Professional Documents
Culture Documents
Merts Satellite 2017
Merts Satellite 2017
André-Jan Merts
March 2017
Stellenbosch University https://scholar.sun.ac.za
Declaration
By submitting this thesis electronically, I declare that the entirety of the work contained
therein is my own, original work, that I am the sole author thereof (save to the extent
explicitly otherwise stated), that reproduction and publication thereof by Stellenbosch
University will not infringe any third party rights and that I have not previously in its
entirety or in part submitted it for obtaining any qualication.
Date: . . . . . . . . .March
. . . . . . .2017
....................
i
Stellenbosch University https://scholar.sun.ac.za
Abstract
ii
Stellenbosch University https://scholar.sun.ac.za
Uittreksel
AJ. Merts
Departement Elektriese en Elektroniese Ingenieurswese,
Universiteit van Stellenbosch,
Privaatsak X1, Matieland 7602, Suid Afrika.
Namate die ruimte meer toeganklik word, verhoog die lewensvatbaarheid van die ge-
bruik van satellietnetwerke en toegewyde dienste wat op daardie netwerk gerig is. Sa-
telliete is steeds relatief duur om te lanseer en kan nie maklik onderhou of herstel word
na dit in 'n wentelbaan geplaas is nie, gevolgelik kan die simulasies van die netwerk voor
lansering baie nuttig wees. Daar is verder ook heelwat navorsing oor netwerkprotokolle
vir aardnetwerke gedoen wat nog nie volledig bestudeer is vir satellietnetwerke nie en wat
dus 'n behoefte skep vir 'n satellietnetwerk simulator.
Enige netwerk se eienskappe kan op 'n verskeidenheid maniere geïnterpreteer word,
wat daartoe lei dat alle netwerkmaatstawwe op so 'n wyse gedenieer moet word dat daar
geen dubbelsinnigheid is nie. Roeteringsprotokolle kan op verkeie manier geklassiseer
word wat hul inherente voordele, nadele, vereistes en beperkings teen toon stel. Deur
die roeteringsprotokolle se eiensakppe saam met inligting oor die komponente van roete-
ringsprotokole te gebruik , kan 'n analise van roeteringsprotokolle uitgevoer word en twee
kandidaat roeteringsprotokolle is gekies vir toetsing. Vir hierdie navorsing is AODV gekies
om die stelsel te help verieer en AntHocNet is gekies om aan te toon dat die simulator
buigsaam genoeg is om verkeie roeteringsprotokolle se funksionaliteit te simuleer.
Satellietnetwerke is handmatig ontleed en die kernverskille tussen aards-en satellie-
tenetwerke wat spesiale aandag geniet het. Deur hierdie informasie saam met vorige
navorsing en aanbevelings te gebruik is 'n verspreide netwerk soortgelyk aan die Iriduim
netwerk gekies om as verikasie van die simulator te dien. 'n Voorheen voorgestelde net-
werk, bekend as die OuterNet, is gekies om verdere navorsing oor die toepasbaarheid van
AODV en AntHocNet op ander satellietkonstellasies uit te voer.
Die uitbreiding van SatSim vanaf 'n basiese punt-tot-punt simulator tot 'n netwerk-
simulator wat in staat is om komplekse roeteringsprotkolle te simuleer, is volledig gedo-
kumenteer, met alle ontwerpskeuses verduidelik en geregverdig. Die hoof ontwerpsmaat-
stawwe van gebruikersgerief, modulariteit en herbruikbaarheid is gebruik, met die kode
wat gestruktueer is om uitbreiding van die simulator te vergemaklik. Weens beperkte
iii
Stellenbosch University https://scholar.sun.ac.za
UITTREKSEL iv
herhaalbare resultate is verdere toetsing vir die MAC laag kanaaltoegang en vir AODV
en AntHocNet uitgevoer.
SatSim is suksesvol uitgebrei om satellietnetwerke te simuleer. Voorlopige resultate vir
die toepaasbaarheid van AODV teenoor die van AntHocNet vir twee satellietkonstellasies
word gegee.
Stellenbosch University https://scholar.sun.ac.za
Acknowledgements
I would like to thank the following people, without whom this thesis would not have been
possible.
Firstly, my parents Chris and Hilda Merts for their unwavering support and moti-
vation in completing this thesis, especially in the face of many challenges.
My supervisor Arno Barnard for his valued insights and contributions as well as
keeping me on the correct track and consistently motivated.
My friends, especially my comrades in the ESL who were always cheery and striving
hard on various projects. Also my close friends Hendrik, Anita and James, whose
company and conversation I truly cherish.
Lastly, I would like to thank all the people who contribute to sites such as StackOver-
Flow in their own free time and made it one of the best resources on the internet.
v
Stellenbosch University https://scholar.sun.ac.za
Contents
Declaration i
Abstract ii
Uittreksel iii
Acknowledgements v
Contents vi
List of Figures viii
List of Tables ix
Acronyms xi
Scientic Constants . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xiii
2 Literature Review 8
2.1 Denitions of Network Metrics . . . . . . . . . . . . . . . . . . . . . . . . . 8
2.2 Routing Methodologies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
2.3 Routing Protocols Frequently Used for MANETS . . . . . . . . . . . . . . 18
2.4 Components of Routing Protocols . . . . . . . . . . . . . . . . . . . . . . . 20
2.5 Overview of Satellite Networks . . . . . . . . . . . . . . . . . . . . . . . . . 21
2.6 Review of SatSim V.1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
vi
Stellenbosch University https://scholar.sun.ac.za
CONTENTS vii
4.2 Simulator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
4.3 SimObjects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
4.4 Testing of Channel Access . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
4.5 Testing of AODV . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
4.6 Testing Of AntHocNet . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
5 Protocol Evaluation 67
5.1 Use of a Distributed Constellation . . . . . . . . . . . . . . . . . . . . . . . 67
5.2 Baseline Scenario . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
5.3 Comparison of AODV and AntHocNet for the OuterNet . . . . . . . . . . 73
Appendices 81
A SatSim Supplementary 82
A.1 Installation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82
A.2 Conguration and Use . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
Glossary 88
Bibliography 90
Stellenbosch University https://scholar.sun.ac.za
List of Figures
viii
Stellenbosch University https://scholar.sun.ac.za
List of Tables
ix
Stellenbosch University https://scholar.sun.ac.za
LIST OF TABLES x
Acronyms
AODV Ad Hoc On-Demand Distance Vector Routing. 3, 5, 6, 15, 17, 1921, 26, 27, 31,
37, 38, 4043, 50, 55, 59, 60, 63, 67, 6973, 7577, 79, 85
ISL Inter Satellite Link. 18, 22, 29, 55, 62, 76, 79
LOS Line of Sight. 3, 13, 14, 19, 26, 29, 32, 33, 36, 48, 7375
MAC Media Access. 10, 20, 29, 3133, 37, 47, 57, 6062, 67, 70, 7679, 8486
MANET Mobile Ad Hoc Network. 3, 7, 8, 1315, 1721, 24, 37, 38, 42, 57, 77, 78
xi
Stellenbosch University https://scholar.sun.ac.za
ACRONYMS xii
RF Radio Frequency. 54
WSN Wireless Sensor Network. 13, 15, 21, 24, 25, 54, 57, 58, 80
ACRONYMS xiii
Scientic Constants
Standard gravitational parameter (Earth) : µ = 3.986004418(9) ∗ 1014 m3 s−2
ACRONYMS 1
Stellenbosch University https://scholar.sun.ac.za
Chapter 1
The Need for a Satellite Network
Simulator
1.1 Background
Space is becoming accessible to more users, due to the decreasing relative cost of building
satellites, more adaptive markets and the ever increasing knowledge base of the eld.
Current trends suggest the increase of satellite launches will continue in the future, with
more space applications appearing and becoming viable as more users gain access to space.
In order to eciently use these satellites, especially in networks, there must be a platform
to study the optimal way for satellite networks to communicate.
When approaching the problems inherent to satellite communication and networks,
some novel solutions have become more viable due to more users being involved in space.
One such idea that was proposed by the Electronic Systems Laboratory (ESL) at Stellen-
bosch University is the use of an OuterNet [1], however for such an idea to be nancially
viable it would need a solid subscription base. In order to justify the cost of subscription
users must clearly be presented with the benets of such a system. While many cong-
urations and routing protocols have been investigated, the vast eld of routing protocols
has not been explored in the same detail for satellite networks as for terrestrial wireless
sensor networks.
It becomes desirable to use this wealth of accrued knowledge, make the necessary
adjustments to proven routing protocols and then attempt to use and evaluate them
in a satellite environment. However, for many of these routing protocols little to no
information is available for their performance in a satellite environment.
As such, there is a need for a new tool that can be used to easily test existing routing
protocols in a space-based environment, whilst studying and highlighting the dierences
between the satellite and terrestrial networks. A key feature of many routing protocols is
that values were chosen with specic terrestrial conditions in mind. In order to optimally
use these routing protocols with satellites, these values must be evaluated and adjusted
if the core assumptions which lead to them are signicantly dierent in a satellite envi-
ronment.
This research will therefore attempt to implement two routing protocols, each on
two dierent satellite constellations, and evaluate their performance. The research is not
primarily aimed at nding an optimal networking solution for a given satellite constellation
but rather towards creating a simulation platform on which satellite network protocols
can be evaluated. As a secondary objective preliminary performance observations for
2
Stellenbosch University https://scholar.sun.ac.za
research [6], that was aimed at the selection of modulation schemes and channel access
methods for Low Earth Orbit (LEO) satellites. However, an entirely new simulator was
developed by Le Roux[5], making use of SimPy in Python rather than DEMSMO-J in Java
as Bezuidenhout did. The work would focus heavily on optimising network throughput
on a single link basis through the development of a platform called SatSim.
SatSim was veried as an discrete events simulator that could give accurate single
link results for a point-to-point connection. This was achieved via the use of dynamic
Bit Error Rate (BER) calculations on a packet by packet basis. The AX.25 and FX.25
protocols were tested on the link and a hybrid protocol proposed in order to optimize the
communication throughput for that specic communications link.
The nal contribution of the expanded tool was an evaluation of OuterNet and the
performance of AODV and AntHocNet in such a network as well as evaluating the per-
formance of AntHocNet in an Iridium like network.
This research shows SatSim has been expanded to be a useful and easily expandable
tool. Most satellites networks have such unique operating conditions that a general solu-
tion can rarely be found. The main benets of SatSim is that it can be used to assist in
discovering the optimal communication strategy for a given scenario.
Chapter 2 - Introduces the network denitions which are used to evaluate network
performance. The chapter gives and overview on dierent routing classications
and their inherent properties. Using this information along with a short section
describing currently used routing protocols two candidate protocols are chosen for
this research. Satellite networks and their salient properties are investigated so that
two candidate networks can be selected for this research. The chapter concludes
with a review of the previous work performed on SatSim and outlining the design
principles which will be used to expand it further.
Chapter 4 - Details the expansion of SatSim, describing and justifying the design
decisions made. The Simulator structure is covered in detail along with descriptions
of the various subcomponents that are used thereby. The SimObject class and
its expanded behaviour is also explained in details. The testing scenarios for the
Channel Access , AODV and AntHocNet are also presented along with their results
and some observations.
Chapter 5 - Gives the results for the verication of SatSim on an Iridium like
network. The testing procedure and results are described and an analysis of the
performance of AntHocNet in such an environment is made, along with some general
observations. A Baseline scenario is then tested for both AODV and AntHocNet,
which is then compared to their performance when the OuterNet is introduced.
The two networks are evaluated against each other for this specic network and
observations and suggestions are made.
Chapter 6 - Concludes this research by describing what has been achieved and
what could not be achieved. Several suggestions are made towards future work,
detailing both the suggest further expansion of SatSim and new research which
could be conducted on the SatSim environment.
Stellenbosch University https://scholar.sun.ac.za
The appendices contain supplementary information on the use of SatSim and routing
protocols.
The appendices in this thesis are structured as follows:
Chapter 2
Literature Review
The following chapter gives an overview of the the denitions of metrics needed to evaluate
and describe MANET networks, as well as the information needed to fully design and
evaluate routing protocols and how they are classied. The main classes of decentralised
routing is then expanded upon.
In order to fully cover the details of decentralised routing reactive, proactive and
hybrid routing is discussed and evaluated. Attention is given to the inherent properties,
advantages , disadvantages, requirements and limitations of the dierent types, which will
allow better understanding of when a specic routing protocol is preferable.
A selection of routing protocols is presented and their suitability to satellite networks is
briey discussed, along with a short section which details the typical routing components
which are used by these routing protocols. The discussion of routing protocols concludes
with a discussion on why the routing protocols which were chosen for this research were
used.
To further describe both networks and protocols in terms of the unique needs and
characteristics of satellite networks , satellite networks and their operation must be un-
derstood. To do this the research presents an overview of satellite networks, which includes
a concise summary of some existing and planned satellite networks and a discussion on
the salient characteristics of networks and how these change in a satellite environment.
The section concludes with the choice regarding which networks will be investigated by
this research.
Finally, the chapter concludes with an analysis of the previous work which led to the
development of SatSim and the research which was performed on the simulator in [5]. The
main design drivers, limitations and decisions are highlighted and their reasons for being
maintained or removed are given, which will be used to guide the expansion of SatSim
towards a simulator that will be useful to the satellite community.
8
Stellenbosch University https://scholar.sun.ac.za
A key point raised by [8] p.8 is that the measurement for certain metrics should be
done externally to the perspective of the individual nodes, which is why the DataManager
class aref used. It was also noted that it should always be stated whether means, variances
or distributions is used.
2.1.2 Throughput
Throughput is used to indicate the rate at which data is eectively being transmitted from
the original source node to the nal destination node. This value is directly obtained by
dening two points (usually referred to as the data source and the data sink), and dividing
the amount of data which passed between them by the amount of time it took the data
to move from point A to point B.
thorough description of this see Media Access (MAC)) as well as the additional overhead
costs of retransmission (which can be highly signicant over lossy links).
When these distinctions aren't made, the eective throughput measured by an appli-
cation might be considerably lower than the given throughput specied by the protocol.
In this study only application data being transferred is used when analysing the through-
put - which implies a much lower throughput value will be calculated than other studies,
which will be closer to that measured by the applications on the communicating devices.
2.1.3 Latency
Latency is used to quantify the time it takes a transmission to propagate from node A
to node B. The rst time interval starts at the moment of transmission and the second
time interval is taken when the packet has been fully received by the destination. This
will usually be given in milliseconds, even in the satellite environment. The latency for a
given network can make it unsuitable for certain applications (such as warning systems
or interactive peer to peer systems), or otherwise highly undesirable(such as the use of
peer to peer interaction).
Latency can be approximated as the time dierence between the packet being trans-
mitted from the source and it being received by the sink. However, this time dierence
can be approximated using the equation below.
2.1.4 Jitter
Jitter is the dierence in time between the arrival of one packet and the next.While
jitter is aected by the network topology, it can also be aected by various other factors,
some of which aren't modelled in this simulation, such as the relevant processor overhead
associated with the buering and sending of data and random events.
The equation to calculate jitter is given below, with N being the number of packets
transmitted.
PN
i=2 tn − tn−1
Jitter = (2.1.3)
N
High jitter may indicate a link is operating a lossy or highly contested channel, ren-
dering it unsuitable for certain applications.
Stellenbosch University https://scholar.sun.ac.za
Corrupted Bits
Bit Error Rate = (2.1.4)
T ransmitted Bits
A bad quality of service will result in more packets being unusable which will reduce
throughput, increase jitter and eectively reduce average latency. For more information
on how the bit errors are assigned to packets, refer to subsection 4.3.3.3.
From a user's viewpoint, QoS can be a deciding factor in deciding whether they wish
to participate in the satellite services, especially for services such as satellite internet or
global communications.
2.1.6 Overhead
Overhead is all additional data needed to ensure application data can safely be sent from
the source and received at the sink. This overhead can be headers which enclose the
application data in the frame structure or are needed to create and maintain routing
tables. It also includes additional handshaking frames used to control channel access.
Route discovery time is also referred to as route acquisition time in literature [8].
2.1.9 Coverage
Coverage is the amount of surface area in which end-users can access the network. Typi-
cally a single satellite's altitude will be a deciding factor in how much ground coverage it
gives, with a LEO satellite covering a few square kilometres while three geosynchronous
satellites can provide near global coverage. However, increased coverage via altitude will
have a negative eect on latency, as this will increase propagation delay. The antenna
type and beam width can also have a very signicant eect on the footprint of the net-
work. The amount of coverage a network gives can dier vastly based upon the network
conguration and has to be matched to the needs of the network.
Coverage is also referred to as the footprint of the network, especially when referring
to satellite networks.
2.1.12 Hops
When a packet has to move across multiple nodes in order to reach a destination, the
transmission towards a relay node is counted as a hop. Hops are often viewed as a cost to
a route since the additional switching delays and processing delays are often very costly.
For this research, a path refers to a multi-hop route from a source to a destination
while a link refers to a direct connection between two nodes.
explained to ensure the reader will be able to correctly associate the classication with
what is currently being used in literature.
2.2.1.2 MANET
For this study, a similar denition of MANETs as the following adapted from [10] is used:
MANETS : A Mobile Ad-Hoc Network is a network of mobile nodes whom commu-
nicate with wireless interfaces to exchange data. The network's topology changes with no
form of central administration or prior planning.
This leads to the network having a dynamic topology, with the state of the network
at a given instant being referred to as a snapshot. A dierence between MANETs and
Wireless Sensor Network (WSN) is that most MANETs rely on intermediate nodes to
communicate with nodes they do not have direct LOS with or that are too far away for
eective communication [11]. In the MANETs scenario, it often becomes necessary for
nodes to organise themselves in order to compensate for the lack of xed infrastructure
or central administration. The OSI protocol layers, initially designed for wired networks,
also faces unique challenges as shown in 2.2.
Table 2.2: Protocol Layer Challenges for MANETs. Adapted from [11], p3
Protocol Layer Challenge
Physical Layer Rapidly Changing Links
Channel Access Collisions and Fair Access, see B.1.2, B.1.3
Network Layer Handle Complex Paths
Transport Layer Packet Loss and Delay
Application Random Disconnection and Reconnection
2.2.2.4 Summary
When analysing table 2.3 , some of the disadvantages have a far heavier weight than
others. The diculty of putting some form of central administration in place for a large
scale satellite network is the main deciding factor. While this can be oset using the
hierarchical routing strategy, the diculty of creating the necessary system and the lim-
itations placed upon throughput make distributed routing a more attractive alternative.
This research will therefore only consider decentralised routing protocols as viable for
satellite networks.
certain scenarios than others. In order to be usable for a wide variety of scenarios, SatSim
needs to be able to simulate reactive, proactive and hybrid protocols.
2.2.3.4 Summary
As can be seen in table 2.4 , reactive and proactive routing will favour dierent applications
and scenarios at dierent times. As such, in order to allow SatSim to be an ecient tool
for simulating a wide variety of satellite networks, it needs to be able to simulate reactive,
proactive and hybrid protocols.
Table 2.4: Comparison of Reactive and Proactive Routing Protocols for MANETs
Property Reactive Proactive
Routes On Demand Always
Power Consumption Lower Higher
Overhead Lower Higher
Route Discovery Time Critical [8] Minimal Eect
Latency Slightly Higher Slightly Lower
Jitter Slight Higher Slightly Lower
Route Maintenance Negligible Constant
Quality of Service Slightly Lower Slightly Higher
Route Repair Lower Higher
2.3.1 Summary
Table 2.5 provides a summary of dierent routing protocols and what they can be classied
as. An ideal choice for the evaluation of SatSim would cover most routing functionality.
2.3.2 ANSI
Ad Hoc Networking with Swarm Intelligence (ANSI) is a hybrid routing protocol proposed
for hybrid ad hoc networks. ANSI makes use of elements from both AODV and AntHocNet
[18].
Pure MANET notes in ANSI use only reactive routing and choose routes determinis-
tically while nodes belong to the more infrastructures, xed network makes use of both
proactive and reactive routing to perform stochastic routing.
ANSI has many of the same routing elements as AntHocNet.
Stellenbosch University https://scholar.sun.ac.za
2.3.3 AntHocNet
AntHocNet is a hybrid routing protocol which was proposed by [19] as an extension of
the ant colony optimization strategy specically for MANETs. Refer to 3.4 for a detailed
description of this protocol and its implementation in the SatSim environment.
One variation which may solve some of the issue caused by a quasi-static network is
HopNet , a modied protocol which combines AntHocNet with aspects of Zone Routing
Protocol (ZRP).
2.3.4 AODV
AODV is a reactive routing protocol which was proposed by [20]. Refer to section 3.2 for
a detailed description of this protocol and its implementation in the SatSim environment.
There has been extensive research regarding the performance of AODV and it is often
used as the benchmark to compare new routing protocols , with [19] refering to it as the
de facto standard.
2.3.5 DSDV
Dynamic Destination Sequenced Distance-Vector Routing (DSDV) [17] is a distance based
routing protocol where each node broadcasts information to the other nodes to allow all
nodes to maintain routing tables. Distance is the main metric to determine routes, which
is not often the case for satellite networks, where LOS is often more important. DSDV
does not perform well in a highly dynamic network.
2.3.6 DSR
Dynamic Source Routing Protocol (DSR) [21] is a hybrid routing protocol which makes use
of on-demand elements in order to nd and maintain routes. It makes use of a discrete
Route Discovery and Route Maintenance processes. Both of the processes operate on
demand. Multiple routes are stored towards a destination, which means that upon a link
failure another route can be tried. However this also presents a new potential problem,
known as Route Reply Storms (as opposed to Route Request Flooding) in which multiple
nodes keep sending Route Replies. DSR is similar to AODV but has less additional
information available.
2.3.7 OSLR
Optimized Link State Routing Protocol (OLSR) [22] is a proactive algorithm that period-
ically exchanges messages to allow all nodes to correctly estimate topology information.
Some variations of this protocol include OSLR-H and OSLR-SAT, which makes use
of satellites to broadcast their terrestrial networks messages. The terrestrial network is
vastly helped by the satellites coverage, covering far more nodes than a single terrestrial
node would be able to on its own.
2.3.8 TORA
Temporally-Ordered Routing Algorithm (TORA) is a routing protocol that favours multi-
ple, loop free routes that can be easily found above locating shortest path routes. Steps are
Stellenbosch University https://scholar.sun.ac.za
also taken to minimize the communication overhead. Each node only explicitly initiates
a query when it needs to send data towards a specic destination.
Its advantages include multiple paths being created, good performance in dense net-
works and disadvantages include not being scalable and being outperformed by both
AODV and other protocols.
2.3.9 ZRP
ZRP [23] is a reactive routing protocol that works in part by grouping the nodes in the
MANET into zones that can then communicate with each other on a zone by zone level.
Several strategies can then be implemented in order to increase the eectiveness of the
communication, such as using proactive methods between the various zones and reactive
methods within individual zones.
to explore new routes or to ensure certain connections are still valid. Proactive elements
are usually periodically sent regardless of network activity.
In order to evaluate the performance of certain protocols for a specic scenario, the
unique properties of a specic network must be analysed and dened rst. Typical network
performance analysis metrics were discussed in section 2.1, while the design parameters
for networks will be described after giving a summary of satellite constellations.
Satellite constellations are very expensive to implement and require a large amount
of resources both in terms of manpower to create operate and maintain. Orbital space
is also a valuable resource, with certain orbits being more desirable than others. The
increasing prevalence of space debris further increases congestion for optimal orbits and
severely limits amount of favourable orbits available whilst also also raising de-orbiting
as a critical issue.
Each satellite constellation needs a signicant amount of start up capital, a dedicated
communication band and operational costs, which needs to be favourable compared to its
benets.
It can be noted that satellite constellations are often congured and launched as planes
with set ISL .
In order to illustrate the currents needs and expectations from satellite constella-
tions the section below includes typical uses of satellite constellations, including scientic
constellations (A-train)[24], successful commercial constellations(Globalstar) and failed
commercial constellations that managed to reinvent their business model and become
successful using modern technology(Iridium).
2.5.1.1 Summary
2.5.1.2 A-train
The A-Train(Afternoon constellation) consists of six active Earth Observational Satellites
placed inside a sun synchronous orbit on the same orbital track [24]. The satellites
are placed so that their observations can be combined to construct high-denition three
dimensional images of the Earth's atmosphere and surface.
Some of the main goals of the constellation include studying atmospheric composition,
the carbon cycle, climate change, water and energy cycles, weather, changes in the global
Earth system and the earth surface and interior.
Stellenbosch University https://scholar.sun.ac.za
2.5.1.3 Iridium
Iridium is meant to provide data connection to various mobile users, such as voice and
text data for cellular phones. It was developed by Motorola. The networks consists of 66
satellites, which is enough to provide global earth coverage. The 66 satellites are placed
within 6 orbital planes with each plane containing 11 satellites and being 30 degrees apart.
An issue with Iridium was that the entire constellation had to be in place before its
functionality could be properly utilised, several ramping up capital costs before any return
on investments can be achieved.
2.5.1.5 Globalstar
Globalstar [26] is a satellite constellation that oers users telephonic communications. It
combines satellites with ground stations, known as GlobalStar Gateways. Users have to
be in range of a GateWay in order to make use of the network.
What is noteworthy of the Globalstar System is that parts of the constellation has
become inactive, causing users to only have functionality when certain satellite are in
range. To help users with this Globalstar store created an Optimum Satellite Availability
tool which will tell a user when a satellite will be well positioned for a specic geographical
position.
2.5.1.6 Inmarsat
Inmarsat [27] provides global mobile and telephone coverage using 12 geostationary satel-
lites, of which 11 are still operational. Inmarsat oers a a wide variety of services using
dierent satellite constellations , including the Global Xpress service, which is a global
high-speed broadband network. It also oers M2M, a global machine-to-machine service
to monitor and manage remote assets.
2.5.1.7 Orbcomm
Orbcomm [28] is a company which oers various data communication solutions and has a
network of 31 LEO satellites and 16 ground stations that act as additional gateways for
communication. Many of the satellites launched by the company are no longer functional.
The ORBCOMM systems have an inherent latency that makes them unsuited for timing
critical applications.
adapted for the Satellite environment and details of the adaptations made are discussed
in the following paragraphs.
1. Dynamic Topologies: the nodes will move in xed, arbitrary orbits, leading to
a changing network topology. This movement is less random and more predictable
than in most terrestrial MANETs meaning snapshots of the network can be cal-
culated once and considered constant for multiple runs while other parameters are
being adjusted.
3. Energy constrained operation: the satellites in the networks will rely partly on
exhaustible means for energy, as well as having limited recharge capability via the
use of solar panels. Energy conservation can therefore be considered a signicant
factor for optimization.
Any frame that is broadcast (such as a Reactive Ant (RA) or a Route Request
(RREQ)) will also result in many more broadcasts, leading to a snowball like eect often
referred to as ooding the network.
Densely populated networks may still be desirable for satellite networks, as they can
provide coverage over larger areas and better serve large numbers of user , provided that
ecient load balancing techniques have been implemented. Although satellites with higher
altitudes may be larger and have superior computational power and memory, allowing
them to better service multiple users, the increased altitude will have a negative eect on
latency.
the interval will reduce overhead , but increase the time it takes for new routes to be
discovered or routes to be repaired.
Many of the parameters have to be carefully set in order to avoid protocol specic
pitfalls. As an example, AODV can use the NETWORK DIAMETER parameter to
avoid ooding a network with broadcasts. If this parameter is set too high, ooding can
readily occur. If it is set too low, the destination may be viewed as unreachable. All
parameters have trade-os which must be considered on a network by network basis.
Spherical Earth Model - Kept since it has minimal inuence on Satellite Networks
and is not justied in terms of increased runtime.
Communication - only relies on LOS. Signal diraction and other phenomena are
not simulated. This was maintained since ISL have even less attenuation.
Burst Errors are not simulated since they are caused by random events not deter-
mined by protocol choice. Implementing random burst errors may therefore have
created an unfair data bias against certain protocols. There may still be interest
to see how protocols react to them, but it should not be used when comparing two
protocols.
Channel access model - An environment class and MAC layer simulations have
been added in order to control access to the transmission medium.
Weather Model - this has not been considered since its cost in terms of execution
would most likely render a network simulation near unusable, while not adding
signicant value to the simulation.
Circular Orbits - Satellites can only orbit in circular orbit paths - changed due to
what is being simulated. With sucient passes the issues will average out.
2.6.6 OOP
As stated by [31], all computer programs consist of code and data. A program can
be conceptually structured around currently occurring, or what data is currently being
aected. The rst paradigm is a process-orientated model, where code acts on data.
The second paradigm was proposed to manage increasing complexity of computer pro-
grams and was named Object Orientated Programming (OOP). The main characteristic
of an object orientated program is that data controls the access to code.
A key advantage of OOP is abstraction, which is very useful for a simulator environ-
ment since it helps users better interpret what is actually happening in a given simulation.
By using OOP correctly the simulation becomes understandable in human terms.
Stellenbosch University https://scholar.sun.ac.za
In order to best serve the design drivers given earlier, it is desirable that code is easy
to maintain and understand. The complexity of the simulation code is therefore best
approached by using OOP principles.
Stellenbosch University https://scholar.sun.ac.za
Chapter 3
Routing Protocols and Channel Access
This chapter describes the addition of crucial functionality needed for a network simulator,
namely some method of regulating and controlling channel contention and access. To
achieve this an implementation of the MAC 802.11 layer was implemented in the SatSim
environment.
This chapter also details the fundamental descriptions of the two routing protocols
currently implemented to assist with the verication and testing of the SatSim environ-
ment, namely AODV and AntHocNet. This section will give an overview of both protocols
as well as including and justifying any changes that were made to allow the protocols to
function in the SatSim environment.
31
Stellenbosch University https://scholar.sun.ac.za
These diculties arise since many of these values were chosen for terrestrial networks
with distances no greater than 300 meters separating the nodes. The biggest dierence
between the terrestrial and space-based networks is the size of the propagation delays.
The massive dierence in distance between nodes suggests that the interval times will be
far too short for space-based application.
Additional testing and the research from [32] and [33] suggested various techniques
to deal with the far greater propagation delay, including variable contention windows.
However, for this research the more basic approach was taken of increasing the slotTime
to better match the average propagation delays experienced in a satellite environment.
The idea of a variable contention window is very appealing however, due to the massive
dierence in distances which can occur in a satellite environment.
A customised MAC 802.11SAT implementation was designed and used for this research
and its conguration is detailed in table 3.2.
sensing mechanism can prevent some viable data transmissions from occurring as e-
ciently as they would occur if channel sense and collision avoidance wasn't implemented,
especially in scenario's such as the Exposed Terminal problem, described in B.1.3.
1. Node A creates a RTS frame and transmits it towards Node B. The RTS frame
contains enough information to allow other nodes that have vision of Node A to
defer for the duration of the transmission.
2. Node B will receive the RTS. If it can receive data, it will create and send a CTS
frame towards node A. Otherwise it will do nothing, causing Node A to wait until
it determines no CTS is coming. Node A will then attempt to restart the process.
4. Node B will then analyse the received DATA frame. If it was correctly received, it
will reply with an ACK frame. Otherwise it will do nothing.
5. Upon receiving the ACK frame, Node A will consider the transmission a success
and wait a short interval before contending for the channel again. If no ACK frame
is received it will consider the transmission a failure and attempt to restart the
process.
Send RTS
Wait 1 Sifs
Send CTS
Wait 1 Sifs
Send DATA
Wait 1 Sifs
Send ACK
Wait 1 DIFS
Continue
control packet, as well as one CTS frame, plus one ACK frame, plus three Short Interval
Between Frames (SIFS) intervals. In SatSim this value is set to the instance of simTime
the channel should be available again, in order to reduce computational complexity.
Bits T ransmitted
T ransmitter Delay = (3.1.3)
BitRate of N ode
It is important to note that the SIFS is set to the maximum expected time for the
transceiver switching time, physical propagation delay and the mac delay. Of these,
Stellenbosch University https://scholar.sun.ac.za
the physical propagation delay can be signicantly dierent to that expected within a
terrestrial network. The virtual sense mechanism was adjusted to use this new value for
duration , reducing the amount of calculation needed by the mechanism.
The padding factor of 1.2 is used to ensure that the wait doesn't time out too early.
This means the wait can only perform its checks when a new transmission has started.
Therefore sequence numbers have been added in order to ensure that the waits are syn-
chronous.
A node will unicast a RREP back towards the original source of an RREQ, given
that it has sucient or fresh information regarding the destination. Multiple nodes can
therefore reply to a a given route discovery procedure, however only the rst received
RREP will be utilised by the original node.
RREP have additional ags that can be set, based upon the needs of the network. For
example, if unidirectional communication is required, the acknowledge ag on the RREP
can be set to indicate it. The RREP frame is shown in table 3.9, with eld sizes indicated
in bits.
Destination IP - A unique address for the destination of the route that the entry
supplies
Next Hop - The next node on the route towards the destination
Hop Count - The amount of hops from this node towards the destination
Route Active - Flag that indicates whether or not the route is currently active
Sequence Number Valid - Flag that indicates whether the sequence number is
currently valid
Precursors - List of all nodes that may be dependant on this route and need to be
informed if a link breakage occurs.
Precursors are all the other nodes that may make use of the route. These nodes need
to be informed when a link break occurs, since it changes the validity of their route entries.
It's important to note that the only information a node has to forward data towards
a specic destination, is the next node it has to forward the data to.
AODV, however the network topology and behaviour was very dierent from what is
expected in a typical satellite environment.
The Rij
d
represents a real pheromone value, indicating a cost or goodness of moving
towards node d via a hop over neighbour j from the source node i.
Another cost value could be the average number of hops, given as hdij . It's important
to note that the frame has no further control or information about what exactly will
happen at node j, only that it has recently been indicated that the value for the route is
true.
The average number of hops is calculated using a moving average, where α determines
how quickly the value adapts to new information. The equation, adapted from [19], is
given below:
d
Rij d
= γRij + (1 − γ)Rnew (3.5.2)
node regarding its neighbours. If a node receives a hello message it replies and notes that
it has a connection with the source node. A node that receives the reply will then also
list the responding node as one of its visible neighbours, before updating its pheromone
table with the information included for pheromone diusion. Typical data elds for a
hello message is given in table 3.12, with the value of k being the number of pheromone
diusion entries which are included per hello message. The data sizes of the elds are in
bits.
3.5.3 Ants
The ants of AntHocNet use pheromone tables, pheromone diusion and broadcasting in
order to create, maintain and explore new routes. As the name implies, RA are created
on demand, whenever a transmission towards a then unknown destination is scheduled.
Proactive Ant (PA) are created periodically while data is being transmitted, but may be
set to only be sent out in favourable conditions(usually when the virtual pheromone table
contains far more favourable values than the real pheromone table), in order to reduce
Stellenbosch University https://scholar.sun.ac.za
overhead. Repair Ant (RPA) are sent from nodes that attempt to repair link breaks upon
detection of a link failure.
All ants are called Forward Ants when they are moving from a source towards a sink
and Backward Ants when they are moving from a sink back towards a source. In order to
reduce overhead and allow the shortest possible path to be found, a destination will only
convert the rst Ant for a given session back towards the source while discarding the rest.
A crucial fact about ants is that they collect a list of all the nodes they have visited
on their path from their source to their destination. This means the size of the Ants, and
therefore their contribution to overhead, changes over time.
The implemented base ant class, that all ants inherit from is presented in gure 3.2
along with the details of the other ants.
Reactive_Ant
+broadcast: boolean = True
+app_id: String = Reactive_Ant
+unique_id: String
+β1
+return_copy()
+get_new_path(path:list)
+get_new_cost(cost:list)
Ant
+c_src: SimObject
+p_src: SimObject
+o_src: SimObject
+n_dest: SimObject
+f_dest: SimObject
+broadcast: boolean
+type: integer = 0 Proactive_Ant
+ptr: integer
+state: integer +β2 : integer
+path: list
+ +app_id: Proactive_Ant
+calculate_size()
+set_current_source()
+set_next_destination()
+post_transmission(cost)
+update_forwards(node:SimObject)
+update_backwards(node:SimObject)
+update_destination(node:SimObject)
+display_path()
Repair_Ant
+max_broadcasts: integer
+app_id: String = Repair_Ant
+unique_id: String
+α: oat
+get_new_path(path:list)
Figure 3.2: Ant Class Diagram: Showing Base Ant Class and Derived Ants
Stellenbosch University https://scholar.sun.ac.za
Pin
d
Probability of using n as the next hop towards d from current node i
Rin
d
Pheromone value associated with the path to d via n from i
β1 Route discovery variable - high implies the best routes are more likely to be used
Nid Set of all visible neighbours, with 1 for visible and 0 for no LOS
Pnd - Probability of using n as the next hop towards d from current node i
Rin
d
- Pheromone value associated with the path to d via n from i
Nid - Set of all visible neighbours, with 1 for visible and 0 for no LOS
Pin
d
- Probability of using n as the next hop towards d from current node i
Rin
d
- Pheromone value associated with the path to d via n from i
Nid - Set of all visible neighbours, with 1 for visible and 0 for no LOS
Chapter 4
Expansion and Testing of SatSim
50
Stellenbosch University https://scholar.sun.ac.za
4.2 Simulator
The simulator is the top object that controls the entire simulation. It contains a SimPy
environment and a list of simObjects which have various interactions with each other.
The simulator is created at the start of any simulation and schedules events to occur
within the simulation.
The simulator also contains a large number of utility functions that can be used by
the protocol implementations. The following section details the components used by the
simulator in order to eciently run a simulation.
List
1 Condition
Table 1 1
Legend 1
New 1 Wizards 1
Minor Changes 1
Major Changes
1
DataScheduler 1 1 Simulator 1 1 DataManager
1
n
Receivers n 1 SimObject 1 n Transmitters
1
1
1
Network Stack 1 Propagation Model
OBC
wizards make use of the deterministic nature of the satellite orbits and the lack of random
propagation in their orbits.
4.2.4 ResultManager
The Result Manager is used to gather overall simulation results, rather than summarised
results on a link by link basis. The Result Manager will therefore keep track of metrics
such as total generated overhead and average route discovery times.
4.3 SimObjects
SimObjects act as the principle actors in the SatSim environment. In order to facilitate
network capabilities, the SimObject had to be expanded to include an implementation of
the OSI model and be capable of storing state information. This information is stored in
the OBC, which has several subclasses including memory, strategy and a data buer.
OSI model
Application
Session Upper
Transport Transport
Network Network
Physical Physical
then determined for the link at that given instant using a link budget equation and used
to randomly assign bit errors to the packet.
When a packet is broadcast, such a link pair is created between the source nodes
transmitter and every receiver that can receive the packet. To achieve this, the simulation
can either make use of dened xed links, as was described by [7] with some variable
links or make use of a completely dynamic topology, that is far more computationally
intensive. The issue of constant recalculations could be mitigated using a central visibility
table, that provision has been made for in the Simulator. This table should contain the
visibility information for the network at a given time, acting as an eective snapshot of
the network which would be especially useful for the calculations regarding collisions and
channel sensing.
4.3.3.1 Environment
The Environment class is used to control data transmissions and to keep track of which
nodes are transmitting data and which nodes are receiving data. When centralised routing
or specic quasi-network congurations are used the environment also denes the ISL and
other links. Two examples included in the research are the constellation implementation
given in [7] for both AODV and AntHocNet.
When a frame is unicast, the environment performs some preliminary calculations and
updates its state variables before handing the frame to the RFChannel class, establishing
a link.
When a frame is broadcast, it is used as a base case. All the visible objects are
detected, an a copy of the frame is handed to the RFChannel for each one. This has the
Stellenbosch University https://scholar.sun.ac.za
same eect as an actual broadcast, with each transmission being handled uniquely and
the frames diering (at that instant) only in some meta-information.
4.3.3.2 RFChannel
The RFChannel represents an instantaneous link between two nodes and all the func-
tionality necessary to simulate the link. The RFChannel class can also be congured in
various ways in order to determine how links are simulated, such as using dynamic or
static BER calculation.
By default the BER calculation is set to dynamic, however it can be set to some static
value or supplied with a constant oset to simulate worst case scenario conditions. The
RFChannel can have multiple features active which can make a simulation more realistic
at the expense of additional computations.
correct order at the destination. Most WSN use User Datagram Protocol (UDP) for their
transport layer, although several custom transport layers are provided with the current
implementation of SatSim.
4.3.8 OBC
Each SimObject includes an OBC model which consists of two discrete classes: strategy
and memory. The OBC can also be shut down or rebooted to simulate network failure
and to analyse the eects of a node suddenly being dead to the network, as often can
happen within satellite networks. The Unied Modelling Language (UML) class diagram
for the OBC is given in gure 4.3. Each protocol will require its own specic strategy and
memory class.
On Board Computer
+node: SimObject
+status: boolean 1
1 +strategy: Strategy
+memory: Memory
+initStrategy(strategy:string)
+initMemory(memory:string)
1 1
Protocol Specic Memory Protocol Specic Strategy
4.3.8.3 Parameters
The parameters le stores the protocol specic variables which are used to congure a
routing protocol. These parameters are stored in a single class for ease of use and to allow
them to be adjusted as needed and the corresponding changes in protocol performance
measured. These parameters should always be congured with default values and allow
the user to set them as needed. If the eect of these parameters are being investigated
they should be set at the beginning of a given simulation, otherwise default values will be
used.
4.3.9 Frames
The frames section contain descriptions of the frame or packets used by the various com-
munications protocols. Each frame consists of its actual data, in other words the data
the real frame will consist of as well as its meta data. The meta data contains informa-
tion needed by the simulation itself and will not count towards the size of the frame in
calculations.
bit-by-bit basis, which would have allowed some specic errors to be simulated but have
a drastic eect on execution time.
Most frames will be protocol specic and the constructors will have to be custom
written for each one, however a wide variety of general purpose frames have been provided,
including all the frames needed for AntHocNet, MAC 802.11 and AODV.
Each frame also contains a type parameter, dened in table 4.1 that is used by the
data link layer for dierentiation.
Each frame will also contain meta-information shown in table 4.2 to aid debugging,
allow functionality for certain protocols, and for determining the results of a given simu-
lation. These elds do not contribute to the eective size of the frame, unless the protocol
explicitly uses it.
Each frame has to contain a signature(for dierentiation) and a size(for error checking
and the calculation of processing and delay).
Any frame which is broadcast needs its own copy function, since the direct copy class
cannot be used inside a generator.
AntHocNet Frames
BaseFrame
+size: integer
+sig: String
+o_src: SimObject
+p_src: SimObject
+c_src: SimObject
+broadcast: boolean
+type: integer
+f_dest: SimObject
+n_dest: SimObject
+mac_attempts: integer
+set_current_src(SimObject:SimObject)
+post_transmission()
+set_next_destination(node:SimObject)
+set_nal_destination(node:SimObject)
results. Tests were only concluded to be successful if their results were close enough
to the analytical values. The following functionality is therefore critical towards a well
implemented MAC layer:
1. Nodes will physically sense the channel before transmitting, and defer or transmit
accordingly (physical sense). If a node has visibility of a transmitting node, it senses
the channel as busy.
2. Nodes will gather NAV data and defer or transmit accordingly (virtual sense). Nodes
will gather NAV data from the appropriate headers and then defer the channel usage
until such a time has elapsed.
3. Nodes will, given a large enough frame to transmit, rst attempt and nish a hand-
shake procedure before transmission. Retries as needed (handshake)
Stellenbosch University https://scholar.sun.ac.za
4. Nodes will wait the appropriate amount of time, including inter frame spaces and
back o interval, before attempting to utilize the channel.
5. Nodes will defer the channel with the correct periods of times, and should happen
fairly - no channel domination by a single node (back o, fairness)
Even a small scale simulation triggers a very large amount of (fortunately short) events
within this context. Dierent issues will also arise based on dierent topologies (see the
Hidden Node Problem and Exposed Node Problem). In order to ensure that the correct
behaviour occurs as expected, multiple testing scenarios were created and tested with
special attention given to the worst case scenarios.
These scenario's made use of Satellites in perpendicular orbits whom communicate
with ISLs links.
1. Route Discovery - Ensure that only one path is active at a given time for a given
logical connection.
2. Route Recovery - Ensure that data will be buered correctly and routes will be
searched for whenever possible.
3. Route Repair - Ensure that routing entries are correctly updated upon a link
break and the appropriate action taken.
In order to test these conditions, scenarios were created where the results can be easily
deducted via analysis and conrmed via test scripts. This will also serve as a test of the
ability of SatSim to evaluate the routing metrics, since they can also be easily calculated
and compared to what the simulators generates.
Short Link
Sink
Source
Long Link
The results given by the simulation were as expected and key values were analytically
veried with test scripts. The path was chosen over the short link, as expected and all
data was transmitted via that link.
Stellenbosch University https://scholar.sun.ac.za
Short Link
Sink
Resets
Sources
Long Link
Once the node resets has occurred, the protocol was veried to follow the appropriate
route repair process. The old routing data expires and a new route request is generated,
this time with the only remaining, less ecient route. Data transmissions were then
veried to continue normally.
Route Discovery - It needs to be veried that real pheromone routes are created
upon route requests, and reactive ants are correctly broadcast and unicast back
towards their source.
Route Repair - Upon detection of a link break repair ants should be created and
the route should be repaired if possible.
Balanced Data Forwarding - Given multiple paths for data to travel, the path
with better pheromone values should have the majority of the data being transmit-
ted via it.
Source Sink
In gure 4.7, the red arrows indicate the path we expect data to take, with the shorter
paths being favoured. Since the top path is shorter, we expect the majority of the data
to travel via that path. Furthermore it is expected that the Source will have virtual
pheromone values for every node in the system, but only real pheromone values for its
neighbours and the sink.
The output of the test was analysed, and the results were as expected. The pheromone
diusion process led to the nodes being aware of all the other nodes in the network. The
source node had real pheromone for the routes with its neighbours, that it is aware of due
to the hello messages. A path is created towards the sink over the shortest path and data
is transmitted as expected.
Source Sink
Node Resets
Stellenbosch University https://scholar.sun.ac.za
The same behaviour as test 1 is seen, up until the node resets. The resulting link breaks
cause the expected reactions of link failure notications and a route repair process. The
middle node learns that the route is broken and send out repair ants to establish a new
route. Once this occur data keeps being forwarded along that path. The source node,
is informed of the broken link but due to the link repair has to take no further action
itself. The resulting pheromone tables correctly represent the new network topology at
simulation end.
Stellenbosch University https://scholar.sun.ac.za
Chapter 5
Protocol Evaluation
This chapter describes the congurations and results for a variety of dierent network
scenarios that were simulated on SatSim.
The rst simulation describes an Iridium like network, using a large number of satellites
to enable communication between two ground stations. This simulation was used for
verication of the simulator as well as for testing its operation under very high data
loads. However, it made no use of a MAC layer and raised some issues regarding the
functionality of AntHocNet in a quasi-static network.
The second scenario discussed is a baseline case, involving a single satellite making
passes over a single ground station. The data gathered will allow users to compare the
benets of a satellite network to the cost of launching and maintaining them.
The third simulation simulates the OuterNet in order to evaluate the performance of
AntHocNet compared to AODV for a unique network. It is important to note that neither
of these routing protocols were intended for a scenario such as the OuterNet, implying
far more optimizations are possible.
67
Stellenbosch University https://scholar.sun.ac.za
also aected by Doppler Shift, which leads to serious link breaks which occur whenever
the satellites go above 75 degrees or below -75 degrees.
The orbital elements for the distributed network satellites are given in table 5.1.
The end-users of the network are two ground stations, whose locations are given in
table 5.2. The rst ground station will send a ood of packets towards the second for the
duration of the simulation.
Stellenbosch University https://scholar.sun.ac.za
The MAC layer was not tested for this constellation since it was not used in the
original study and there is a very low likelihood of collisions due to the nature of the inter
satellite-links and the ground stations only pointing towards one satellite at a time.
Another important aspect of this scenario is the large network size. The simulator did
struggle to an extent with the memory consumption of the simulation, indicating that
some programming optimizations may be benecial to the implementations of AODV and
AntHocNet on SatSim.
5.1.5 Results
The following subsection describes the results of the above simulation as well as giving
an analysis of why the results occurred as they did. These results also appear in [38].
upon route discovery and RREQ, RREP and RERR upon a link break. As a Hybrid
Protocol AntHocNet will send Reactive Ants upon route discovery and Proactive Ants
while data is being transmitted. Repair Ants and Link Failure Notications will be sent
after a link break and Hello Messages will be periodically sent.
Table 5.5 compare the SatSim results to the Liu et al results. As can be seen the
results for the amount of RREQ compare very favourably, although the amount of RERR
and RREP generated dier signicantly. It is assumed that this is due to a dierent
method of counting the overhead frames, although it is also possible that there were some
dierences in the congurations of the networks, since some conguration settings had to
be extrapolated or approximated from [7].
The overhead generated by the AntHocNet implementation is given by 5.6 for the
reactive elements of AntHocNet and by 5.7 for the proactive elements of AntHocNet. The
overhead generated by AntHocNet is far more signicant than that generated by AODV,
eectively congesting the network unless specic changes were made to the protocol.
These are detailed in the analysis of the results.
an alternative method of handling link breaks was implemented, which cleared fully and
reacted better to link breaks in a quasi-static network. Essentially for this alternative
implementation of AntHocNet, the pheromone entries were manually cleared upon de-
tection of a link break, allowing the network to rapidly rebuild itself without the use of
excessive link failure notications or unicast warning messages. This implementation of
AntHocNet performed slightly better than SatSim in terms of latency and route discovery
time, however it is only applicable to this specic network.
Access T ime
Satellite Access Ratio = (5.2.1)
Orbit T ime
Stellenbosch University https://scholar.sun.ac.za
Using the above equation we get an access ratio of roughly 14 percent, as we only have
communication for 13 and a half minutes of the 1.62 hours orbit.
The results obtained for the baseline case are summarised in 5.9, which shows slightly
better results for AntHocNet.
OuterNet. As [7] suggested, a custom routing protocol can have vastly superior results
compared to a general case routing protocol and considering the resources used to get
a satellite network operational, the additional time to choose the best routing protocol
during mission planning phase is justied.
The OuterNet satellite will not deviate from a path more or directly above the equator,
which limits the viable ground stations for such a network. [1] suggests the following
potential ground stations, described by 5.11.
For this research the European Space Centre was used as the ground station and
eective sink for the data.
OuterNet Relay
5.3.3 Results
For this section of the research, it is of interest to see how easily the OuterNet can be
implemented on the SatSim environment and if it provides a good platform to analyse
the performance dierences for AODV and AntHocNet for the OuterNet.
Table 5.12 shows the results for a simulation of the OuterNet for a single client satellite.
When compared with table 5.9 it can be seen that the total throughput for the simulation
decreased. The reasons therefore were analysed using SatSim and are given in the next
subsection.
Table 5.12: Baseline Client connected to OuterNet for AODV and AntHocNet
Protocol Data Frames Transmitted
AODV 2480
AntHocNet 4408
This issue does not occur with AntHocNet. Data is stochastically forwarded when
multiple paths are available, allowing the better paths to be utilised. Multipath algorithms
therefore seem to perform better in an OuterNet type scenario.
However, both protocols suer due to path discovery time being high relative to the
size of the communication window, and extremely long retransmission times due to the
adjusted MAC layer. Collisions therefore have an extremely signicant eect on this
simulation, and avoiding channel contention and collision via specic schemes such as
special band for ISLs should be highly benecial.
Chapter 6
Conclusion and Recommendations for
Future Work
6.1 Conclusion
The three objectives for this research were the expansion of SatSim as a full satellite
network simulator, the analysis and implementation of two complex distributed MANET
routing protocols and the evaluation of those two routing protocols in the same network
scenario.
77
Stellenbosch University https://scholar.sun.ac.za
still additional functionality that can be implemented as well as modications which can
be added for space-based applications.
6.2.1.1 Antennas
SatSim currently has a functional antenna system but this can be improved to allow
for beam shaping and more complex antenna patterns to be simulated. The current
assumptions of omnidirectional antennas can also be replaced by modelled pointing, as
well as more accurately simulating switching times. A system which read in and implement
antenna patterns can also be highly benecial.
receiving. Each transmitter and receiver would then have its own instance of a MAC
layer, allowing communication to occur in dierent bands.
6.2.2.1 ZRP
ZRP can be useful to work with some of the quasi-static satellite constellations that can
be achieved by correctly choosing orbits. ZRP will be well suited to these networks by
correctly choosing the zone areas.
6.2.3 Investigations
The two investigations detailed below do not directly tie in to the research performed in
this thesis, but act rather as suggested uses for the SatSim platform.
Stellenbosch University https://scholar.sun.ac.za
Appendices
81
Stellenbosch University https://scholar.sun.ac.za
Appendix A
SatSim Supplementary
The following Appendix Cover the supplementary information for the installation and
use of SatSim. It also includes additional diagrams and information which will allow the
user to easily congure SatSim for unique simulations and add their own protocols and
additional funcionality.
A.1 Installation
In order to allow easy cross platform use, SatSim makes use of custom functions wherever
possible and only uses external libraries where absolutely needed. The following appendix
details which packages have to be installed in order to use SatSim and the reasoning behind
using them.
The core language SatSim is written in is Python, due to it being highly reliable. Most
of the computationally expensive functions or high loop functions have been executed
using C based procedures in NumPy or PyEphem.
A.1.1 PyEphem
PyEphem is used to calculate the positions of the satellites. The PropogationModels class
makes use of this module to determine the position of the simulation nodes as needed.
The use of the module was compared to the freely available SGP4 [39] module in 2015,
but while similar results were generated PyEphem had far better execution times. Upon
analysis it was conrmed this was due to SGP4 being written mostly in Python while
PyEphem made extensive use of C.
A.1.2 NumPy
NumPy was extensively used due to it's popularity and fast execution time. All matrices
in SatSim are NumPy matrices, optimized according to their likely or expected data type.
Furthermore all NumPy matrices and their operations are highly optimized and should
be fairly familiar to any user who often uses them in Python.
A.1.3 SimPy
SimPy as used for the discrete event driven simulation. This was due to it's use being
established in previous versions of SatSim and it's extensive resources. The expansion
82
Stellenbosch University https://scholar.sun.ac.za
of SatSim to allow networks also created the need for far more timers and events, which
SimPy was capable of easily facilitating.
A.1.4 cPickle
Used for the storage of the large amount of output data generated by the SatSim envi-
ronment as well as the storage of network tables. cPickle is used since the stored data
will be either single values or matrices rather than complex objects.
A.2.4 Protocols
The following protocols have been implemented in the SatSim environment:
AntHocNet
AODV
Appendix B
Routing Supplementary Information
This section briey describes some of the issues inherent to routing in MANETS. These
issues frequently occur while implementing routing protocols for MANETS and while most
strategies cannot fully solve these problems but some forms of mitigation are possible.
85
Stellenbosch University https://scholar.sun.ac.za
Collision Occurs!
Transmit Transmit
LOS RADIUS
LOS
Wants to Transmit
Transmitting
Glossary
AppID A unique identier which groups data packets as belong to a specic stream. 58
backo A set amount of time a node waits before reattempting some failed transmission
or to use the channel again. 57, 78
broadcast A Data transmission from a node which is sent to every node which has vision
of it. Can lead to ooding. 16, 20, 25, 38, 41, 44, 5456
collision Occurs when a node receives data from multiple sources at once, resulting in
the relevant packets being dropped. 31
DataManager The SatSim class which is used for the storing and analysing simulation
results.. 9
Hello Message A short message sent to neighbouring nodes, usually to determine which
nodes are currently visible from the node. 16, 17, 20
HEO A High Earth Orbit(HEO) satellite has an orbit above 35786 kilometres. 15
Hybrid Protocol A protocol which makes use of both proactive and reactive elements
in its functionality. 42, 71
Jitter Jitter is the dierence in time between the arrival of packets. Also called inter-
packet delay.. 10
LEO A Low Earth Orbit(LEO) satellite is a satellite whose orbit is lower than 2000
kilometres. 15, 23
Lifetime Refers to the expected amount of time for which routing data is valid. 37
MEO A Medium Earth Orbit(MEO) Satellite stays in orbit in an altitude between 2000
and 35786 kilometres. 15
Neighbour A node which currently has line of sight with the current node. 13, 20, 44
OuterNet A proposed ring of equatorial satellites which will act as an eective relay
network for sun synchronous satellites by providing additional access windows.. 2,
6, 27, 67, 7276, 78, 79
88
Stellenbosch University https://scholar.sun.ac.za
GLOSSARY 89
precursor In AODV, refers to all the nodes which may use or be dependant of a specic
routing entry. 40
Reactive Protocol A protocol which only updates routing information (and therefore
generates overhead when explicitly needed for transmissions. Often called on-
demand.. 37, 70
RFChannel The class in SatSim which is used to simulated communication links be-
tween two nodes. 56
Route Discovery The process a node undertakes to nd a routing path towards a spe-
cic destination. 19
Routing Freedom A term which refers to a routing protocols ability to prevent mes-
sages from looping indenitely in the network. 21
SatSim A Satellite Simulator developed at Stellenbosch University. 38, 11, 1619, 21,
2629, 31, 33, 34, 37, 42, 43, 50, 51, 53, 54, 5659, 63, 67, 6972, 7580, 82, 83
SimObject A node in the satellite environment which will either be a Ground Station
or a Satellite. 53, 58, 59
simTime Refers to the amount of time(usually in seconds) which has elapsed in the the
simulation. 34, 53, 59
slotTime The size of the intervals the MAC layer waits in before sensing the channel
again. 31, 32, 36, 62
switching times Time it takes for a node to go from receiving mode to transmit mode.
78
Throughput The Rate at which data is send from an source node to a destination node.
9
unicast A Data transmission from a node which is sent to only a single visible node. 39
Stellenbosch University https://scholar.sun.ac.za
Bibliography
[1] Kearney, M., Botma, P., Jordaan, H., Gerber, J., Thesnaar, E., Nolte, F., Erlank, A.,
Groenewald, C. and Barnard, A.: The outernet: A novel satellite communication relay
constellation. 2013.
[2] Al, D.M.E.: Qb50 is an european fp7 project for facilitating access to space. 2016.
Available at: https://www.qb50.eu/index.php
[5] Le Roux, J.-H.: Development of a satellite network simulator tool and simulation of AX.
25, FX. 25 and a hybrid protocol for nano-satellite communications. Master's thesis, Stel-
lenbosch: Stellenbosch University, 2014.
[6] Bezuidenhout, Q.: Satellite Communications Strategy selection for optimal LEO satellite
communication. Master's thesis, Stellenbosch: Stellenbosch University, 2012.
[7] Liu, X., Gokhale, D., Mizas, T. and Springston, P.: Analysis of ip routing approaches for
leo/meo satellite networks. 2010. Available at http://enu.kz/repository/2010/AIAA-2010-
8685.pdf , 28th AIAA International Communications Satellite Systems Conference , 2010.
[8] Corson, S. and Macker, J.: Mobile ad hoc networking (manet): Routing protocol perfor-
mance issues and evaluation considerations. Tech. Rep., University of Maryland , Naval
Research Laboratory, 1998.
[9] Johnson, D.B. and Maltz, D.A.: Dynamic source routing in ad hoc wireless networks. In:
Mobile computing,pp. 153181. Springer, 1996.
[10] Johansson, P., Larsson, T., Hedman, N., Mielczarek, B. and Degermark, M.: Scenario-based
performance analysis of routing protocols for mobile ad-hoc networks. In: Proceedings of
the 5th annual ACM/IEEE international conference on Mobile computing and networking,
pp. 195206. ACM, 1999.
[11] Hoebeke, J., Moerman, I., Dhoedt, B. and Demeester, P.: An overview of mobile ad hoc
networks: applications and challenges. Journal-Communications Network, vol. 3, no. 3, pp.
6066, 2004.
[13] Chen, C. and Ekici, E.: A routing protocol for hierarchical leo/meo satellite ip networks.
Wireless Networks, vol. 11, no. 4, pp. 507521, 2005.
90
Stellenbosch University https://scholar.sun.ac.za
BIBLIOGRAPHY 91
[14] Lee, J. and Kang, S.: Satellite over satellite (sos) network: A novel architecture for satellite
network. In: INFOCOM 2000. Nineteenth Annual Joint Conference of the IEEE Computer
and Communications Societies. Proceedings. IEEE, vol. 1, pp. 315321. IEEE, 2000.
[15] Boukerche, A., Turgut, B., Aydin, N., Ahmad, M.Z., Bölöni, L. and Turgut, D.: Routing
protocols in ad hoc networks: A survey. Computer networks, vol. 55, no. 13, pp. 30323080,
2011.
[16] Marina, M.K. and Das, S.R.: On-demand multipath distance vector routing in ad hoc
networks. In: Network Protocols, 2001. Ninth International Conference on, pp. 1423.
IEEE, 2001.
[17] Perkins, C.E. and Bhagwat, P.: Highly dynamic destination-sequenced distance-vector rout-
ing (dsdv) for mobile computers. In: ACM SIGCOMM computer communication review,
vol. 24, pp. 234244. ACM, 1994.
[18] Rajagopalan, S. and Shen, C.-C.: Ansi: A swarm intelligence-based unicast routing protocol
for hybrid ad hoc networks. Journal of Systems Architecture, vol. 52, no. 8, pp. 485504,
2006.
[19] Ducatelle, F.: Adaptive routing in ad hoc wireless multi-hop networks. Ph.D. thesis, Uni-
versità della Svizzera italiana, 2007.
[20] Perkins, C., Belding-Royer, E. and Das, S.: Ad hoc on-demand distance vector (aodv)
routing. Tech. Rep., University of California , University of Cincinatti, 2003.
[21] Johnson, D.B., Maltz, D.A., Broch, J. et al.: Dsr: The dynamic source routing protocol for
multi-hop wireless ad hoc networks. Ad hoc networking, vol. 5, pp. 139172, 2001.
[22] Clausen, T. and Jacquet, P.: Optimized link state routing protocol (olsr). Tech. Rep.,
Network Working Group, 2003.
[23] Haas, Z.J., Pearlman, M.R. and Samar, P.: The zone routing protocol (zrp) for ad hoc
networks. Not Specied, 2002.
[25] Inc., I.C.: Iridium next:the bold future of satellite communications. 2016.
Available at: https://www.iridium.com/network/iridiumnext
[28] Orbcomm: Orbcomm is a leading global provider of machine-to-machine (m2m) and internet
of things (iot) communication solutions that remotely track, monitor, and control xed and
mobile assets. . . . 2016.
Available at: http://www.orbcomm.com/
[29] Rhodes, B.: Pyephem: Basic astronomical computations for python. 2016.
Available at: http://rhodesmill.org/pyephem/
[30] Group, O.: Outernet: Radio for the information age. 2016.
Available at: https://outernet.is/
Stellenbosch University https://scholar.sun.ac.za
BIBLIOGRAPHY 92
[31] Schildt, H.: Java the Complete Reference 7E. McGraw-Hill, 2007.
[32] Marszalek, M., Rummelhagen, M. and Schramm, F.: Potentials and limitations of ieee
802.11 for satellite swarms. In: 2014 IEEE Aerospace Conference, pp. 19. IEEE, 2014.
[33] Ambartani, P.R., Patil, P. and Devi, U.: Analysis and simulation of networking protocols
in constellation ying system for inter-satellite communication.
[34] Society, I.C.: Part 11: Wireless lan medium access control (mac) and physical layer (phy)
specications. 2012.
Available at: http://standards.ieee.org/about/get/802/802.11.html
[35] Van Glabbeek, R., Höfner, P., Tan, W.L. and Portmann, M.: Sequence numbers do not
guarantee loop freedom: Aodv can yield routing loops. In: Proceedings of the 16th ACM
international conference on Modeling, analysis & simulation of wireless and mobile systems,
pp. 91100. ACM, 2013.
[36] Di Caro, G., Ducatelle, F. and Gambardella, L.M.: Anthocnet: an adaptive nature-inspired
algorithm for routing in mobile ad hoc networks. European Transactions on Telecommuni-
cations, vol. 16, no. 5, pp. 443455, 2005.
[37] Zimmermann, H.: Osi reference model-the iso model of architecture for open systems inter-
connection. IEEE Transactions on communications, vol. 28, no. 4, pp. 425432, 1980.
[38] AJ, M. and A, B.: Simulating manets:a study using satellites with aodv and anthocnet,
2016. Submitted.
[39] et al, V.: Track earth satellite tle orbits using up-to-date 2010 version of sgp4. 2010.
Available at: https://pypi.python.org/pypi/sgp4/