiveness of an operation by effective use of 
power, which is called Effective Based Operation (EBO). The meaning of effective use of power 
45 
includes faster, fewer, and lighter weapons, fewer losses, higher accuracy, etc. Commanders can 
decide faster and provide optimal power in a timely manner and control the deployed forces 
better with accurate situational information. Consequently, they can go one step ahead of enemy 
forces and perform combat under more advantageous conditions [64].  
EBO is a trend in military operations and is possible only by network-centric operation. 
In [94], Smith presents that effective-based operations are ?sets of actions directed at shaping the 
behavior of friends, neutrals, and foes in peace, crisis, and war.? OFT also describes that ?EBO is 
a methodology for planning, executing, and assessing military operations designed to attain 
specific effects that achieve desired national security outcomes.? As a result, network-centric 
operations are really about optimizing combat power for effective use [95]. The real payoff in 
network-centric operations is minimizing combat by causing the enemy to yield, or forestalling 
the foes by effective and accurate attacks on the enemy?s critical targets. This efficiency revolves 
around the ability of network centric forces to perform precise EBO which focus on enemy 
behavior. Therefore, these operations are psychological rather than physical. That is why the 
enemy?s decision-making process and ability to take action should be primary targets for attack.  
U.S. forces experienced and proved the importance of effective-based operations based 
on information superiority through two recent combats, Operation Enduring Freedom (OEF) in 
Afghanistan and Operation Iraq Freedom (OIF) in Iraq, as mentioned before. U.S. armed forces 
were networked better based on the relatively advanced technologies than enemy forces in those 
conflicts. A well established network provided U.S. forces with shared awareness, supporting 
forces to be effective and protecting forces from external risks. It eventually enhances their 
lethality and survivability in those operations. That is, a commander of a well networked force 
46 
can quickly develop an operational picture, distributing critical information to his own forces, 
and use power to the maximum effectiveness [64].  
The most important change with NCW is precision and real time data distribution through 
the network. Many advantages for forces operating on the basis of this concept were created. 
Therefore, military forces waging modern war should be able to adapt to new environments 
enabled by the network-centric warfare concept.  
 
3.2 Tenets and governing principles for NCW 
The Office of Force Transformation under the U.S. DoD presents four tenets for NCW 
[64]:  Information sharing based on robust networking, Situational awareness, increased 
collaboration and self-synchronization, and Mission effectiveness. In addition to these four tenets, 
the following principles are also proposed in order to guide the application of NCW: 
1) Fight first for information superiority: All combat starts from information warfare. 
Therefore, regardless of conflict or peace time, information superiority should be 
maintained over the enemy or potential adversary. By using all sources, our information 
capability should be maximized to reduce our own information needs. Conversely, 
increasing the enemy?s information needs, reduce his ability to access information and 
raises his uncertainty.  
2) Shared Awareness: All forces within a network should be able to obtain a common 
understanding and situational awareness. To do so, all information users are responsible 
for posting correct information without delay as suppliers. 
3) Rapid decision, command, and effective operation: Make faster decisions and commands, 
and enhance the capabilities for effective operation on the basis of information superiority. 
Information superiority assured by shared awareness eliminates procedural boundaries 
47 
between services and within processes. It eventually enables the lowest organizational 
levels to be able to conduct joint operations and to achieve rapid and decisive effects.  
4) Self-Synchronization: Adapt rapidly to changes in the battlefield without performing the 
slow functional steps from the traditional military operations. It eventually increases the 
operational tempo by improving the subordinate level of the commander?s common 
understanding about the operational situation.  
5) Dispersed forces and demassification: Move combat power from the linear battles pace to 
non-contiguous operations, and also move from mass centricity to information centricity 
for effectiveness. That is to say, modern combat is conducted simultaneously in many 
locations and maximize the effect with minimum power use. 
6) Deep sensor reach: Detect high quality and valuable information on items or enemy 
forces of interest to achieve decisive effects.  
7) Alter initial conditions at higher rates of change: Change operation conditions at a high 
rate by using information advantage. It confuses the enemy and forces it to redo all 
operation decisions and plans. As a result, the operation speed of enemy forces is delayed. 
 
3.3 Domains of conflict in NCW 
The office of Force Transformation developed a construct for NCW in which four 
domains are recognized, as follows: 
1) Physical domain 
As a traditional domain of warfare, there are four different environments in the physical 
domain, which are the land, sea, air and space. That is, the physical platforms and 
48 
communication networks that connect military forces reside in one of the environments 
of physical domain.  
2) Information 
Information collected, manipulated and shared by networked forces, the commander?s 
intent and command and control of military forces belong to this domain.  
3) Cognitive 
There are many abstract, invisible and intangible components in a military operation such 
as: leadership, training and experience level, fighting spirit, cohesion, and the 
commander?s intent, doctrine, tactics, techniques and procedures. Although these are 
spiritual components, they directly affect combat. That is, the victory or defeat is decided 
by the cognitive level of a military force. The cognitive domain consists of the mind of 
warfighters. 
4) Social 
The necessary human elements for NCW belong to this domain. In this social domain, 
humans interact to exchange information, get situational understanding, and make 
collaborative decisions. A commander?s intent and will is conveyed to the subordinate 
forces.  
49 
 
 
Figure 3-1 Domain of conflict [65] 
 
The information and cognitive domains intersect to form shared awareness and the 
commander?s intent is conveyed at this intersection. Speed and access at the intersection between 
the physical domain and the information domain enable a precision force which is very important 
to the conduct of successful joint operations. The intersection between the physical and the 
cognitive domain enables compressed operations. As shown in the figure all four domains 
intersect to form NCW at the very center. 
In the past most innovations that created significant warfighting advantages were 
concentrated on the physical domain, and were translated into the tactical level of advantages. 
Even though those ideas were considered as platform-centric warfare, there were also many 
innovations focusing on information advantages. Today, with the development of communication 
and information technologies, the military warfighting paradigm is shifting from platform-centric 
50 
to network-centric warfare. The implementation of NCW brings about information advantages in 
military warfighting. The common tactical picture created by the combination of advanced 
technologies and networking reduces the uncertainty about the operational situation. As a result, 
successful employment of information under the new war paradigm increases combat power. We 
will discuss more about this in the next section.  
 
3.4 Information warfare 
The physical strength and skills of the players in a sports team decide the game. For 
example a quarterback?s fast and precise passing skill in a football team and a boxer?s agility and 
punch power may directly affect the result of the match. Similarly, the advanced weapon system 
of a military force which has long range and strong power is one of the most important 
components to assure victory in combat. However, this physical level of power strength cannot 
guarantee that they will always defeat their adversary. Actually, we have known some historical 
warfare in which powerful force is defeated by a weaker force. The primary reason of its defeat 
is that it neglected collecting information about its adversary and implementing it, on the other 
hand, the other force used information effectively to overcome its physical weakness.  
Using information in warfare is not new. Sun Tzu [98] and Clausewitz [20] both were 
very emphatic about using it in warfare. Sun Tzu states that if you know your enemy and you 
know yourself, you will win a hundred battles. Also, Clausewitz describes warfare by using the 
term ?fog and friction?. Fog refers to the commander?s lack of clear information and friction 
means physical difficulties to military action. That is, uncertainty and impediments to the 
effective use of military force due to unforeseeable incidents in combat tend to lower the level of 
performance. The traditional role of information in military combat was to identify the enemy?s 
51 
location and power strength. That is, it was just knowledge at the level of intelligence. However, 
thanks to the development of information and communication technology, its role has changed. 
Information itself has become a target or weapon. The identification of the change has given rise 
to the term of information warfare.  
Information warfare is not about technology but about using information. The purpose of 
using information in a conflict is to influence the adversary?s decision making and operational 
capability, and public opinion, in order to achieve specific objectives. 
During the conflict, each side tries to take actions to affect adversary information and 
information systems while protecting its own information and information systems. Today, many 
different types of attack on information systems are available. The most common attack, as an 
external attack, is to spread viruses through the enemy?s network in order to deny his information 
processing. There exist many solutions such as firewalls, intrusion detection and prevention 
systems, anti-virus software, anti-spyware software, cryptographic protocols, and access control 
mechanisms to protect the information systems from these kinds of attacks. However, internal 
attacks by compromised users are always difficult to handle because they bypass traditional 
security solutions. The Libicki model that is referred to most commonly presents seven 
information types of warfare: Command and control warfare (C2W),  Intelligence based warfare, 
Electronic warfare, Psychological warfare, Hacker warfare, Economic information warfare, and 
Cyber warfare [57]. 
 
3.5 Decision making and security in NCW environment 
Schechtmann (2002) analyzed information warfare according to the OODA loop. U.S. Air 
Force Colonel John Boyd developed the OODA loop based on his previous work on a fast-
transient brief which suggests keeping a faster operation tempo than the enemy to assure a 
52 
victory from conflict. OODA is an abbreviation of Observe, Orient, Decide, and Act. The brief 
description of each component is as follows [89]: 
1) Observation: It is the behavior of becoming aware of the operational situation and 
environment by careful and directed attention. 
2) Orientation: It is the process of converting the collected raw data to valuable information 
supporting the forces by using analysis and synthesis. The human brain forms a mental 
image of the environment.  
3) Decision: The information from the orientation phase is analyzed and various options of 
action are formed. The options are considered to the extent allowed by time, and a choice 
is made. 
4) Act: The decision is implemented. It leads to unfolding interaction within the 
environment. 
The OODA is a model to address human decision making. As the first step of the OODA 
loop, the operational environment is scanned for data. This is possible by robust networking of 
forces in the area. All forces within the network share what they know at the current time and 
location. This collected data, or intelligence, is processed further for creating information and a 
mental image of the situation. Then, a decision is made among alternatives identified through the 
orientation, and it is carried out.  
In short, it is possible to perform compressed operations by OODA. The force can cause 
the enemy, to be confused and disoriented by changing the environment rapidly. This confusion 
and disorientation causes the enemy to pause frequently for reprocessing. Consequently, the 
enemy is more delayed in making accurate decisions and is eventually defeated. 
53 
Each party in a conflict tries to confuse the opposite party as much as it can by using all 
possible means. Fog and friction increased by frequent environment change affects the operation 
tempo of the enemy. One of the most practical ways is to destroy the enemy?s communication 
and information system so that its network cannot function properly any more. At the same time, 
our decision making process should be protected from this kind of attack. 
Needless to say, the decision making process is very important, so it always becomes a 
primary target of the enemy. To secure the decision process and its accuracy, it is crucial that the 
network is not vulnerable to friction and fog by the enemy. For example, data collection in the 
observation phase can be obstructed by offensive actions such as physical attack or deceptive 
actions. And the input from observation will have direct effects on the orientation phase and the 
decision phase. Especially, action could be impaired by disrupting communications using various 
negative actions such as jamming the network, delaying and modifying packets, and injecting 
erroneous packets into the network. These security problems have been an issue in both 
commercial and military networks since the internet was commercialized for network 
communications [89].  
Network security is divided into three categories: Content, Communication, and Network 
security. Content security is to protect the content between the two communicating peers, and 
communication security is about protecting the data between each source and destination over 
the network. The difference between content and communication security is that the end user in 
communication security can be more than two users while there is only one in content security. 
Network security is to protect the network itself so that it can perform properly. The main 
purpose of the network is to function as a tool for command, control, and communication. 
54 
Therefore, the network must be secured to protect the decision making process from enemy 
attacks in a network-centric environment.  
In the past, military technologies in most fields have led commercial ones. However, this 
has been reversed and no barrier between the two fields exists. Many IT equipment and weapon 
systems for military use are also used in civilian organizations. Due to this fact, military 
networks are vulnerable to attack since they use open standards and commercially off the shelf 
(COTS) products. Furthermore, the distinctive characteristics of military operation environments 
such as terrain, weather condition, wireless, and enemy forces increase the burden of security.  
Failure in securing the network from various negative activities by the enemy can lead to 
defeat in warfighting. The enemy can easily access the network when security measures fail. As a 
result, the decision making process, information sharing and situational awareness are more 
vulnerable.  
Future military systems under the new paradigm of NCW will be based on a network of 
heterogeneous networks wired and wireless access networks, sensor networks, ad-hoc networks, 
and fixed and temporary backbone networks. Among these, military networks have a close 
connection with wireless networks. We will discuss wireless networks in the following section. 
 
3.6 Wireless networks 
Today, military and commercial networks are increasingly mobile and wireless. 
Furthermore, in the military operation, ad-hoc networks are relied on to ensure communication in 
difficult environments such as a battlefield. Both the military and civilian environments depend 
on the same technology. However, many military requirements are different due to its nature. 
Wireless networks introduced by innovated communication and information technology have 
55 
changed our living pattern and method of waging. Now military forces can operate in a 
distributed manner while maintaining communication. The shared awareness system based on 
the wireless networking brings many remarkable benefits to forces operating under a high level 
of uncertain battle situations. Candolin [15] defines wireless networks in five categories based on 
the communication coverage. The wireless network spectrum is specified as seen in Figure 3-2 
below. 
  
 
Figure 3-2 Wireless networks and their coverage [15]. 
 
Satellite communication at the highest level is considered as the most important 
communication network in modern warfare. The dream of getting connected "anywhere and at 
any time" is able to come true by the full implementation of satellite communication. Especially, 
the direct connection between satellites and networks can maximize the flexibility of military 
operations by information flow in real time. Among those in the lower levels of networks, 
56 
Wireless LAN is useful for a small military force operation and practical for fighting with 
terrorists currently. 
Wireless LAN is a network linking two or more computers without cables in a limited 
local area such as a company or home. WLAN gives users the mobility to move around within a 
broad coverage area when connecting to the network. The transmission range of WLAN is about 
100m, and the rate is specified between 1 Mbps and 108Mbps. WLAN is defined in the IEEE 
802.11 standard and specified into infrastructure and ad-hoc modes. Network nodes in the 
infrastructure mode are connected to a wireless access point which typically is connected to a 
wired network, while in the ad-hoc mode they communicate in a peer-to-peer fashion.  
Wi-Fi, which stands for Wireless Fidelity, is a WLAN technology defined in the IEEE 
802.11 standards and promoted by the Wi-Fi forum, an online community dedicating to the 
exchange of technical or non-technical ideas related to Wi-Fi. One of the benefits of Wi-Fi is that 
it is able to provide high speed internet access with low requirements for transmission power. 
Thus, it is a feasible technology even for small, battery driven devices. It also adds flexibility to 
local area networking, as it supports user mobility. Wi-Fi is the most cost-efficient wireless 
technology today [15, 106]. The disadvantage is the lack of support for QoS, and the lack of 
privacy protection for users.  
Satellites are important communication assets for enabling mobile communications in 
remote areas, as well as for providing imagery, navigation, weather information, missile warning 
capability, etc. A communication satellite is stationed in space for the purposes 
of telecommunications. Modern communications satellites use a variety of orbits 
including geostationary orbits, Molniya orbits, other elliptical orbits and low (polar and non-
polar) earth orbits. As we know, the coverage by satellite communication is huge and even the 
57 
whole earth can be covered. At first satellite communications could not be available to all users 
because of high usage fees and the limitation of two-way high-speed communication requiring 
only short time delay, but the decrease in cost and developed technology has made it possible 
even for individuals to take advantage of satellite communications. Therefore, it is no longer 
limited to governments, military, and large corporations. INMASAT is one of examples which 
have been used for military operations and provides communications over geostationary 
satellites. 
The U.S. Armed forces have used 28 satellites for global positioning support and six 
orbital constellations for Intelligence, Surveillance, and Reconnaissance (ISR): one for early 
warning, two for imagery, and three for signals intelligence [102]. Despite the growing number 
of military satellites, 84 percent of satellite communications bandwidth in the Operation Iraqi 
Freedom (OIF) Theater was provided by commercial satellites. Communication services by 
commercial satellites may face problems related to interoperability and security. The U.S. DoD 
has made efforts to build a satellite-based military internet in the future to reduce or remove 
currently identified network problems. As a part of this plan, the Transformational Satellite 
Communications (TSAT) program is being run by the Air Force. With this program, five 
geosynchronous orbit satellites will be launched and provide warfighters worldwide with high-
speed, high-capacity communications [21, 42, 102].  
 
3.7 Tactical battlefield network 
Speed and precision are vital factors in the modern warfighting paradigm as discussed 
previously. A scout from a remote place, which is very close to an enemy?s camp, is observing 
the camp and flies a small unmanned aerial vehicle (UAV) to collect more detailed information. 
58 
The small UAV flies in the sky over the enemy?s camp and scans targets of interest without being 
detected by the enemy. This data from the critical area is directly sent to a control center in real 
time, through satellite communications, where a decision maker and his staff are located. They 
analyze the data to get exact information on the target and discuss what to do. They are looking 
for the leader of a terrorist organization at the moment. At last, they find the target, terrorist 
leader, and decide to kill him. The mission is assigned to a responsible force in that area. A 
predator UAV charged with a precision guided missile takes off and launches a missile to the 
target area.  
This is a scene from a recent movie. This simply summarizes the important 
characteristics of modern combat. As we identified from the movie scene, the place where the 
scout was operating is a deep location close to the enemy camp but far away from his camp. 
Naturally, no predefined infrastructures are available there. Only satellite communications or 
another wireless network can connect the scout with the control center. This is an ad-hoc network 
which is a collection of nodes that do not need to rely on a predefined infrastructure to establish 
and maintain communications [14]. Therefore, the coverage of an ad-hoc network depends on the 
locations of the network participants. An ad-hoc network is also referred to as a tactical 
battlefield network in the military. Although nodes in the ad-hoc network may be connected to a 
wired infrastructure, ad-hoc network nodes are most likely wireless. 
Military forces in many different operations have experienced the fact that networking is 
always constrained by time. That is, the speed of establishing a network in a tactical space cannot 
catch up with the operational movement of forces in the area. Military forces have already 
moved ahead when network infrastructures are ready for use. This had motivated the military to 
innovate the networking technology for their forces so they can be networked robustly where and 
59 
when needed. The DARPA Packet Radio Networks and the Survivable Adaptive Networks 
(SURAN) programs sponsored by the U.S. government in 1970s and 1980s were the 
predecessors of today?s mobile ad-hoc networks. 
 The forces of a network and given mission usually determine the required network 
technology and size. Members of a tactical battlefield network are diverse in the capability of 
moving speed and transmission range. Human soldiers, trucks, air fighters, and battle ships as 
network nodes have different capabilities. This heterogeneousness is a trend in military mobile 
ad-hoc network.   
 
Figure 3-3 MANET interoperability [82] 
 
In summary, there exist networks of various size and technology in military battle fields. 
The real-time free flow of information vertically and horizontally among networked forces in a 
theater are the first priority to assure victory in waging modern war. NCW focuses on the combat 
60 
power that can be generated from the effective networking of the combat forces. Shared 
awareness and collaboration in a battlefield, which are given by robust communications and 
rapid exchange of data via one or more networks, provide the forces in battle with potential 
power. Well networked forces show the NCW characteristics of speed of command, high tempo 
and responsiveness, massing of effects, cooperative engagement, and self-synchronization to a 
degree that cannot be matched by any non-NCW capable opponent.  
All of the individual warfighting entities such as satellite, aircrafts, UAVs, ships and 
unattended sensors, and all tactical warfighting nodes in Figure 3-3 must be integrated into the 
grid to achieve the goals of NCW. So, all forces and networks are able to maintain the 
connectivity with each other and minimize the uncertainty. Consequently, this NCW-compliant 
integration can provide a common picture of the battlefield to all network nodes, even to the 
tactical edge nodes. Interconnecting these edge nodes will rely upon mobile ad-hoc networking 
(MANET) technologies [69]. Achieving this tactical edge connectivity will depend on the 
development of significantly improved MANET technologies. 
The tactical battle field is the basic level in an entire network. A top command network is 
made up of several tactical battlefield networks. Therefore, as a basic network, each tactical 
battlefield network is required to maintain robust communication with other networks under 
varied military operation environments.  
61 
 
Chapter 4 
Military MANET Description and mathematical model 
 
There are many different types of obstacles in military combat and operation areas which 
limit network performance naturally or artificially. Unlike obstacles due to natural causes such as 
weather conditions or terrain shape some obstacles are intentionally produced by enemy forces. 
We call those kinds of obstacles ?Enemy? in this research. 
A military MANET model representing military operation scenarios including enemies? 
hostile activities in the operation area is developed in this chapter. The optimizer under combat 
conditions should be able to effectively deploy the agent nodes to the best locations so that 
network performance can be maximized every time step. Two decision variables are the agent?s 
direction and velocity. The network performance is evaluated at each time step during the 
operation. The network objective function can be evaluated when the node coordinates of agents 
are known.  
 
4.1 Proposed military MANET operation system 
The mobile agents play key roles in maintaining network connectivity in a network, 
particularly when the client nodes are positioned beyond their communication range with the 
control node. The network performance in this research is mainly evaluated by the number of 
connected user nodes to the control node and the network bandwidth among connected users at 
each time step. Therefore, the proper positioning of the mobile agents to maximize those 
62 
performance parameters is crucial in the optimizing process of military MANETs, since it 
determines the links among nodes in a network.  
The basic operation method for deploying agents is to control their directions and 
magnitudes of velocities. This is possible when the location information of the MANET nodes is 
given. A global positioning system (GPS), which is implemented by many mobile phones or 
mobile network devices, provides all network nodes with valuable location information. The 
mobile network nodes change their locations to perform the given mission in the tactical theater 
over time. However, we assume that paths of users are not known in advance. The military 
MANET should be able to respond to these changes of location of the network nodes. That is, the 
required system should be able to use the location information from previous environments and 
continuously adapt to the change of network environment conditions as well. This is a non-linear 
problem which requires a heuristic algorithm with a continuously changing objective function. 
As a result, the PSO algorithm, which is a population based heuristic with a time varying fitness 
function, is applied to solve the military operation problem.  
We assume that the location information is always available if the network nodes are 
connected with the control node since a GPS can be mounted on the network nodes. And this 
could technically be possible by broadcasting the location information provided by the GPS. 
Also an ESM sensor to detect the direction of jamming coming from an unknown location of 
enemy is available for each network node. 
As we addressed before, four different types of nodes are used in the study to represent 
real military operation scenarios, which are different from commercial ones. Especially, the 
control and agent nodes are considered as service nodes which should be controlled by the 
63 
MANET control system throughout the operation period to maintain network connectivity. These 
MANET nodes are discussed in the following section. 
 
4.2 Military MANET nodes 
For military MANETs, we functionally divide military operation units into three groups: 
command/control, execution, and support. The command/control unit, which is represented by a 
control node, manages all of its assigned subordinate units. Before performing the 
command/control function, it is important to collect data by using available information 
acquisition assets, analyze the data within a given time frame, and distribute information through 
a network for proper and quick decision making. Consequently, effective command/control 
would be possible over subordinate units only through a robust networking of forces in the 
network.  
Execution units, which are represented by user nodes, perform the commands received 
from the control center or the higher unit in the command/control hierarchy. They also report 
operation situations at their current locations to the control center to support drawing a common 
picture of the tactical space, which will be provided for all networked forces. Of course, the 
execution units could exchange valuable information with each other over the network, but the 
decision making of important actions directly related to a given mission is performed through 
communication with the control center.  
Finally, support or service units, which are represented by agent nodes, seek to provide all 
units in an operation network with a tight communication relay service for enabling military 
operation functions such as command/control, information acquisition and exchange, and 
execution required for mission accomplishment. 
64 
The nodes representing network devices in the mobile ad-hoc network are generally 
divided into user (client) and agent (server) nodes in the literature reference. But considering the 
nature of military operation discussed above, those nodes need to be further specified to 
represent realistic scenarios. So, we have extended the general classification with four different 
types of node: control, agent, user and priority node. The classification and brief descriptions of 
each category are given as follows: 
1) Agent node 
The support unit described above is represented by an agent node. The agent node is 
responsible for connecting network nodes positioned beyond their communication ranges in a 
network to maintain the network at the best condition possible. For this purpose, agent nodes 
are repositioned based on the information about the location of the network nodes and the 
communication capabilities using the distances between nodes.  
2) Control node 
The command/control center is represented by the control node. The command/control center 
in military operations plays a key role of controlling agents in the network to perform a given 
mission. Particularly, the control node in this research performs two additional tasks. It is 
responsible for supporting user nodes like other agent nodes and it designates any user node 
required for a tighter connection with the control node as a priority node, which has priority 
for network service. 
 
Unlike traditional MANET models, user nodes are divided into two types, user and 
priority in this military MANET. The user and priority nodes represent a variety of combat or 
forward military units from individual soldier to battalion military unit level or greater. These 
65 
user nodes are free to move according to their mobility characteristics within a tactical theater.  
3) User node 
User nodes represent an individual soldier or a military combat unit at a variety of 
organizational levels. The user nodes require network service while moving around the 
tactical space.  
4) Priority node  
Because of the dynamic and unpredictable nature of military operations, it is impossible 
to guarantee full connectivity among nodes all of the time. So, this unavoidable limitation forces 
the operation planner or manager (military commander) to classify users into priority based on 
the tactical situation. Any user node which is considered as an important node related to the 
tactical situation can be designated as a priority node by the control center. Priority nodes should 
have preference over user nodes to be connected to the control node since MANET has a limited 
number of agent nodes. 
 
4.3 Network states 
The connections between network nodes are decided by their signal strengths and the 
network connection for a network node in this research means that the network node is able to 
communicate with the control. Therefore, any network node which is disconnected from the 
control node is referred to as an isolated node. However, in order to represent effectively the 
network states corresponding to an environment change, more distinctive definitions for possible 
cases are required.  First, the exact definitions regarding isolation are as follows: 
? Isolated user: Disconnected with the control and all agents 
? Isolated agent: Disconnected with the control and all users 
66 
? Isolated user-agent cluster: One or more agents and/or one or more users connected 
together, but disconnected from the control node 
Based on these definitions of isolated nodes, two network states, State 1 and 2, are 
defined in this research. If there is no isolated user in the network, the network is State 1. 
Otherwise, it is categorized into State 2.  
There are two different ways to evaluate the network performance in this research, and 
these depend on the network state. According to the network state, the objective function 
required for the fitness evaluation is decided. That is, objective function 1 in equation (4-7) is 
used to evaluate the network under the network state 1 and the objective function 2 is used in 
equation (4-8) under the state 2. We will discuss more about the objective functions in Section 
4.4.  
 
4.4 Military MANET optimizer 
The movements of the mobile agents are under the control of the location optimizer that 
is responsible for maximizing the performance objectives as a function of the current coordinates 
of the nodes and enemy?s hostile activities in a MANET operation space. The objective function 
gauges the network performance based on predefined parameters such as number of connected 
user nodes, number of connected agent nodes and network bandwidth. These performance 
parameters can be computed using detailed performance measures such as hop-count or 
bandwidth. 
The network performance measures typically show flat behavior over large portions of 
the search space. However, the measures may also show sudden changes in certain regions of the 
search space depending on the network states. In order to overcome this issue, the objective 
67 
function on which network performance is evaluated, should be able to respond to any small 
changes in the network environment. In this research, this is accomplished by using a hop-count 
based algorithm with additional performance metrics which can help improve network 
performance.   
 
4.4.1 Performance metrics 
As we discussed above, the objective function for the evaluation of a network should be 
made up of proper performance metrics which represent the network characteristics and respond 
well to the network changes. In order to do that, some different parameters are employed to 
evaluate the military MANETs in this research and also two different objective functions are 
used for evaluating the networks under different network conditions. The evaluation of network 
performance is conducted by sequential filtering based on the metrics, which are divided into 
two categories, primary and secondary.  
The primary metrics category evaluates the networked level (NWL) of a network. NWL 
includes four metrics, the Number of Connected Users (NCU), the Number of Connected Agents 
(NCA), the Pre-deployed Agent Level (PAL), and the Number of Agent-User Links (NAUL). 
These primary metrics are common in both objective functions.  
On the other hand, the network evaluation by secondary metrics is divided into two and 
depends on the network states mentioned previously. There are two different network states as 
discussed before previously, State 1 and State 2, in this study. The network state at every step is 
decided by whether there is any isolated user node in the network. If any user node is isolated, it 
is State 2, otherwise, State 1. Under the desired network state (State 1) the network is evaluated 
for network quality (NQ) using bandwidth (BW) and control centered (CC) metrics, while the 
68 
network evaluation for State 2 is performed by the Search Range (SR). The definitions of these 
metrics used here are given as follows: 
1) NWL: Networked Level which is computed by sum of NCU, NCA, PAL, and NAUL. 
The priority among these metrics is represented by the assigned weight to each metric. 
The highest weighted parameter is NCU, the lowest one is NAUL. NCU is used as a 
primary performance measure in this study due to its importance in the military. PAL 
is also an important parameter, but it is weighted by a relatively smaller number than 
NCU and NCA since it is just to maximize network connection.  As a result, the 
optimizer can maximize network performance using these weighted differentiated 
metrics. 
2) NCU: Number of Connected Users. The number of connected user nodes which 
maintain connection with the control node at the moment. NCU includes both 
connected user nodes and connected priority nodes (NCP). 
3) NCA: Number of Connected Agents. The number of connected agent nodes which 
maintain connection with the control node at the moment. 
4) PAL: Pre-deployed Agent Level. PAL is the indication of the agent nodes? 
performance. Under PAL, each agent is required to support at least two user nodes 
and the user nodes being supported by that agent are different from those being 
supported by other agents. Similarly, a user node is required to have at least one agent 
support and the agent node supporting a user node should be different from those 
supporting other user nodes. PAL enables the agents to be distributed through the 
network to support more user nodes over time. Each network node meeting above 
69 
conditions represents one unit of PAL. The mathematical formula of PAL is given in 
Section 4.5.?
5) NAUL: Number of Agent-User Links. NAUL is a direct link between agent and user. 
This parameter is used for maintaining a high quality connection within a cluster 
which might be isolated from the main network.?
6) BW: Total bandwidth of a network. Bandwidth represents the signal strength between 
a network node and the control node. It is computed based on the distance between 
the two nodes. We will discuss this in more detail in the next section.?
7) CC: Control Centered. The sum of the distances between the agents and the control 
node. By this metric all agents can be controlled to stay close to the control node as 
much as possible under a given network condition. The purpose of this metric is to 
respond quickly to network needs. For example, an extra agent is needed to be 
deployed to other location to improve network connections when the user has been 
supported by the agent is killed or disconnected for long time. Under this kind of 
situation, CC metric help deploy the agent to any proper location.  
8) SR: Search Range. The distance between a messenger and the estimated destination 
of an isolated user. The messenger is an agent that is designated for isolated user node 
search by the control node. We will discuss this later in this chapter.  
The optimizer continues to evaluate the network based on the above performance metrics 
and optimize the location of agent nodes to keep maximizing the network connection during the 
operation. In the following section, we will discuss a hop-count based algorithm and how to 
compute the network bandwidth. Bandwidth is used to check the connectivity and quality of the 
path between two nodes in a network.  
70 
4.4.2 Hop-count based algorithm 
Finding the shortest path from a source to a destination in a network is a common 
problem and there are many ways to do it. Here, the meaning of ?shortest path" may be the least 
number of hops (links or arcs) or the least total weight [58, 96]. In [68] the shortest path problem 
is simply defined as the problem of finding a path between two nodes such that the sum of 
the weights of its constituent edges is minimized.  
For finding the shortest path between two designated nodes in a network, the minimum 
hop-count is used in this research. The hop-count approach considers only the minimum number 
of hops, or links, required to establish a path between two nodes regardless of the link quality or 
stability or distance. As described before, it can minimize the waste of limited resources in the 
longer term. However, it may select an inferior route when there are more than two paths 
available, since it considers only the minimum hop. In order to improve this measure we 
introduce other metrics representing network quality and stability as subsidiary ones. The 
shortest path is evaluated by the primary metric, hop-count. If there are two or more shortest 
paths which have same number of hops, the tie among the paths is broken using the second 
metric based on the bandwidth. Each link?s bandwidth for fitness evaluation is different from the 
one measured by the distance alone. In order to measure the level of the quality and stability of a 
link, the jamming Effect Level is also used. The description of these metrics will be given in the 
following section. 
Bandwidth (not including Jamming Effect Level) is the most common metric to measure 
the network performance and is measured as a bit rate expressed in bits/s or multiples of it (kb/s, 
Mb/s etc.). We use bandwidth to measure the quality of a path. Connections among nodes in a 
network depend on the transmission ranges based on each other?s signal strength. That is to say, 
71 
the transmission range of a link between two nodes is decided by the one having the smaller 
transmission strength. 
The basic goal of the links in a network is to deliver sufficient signal strength from a 
source node to a destination node to achieve some performance requirements. However, signal 
strength could be varied by environmental conditions. The estimation of the signal power under a 
great variety of existing conditions is not easy work and it may be impossible to predict exactly. 
Also, this is beyond our research scope. Therefore, we assume that the estimation of the path loss 
in this research follows the propagation model for free space scenario. However, this propagation 
model could be replaced by another propagation model corresponding to the application. 
To represent the tactical battlefield network in the lower level of military mobile network, 
the wireless local area network is considered in this research. The wireless IEEE 802.11 is a set 
of standards for wireless local area network computer communication. It is technically possible 
to create a multi-hop network that covers several square kilometers and operates in the 5GHz and 
2.4GHz radio band [26]. The free space path loss equations are given in two ways as seen in 
following equations depending on the distance measure unit, km or mile. 
 
 km)in   MHz,in  (   log20log204.32 ijijp dfdfL ++=  (4?1) 
       or miles)in   MHz,in  (   log20log206.36 ijijp dfdfL ++=  (4?2) 
 
Where Lp is the path loss in decibels (dB), dij represents the distance between two nodes 
and f is the signal frequency. 
For the path loss model, Dengiz (2007) applies equation (4-1) to an industry company?s 
product specification sheet to compute the data transfer rate versus distance [26]. The 
relationship between path loss and data rate is shown in Table 4-1. 
72 
Table 4-1 Path loss vs. data rate 
Data rate 
(Mbps) 
        Receive Sensitivity 
(dBm) 
54
48
36
24
18
12
9
6
2
-75
-76
-80
-84
-88
-90
-90
-93
-93
 
For a normalized wireless transmission range, he uses equation (4-3) to estimate the 
normalized data rate from the Euclidean distance (dij) between two nodes, i and j. To calculate 
the normalized bandwidth of each link the following equation is used in this research as well. 
 
 
           
( )
()
1
5.010
1),(
?
??
+=
ij
d
ejiDataRate  
(4?3) 
 
So, once we get the Euclidean distance between nodes, the available data rate can be easily 
computed using equation (4-3). 
There is another factor affecting network bandwidth in military MANET scenarios, and 
this is an electrical attack (jamming) by enemy forces existing in the operation area. The 
communication capability of the network nodes are constrained by this jamming effect. In 
addition, nodes which are located near jammers could be attacked and killed by their attack. This 
73 
is referred to as the Kill Effect Zone (KEZ). We will discuss enemy effects in more detail in 
Section 5.4.1. 
 
4.5 Mathematical formulation 
In order to construct the network graph G = (N, E) at each step, the Euclidean distances 
among network nodes are measured.  
 
 22
)()( jtitjtitijt yyxxd ?+?=  
(4?4) 
If the Euclidean distance between two nodes in a free space environment is less than the 
minimum possible communication range of those, the nodes are connected to each other as 
shown in the following equation. 
 
 
( )
?
?
? ?
=
otherwise  0
  dTR,TR inm if   
e
ijtjtit
ijt
1
 
(4?5) 
 
In the real world, the transmission capabilities of network nodes can be limited by 
environmental conditions as discussed. Although there are many possible factors causing this 
limitation such as terrain shape, weather condition, and electrical attacks by enemy forces. All of 
these are not studied in this research. We consider only the jamming effect by enemy forces. The 
possible bandwidth for a link is computed based on the Euclidean distance and jamming 
effective level. And then if the possible bandwidth for the link is larger than the minimum 
bandwidth required under current environmental condition, they can be connected with each 
74 
other. However, the minimum bandwidth is   assumed as one distance unit here since we do not 
consider reduction of bandwidth by environmental conditions in this research. 
 
Notation 
r
it
   Rotation angle from x-axis of the i
th
 agent node at time t      
v
it
   Speed of the i
th
 agent node at time t  
v
max  
Maximum speed of the agent node 
(x
it,
 y
it
)  x and y coordinates of the i
th
 agent node at time t 
(X
min,
 X
max
) x-axis boundaries 
(Y
min,
 Y
max
) y-axis boundaries 
M Weight for the swarm having at least one of its particles within the kill effect 
range. If all particles are out of the kill effect range, M is 0  
NN   User and agent node set 
AN  Agent node set 
UN  User node set 
CN
  
Control node 
IU
t  
Isolated user node set at time t 
PN
t  
Priority node set at time t  
MA
t  
Messenger agent set at time t 
TN
t
  Total number of alive user or agent nodes at time t 
TU
t  
Total number of alive user nodes at time t 
TA
t  
Total number of alive agent nodes at time t 
TP
t  
Total number of priority nodes at time t 
75 
d
ijt
  Euclidean distance between node i and j at time t 
TR
it
  Transmission range of node i at time t 
Pn      Weight for priority node 
e
ijt
        e
ijt
 is 1 if there is a link (single hop) between node i and j at time t, otherwise is 0 
P
t 
(i, j)
 
P
t 
(i, j) is 1 if there is a path between nodes i and j at time t, otherwise is 0 
SU
it 
The set of user nodes supported by agent node i at time t 
SA
it 
The set of agent nodes supporting user node i at time t 
BW
ijt
 Bandwidth of a link between node i and j at time t, i?j  
JL
ijt
 Jamming effect Level for the link between node i and j at time t, i?j 
ShortestPathBW
t
(i,j)  Shortest path bandwidth between node i and j at time t  
)j,i(PathBW
k
t
 k
th 
path bandwidth between node i and j at time t
 
)j,i(N
k
t
  A set of nodes on the k
th
 path between nodes i and j at time t 
 
The purpose of the optimizer is to deploy the agents into the best locations in order to 
provide MANETs with maximum communication quality. To meet this goal, different metrics 
representing military MANET characteristics are employed to evaluate network performance; 
these are hop-count, bandwidth, and jamming effect level. 
For instance, let?s say we want to find a path between nodes i and j. First, the link matrix 
is constructed based on the location information. Two different matrices are required to compute 
the cost for a link or path in the network, which are one hop-count matrix and bandwidth matrix.  
The hop-count matrix is used to search for the shortest hop path, while the bandwidth 
matrix is used to find the highest quality or cheapest cost path among the shortest paths which 
have the same number of hops. The link between two nodes is represented by the binary value (1 
76 
or -1) in the hop-count matrix. On the other hand, the cost of each link in the bandwidth matrix is 
represented by the computed bandwidth based on Euclidean distance and jamming effect level. 
As a result, each cost in the bandwidth matrix is real number representing the possible bandwidth 
of the link for fitness evaluation.  
The hop-count based algorithm first finds the possible paths between nodes i and j using 
minimum hop-count. The path which has the smallest number of hops is selected as the shortest 
path. If there is only one path, it will be the best route without requiring further consideration. 
However, in many cases, there exist two or more shortest paths which have same number of hops. 
So, the tie among the possible paths should be broken by the bandwidth of each path. That is, the 
path with the fewest hops and which has the largest bandwidth is selected as the best path. 
Each link?s bandwidth is calculated by equation (4-3). Euclidean distance between the 
nodes is first measured. Then, for the actual possible bandwidth calculation for fitness evaluation, 
jamming effective level, which is changed over time, is added to the computation. A path 
between any two end nodes could be composed of several different links. Among those, the link 
which has the smallest bandwidth is chosen for the path?s possible bandwidth. Finally, the path 
that has the smallest hop count and the largest actual bandwidth will be the best route for the 
connection between two end nodes, source and destination.  
Network performance depends on the location of the agent nodes, so the optimizer 
continually relocates the agent nodes to the locations which best improve the network 
performance. Consequently, the agent nodes? coordinates are decided by the direction and 
velocity randomly generated by PSO. So, these direction and velocity of agent are decision 
variables. Particularly, in the measurement of network performance, a priority node path is 
77 
weighted by a positive value (Pn) to represent its priority to the military operation, as described 
earlier.  
As we discussed, several network parameters are used to evaluate network performance 
and these parameters need to be differentiated by their importance in a network. So, different 
weights are introduced to represent these priorities for network connection as seen in equation 4-
6. Network parameters are NCU, NCA, PAL and NAUL in order of importance and weighted by 
?
1
, ?
2
, ?
3
 and ?
4
, respectively. Also, while optimizing a network if any particle in a swarm enters 
the kill effect range of an enemy, network performance at that moment will be penalized a big 
positive number (M). As a result, agent nodes would not be deployed into the kill effect range. 
The mathematical model based on the above concept is given as follow: 
 
 
(){}     : MNAULPALNCANCUmaxO ??+?+?+? 43211 ????  
(4?6) 
 
) (if   12 ?=
?
?
?
?
?
?
? tIU
CC
BW
max:SO
 
(4?7) 
 () ) (if       22 ??? tIUSRmin:SO
 
(4?8) 
 
??
??
?+=
UNiPNi
t  t
t
Pn)CN,i(P)CN,i(P NCU
 
?
?
=
ANi
t )CN,i(P NCA
 
(4-9) 
 
 
?
?
?
?
?
?
+=
??
==
ttTU
i
TA
i
ii aPALuPAL PAL
11
 
(4-10) 
 
?
??
=
UNjUNi
ijteNAUL
,
 
(4-11) 
 
?
?
=
NNi
t )CN,i(thBWShortestPaBW  
(4-12) 
78 
 
        
{ }     )CN,i(PathBWmax)CN,i(thBWShortestPa
k
t
k
t =    
 
 
        
{ }mntmnt
)CN,i(Nn,m
k
t
JLBWmin)CN,i(PathBW
k
t
?=
?
  
 
 
?
=?
=
CNj,ANi
ijtdCC
 
(4-13) 
 
?
??
=
tt MAjIUi
ijtdSR
,
 
(4-14) 
 
Subject to 
 
ANr
ANvv
iit
imaxit
        
    
????
????
?20
0
 
(4-15) 
   
 
?
?
?
?
?
???+
??>?+
??<?+
=
?
?
?
?
?
???+
??>?+
??<?+
=
+
+
AN,vrsiny
AN,Yvrsiny,Y
AN,Yvrsiny,Y
y
AN,vrcosx
AN,Xvrcosx,X
AN,Xvrcosx,X
x
iititit
imaxitititmax
iminitititmin
ti
iititit
imaxitititmax
iminitititmin
ti
                                                                
         
          
)1(
                                                                
     
      
)1(
otherwise   )(
)(         if                    
)(         if                    
otherwise   )(
)(          if                    
)(         if                    
 
(4-16) 
   
 
?
?
? ?
=
otherwise        1  
  node,agent any for   toidentical is  if        0   j   iSUSU
aPAL
jtit
i
 
?
?
? ?
=
otherwise        1  
node,user any for    toidentical is  if        0  j iSASA
uPAL
jtit
i
 
(4-17) 
 
  
The performance evaluation of a network is conducted by two different objective 
function sets based on the network states, as discussed before. The objective function set 1 
79 
80 
consists of O1 and O2-S1 and set 2 is made up of O1 and O2-S2. The objective function set for 
the evaluation is determined by the network state at the moment. That is, network performance is 
evaluated by set 1 if no isolated user node is in the network. Otherwise, it is evaluated by set 2.  
The sequential filtering method is used for the network evaluation in this study. First, 
network is evaluated by O1 and is compared with the previous fitness evaluation. If it is 
improved by O1, then O2-S1 or O2-S2 does not matter. However, if the current fitness evaluation 
by O1 is the same as previous one, the tie should be broken by O2-S1 or O2-S2 based on 
network state. O2-S1 in Equation (4-7) can be maximized by minimizing the distance between 
agents and the control node (computed by Equation (4-13)) and maximizing the network 
bandwidth (computed by Equation (4-12)). By this objective function, the idle agents can be 
deployed to other location to support other network connections.  
If any user node is isolated from network, the messenger is operated to reconnect the 
disconnected user. This is represented by Equation (4-8). The NCU, NCA, PAL  and NAUL 
metrics are computed using the equations, (4-9) through (4-11). Especially, to compute the PAL, 
first aPAL
i
 or uPAL
i
 should be computed by Equation (4-17). Equation (4-11) is used for the 
isolated cluster, by which agents in the cluster support the users in that cluster instead of moving 
toward the control node to be deployed other locations. 
Finally, the location of each agent every time is decided by the decision variables, their 
velocity and direction, and constrained by the boundaries of operational area as described in 
Equation (4-16). 
 
 
Chapter 5 
Military MANET System and simulation environment 
 
The optimization process for routing agents to the best locations is divided into three 
main phases. The first phase is to read location information. The control center collects 
information about the tactical operation in the assigned area such as the network nodes? locations 
and enemies. The connection matrix is constructed based on this information. Second, the 
optimizer performs the network performance evaluation using the predefined objective functions. 
Finally, the agent nodes will be deployed by either the PSO or a self deploy rule. If the agent 
node is connected to the control node, it will be deployed by the PSO. Otherwise, it will self 
deploy toward the center of future estimated locations of the last connected users or to support 
the users in a cluster.  
 
  
 
 
 
Phase I 
Read Location Information 
Phase II 
Fitness Evaluation 
Phase III 
Deploy Agents 
Figure 5-1 Main optimization framework 
 
81 
The main decision making processes for deploying agents are shown in Figure 5-1. The 
loop of these three phases of optimization process will be repeated during the simulation run. 
The detailed control logic is executed based on this main framework.  
 
5.1 Routing by network state 
There are two different network states in this study and the optimization process is 
differently applied by network state. As we discussed before, network performance is evaluated 
differently when there is an isolated user. This is the search effort to reconnect an isolated user in 
a real operation and thus to improve network performance. 
First, it is checked if there is any isolated user node in the network. If all user nodes are 
connected with the control node, the network is evaluated by the objective function O2-S1, 
otherwise it is evaluated by objective function O2-S2 again. Then, all candidate locations for the 
agent nodes are evaluated and agents are deployed to the best location of those candidates. 
Different deployment rules are applied according to the agent status at the moment. That is, if an 
agent maintains connection with the control node, it will be deployed by the PSO logic. If the 
agent does not meet above condition, it will be deployed by the isolated agent deployment rule. 
We will discuss this in more detail in Section 5-3. The optimization framework by the network 
state is given in the following Figure 5-2. 
82 
 
Figure 5-2 Optimization by network states 
?
5.2 User node mobility model 
The movement pattern of user nodes in a network depends on the given mission. The 
random waypoint model is very suitable for representing mobile phone users who move freely 
around the service area. It has also been implemented for military applications.  However, some 
military operations, such as search & rescue and convoy, require more suitable to these 
environments.  
All user nodes move toward their assigned destinations from the initial locations with a 
random velocity.  In a real military operation, combat units cannot directly move to their 
83 
destinations due to the dynamic surrounding conditions which may limit its movement. To 
represent this two parameters, the perturbation level (?) and the rotation angle (?) for a user node, 
are employed so that the movement of users in the network can simulate a real world operation. 
The user direction is calculated by the direction to current destination and the 
perturbation direction.  First, the direction to the assigned destination from the current location 
( ) is calculated by equation (5-1).  
?
itTarDir
 
 
)().,.(
)().,.(
itityitit
itityitit
it
yxnDestinatioxnDestinatio
yxnDestinatioxnDestinatio
TarDir
??
??
=
?
      
(5-1) 
 
All user nodes under dynamic environmental conditions cannot directly move to their 
destinations, but take random paths every step. That is, a user?s direction is perturbed by a 
uniformly distributed random number between (-?/2, ?/2) (Rotation angle (?) = U(??/2, ?/2) and 
the perturbation level weight ?. These parameters can be adjusted for a given application. The 
perturbation direction for a user is created by following processes. 
 
 
?
?
?
?
?
?
?
=
??
??
cossin
sincos
RotMatrix
      
(5-2) 
 
RotMatrixTarDirPerDir itit ?=
??
    
(5-3) 
 
As shown in equation (5-3), the perturbation direction is computed by the direction to the 
assigned destination and the rotation matrix. After randomly generating a rotation angle ?, the 
perturbation direction is computed using the destination direction already computed and the 
84 
rotation matrix. Finally, the user node direction at each time step is calculated by following 
equation: 
 
 
?? ?+??=
???
ititit PerDirTarDirUserdir )1(
      
(5-4) 
 
Where, ? is probability weight for the perturbation direction and 1- ? is the weight for the real 
destination direction. Although ? is set by 0.1 in this study, it can be changed by user?s 
application. The location of a user from the previous time step is updated by the new user 
direction and the user speed as randomly generated by equation (5-5). The user node?s velocity is 
also generated by a randomly distributed number between a preset minimum and maximum.  
 
 max)(min,UUserSpeed =
 
).sin(..
).cos(..
)1(
)1(
?
?
?
?
?+=
?+=
ittiit
ittiit
yUserDirUserSpeedyUserlocyUserloc
xUserDirUserSpeedxUserlocxUserloc
 
 
(5-5) 
Where,                              
   : Destination direction for user i at time t 
?
itTarDir
   : Perturbed direction by Rotation Matrix for user i at time t 
?
itPerDir
   : Direction for user i at time t 
?
itUserDir
 
After updating the user?s location at each time step, a check is made to see if the user 
node has arrived at the assigned destination. If so, another new destination will be generated for 
the user node. There exist only two cases that require reassigning a user?s destination. One is 
arrival to the destination and the other is arrival to the boundary of the problem space. The user 
85 
node mobility in this study is represented by either random waypoint or directed movement 
pattern which is constrained by predefined destinations or by movement within a responsibility.  
 
5.3 Isolated agent behavior 
During MANET operation, an agent node could become disconnected with the control 
node and if it is not a member of an isolated user-agent cluster, this node is referred to an 
?Isolated agent?. These isolated agents cannot be controlled by the control node. Therefore, an 
instruction for reconnecting the isolated agents to the network as soon as possible is required for 
better network operation. Under this rule, an agent node can think and behave intelligently to 
find the right location to contact the network again by itself when it is isolated from the network. 
This kind of agent behavior is called ?semi-intelligent agent behavior? [26].  
An agent keeps track of the location of users which are connected with it by a single hop 
and this accumulated data is used to estimate the users? locations at four time steps later. The 
center of the predicted location of the last connected users will be the most attractive direction 
for the isolated agent in this research.  
The isolated agent moves toward the center of the last connected users? location at four 
times in advance with maximum velocity by this self deploy rule until it connects with any user 
node in the network. 
 
5.4 Enemy behavior 
There may be many obstacles limiting MANET performance in a military operation area. 
Because of obstacles, the transmission ranges and movements of network nodes may be 
constrained. However, it is very difficult to measure the impact of these obstacles accurately and 
86 
it may be impossible since it is the result of mixed effects of weather condition, terrain shape, 
jamming by hostile forces and other surrounding conditions. There are also many different 
obstacles in a military combat operation area which limit the network performance naturally or 
artificially. Unlike obstacles due to natural causes such as weather condition or terrain shape, 
some obstacles are intentionally installed or deployed to those areas by enemy forces. The 
obstacles in the artificial obstacle category are employed in this study and termed Enemy. 
 
5.4.1Enemy effective zones 
Enemies are fixed at the initially assigned locations without moving throughout the 
simulation time and its mission is to degrade the communication capability of the network nodes 
and destroy those positioned within the kill zone. These effects are shown in Figure 5-3. Two 
different enemy?s effective zones based on the distance between a MANET nodes and an enemy 
are introduced to represent enemy?s effects in this study: the Jamming zone and the Kill zone, 
and the distances, d
1
 and d
2
, are uniform widths. The effective zones are defined with a radius 
circle from an enemy and jamming zone includes the kill zone. That is, any network node 
entering the effective zone first receives the jamming effect and then it may get physical attacks 
within the kill zone by the enemy as it nears to the enemy. The jamming effect level is 
determined by the Euclidean distance between the network node and the enemy, while the kill of 
a network node in the kill zone occurs randomly using a predefined probability based on a 
potential combat power. 
 
87 
 
Figure 5-3 Enemy effect zones 
 
As combat units, users can also attack and kill the enemies. The combat powers of the 
user and the enemy are represented by probabilities. We will discuss this more in detail in 
Section 5.7. The effective range and relative combat power of a resource (network node or 
enemy) could be adjusted accordingly to the military application. 
 
5.4.2 Jamming effect 
The signal strength represents the communication capability that a MANET node can 
communicate with other in a network and it is reduced by the jamming effect of enemy. The 
communication capability over a link between two nodes in a network decreases as the distance 
between them increases. In order to measure the jamming effect level for a link we use the 
relationship between signal strength and the distance between nodes constructing the link. The 
equation (5-6) shows the relationship between the signal strength and the distance under a free 
88 
space environment. This is not exact method but it is simple and a good approximation for 
appropriate the jamming level over a link [35].  
 
 
                                  
2
)/4( ??dPloss =
(5-6) 
 
As seen in above equation, the loss of signal power (P
loss
) is decreased as the distance (d) 
between two objects increases. This relationship is applied for estimating the jamming effect 
level in the research. 
In order to limit a network device?s capability, the jamming strength should be greater 
than its signal strength and this is common in a real military operation. For example, let?s say the 
ratio of signal strength of jammer to network device is 3:2. That is, if the communication range 
of a network device is 1, the jamming effect range is 1.5. The jamming effect range is usually 
much longer than the network devices transmission range. Accordingly, if any node of a network 
link is within the jamming effect range, the performance of the node and link is degraded as 
much according to the distance to the jamming obstacle. The closer network device approaches 
the jammer, the stronger the jamming effect. 
As a result, a node?s communication capability is decreased and a poor quality of service 
by the reduced bandwidth. The jamming effect is normalized by the distance between the jammer 
and the network device as shown in Table 5-1. The table is developed under the assumption that 
the signal strength ratio between the jammer and the network device is 3:2. This ratio is applied 
to the scenarios in the study.  
 
 
89 
Table 5?1 Distance vs. Jamming effect level 
d d
2
 
normalized 
jamming effect 
0.1 0.01 1.00 
0.2 0.04 0.98 
0.3 0.09 0.96 
0.4 0.16 0.93 
0.5 0.25 0.89 
0.6 0.36 0.84 
0.7 0.49 0.78 
0.8 0.64 0.72 
0.9 0.81 0.64 
1.0 1.00 0.56 
1.1 1.21 0.46 
1.2 1.44 0.36 
1.3 1.69 0.25 
1.4 1.96 0.13 
1.5 2.25 0.00 
 
 
5.5 Prediction of user future location 
MANET users are free to move within the given operation space. This causes some users 
to disconnect from the network during operation due to the limited number of agents. In order to 
maximize the chance for reconnecting an isolated user, we need information about the likely 
location of the isolated user.  
Dengiz (2007) used the laws of kinematics to predict the future location of users in his 
research. Kinematics has commonly been used to estimate the future location of an object in the 
literature [26]. The location of a network node is described by its x and y coordinates. The 
change of location can be described by velocity and acceleration. That is, velocity is described by 
the magnitude of the position change and acceleration is represented by the rate of change of 
90 
velocity. These two kinematic factors are used to predict the future location of an isolated user in 
this research as well. 
In this method, the only data required for the estimation of an isolated user?s future 
location is its position history for three previous time steps. The basic mechanism of future 
location prediction based on the kinematics is as follows [26]: 
 
 
            
      
1
211
322
),(),(
),(),(
),(),(
?
???
???
?=
?=
?=
ttt
ttt
ttt
yxyxv
yxyxv
yxyxv  
(5-7) 
 
Velocity, the magnitude of the position change for three time steps, is first computed by 
equation (5-7).  v
t
 indicates the velocity at time t and similarly (x, y)
t
 means the location of a user 
at time t.  
 
 
            
t
vv
a
t
vv
a
tt
t
tt
t
?
?
=
?
?
=
?
?
??
?
1
1
21
2
      
 
(5-8) 
 
            
t
aa
a
tt
t
?
?
=?
??
?
21
1
      
(5-9) 
 
Based on the computed velocities, the acceleration of a user is calculated as shown in 
equation (5-10) and (5-11).   
91 
 
 
                  
taaa ttt ??+= ??? 11?
(5-10) 
 
                  
tavv ttt ??+=+ ?? 1
(5-11) 
 
            
tavxy ttt ???+=? ?
2
1
      
(5-12) 
 
Equation (5-10) shows the computation of predicted acceleration ( ) at the next time 
step. Then, the estimated velocity ( ) and the magnitude of the position change ( ) can be 
computed by Equation (5-11) and (5-12), respectively. 
ta?
1? +tv txy?
 
 
              
tt
P
t
xyXYXY ?+=
+1
(5-13) 
 
              
HxyXYXY tt
P
Ht
??+=
+
(5-14) 
 
Finally, the predicted location of a user at the next time step ( ) is calculated by 
equation (5-13). If we want to predict a user node?s location at the t+10 time step, it can be 
computed by multiplying by the prediction horizon (H) 10 as seen in equation (5-14). 
P
t
XY
1+
txy?
 
5.6 Messenger behavior 
 During MANET operation, user nodes may be isolated from the network as we discussed 
before. When a user node is isolated, the control center tries to reconnect the isolated user if the 
network has any remaining capability, with which agents can perform other missions without 
degrading current network performance, for supporting the isolated user.  
92 
The control center designates an agent to search for the isolated user by choosing the 
closest agent from the last contact point and multiple messenger agents could be possible 
depending on the number of disconnected users. The pseudo code for a messenger agent is 
shown in Figure 5-4. This designated agent is referred to as a messenger. The messenger searches 
for the isolated user by using its predicted location. However, this search activity is permitted 
only for given time by the control center. Therefore, after the allowed time for search, the 
messenger is deployed to other location to support other network nodes if the isolated user is still 
missing. 
 
 
Start {   
// for any disconnected user node j 
Set  SearchTime = st 
Set  NetState = 2         
Set MinD  = Big Value // Initial minimum distance 
for (AN
t
=1 to AN
t
=n) { 
 Dist
n
 = distance (UN
j
, AN
n
) 
 if (Dist
n
 < MinD) { 
  MinD = Dist
n
     
  Messenger = n 
 } 
} 
while  (Search Time Steps < st) { 
 Fitness Evaluation by O1 and O2-S2 
} 
} End 
Figure 5-4 Pseudo code for a messenger 
93 
We expect that the messenger helps improve network performance by reconnecting the 
isolated user. However, it may worse the network performance since it requires operate some 
messengers among the limited number of agents under the circumstance that cannot even 
guarantee the reconnection of the isolated users. As a result, the search time for a isolated user 
should properly be determined by considering the mission and network performance. 
 
5.7 Network node behavior in an enemy effect zones 
In order to represent the different behavior of network nodes depending on the node type, 
we divided network nodes into four categories as discussed in Chapter 4.2. 
All network nodes receive a jamming effect once they get into the jamming zone. Even 
though each network node can recognize its reduction of communication capability as it 
approaches an enemy, it is difficult to identify the location of enemy. However, the enemy 
location can be estimated using tactical information about the enemy and the jamming effect 
level. In this research, it is assumed that MANET node has an ESM sensor to detect the bearing 
of a jammer and two networked nodes within the jamming effect zone of an enemy are at least 
required to identify the location of an enemy. Also, any network node entering the kill effect 
zone of an enemy without knowing it can automatically detect the enemy. 
User node?s movement in the kill zone is not stopped by the enemy, that is, user node 
tries to fight against the enemy on the way its destination. On the other hand, agent nodes try to 
escape from the zone because of their weak combat power even though they have a combat 
capability. The combat power of each resource in this study is assumed as shown in Table 5-2. 
The kill probability of resources is determined by the distance between them, a user and an 
enemy or an agent and an enemy. That is, the kill probability of a user within a kill effect zone 
94 
increased as the user approaches an enemy. This kill probability is a function of index probability 
and distance to an enemy. The index probability for both user and agent node is differently 
defined here, which would determine the combat level in a battlefield.  
 
Table 5-2 Combat power of resources and Index probability 
Resources User Agent Enemy 
Combat power 1 0.1 1 
Index probability 0.05 0.2 - 
  
By using the index probability and the distance between resources, the kill probability 
can be computed by the equation below. 
 
        Kill probability (p) = 
?
?
?
?
?
?
?
?
? yprobabilit Index  
)(d resources between cetandis
Range Effect Kill
  ,min
ij
1  
(5-15) 
 
Figure 5-5 shows the kill probability between a user and an enemy resource when they 
are within a kill effect range together. Particularly, within 0.05 distance, one of resources must be 
killed by the other one. 
 
95 
 
Figure 5-5 Kill probability of resources in a kill effect zone 
 
 Although the chance that agents enter the kill zone is very low since the location of 
enemies can be detected by ESM sensors in advance, the possibility still exists. The movement of 
agent nodes in the kill zone is shown in Figure 5-6: 
 
 
Figure 5-6 Agent behavior in the kill zone  
 
96 
97 
Although many ways are available in real operation, to simplify there are only two 
possible options for agents to escape from the enemy in the kill zone. Both options are 
perpendicular to the user?s original moving direction as shown in Figure 5-6. The user node will 
take the option which enables it to get out of the dangerous zone most quickly. That is, agent 
node in the kill zone compute the distance to the enemy from both possible locations it can reach 
from its current location with a maximum velocity. The option which enables the agent to be 
further distant from the enemy is the evasion direction. The agent in Figure 5-6 will take the 
option 2 to get out of the risk area quickly. 
 
 5.8 PSO parameters 
 The PSO performance can be improved by inertia weight and cognitive coefficients [34]. 
The inertia weight gets important effect on balancing the global search and the local search in 
PSO. The selected inertia weight in this study decreases linearly from 0.9 to 0.4 during a 
simulation running and computed using Equation 2-6. This selection has provided improved 
performance in many applications. Also, the cognition parameters are set by the common 
practice, 2.05 each.  From the preliminary experiments the population size and the maximum 
number of iterations are set by 20 and 100, respectively.  
 
 
Chapter 6 
Simulation experiments and analysis 
 
For the representation of a realistic military MANET operation, an enemy force?s hostile 
activities are employed, with assumptions in this study. Several important performance metrics 
are created and used to improve network performance under the vulnerable conditions caused by 
the enemy obstacles in a tactical military operation area. Computerized simulations are 
performed to analyze network performance under the different scenarios considering the number 
of enemy obstacles, number of MANET nodes, and user mobility.  
The simulation experiments are first performed to verify the effect of metrics used in this 
thesis. This verification is conducted based on three different user mobility models using a small 
or medium sized problem: random waypoint (RW), search and rescue (SR), and convoy and 
defense (CD). Those user movements are created for testing a more realistic military MANET 
environment, which are briefly discussed in the following section and shown with initial and 
operation formation figures. Then, the metrics effects are tested with medium and large sized 
problems for only RW again. The three different sizes of problem are represented in Table 6-1. 
The problem size is determined by the number of network nodes and the number of enemies 
involved. 
 
 
 
 
98 
Table 6-1 Test problems 
 Small Medium Large 
Users 5 10 20 
Agents 4 8 16 
Enemies 3 4 4 
 
Also, the efficiency of the developed heuristic algorithm in terms of the number of hops 
required for a user node connection is tested by comparing it with the shortest distance based 
algorithm. To properly gauge the effect of metrics, destruction of network nodes or enemies is 
not included into the metric verification experiments, but it will be simulated for the cost benefit 
analysis later.  
The cost benefit analysis is also conducted based on different sized problems, but the 
problem size differs by the mobility model. That is, the experiment with RW is conducted on the 
three sized problems as in the experiment for verification of metrics effect, however, the tests 
with SR and CD are performed only for the small and medium sized problems because of 
computation time with large sized problems. 
The heuristic algorithm is coded and run in a C++ environment and the simulation results 
can be animated using a MANET simulation test bed especially developed. It can show the 
dynamic movement of MANET nodes including the enemy effects in the problem space. 
 
6.1 Simulation environment 
 At the initial time step each MANET node is generated at its assigned location. The 
network is in the connected state. The user nodes start to move toward their destinations which 
are randomly given or predefined depending on the assigned mission or mobility model. User 
99 
nodes? velocities and directions are also randomly assigned at each time step. However, to 
evaluate network performance under the same test conditions, the target destination, velocity and 
direction of a user node are generated using the same random number seeds for the different 
cases. As a result, the user nodes in different scenarios take the same paths.  
The simulation area is a two dimensional 6?6 distance square unit. The communication 
range of all MANET nodes is 1.0 distance unit. The velocity of MANET nodes is described by 
Euclidean distance unit traveled per unit time step, which is constrained by a minimum and a 
maximum velocity. The velocity range of a user node is between 0.02 and 0.05, whereas for an 
agent node, including the control node, it is between 0 and 0.07. However, enemy locations are 
not changed through the simulation time. This is true for all user movement patterns employed in 
this study. 
 
6.2 Test mobility models 
 The metrics verification experiment is conducted based on three different mobility 
models in order to compare network performance under different operation environments. Each 
mobility model represents a different movement pattern of military operation. The random 
waypoint is very common in MANET research as a default mobility model. User nodes with the 
random waypoint randomly move around the operation area without being directed by a control 
unit or an operational plan. User nodes? movements are directed in the SR and CD mobility 
models by a predefined operational plan. These movement patterns are described below. 
 
 
 
100 
6.2.1 Random Waypoint (RW) 
This movement pattern is used to represent military operations with a high random 
movement of users within a given operation area. All target directions for users are randomly 
generated without any predefined movement schedule. All MANET nodes are connected with 
each other at the initial time step and they start to spread out according to their random assigned 
destinations, velocities and directions. Figure 6-1 and Figure 6-2 show the initial formation and 
movement of network nodes and enemies.  
 
 
Figure 6-1 Initial formation with RW 
 
101 
Black balls surrounded by a blue and a red circle are enemies and the group of nodes in 
the center in Figure 6-1 includes the MANET nodes. User nodes (red balls) start to move toward 
their destination and are followed by agent nodes (green) and the control node (blue).  
 
 
Figure 6-2 Operation formation with RW 
 
6.2.2 Convoy and Defense (CD) 
 This movement pattern represents a military operation convoying a core body to a 
destination. A core body, such as a command unit, a VIP or war supplies, is located at the center 
of the formation and users surround it while maintaining a predefined distance. The purpose of 
this operation is to protect the important resources from enemy forces and convoy it safely to a 
102 
target point. Each user is assigned a responsible area, each rectangular box in Figure 6-3, to 
perform this mission and the user?s movement is constrained by the assigned area. There are four 
pre-assigned patrol boxes in which two user nodes operate and stay inside during the simulation.  
 
 
Figure 6-3 Initial formation with CD 
  
Once a simulation is run, user nodes move into their responsible box. However, the 
control node is fixed at the initial location through the simulation span. Figure 6-4 shows the 
movements during the operation simulation. 
 
103 
 
Figure 6-4 Operation formation with CD 
 
6.2.3 Search and Rescue (SR) 
One of the most important considerations in a search operation is that a given 
responsibility area can be covered as much as possible to accomplish an assigned mission 
efficiently. To meet this requirement, the responsibility area is uniformly divided into several 
sub-sectors. The number of sub-sectors is generally equal to the number of available users. Each 
user assigned to an individual sub-sector to search and rescue targets.  As shown in Figure 6-5, 
user nodes start to search from initial points where the network is fully connected, and follow 
paths which are predefined to cover the given search area in Figure 6-6. 
 
104 
 
Figure 6-5 Initial formation with SR 
 
The purpose of this operation is to search or rescue some targets such as mines, lost 
friendly forces, etc. The performance of a search and rescue operation depends on the effective 
use of limited resources. For effective deployment of users in this operation, users need to be 
directed by an operational plan. First, each user node travels through predefined target points. 
This enables users to uniformly cover the responsibility area without wasting limited resources 
by duplicated deployment of users.  Figure 6-6 shows the characteristics of this movement. This 
is the least random movement among those mobility models used in this study. 
 
105 
 
Figure 6-6 operation formation with SR 
 
6.3 Performance measures 
The primary goal of a military MANET in this study is to maximize the connection of 
user nodes, with the control node and to provide users in an operation space with a common view 
of the battle space. Military MANETs are usually operated under vulnerable conditions like 
enemy?s hostile activities and the available resources for MANET operation are limited. So, the 
efficient operation of given service resources becomes an important consideration for military 
operations. Network performance is compared based on the following metrics: 
 
106 
1) The average number of connected user nodes with the control node is computed by 
Equation (6-1).  
 
                         
ts
)CN,i(P
PM
ts
tUNi
t
??
=?
?
?
?
?
?
?
=
1
1
 
 
(6-1) 
 
Where, ts is the simulation time span in the given problem scenario. The indication of 
a path connection,  is 1 if there is a path connection between a user nodes i 
and the control node j, otherwise it is 0. Network performance is evaluated by PM
1
 as 
the primary performance measure. 
)j,i(Pt
2) The average number of hops per user node connection with the control node, PM
2
, 
given in equation (6-2). In particular, this metric is used to evaluate the efficiency of a 
MANET optimizer. 
 
 
                    
 
ts
)CN,i(P
Hop
PM
ts
t
UNi
t
CNj,UNi
ijt
?
?
?
=
?
=?
?
?
?
?
?
?
?
?
?
?
=
1
2
 
 
 (6-2) 
 
Here, Hop
ijt
 represents the number of hops between user node i and the control node j. 
3) The average total propagation loss by jamming, PM
3
, can be computed by subtracting 
actual bandwidth (aBW
ij
) from the possible bandwidth (pBW
ij
) which is calculated by 
the distance between node i and j as given in equation (6-3). By this metric, the 
average of jamming by enemies in the operation area can be measured. This metric is 
used only for the enemy case scenarios. 
107 
 
 
 
                    
()
ts
aBWpBW
PM
ts
tCNj,UNi
ijtijt
??
==?
?
?
?
?
?
?
?
?
?
=
1
3
 
 
(6-3) 
 
4) The last performance measure of this study is the Mission Completeness Rate (MCR). 
This is computed by dividing total number of connected priority nodes by total 
number of priority nodes as shown in Equation (6-4).  
 
 
 
                    
 
ts
TP
)CN,i(P
PM
ts
t
t
PNi
t
?
?
=
?
?
?
?
?
?
?
?
?
?
?
=
1
4
 
 
(6-4) 
 
Any user node entering the priority node assigning zone is defined as a priority node. 
TP
t
 represents the total number of priority nodes at time t.   
By using these performance measures listed above, we will evaluate a MANET 
performance. 
 
 
 
 
 
 
108 
 6.4 Effect of metrics under different mobility 
The implemented metrics in this study, PAL, messenger, and Priority node are expected 
to improve MANET performance from considering military aspects. These effects are verified 
under different test environments by the mobility models and number of network nodes and 
enemies. As we mentioned before, the kill scenario is not included in the experiments for metric 
verification. The test problems for RW and SR include 9 network nodes (5 users, 3 agents, 1 
control) and 3 enemies, while that for CD includes 17 nodes (8 users, 8 agents, 1 control) and 3 
enemies. Each problem is tested with 100 replications and each replication is simulated for 200 
time steps for RW and CD and for 180 time steps for SR mobility. 
  
6.4.1 PAL effect 
 PAL, which is newly developed to support the weakness of the hop-count based 
algorithm for military MANET, is the most important metric in this study. The main function of 
this metric is to properly distribute agent nodes within a given tactical area to support the user 
nodes moving toward their randomly assigned or predefined destinations in a timely manner.  
The simulation results in Table 6-2 show the effect of the PAL metric. The PAL effect is 
compared with the result of the No PAL case. The difference between No PAL and PAL is 
verified by a paired-T test with the primary performance measure (PM
1
).  
Network performance by PAL is much improved in all cases. For each mobility pattern, 
network connectivity is improved by 2.63%, 10.5% and 12.6%. For all mobility models by 
increasing the number of enemies, the network performance is degraded by small successive 
decrements. The PAL metric is the most effective for SR mobility.  
 
109 
Table 6-2 PAL effect with the hop-count based algorithm 
Mobility 
Num 
Enemy 
No PAL PAL 
p-value 
Avg NCU Avg Jamming Avg NCU Avg Jamming 
RW 
0 3.850 - 3.970 - 0.006 
1 3.792 0.313 3.944 0.248 0.000 
2 3.672 0.653 3.819 0.534 0.000 
3 3.540 0.981 3.647 0.801 0.019 
Avg 3.714 0.649 3.845 0.527  
CD 
0 6.456 - 7.400 - 0.000 
1 6.453 0.308 7.314 0.242 0.000 
2 6.420 0.719 7.213 0.561 0.000 
3 6.449 1.109 7.226 0.793 0.000 
Avg 6.445 0.712 7.288 0.532  
SR 
0 3.702 - 4.401 - 0.000 
1 3.547 0.075 4.223 0.167 0.000 
2 3.379 1.591 3.999 1.281 0.000 
3 3.314 1.773 3.830 1.522 0.000 
Avg 3.486 1.14633 4.113 0.99  
  
 
110 
 
Figure 6-7 Comparison of PAL and No PAL (RW, 1E) 
 
The PAL effect with RW is a little smaller than in the other two mobility models. 
However, the average plot line of the PAL case is always positioned above the No PAL case as 
shown in Figure 6-7. Also, the performance difference by the paired-T test is statistically 
significant. The p-values in the far right column in Table 6-2 are the paired-T test results between 
the two cases. 
 On the other hand, this performance improvement by PAL is much clearer with CD and 
SR than with RW as shown in Figure 6-8 and Figure 6-9. The 95% confidence interval plot for 
the two cases shows this difference more clearly.  There are two groups of line in the figures. 
From the top, each line indicates the mean, upper and lower confidence limits of a 95% 
confidence interval on the mean. 
 
111 
 
Figure 6-8 Comparison of PAL and No PAL (CD, 1E) 
 
 
Figure 6-9 Comparison of PAL and No PAL (SR, 1E) 
 
 The step?shaped plot of the PAL case in Figure 6-9 is related to the characteristics of SR 
movement. Around the time range between 95 and 110, one of user nodes at the end of both sides 
112 
of search space is disconnected from the network due to the shortage of agents. So, NCU is 
decreased to 4 from 5. At about 110 time, NCU is again decreased to 3 when the user node in the 
other side of search space is disconnected. It takes about 40 time steps for the disconnected users 
to start to recover the connection with the control node again. Messenger is not employed in 
either scenario in Figure 6-9. If messenger agents were employed, the plot would change since 
NCU can be improved by their use. 
Table 6-3 PAL effect with the shortest distance based algorithm 
Mobility 
Num 
Enemy 
No PAL PAL 
p-value 
Avg NCU Avg Jamming Avg NCU Avg Jamming 
RW 
0 3.890 - 3.997 - 0.003 
1 3.827 0.418 4.003 0.348 0.000 
2 3.753 0.789 3.878 0.711 0.001 
3 3.566 1.239 3.748 1.015 0.000 
CD 
0 6.575 - 7.438 - 0.000 
1 6.877 0.346 7.429 0.288 0.000 
2 6.777 0.693 7.493 0.555 0.000 
3 6.777 1.210 7.422 0.873 0.000 
SR 
0 3.854 - 4.3999 - 0.000 
1 3.546 0.116 4.233 0.146 0.000 
2 3.377 1.858 4.039 1.158 0.000 
3 3.322 1.856 3.886 1.285 0.000 
 
113 
This PAL effect is also effective for the shortest distance based routing algorithm. 
Network performance is improved for all cases and mobility models as shown in Table 6-3. 
Based on these results, we can postulate that PAL effect depends on the randomness of user 
mobility. Among user mobility models used in this study for experiment, RW is the mobility 
model with the most randomness, whereas SR has the least randomness user node directions are 
predefined.  
 
6.4.2 Messenger effect 
 The purpose of operating messengers is to reconnect isolated users from the network. 
Messengers employed in this study are expected to improve network performance by 
reconnecting disconnected users. To accomplish that, the proper duration for operating 
messengers is crucial to manage the limited number of agents effectively in a military MANET 
operation. So, the operation time of messenger for each mobility model needs to be considered 
first. 
 
6.4.2.1 Random waypoint (RW)  
For all cases in Figure 6-10, both no enemy and enemy cases, the user node connectivity by 
messenger is approximately increasing in messenger time. The operation time of a messenger 
with RW mobility is not limited if user nodes are not killed by enemy attack or any other reasons. 
That is, the better network performance can be accomplished by operating messengers for longer 
time. The 95% CI plot in Figure 6-11 represents the advantage of using messenger visually. The 
messenger case is always better than the no messenger case in the plot. Disconnected user nodes 
114 
might be reconnected by either messenger agents or other network nodes (users or agents). The 
average percentage of reconnected users reconnected by only messenger nodes is about 20.05%. 
 
 
Figure 6-10 Messenger effect (RW) 
  
 
Figure 6-11 Comparison of the Messenger and No messenger (RW) 
 
115 
6.4.2.2 Convoy and defense (CD) 
Figure 6-12 and Figure 6-13 summarize the effect of the messenger with CD mobility. 
CD mobility does not show full random mobility like the random waypoint. However, users are 
free to move within their responsibility area. That is, each user?s mission area has been shifted 
from the entire operational space to a small patrol box in CD mobility. The messenger effect in 
CD is a little greater than in RW. The average improvement for RW is about 1.82%, but is 2.5% 
for CD. 
 
 
Figure 6-12 Messenger effect (CD) 
 
  
116 
 
Figure 6-13 Comparison of the Messenger and No messenger (CD, 1E) 
 
6.4.2.3 Search and Rescue (SR) 
 SR mobility is the most directed movement among the mobility models used in this study 
because of its predefined operational plan. User nodes in SR go through their predefined paths or 
destinations. It has relatively low randomness. However, the effect of the messenger is the largest 
in this lowest random mobility model as shown in Figure 6-14 and Figure 6-15. The average 
improvement with SR is about 4.94%, almost twice of that in CD and triple of that in RW. The 
movements of disconnected users in SR operation are limited to their assigned search areas, so 
using a messenger can improve the connectivity significantly. 
 
117 
 
Figure 6-14 Messenger effect (SR) 
 
 The difference between the messenger and no messenger cases is very clear in SR as 
shown in the 95% confidence interval plot in Figure 6-15. For first 90 time steps, a messenger is 
not required since all user nodes are connected with the control node. But, after this time range, 
messengers play a key role for network connections. The confidence interval plot in Figure 6-15 
shows the definite advantage of using messengers. 
 The no messenger case maintains the three user nodes connected state between time 120 
and 150 without reconnecting the disconnected users. This is because the agents move back to 
the control node when they lose connection with the users. However, in the messenger case, 
agents are converted to messengers when they lose connection with the users which they have 
served. The messengers try to reconnect the disconnected users for a given time by moving close 
to the estimated location of disconnected user as much as possible. As shown in Figure 6-15, the 
disconnected users start to be reconnected right after disconnecting from the network. 
118 
 
 
Figure 6-15 Comparison of the Messenger and No messenger (SR, 1E) 
 
6.4.3 Priority node effect 
 Unlike the other metrics, a priority node does not guarantee improved network 
performance. The purpose of the priority node in this study is to represent a more realistic 
military MANET operation scenario. A user node that requires the priority service of agents 
because of the operational situation (such as combat, positioning close to interested target) is 
defined as a priority node, as addressed in Section 4.2. In this study, the connectivity of priority 
nodes is needed for mission success and a priority node is weighted by a higher positive value to 
increase its connectivity. Another parameter, Mission Complete Rate (MCR), is employed to 
measure the effectiveness of the priority node concept.  The distance to an enemy is used in this 
study to assign a priority node. 
 A radius circle from an enemy is set as a priority assigning zone (PAZ), similar to the 
jamming zone and the kill zone. Any user node entering this circle is assigned a priority node 
119 
which requires service before other user nodes outside the PAZ. This zone is set to 0.7 unit 
distance for this experiment. 
 
6.4.3.1 Random waypoint 
 
 
Figure 6-16 MCR changes by priority node weights (RW) 
 
 As shown in Figure 6-16, MCR (%) is improved for all enemy cases by assigning priority 
nodes, but this improvement of MCR may require some tradeoff of NCU as shown in Figure 6-
17. Network performance is decreased as the priority weight increases.  
For one enemy case with RW, the best weight looks seems to be 4 from Figure 6-17 
considering both NCU and MCR. However, as the number of enemies is increased the best 
weight for a priority node is changed to 1.1 as shown in Figure 6-18 and Figure 6-19. 
 
120 
 
 
Figure 6-17 Priority node effect (RW, 1E)  
 
 
Figure 6-18 Priority node effect (RW, 2E) 
121 
 
Figure 6-19 Priority node effect (RW, 3E) 
 
Especially, in the three enemy case (3E), both NCU and MCR are increased when the 
priority is weighted by 1.1 as shown in Figure 6-19. As a result, the best priority weight which 
can maximize MCR and minimize the reduction of NCU at the same time is 1.1. However, the 
number of agents in the small sized problem are not enough to cover the users through the given 
operation area. So, assigning an agent to support a priority node degrades the network 
performance since other users may not be served.  
 
6.4.3.2 Convoy and Defense (CD) and Search and Rescue (SR) 
The priority node metric is not as effective in CD and SR as in RW. Figure 6-20 through 
Figure 6-23 do not show any critical difference from the no priority node case. MCRs in Figure 
6-20 maintain almost steady state for the three enemy cases even though the priority weight is 
increased.  
 
122 
 
Figure 6-20 MCR changes by priority node weight (CD) 
 
 Although network performance fluctuate a little compared with MCR chart, it is almost 
constant too. This is related to the randomness of user mobility. As discussed before, CD and SR 
have lower randomness than RW.  
The network nodes in CD and SR move and spread out by their predefined movement 
plans. Agent nodes are matched with their responsible user when users reach their mission area 
or are distributed enough from other users. Consequently, there are only a few cases where agent 
nodes should give up supporting user nodes to connect priority nodes because of the distribution 
of network nodes through the operation area by a predefined plan. As a result, the use of the 
priority node in CD and SR does not make any critical change with network performance.  
 
123 
 
Figure 6-21 NCU changes by priority node weights (CD) 
 
 
Figure 6-22 MCR changes by priority node weights (SR) 
 
124 
 
Figure 6-23 NCU changes by priority node weights (SR) 
 Until now, we have tested the effects of metrics employed in this study based on different 
mobility models. In summary, PAL and messenger have improved network performance for all 
mobility patterns even though there exist some differences in effect according to the randomess 
level. The effect of PAL and messenger is greater with a low random mobility model. On the 
other hand, the priority node metric is effective with a high random mobility model like the 
random waypoint model 
 
6.4.3.3 Efficiency of employed algorithm 
 As discussed before, in this section we compare the efficiency of hop-count based routing 
with the shortest distance based routing approach in terms of the number of required hops for a 
user node connection.  
Table 6-4 shows the results. Considering the number of hops used for a user connection, 
the hop-count based algorithm is better than the shortest distance based algorithm for the 
125 
experiments with all mobility models used. The efficiency difference between hop-count and 
shortest distance based routing is about 10.4% with RW, 24.4% with CD and 4.01% with SR. 
The efficiency is the largest in the CD and the smallest in the SR. As we know, the small effect 
with SR is related to the randomness of mobility as mentioned earlier. 
Table 6-4 Efficiency of algorithms as measured by number of hops per connected user 
Mobility Enemy Hop-count Shortest distance Difference p-value 
RW 
0 2.698 2.871 0.173 0.000 
1 2.692 3.018 0.326 0.000 
2 2.718 3.103 0.385 0.000 
3 2.785 3.163 0.378 0.000 
Average 2.723 3.039 0.316 (10.4%)  
CD 
0 4.692 5.110 0.418 0.000 
1 4.664 6.648 1.984 0.000 
2 4.665 6.490 1.825 0.000 
3 4.554 6.261 1.707 0.000 
Average 4.644 6.127 1.484 (24.2%)  
SR 
0 3.117 3.133 0.016 0.000 
1 3.093 3.175 0.082 0.000 
2 3.113 3.313 0.2 0.000 
3 3.093 3.315 0.222 0.000 
Average 3.104 3.234 0.13 (4.01%)  
126 
 The 95% CI plots in Figure 6-24 show this result visually and suggest that the hop-count 
based routing is much more efficient than the shortest distance based routing. 
 
 
Figure 6-24 Comparison of number of hops used for a user connection (RW, 2E) 
 
 
Figure 6-25 Comparison of number of hops used for a user connection (CD, 2E) 
 
127 
 
Figure 6-26 Comparison of number of hops used for a user connection (SR, 2E) 
 
As shown in Figure 6-26, the efficiency gap between the two routing approaches is a little 
smaller in SR than in other mobility models. This is also related to the movement characteristics 
and randomness of the mobility model. For a few time ranges, the shortest distance based routing 
looks more efficient. This is because network performance by the hop-count based routing is 
higher than the shortest distance based routing and the connection at the moment requires many 
hops.  
 We have tested the effect of metrics using the small sized problems. These will be further 
tested with medium and large sized problems using random waypoint in the next section. By this 
extended test, the effects of metrics can be verified. 
 
6.5 Effect of metrics for medium and large sized problem with RW 
 The networks for the experiments in this study are tested over the same size of 
operational area. The difference among those is the number of network nodes or enemies. So, 
128 
network density is increased by the network size. SR and CD are not considered in this section 
because of computations involved. 
 
6.5.1 PAL effect 
 PAL is also very effective for both medium and large size problems as shown in Table 6-5. 
Paired-T test p-values in the right column verify the effect of PAL. For test scenarios, about 
8~10% improvement is obtained by PAL. 
Table 6-5 PAL effect with medium and large problem 
Mobility 
Num 
Enemy 
No PAL PAL 
p-value 
Avg NCU Avg Jamming Avg NCU Avg Jamming 
Medium 
0 8.279 - 9.156 - 0.000 
2 8.041 1.613 8.863 1.363 0.000 
4 7.875 3.053 8.756 1.994 0.00 
Avg 8.065 2.333 8.925 1.679  
Large 
0 17.876 - 19.309 - 0.000 
2 17.536 5.147 18.873 4.422 0.000 
4 17.116 8.457 18.699 6.710 0.000 
Avg 17.509 6.802 18.960 5.566  
  
The 95% CI plots of the PAL effect in Figure 6-27 and Figure 6-28 indicate the definite 
difference between the PAL used and No PAL cases. We replicated only 20 times for each 
scenario because of the computational effort involved. So, if the number of replications is 
129 
increased, the confidence interval of each group would be narrowed, and the difference between 
two cases in Figure 6-27 would be clearer.  
 
 
Figure 6-27 Comparison of PAL and No PAL (Medium, no enemy) 
 
 
Figure 6-28 Comparison of PAL and No PAL (Large, no enemy) 
 
 
130 
6.5.2 Efficiency of employed algorithm  
 By looking at the results shown in the table and figures below, the hop-count based 
routing algorithm requires a smaller number of hops to connect a user node with the control node. 
For most of simulation time span, it is better than the shortest distance based algorithm. But, this 
is reversed for a few time ranges as shown in Figure 6-29 and Figure 6-30. The reason is the 
same as discussed in Section 6.4.3.3. 
Table 6-6 Efficiency (# hops/connected user) of algorithms 
with medium and large problems 
Mobility Hop-count Shortest distance Difference p-value 
Medium 3.738 4.057 0.319 (7.9%) 0.000 
Large 4.217 4.407 0.19 (4.3%) 0.000 
 
 
 
Figure 6-29 Comparison of number of hops used for a user connection (Medium) 
131 
 
Figure 6-30 Comparison of number of hops used for a user connection (Large) 
 
6.5.3 Messenger effect 
 The messenger effect shows almost the same pattern by problem size and number of 
enemies. Network performance in medium sized problems is improved by messenger, but its 
improvements are limited by the operation time of messengers. The best messenger is 80 time 
step for the two enemy case and 40 time step for the four enemy case. After these operation times 
of messenger, network performance is not improved further. Messenger operation time is shorter 
in the four enemy case than in the two enemy case. This is related to the number of enemies in 
the operation area. That is, the randomness of network is increased by more enemy numbers and 
it increases the time required for reconnecting the disconnected users from the network. 
Consequently, network performance is degraded by wasting agents. Also, reconnecting a 
disconnected user under a combat situation may be impossible because of its destruction by 
enemy. So, the messenger operation time is shortened by increasing the number of enemies. 
However, if the number of agents in the network is enough to cover the effect of 
increased enemies, the operation time would not be affected. As shown in Figure 6-31 and Figure 
132 
6-32, the messenger operation time for the four enemy case is almost same as the two enemy 
case.  The average percentage of users reconnected by messengers is about 29.1%. It is higher 
than in the small sized problem since it has more extra agent capability. 
 
 
Figure 6-31Messenger effect (Medium, 2E) 
 
 
Figure 6-32 Messenger effect (Medium, 4E) 
 On the other hand, the number of network nodes in the large sized problem is twice that 
of the medium sized problem, but both problems use the same number of enemies and size of the 
133 
problem space. As a result, the only difference between them is network density. That is, the 
density of the large sized problem is much more than that of the medium sized problem. 
 As shown in Figures 6-33 and 6-34, the network performance improvement by 20 
messenger duration is small but notably different from other time values in the figure. So, 20 is 
definitely the best messenger operation time for both scenarios. This improvement is also very 
small compared with the one made in the medium sized problems. The shorter messenger and 
small improvement of network performance are related to the high density in large sized problem.  
 
Figure 6-33 Messenger effect (Large, 2E) 
 
Figure 6-34 Messenger effect (Large, 4E) 
134 
6.5.4 Priority node effect 
 In the small sized problem the best priority weight was 1.1 considering the tradeoff with 
network performance. MCR in the small sized problem could be improved by a small weight, but 
it requires higher weights for both medium and large problems as shown in Figures 6-35 through 
6-38. 
 
 
Figure 6-35 Priority node effect (Medium, 2E) 
 
 
Figure 6-36 Priority node effect (Medium, 4E) 
135 
 
 
Figure 6-37 Priority node effect (Large, 2E) 
 
 
Figure 6-38 Priority node effect (Large, 4E) 
 
 The above figures represent the effect of priority node in enemy scenarios. First, MCR is 
improved by increasing the priority weight regardless of the problem size and the number of 
enemies. However, NCU under increased number of enemy cases for both problem sizes is 
136 
decreased by the higher priority weights as shown in the above figures. That is, although the 
agents in the two enemy case have to support the priority nodes first, NCU is not worsened since 
it still has some extra capability to serve other user nodes. On the other hand, NCU is degraded 
in the four enemy case because of the shortage of agents to serve other user nodes.  
However, the effect of priority node is very small with the large problem with four 
enemies as shown in Figure 6-38. Both NCU and MCR are not significantly changed by 
weighing the priority nodes. From this fact, we guess that the effect of priority node in enemy 
scenarios is related to the density and the number of enemy of a network. 
Based on the metrics verification results with different sized problems and user mobility 
models we conclude the effect of agent behaviors as followings: 
? PAL is effective for all sizes of problem and all user node mobility models. It is most 
effective with low randomness mobility models. 
? Messenger is sensitive to the network density. The effect of messenger becomes 
smaller and the operation time is shortened as network density increases. 
? Priority node is sensitive to both the number of enemies and network density. NCU is 
degraded by increasing the weights and the number of enemies. MCR with few 
enemies continues to increase by increasing the weights. However, Priority node is 
not effective with high density networks.  
 
6.6 Cost benefit analysis 
In previous sections we have verified the effectiveness of devised approaches for military 
MANETs. From the military operation aspect, the most efficient number of agents is an 
important consideration to establish a sucessful operation plan since it is impossible to operate 
137 
enough agents every time. Therefore we have conducted experiments to choose the best number 
of agents under various scenarios. The results obtained from this analysis can be used by 
commanding officers or operation planners for decision making during the operation planning 
step. 
We generated three different sizes of problem for this experiment. RW is used for the 
three different sized scenarios. This was chosen because it enable us to use all approaches 
devised in this study and is suitable for combat scenarios. However, CD is used only for the 
medium size and SR only for the small size because of computation time. In addtion, for SR 
mobility, the physical destruction of network resources by combat is not considered since we 
assume that no combat sitaution is expected during a search and rescue operation.  
First, for all combat scenarios (small, medium, large) an appropriate messenger operation 
time will be found. Then, the most efficient number of agents is assumed by the number of 
agents to a 95% actual user node connectivity goal is identified. This efficient number is 
determined for both the no enemy and the enemy scenarios and compared. 
 
6.6.1 Cost benefit analysis with RW 
 The three different sized scenarios are tested using RW in this section and eight enemies 
are used for the large size scenario to represent an acute combat situation as mentioned above. 
 
6.6.1.1 Combat scenario 1 (Small size) 
There is almost no limitation for messenger time in the small size of networks. However, 
it is different for real militray combat operations since some network nodes could be killed by 
the enemy. That is, our expectation is that the messenger effect would be smaller or the operation 
138 
time be shorter under the kill environment. The messenger time for combat scenario 1 is about 60 
as shown in Figure 6-39. Network performance is improved by 3.6% and it is not changed by 
any other longer messenger time. 
 
 
Figure 6-39 Messenger effect (S1) 
 
 
Figure 6-40 Comparison of network performance by messenger (S1) 
 
139 
Figure 6-40 clearly shows the effect of messenger in combat scenario 1. Network 
performance with messenger is always better than with no messenger case. So, we will use 
messenger for 60 time steps for the experiments to find the most efficient number of agents. Each 
priority node is weighted by 1.1 according to the experiments shown earlier.  
95% network connectivity in scenario 1 indicates 4.75 network node connections with the 
control node if no users are killed. However, some of network nodes could be killed by combat 
with enemies during the operation. So, the 4.75 target connection may be impossible to 
accomplish under a combat environment. As a result, actual connectivity (%) is required to 
measure network performance. If no enemy were in the combat area, fewer agents would be 
required to reach the target connctivity level since network connectivy is not reduced by 
jamming. However, it is true only compared with the enemy scenarios without the kill 
environment. That is, the number of agents required in the enemy case may be smaller than the 
no enemy case by the actual connectivity. Figure 6-41 represents the result under the no enemy 
condition. 11 agents (96.4%) are required to meet the target connectivity in this case and network 
connectivity continues to be improved as more agent nodes are employed.  
 
Figure 6-41 Efficient number of agents in the no enemy case (S1) 
140 
   
Figure 6-42 Efficient number of agents in the enemy case (S1) 
 
As shown in Figure 6-42, 14 agents are required under the enemy case. However, the 
actual connectivity with 10 agents is about 95.6%, 96.8% with 12 agents this and 99.3% with 13 
agents. So, under the enemy condition, the target connectivity can be accomplished using only 10 
agents.  
 
Figure 6-43 Network performance and number of killed resources (S1) 
141 
 In Figure 6-43, the average number of killed users and enmies are 0.45 and 1, 
respectively. But no agent node is killed because the agent nodes follow the users and the 
locations of enemies are usually identified before agents enter the kill zone. So, the possibility of 
an agent being killed is relatively low in this scenario.  
 
6.6.1.2 Combat scenario 2 (Medium size) 
In the combat scenario 2, more network nodes and enemies are included, thus it is more 
complicated than scenario 1. As we do in combat scenario 1, the messenger time for this scenario 
will be found first before performing further experiments.  
 
 
Figure 6-44 Messenger effect (S2) 
 
For combat scenario 2, the best messenger time is 10 as shown in Figure 6-44. This is 
much shorter than in scenario 1. The improvement by messenger is also smaller than scenario 1. 
The difference between the no messenger and the messenger (10) scenarios is only 1.19%. From 
142 
Figure 6-45, network performance with messenger (10) is equal to or better than the no 
messenger scenario through the simulation time span, but the gap is not big. 
 
 
Figure 6-45 Comparison of network performance by messenger (S2) 
  
 The minimum required number of agents meeting the target connectivity under the no 
enemy condition is 10 and network performance can be improved more by adding more agents 
as in scenario 1 as shown Figure 6-46.  
 
Figure 6-46 Efficient number of agents in the no enemy case (S2) 
143 
For the enemy case, the efficient number of agents is 14 and the actual connectivity is 
95.1%. The difference between the no enemy and enemy scenarios is a little larger than in 
scenario 1 and we think that it is related to the interaction with the enemies such as jamming and 
combat since the numbers of enemies and network nodes for scenario 2 are twice that of scenario 
1.  
 
Figure 6-47 Efficient number of agents in the enemy case (S2) 
 
 By adding more agents, network connectivity continues to be improved to certain limit, 
possible network connectivy considered by current network state, as shown in Figure 6-47. 
Figure 6-48 shows the relation between NCU and the number of killed users over time. 
Under the combat 2 condition, about 0.75 user nodes and 1.15 enemies are killed. NCU is 
reduced as the number of killed users is inceased since some of users are killed by enemies.  
 
144 
 
Figure 6-48 Network performance and number of killed resources (S2) 
The movements of network nodes in combat scenario 2 are represented in Figure 6-49. 
Currently, no network nodes and enemies are killed. There are 4 enemies which have jamming 
(red circle) and kill effect zones (blue circle). If network nodes enter the jamming zone, they start 
to lose their communication capability because of the jamming effect. Most MANET nodes are 
inside a jamming zone at the moment. If any node enters a kill zone, it fights with the enemy; the 
combat is simulated by the kill probability and the combat power of network nodes and enemies. 
 
145 
 
Figure 6-49 Animation picture of combat scenario 2 
(Red: user, Green: agent, Blue: control, Black: enemy) 
 
6.6.1.3 Combat scenario 3 (Large size) 
 In this scenario, 36 network nodes are used and the operation area size is the same as in 
scenarios 1 and 2. So, network density is relatively higher than previous scenarios and user nodes 
in this network are expected to help improve the network connection. So, 16 agents in the default 
network size of scenario 3 may be enough to serve 20 user nodes without adding more agent 
nodes. The experiment results in Figure 6-50 confirm this expectation. 
  
146 
 
Figure 6-50 Efficient number of agents in the no enemy case (S3) 
 
Network connectivity of this default scenario 3 (no enemy) is already beyond the target 
performance by 16 agents as seen in Figure 6-50. To simulate an acute combat situation, 8 
enemies are used in scenario 3, so more network nodes and enemies would be destroyed by the 
increased chance of combat. 
First, we will go over the effect of messenger time in this high density, combat operation 
condition. As shown in Figure 6-51 network performance is not improved after it reaches the best 
at messenger time 50.  
 
147 
 
Figure 6-51 Messenger effect (S3) 
 Network performance by messenger is improved by 3.6% with scenario 1 and 1.19% 
with scenario 2 as discussed previously. But only 0.3% of network connection is increased by 
messengers in scenario 3. As we already discussed the messenger effect with a high density 
network in Section 6.5.3, the messenger effect is also very small in this large sized problem 
including the kill event by combat. 
 
Figure 6-52 Comparison of network performance by messenger (S3) 
 
148 
 The average NCU plot in above Figure 6-52 shows a small effect of messenger in 
scenario 3.  
 
Figure 6-53 Network performance and number of killed resources (S3) 
 
The combat between users and enemies is much more severe than in scenario 1 and 
scenario 2. The average number of killed users and enemies is 2.4 and 1.95, respectively, and the 
highest number of killed users is 5 during the simulation. Figure 6-53 shows network 
performance, percentage of killed users and percentage of killed enemies over time. Network 
performance starts to go down along with users? destruction but after time step 100 experiences 
no further deterioration. Under this circumstance, as shown in Figure 6-54 below, deploying 
more agents does not help improve network performance.  The actual user node connectivity is 
between 94.3% and 94.9% by the number of agents more than 16 as shown in Figure 6-54. This 
actual connectivity level could be accomplished by 14 agents. As a result, the number of agents 
in the scenario 3 does not matter. 
 
149 
 
Figure 6-54 Efficient number of agents in the enemy case (S3) 
 
6.6.2 Cost benefit analysis with CD 
The cost benefit analysis with CD is performed only for medium size problems. There are 
four patrol boxes in this test scenario and an enemy is assigned to each patrol box.  As shown in 
Figure 6-55, network performance using messenger agents improves regardless of messenger 
duration. The best messenger operation time among those tested is 10. The number of connected 
user nodes using the best operation time increases by 2.64%. 
 
150 
 
Figure 6-55 Messenger effect in the combat scenario with CD 
 
 The performance of messenger case maintains dominates the no messenger case 
throughout  the simulation time span as shown in Figure 6-56. 
 
 
Figure 6-56 Comparison of network performance by messenger  
in the combat scenario with CD 
 
151 
 
Figure 6-57 Efficient number of agents in the combat scenario with CD 
 
 Figure 6-57 shows the simulation result to find the efficient number of agents meeting the 
operational goal. The actual network connectivity which can be accomplished using eight 
assigned agents for this scenario is about 89.3%, which does not meet the target connectivity, so 
more agents are required to reach the target connectivity. The actual network connectivity using 
nine agents is about 93.6% and about 95.7% using ten agents. So, at least two more agents need 
to be assigned for this operation in order to reach the target. 
 Figure 6-58 shows the percentage of actual network connectivity, the percentage of killed 
users and enemies over time. During operation, about 53% of enemies and 30% of users are 
killed by combat. 
 
152 
 
Figure 6-58 Network performance and number of killed resources 
in the combat scenario with CD 
 
6.6.3 Cost benefit analysis with SR 
A combat situation at the initial planning phase of a search and rescue operation is rarely 
considered since search and rescue is commonly conducted in safe area. There many types of 
search formations for the SR operation; these are determined by the operational situation of the 
assigned tactical space and the given mission. The line formation search implemented in this 
study is the most common type of search operation since it enables the forces to cover the entire 
search space uniformly and quickly for a given time.  
It is assumed that neither network nodes nor enemies are killed in this scenario and the 
movement paths of users are predefined by the network devices? capability. Thus, the number of 
agents required to support the users assigned to the operation may roughly be estimated by 
considering these predefined paths and the number of users. Consequently, the randomness level 
of this operation is relatively low compared with other mobility models. In this section, a 
planned search operation executed by five users is considered using simulation.   
153 
 
Figure 6-59 Efficient number of agents in the combat scenario with SR 
 
Four agents are assigned to this SR operation by considering the radius of action of the 
users. As expected from the initial operation plan, four agents are enough to support the users. 
Network performance by four agents is about 95.02% which satisfies the target connectivity. 
However, the target connectivity cannot be accomplished with fewer agents than four as shown 
in Figure 6-59.  
In summary, the most efficient number of agents needed to accomplish the target 
connectivity of a military MANET depends on the operational situation such as combat level, 
obstacles, and network density. The target network performance could be met by a relatively 
smaller number of agents in a high density network because of the many operating users. Some 
operations are vulnerable or obstructed by obstacles. In this kind of situation, it may be very 
difficult to accomplish target performance because of highly random operations. In particular, 
under a severe combat situation, the operation of agents could be disturbed by the uncertainty 
about whether the disconnected users are alive or killed. That is, agents could be deployed to 
support killed users. This relates to choosing messenger duration for disconnected users. So, the 
messenger approach may not effective in the severe operations. 
154 
 
Chapter 7 
Conclusions 
 
The efficient use of networks for acquisition and sharing of valuable information among 
friendly forces in a tactical battlefield is a crucial factor for an force awareness and it contributes 
to the outcomes of combat in the information age. Especially, under an environment where rapid 
reaction and dynamic mobility are required, the importance of mobile ad-hoc networks has been 
increased. 
In this thesis, a military MANET model that represents vulnerable combat environments 
realistically is proposed and a heuristic algorithm to optimize mobile agents is developed using 
PSO. 
The communication connectivity for military MANETs in this thesis is measured by the 
connection between users and the control node, and the optimization of this communication 
network is done by the hop-count based routing approach to manage limited resources efficiently. 
However, the hop-count based approach shows a lack of response to network changes even 
though it can reduce resource requirements by minimizing use of hops. To compensate, a new 
approach, termed PAL, is developed in this thesis. PAL enables mobile agents to self-configure 
towards the best location in advance, responding to the change of user nodes? topology. This pre-
deployment of mobile agents makes it possible to support users at the needed time and be placed 
effectively without wasting resources.  
The representation of a real network system is very important in order to evaluate the 
performance of a network. The evaluation of network performance is also necessary to plan an 
155 
efficient military MANET operation plan under varied military combat operation environments. 
To accomplish these requirements, we conceptualize a realistic military MANET under combat 
environments by introducing enemy obstacles, by specifying network nodes in more detail, and 
by designing and using new performance metrics representing the characteristics of military 
MANETs. Through hostile activities of enemy obstacles such as jamming and killing, the 
capability of MANET nodes is reduced and MANET nodes may even be killed in combat with 
enemies. In addition, MANET nodes are split into four different types by the function in the 
military unit structure. This classification of network nodes represents the characteristics of 
military MANET which differ from commercial networks. With PAL, two additional approaches 
(messenger and priority node) are employed to support the reality of the proposed military 
MANET model. Messengers cannot only enhance model reality but also improve network 
performance by emulation of a real military operation searching for the disconnected users. 
To test the effect of the new approaches, we use three different user mobility models: 
random waypoint (RW), search and rescue (SR), and convoy and defense (CD). These mobility 
models are used to verify the effect of the approaches over different military movement patterns. 
User nodes? movement in SR and CD is directed and controlled by an operation plan, so the 
randomness level of these is relatively lower than with the random waypoint. The effects of PAL, 
messenger and priority nodes have a close relationship with randomness of the user mobility. 
Also, network density and the combat level (number of enemies) in the network are significant 
factors to the effectiveness of PAL, messenger and priority nodes. 
The proposed heuristic algorithm based on the hop-count routing approach is expected to 
be useful for military operations that require efficient use of limited resources under various 
operational conditions. By adjusting the simulation parameters employed in the proposed model 
156 
such as combat level by the number of enemies, minimum bandwidth required for a connection 
between nodes, enemy effect ranges and the perturbation level (?) in the employed user mobility 
model, it is possible to simulate a great variety of military operational environments and to 
evaluate networks under those varied military operation conditions. The simulation based on this 
proposed model can be useful to estimate the optimal number of agents before deploying an 
actual network in a battlefield. 
For future research, the following can be studied further by extensional or modification to 
the proposed model: 
The proposed military MANET is focused on a tactical battlefield network consisting of 
identical types of network devices. However, this could be extended by introducing different 
types of network devices in terms of velocity, transmission range, etc. For example, network 
devices designed for individual soldier, tank, UAVs, or combat aircraft have different capabilities. 
All these devices can be combined for a strategic level of military network.  
Also, the obstacle behavior can be modified. The obstacles used in this thesis are fixed 
throughout the simulation. A better simulation of military combat could be possible by giving 
mobility to these enemy obstacles.  
Finally, users are free to move independently in any direction within a fixed range of 
space, but this could be modified to have users coordinate their movements with the control node. 
That is, the control node directs the movement of users considering the network capability at the 
moment. 
 
 
 
157 
 
References 
 
[1] Abolhasan, M., Wysocki, T., and Dutkiewicz, E., ?A review of routing protocols for mobile 
Ad-hoc Networks,?  Ad-hoc Networks 2 (1) (2004) 1?22.  
[2] Alberts, D., Garstka, J., Stein, F., Netwrok centric warfare ? developing and leveraging 
information superiority, 2
nd
 edition, CCRP2000, 2000. 
[3] Ammari, H. and El-Rewini, H., ?A location information-based route discovery protocol for 
mobile ad-hoc netoworks,? in: Proceedings of the IEEE International Conference on 
Performance, Computing, and Communications, 2004, 625-630. 
[4] Baburaj, E. and Vasudevan, V., ?An intelligent mesh based multicast routing algorithm for 
MANETs using Particle Swarm Optimization,? IJCSNS International Journal of Computer 
Science and Network Security 8(5) (2008) 214-218. 
[5] Bai, F., Sadagopan, N. and Helmy, A., ?The important framework for analyzing the impact 
of mobility on performance of routing for Ad-hoc Networks,? Ad-hoc Networks Journal - 
Elsevier Science 1(4) (2003) 383-403. 
[6] Bakht, Humayun, Wireless Infrastructure: Understanding mobile Ad-hoc Networks, 2008, 
Accessed on July 17, 2008; Available 
at: http://www.computingunplugged.com/issues/issue200406/00001301002.html 
[7] Belding-Royer E.M. and Toh, C. K., ?A review of current routing protocols for Ad-hoc 
mobile wireless networks,?  IEEE Personal Communications Magazine (1999) 46?55.  
[8] Bluetronix. Autonomous agents, Accessed on  July 6, 2009; Available 
at: http://www.bluetronix.net/autonomous_agents.htm 
[9] Branke, J., E. Saliho?lu, S. Uyar, ?Towards an analysis of dynamic environments,? in: 
Proceedings of the 2005 Conference on Genetic and Evolutionary Computation, 2005, 
1433-1440. 
158 
[10] Branke, J., "Evolutionary approaches to dynamic optimization problems ? updated  
survey," in: Proceedings of the GECCO Workshop on Evolutionary Algorithms for 
Dynamic Optimization Problems, 2001, 27-30. 
[11] Broch J., Maltz, D. A., Johnson, D. B., Hu, Y. ?C. and Jetcheva, J., ?A performance 
comparison of multi-hop wireless ad-hoc network routing protocols,?  in: Proceedings of 
the Fourth Annual ACM/IEEE International Conference on Mobile computing and 
networking, 1998. 
[12] Buruhanudeen S., M., Othman, B. and Ali, M., ?Existing MANET routing protocols and 
metrics used towards the efficiency and reliability- An overview,? in: Proceedings of the 
2007 IEEE International Conference on Telecommunications and Malaysia International 
Conference on Communications, 2007, 14-17. 
[13] Camp, T., Boleng, J. and Davies, V., ?A survey of mobility models for ad-hoc network 
research? Wireless Communications and Mobile Computing (WCMC): Special issue on 
Mobile Ad-hoc Networking: Research, Trends and Applications, 2(5) (2002)  483-502.  
[14] Carlisle, A. and Dozier, G., ?An off-the-shelf PSO.? In: Proceedings of the 2001Workshop 
on Particle Swarm Optimization, 2001, 1-6. 
[15] Catharina, C., Securing military decision making in a network-centric environment, PhD 
thesis, Helsinki University of Technology, 2005. 
[16] Chiang, C.-C., ?Routing in clustered multi-hop, mobile wireless networks with fading 
channel,? in: Proceedings of IEEE SICON_97, 1997, 197?211. 
[17] Chin, K., ?The behavior of MANET routing protocols in realistic environments,? in: 
Proceedings of the Asia-Pacific Conference on Communications, Perth, Australia, 2005, 
906-910. 
[18] Chlamtac, I., Conti, M and Liu, J., ?Mobile ad-hoc networking: imperatives and 
challenge,? Ad-hoc Networks 1(1) (2003) 13-64.  
159 
[19] Chu, T. and Nikolaidis, I., ?Node density and connectivity properties of the random 
waypoint model,? Computer Communications 27 (10) (2004) 914?922. 
[20] Clausewitz, C.; Vom Kriege. Auflage mit historischkritischer Wurdigung von W. Hahlweg, 
Bonn: Dummler. 1980. 
[21] Clay Wilson, Network centric operations: background and oversight issues for congress, 
CRS Reports for Congress, 2007. 
[22] Clerc, M., ?The swarm and the queen: towards a deterministic and adaptive particle swarm 
optimization,? in: Proceedings of 1999 Congress on Evolutionary Computation, 1999, 
1951-1957.  
[23] Clerc, M., and Kennedy, J., ?The particle swarm: explosion, stability, and convergence in a 
multi-dimensional complex space,? IEEE Trans. Evol. Comput. 6 (2002) 58-73. 
[24] Couto, D. D., Aguayo, D., Chambers, B and Morris, R., ?Performance of multi-hop 
wireless: Shortest path is not enough,? in: Proceedings of the First Workshop on Hot Topics 
in Networks (HotNets-I), 2002. 
[25] David Oliver J?rg, Performance comparison of MANET routing protocols in different 
network sizes, Computer Science Project, University of Berne, Swizerland, 2003. 
[26] Dengiz Orhan, Maximizaing connectivity and performance in mobile ad-hoc networks 
using mobile agents, PhD thesis, Auburn University, AL., 2007. 
[27] DoD CIO, Department of defense global information grid architectural vision, 2007, 
Accessed on April 2 2009; Available at:  
http://defenselink.mil/cio-nii/docs/GIGArchVision.pdf.  
[28] Dow C. R., Lin, P. J., Chen, S. C., Lin, J. H. and Hwang, S. F., ?A study of recent research 
trends and experimental guidelines in mobile ad-hoc networks,? in: Proceedings of 19
th
 
International Conference on Advanced Information Networking and Applications, 2005,   
72-77. 
160 
[29] Eberhart, R. C. and Kennedy, J., ?A new optimizer using particle swarm theory,? in:  
Proceedings of the sixth International Symposium on Micro Machine and Human Science, 
IEEE service center, 1995, 39-43. 
[30] Eberhart, R. C. and Shi, Y., ?Evolving artificial neural networks,? in: Proceedings of Int'l 
Conference on Neural Networks and Brain, 1998. Beijing, P. R. China.  
[31] Eberhart, R. C. and Shi, Y., ?Particle swarm optimization: developments, applications and 
resources,? in: Proceedings of Congress on Evolutionary Computation, 2001, Seoul, Korea.  
[32] Eberhart, R. C., & Shi, Y., ?Comparing inertia weights and constriction factors in particle 
swarm optimization,? in: Proceedings of the IEEE Congress on Evolutionary Computation, 
2000,  84?88, San Diego, CA. 
[33] Eberhart, R. C., & Shi, Y., ?Tracking and optimizing dynamic systems with particle 
swarms,? in: Proceedings of the IEEE Congress on Evolutionary Computation, 2001, 94?
100, Seoul, Korea. 
[34] Frodigh, M., Johansson, P., Larsson, P., ?Wireless ad-hoc networking: The art of 
networking without a network,? Ericsson Review 4 (2000) 248-263. 
[35] Green, D. B. and Obaidat, M. S., ?An accurate line of sight propagation performance model 
for ad-Hoc 802.11 Wireless LAN (WLAN) devices,? in: Proceedings of IEEE ICC 2002, 
New York, 2002. 
[36] Grossglauser M. and Tse, D.N.C., ?Mobility increases the capacity of ad-hoc wireless 
networks,? IEEE/ACM Transactions on Networking, 10(4) (2002) 477-486. 
[37] Guerin, R., A. Orda, A., ?Computing shortest paths for any number of hops,? IEEE/ACM 
Transactions on Networking, 10 (5) (2002) 613?620. 
[38] He L., Wei, Y., ?A measure of mobility for evaluating mobile ad-hoc network 
performance,? in: Proceedings of ICMMT2008, 3 (2008) 1528-1531. 
161 
[39] Ishibashi B., Boutaba, R., ?Topology and mobility considerations in mobile ad-hoc 
networks,? Ad-hoc Networks, 3 (2005) 762-776. 
[40] Jack. L. B., Chimento, P. F., Haberman, B. K. and Kasch, W. T., ?Key challenges of 
military tactical networking and the elusive promise of MANET technology,? Communications 
Magazine, 44(11) (2006) 39-45. 
[41] Jacquet, P., Muhlethaler, P., Clausen, T., Laouiti, A., Qayyum, A. and Viennot, L.,  
?Optimized link state routing protocol for ad-hoc networks,? in: Proceedings of the IEEE 
International Multi Topic Conference (INMIC 2001), Technology for the 21
st
 Century, 2001, 
62-68. 
[42] James Canan, (2006). Timing in battle: The T-Sat edge, American Institute of Aeronautics 
and Astronautics, Inc., Aerospace America. 
[43] Jardosh, A., Belding-Royer, E. M., Almeroth, K. C. and Suri, S., ?Towards realistic 
mobility models for mobile ad-hoc networks,? in: Proceedings of ACM MobiCom, 2003, 
217-229. 
[44] Ji, C., Zhang, Y., Gao, S., Yuan, P. and Z. Li., ?Particle swarm optimization for mobile ad-
hoc networks clustering,? in:  Proceedings of IEEE International Conference on, vol.1, 
2004, 372-375.  
[45] Jin, Y., Branke, J., ?Evolutionary optimization in uncertain environments ? a survey,? IEEE 
Transactions on Evolutionary Computation 9 (3) (2005) 303-317. 
[46] John J. Garstka, Network centric warfare: an overview of emerging theory, 2000, Accessed 
on September 17, 2008; Available 
at: http://www.mors.org/publications/phalanx/dec00/feature.htm. 
[47] Johnson, D.B., Maltz, D. A., Dynamic source routing in ad-hoc wireless networks, in: H.K. 
T. Imielinski (Ed.), Mobile Computing, Kluwer, Boston, 1996, 153?181.  
[48] Kalaba, R. and Tesfatsion, L., ?Exact sequential, smoothing and prediction for non linear 
systems,? Non-Linear Analysis 12 (6) (1998) 599-615. 
162 
[49]  Kalaba, R. and Tesfatsion, L., ?An organization principle for dynamic estimation,? Journal 
of Optimization Theory and Applications, 64 (3) (1990) 445-470. 
[50] Karhima, T.   Lindroos, P.   Hall, M.   Haggman, S.-G., ?A link level study of 802.11b 
mobile ad-hoc network in military environment,? in: Proceedings of Military Communications 
Conference (MILCOM 2005), vol.3, 2005, 1883- 1886. 
[51] Kennedy, J., ?Small worlds and megaminds: Effects of neighborhood topology on particle 
swarm performance,? in: Proceeding of the 1999 Conference on Evolutionary Computation, 
1999, 1931-1938. 
[52] Kennedy, J. and Eberhart, R. C. ?Particle swarm optimization,? in: Proceedings of IEEE 
international conference on neural networks, 4, 1995, 1942-1948. 
[53] Kennedy, J., Eberhart, R. C., Swarm intelligence, Morgan Kaufmann Series in 
Evolutionary Computation, 2001. 
[54] Kwak, B.J., Song, N.O. and Miller, L.E., ?A standard measure of mobility for evaluating 
mobile ad-hoc network performance,? IEICE Transactions on Communications E86-B 
(2003) 3236-3243. 
[55] Lee, S-J., Hsu, J., Hayashida, R., Gerla, M. and R. Bagrodia., ?Selecting a routing strategy 
for your ad-hoc network,? Computer Commnications 26(7) 723-733. 
[56] Lekovic, B., Mieghem, P. V., ?Link state update policies for quality of service routing,? in: 
Proceedings of 8
th
 IEEE Symposium on Communications and Vehicular Technology in the 
Benelux (SCVT2001), Delft, The Netherlands, 2001, 123-128 
[57] Libicki, M., What is information warfare?, Institute for National Strategic Studies, 1995. 
[58] Lin, G., Noubir, G. and Rajaraman, R., ?Mobility models for Ad-hoc network simulation,? 
in:  Proceedings of the 23rd Conference of the IEEE Communications Society (INFOCOM), 
2004.  
163 
[59] Luddy, J., ?The challenge and promise of network-centric warfare,? Lexington Institue: 
2005, Accessed on December 12, 2008; Available at:  
http://www.lexingtoninstitute.org/docs/521.pdf . 
[60] Macker, J., Corson, M. S., ?Mobile ad-hoc networking and the IETF,? ACM Mobile 
Computing and Communications Review 2(1) (1998) 7-9. 
[61] Masip-Bruin, X., Yannuzzi, M., Domingo-Pascual, J., Fonte, A., Curado, M., Monteiro, E., 
Kuipers, P., Mieghem, V., Avallone, S., Ventre, G., Aranda-Guti?rrez, P., Hollick, M., 
Steinmetz, R., Iannone, L. and Salamatian, K., ?Research challenges in QoS routing, 
Computer Communications Journal,? Elsevier 29 (5) (2006) 563-581. 
[62] Murthy, S. and Garcia-Luna-Aceves, J. J., ?An efficient routing protocol for wireless 
networks,? Mobile Networks and Applications (Hingham, MA: Kluwer Academic 
Publishers) 1(2) (1996) 183 ? 197. 
[63] NIST, Shortest path, Accessed on July 26, 2008; Available at:  
http://www.nist.gov/dads/HTML/shortestpath.html  
[64] Office of Force Transformation, The implementation of network-centric warfare, Accessed 
on January 22, 2009; Available at:  
http://www.au.af.mil/au/awc/awcgate/transformation/oft_implementation_ncw.pdf 
[65] Oliveira, C. A. and Pardalos, P. M., An optimization approach for cooperative 
communication in ad-hoc networks, Technical report, School of Industrial Engineering and 
Management, Oklahoma State University, 2005. 
[66] ORACLE, Building a network-centric warfare architecture, White paper, 2004, Available 
at: http://www.oracle.com/industries/government/ncwwhitepaperr1.pdf. 
[67] Parsopoulos,  K. E. and Vrahatis, M. N., ?Recent approaches to global optimization 
problems through particle swarm optimization,? Natural Computing: an International 
Journal 1(2-3) (2002) 235-306. 
164 
[68] Pazand, B. and McDonald, C., ?A critique of mobility models for wireless network 
simulation,? in: Proceedings of Computer and Information Science, ICIS, 6th IEEE/ACIS 
International Conference on, 2007, 141-146. 
[69] Peacock, B. A., Connecting the Edge: Mobile Ad-Hoc Network (MANETs) for network 
centric warfare, Blue Horizons paper, Center for Strategy and Technology, USAF Air War 
College, 2007. 
[70] Pearlman, M.R., Haas, Z. J. and Samar, P., The zone routing protocol (ZRP) for Ad-hoc 
networks, Internet Draft, draft-ietf-manet-zone-zrp-04.txt, 2002. 
[71] Pearlman, M.R. and Haas, Z. J., ?Determining the optimal configuration for the zone 
routing protocol,? IEEE Journal on Selected Areas in Communications 17 (8) (1999) 1395?
1414. 
[72] Peng, B., Kemp, A. H. and Boussakta, S., ?QoS routing with bandwidth and hop-count 
consideration:  A performance perspective,? Journal of Communications 1 (2) (2006). 1-11. 
[73] Perisa, D., Allwright, A. and Pourbeik, P., ?Structural dynamics of war game MANETs,? in: 
Proceedings of Communications and Information Technologies, ISCIT '07. International 
Symposium, 2007, 830-835. 
[74] Perkins, C.E. and Bhagwat, P., ?Highly dynamic destination sequenced distance-vector 
routing (DSDV) for mobile computers,? Computer Communications Review 24 (4) (1994) 
234?244. 
[75] Perkins C.E., Royer, E. M., Das, S. R. and Marina, M. K, ?Performance comparison of two 
on-demand routing protocols for Ad-hoc Networks?,  IEEE Pers. Commun  8(1) (2001) 16-
28. 
[76] Perkins, C.E. and Royer, E. M., ?Ad-hoc on-demand distance vector routing,? in: 
Proceedings of 2nd IEEE Workshop on Mobile Computing Systems and Applications, 1999, 
90?100. 
165 
[77] Poli, R., Kennedy, J. and Blackwell, T., ?Particle swarm optimization,? Swarm Intelligence 
1(1) (2007) 33-57. 
[78] Raby, M, Wireless networks jump in popularity: survey, 2007, Accessed on May 24 2009; 
Available at:  
http://www.tomshardware.com/news/wireless-networks-survey,4372.html. 
[79] Quintero, A., Pierre, S. and Macabeo, B., ?A routing protocol based on node density for ad-
hoc networks,? Ad-hoc Networks 2 (3) (2004) 335-349. 
[80] Raghavan, V. N., Venkatesh, T., Peer, M. L., Praveen, D. P., ?Evaluating performance of 
quality-of-service in large networks,? in: Proceedings of World Academy of Science, 
Engineering and Technology, vol.20, 2007.  
[81] Rani, K. N. A. and Sharif, B. S., Particle Swarm Optimization (PSO): Fundamental 
concepts and applications in mobile communications, University of Newcastle upon Tyne, 
United Kingdom, 2008. 
[82] Robert F. Dillingham, Communicating on the Move: Mobile ad-hoc networks, SRA 
International, Inc. the Journal of Defense Software Engineering, Accessed on September 27, 
2008; Available at: 
 http://www.stsc.hill.af.mil/crosstalk/2007/07/0707DillinghamNathans.html. 
[83] Royer, E. M., Melliar-Smith, P. M., Moser, L. E., ?An analysis of the optimum node density 
for ad-hoc mobile networks,? in: Proceedings of the IEEE International Conference on 
Communications, 2001, 857?861. 
[84] Roux, N., Pegon, J. S., Subbarao, M. W., ?Cost adaptive mechanism to provide network 
diversity for MANET reactive routing protocols,? in: Proceedings of IEEE MILCOM, 2000, 
387-391. 
[85] Sahin, C. S., Urrea, E., Uyar, M. U., Conner, M., Self-deployment of mobile agents in 
MANETs for military applications, 2008, Accessed on March 22, 2009; Available at: 
http://asc2008.com/manuscripts/B/BP-22.pdf. 
166 
[86] Salmanian, M., Military wireless network information operation scenarios, Defense R&D 
Canada Ottawa Technical Memorandum DRDC Ottawa TM, 2003, 2003-241. 
[87] Sanchez, M, Mobility models,  2001, Accessed on Feb 7 , 2009; Available 
at http://www.disca.upv.es/misan/mobmodel.htm.  
[88] Satyanarayanan, M., ?Fundamental challenges of mobile computing,? in: Proceddings of 
15
th
 ACM symposium. Principles of dist. Comp, 1996. 
[89] Schechtmann, G., Manipulating the OODA loop: the overlooked role of information 
resource management in information warfare, Master?s Thesis, Faculty of the graduate 
school of logistics and acquisition management of the Air Force Institute of Technology, 
1996. 
[90] Shaikh, A., Rexford, J. and Shin, K., ?Evaluating the impact of stale link state on quality-
of-service routing,?IEEE Trans. Networking 9 (2001) 162?176. 
[91] Shi, Y. and Eberhart, R. C., ?Parameter selection in particle swarm optimization,? in: 
Proceedings of Evolutionary Programming VII (EP 98), 2001, 591-600. 
[92] Shi, Y., ?Particle swarm optimization,? IEEE Neural Networks Society (2004) 8-13. 
[93] Shukla, D, Mobility models in ad-hoc networks, Master's Thesis, KReSIT-IIT, Bombay, 
2001. 
[94] Smith, A. E., Effects-based operations: applying network-centric warfare in peace, crisis, 
and war, DoD CCRP, Washington, DC, 2002. 
[95] Smith, A. E., Network-centric warfare, Naval War College Review, 2001, Accessed on 
March 24, 2009; Available at:  
http://findarticles.com/p/articles/mi_m0JIW/is_1_54/ai_75762213. 
[96] Sridhar, K.N. and Chan, M. C., ?Stability and hop-count based approach for route 
computation in MANET,? in: Proceedings of IEEE ICCCN, 2005, 25-31. 
167 
[97] Stepanov I., Maron, P. J. and Rothermel, K., ?Mobility modeling of outdoor scenarios for 
MANETs,? in: Proceedings of the38th Annual Simulation Symposium (ANSS?05), 2005, 
312-322.  
[98] Sun, T, The art of war, Filiquarian Publishing, LLC., 2006. 
[99] Ted, M., ?Developers of net-centric warfare battle complexity,? Journal of Electronic 
Defense, 7 (2005) 23. 
[100] Walter, M., ?Technology for a naval revolution in military affairs,? Second Navy RMA 
Round Table, Science Applications International Corportation, Tysons Corner, Virginia, 
1997. 
[101] Wang, N-C., and S-W. Chang, ?A reliable on-demand routing protocol for mobile ad-hoc 
networks with mobility prediction,? Computer Commnications 29(1) (2005) 123-135. 
[102] Warren Ferster, Military bandwidth demand energies market, SpaceNews, 2003, Accessed 
on March 1, 2009; Available at:  
http://www.spacecom/spacenews/archive03/militaryarch_090203.html 
[103] Wikipedia, Partical Swarm Optimization, Accessed on July 24, 2008; Available 
at:  http://en.wikipedia.org/wiki/Particle_swarm_optimization  
[104]  Wikipedia, Shortest path problem, wikepedia, Accessed on August 17, 2008; Available 
at:  http://en.wikipedia.org/wiki/Shortest_path_problem#Applications.  
[105] Wikipedia, Telecommunication network Wikipedia.org, Accessed on July 15, 2008; 
Available at: http://en.wikipedia.org/wiki/Telecommunication_network.  
[106]  Wikipedia,Wi-Fi, Accessed on June 27, 2009; Available 
at:   http://en.wikipedia.org/wiki/Wi-Fi 
[107] Wilson J. W, The importance of mobility model assumptions on route discovery, data 
delivery, and route maintenance protocols for ad-hoc mobile networks, Virginia 
Polytechnic Institute and State University, 2001. 
168 
169 
[108] Zheng, Y.-L., Ma, L.-H., Zhang, L.-Y., and Qian, J.-X., ?On the convergence analysis and 
parameter selection in particle swarm optimization,? in: Proceedings of the IEEE 
International Conference on Machine Learning and Cybernetics, 2003, 1802?1807. 
[109] Zhou, B., Xu, K. and Gerla, M., ?Group and swarm mobility models for Ad-hoc network 
scenarios using virtual tracks,? in: Proceedings of Military Communications Confrerence, 
MILCOM2004, IEEE 1, 2004, 289-294. 
[110] Zitlzler E., Laumanns, M., Bleuler, S., ?A tutorial on evolutionary multiobjective 
optimization,? Metaheuristics for Multiobjective   Optimization, 2004, 3-37.