Mô hình đặc tính kênh truyền cho thông tin thủy âm vùng nước nông

INTRODUCTION 1. Overview of the Dissertation Underwater acoustic (UWA) communication systems have been devel- oped for the past three decades [25]. They can be used in potential applications such as environmental monitoring, oshore oil exploration, and military missions. Nevertheless, UWA communications have a plethora of diculties, so they display many challenges for further developments. The reason can be explained by a large demand on high frequency utilization as well as high data rate access under very complexity shallow underwater environments. All these requirements, without doubt, call for intensive research eorts on how to cope with problems faced by current UWA communications, e.g., limited availability of acoustic frequency spectrum, complex time variations in UWA fading channels, and urgent needs for good quality of service. Therefore, this dissertation is devoted to investigate UWA communication systems by considering all these challenges. In particular, two goals are aimed at, which are known as: i) UWA channel modeling and ii) performance analysis of UWA communication systems The design, development, performance analysis, and test of such communication systems, however, call for a deep insight of the most important characteristics of real-world propagation environments. Similar to the other communication fashions, channel modeling is an initial investigation because it provides hints to predict performance of communication systems before doing further high cost implementations as hardware designs [75, 96]. The task of channel modeling is to reproduce the real channel conditions. In other words, the statistical properties of the real channel such as path loss, multipath fading and Doppler eect should be represented by channel modeling. For this reason, this dissertation presents the analysis and modeling UWA channels in shallow water en-

HANOI UNIVERSITY OF SCIENCE AND TECHNOLOGY

1 Overview of the Dissertation

Underwater acoustic (UWA) communication systems have been oped for the past three decades [25] They can be used in potential appli-cations such as environmental monitoring, offshore oil exploration, andmilitary missions Nevertheless, UWA communications have a plethora

devel-of difficulties, so they display many challenges for further developments.The reason can be explained by a large demand on high frequency uti-lization as well as high data rate access under very complexity shallowunderwater environments All these requirements, without doubt, callfor intensive research efforts on how to cope with problems faced bycurrent UWA communications, e.g., limited availability of acoustic fre-quency spectrum, complex time variations in UWA fading channels, andurgent needs for good quality of service Therefore, this dissertation isdevoted to investigate UWA communication systems by considering allthese challenges In particular, two goals are aimed at, which are knownas: i) UWA channel modeling and ii) performance analysis of UWAcommunication systems

The design, development, performance analysis, and test of such munication systems, however, call for a deep insight of the most impor-tant characteristics of real-world propagation environments Similar tothe other communication fashions, channel modeling is an initial inves-tigation because it provides hints to predict performance of communica-tion systems before doing further high cost implementations as hardwaredesigns [75, 96] The task of channel modeling is to reproduce the realchannel conditions In other words, the statistical properties of the realchannel such as path loss, multipath fading and Doppler effect should

com-be represented by channel modeling For this reason, this dissertationpresents the analysis and modeling UWA channels in shallow water en-

1

vironments, which have strong multipath and Doppler effects on signalpropagations [97].

Without discussing the performance of UWA communication systemsunder different propagation environments, this study seems to be unfin-ished In this view point, Orthogonal Frequency Division Multiplexing(OFDM) has been widely applied to acoustic transmission [10, 22, 42, 68,86] since it can mitigate inter-symbol interference as well as has higherspectral efficiency than single carrier systems Thus, for the sake ofcompleteness, we utilize analyses such as the signal-to-interference ratio(SIR), the signal-to-interference-plus-noise ratio (SINR), and the chan-nel capacity to determine the performance of the UWA-OFDM systems

We believe that the performance assessment reported here bridges thegap between the derived UWA channel models and their impact on theperformance of the deployed UWA communication systems

This chapter presents the general concepts in UWA channel ing and a brief introductions to UWA channel characteristics Moreover,the motivations and the major contributions of this dissertation are high-lighted in the remainder of this chapter

model-2 Characteristics of Shallow Underwater Acoustic ChannelsThe physical characteristics of UWA propagation environments are verydifferent from those of terrestrial ones with electromagnetic waves UWAchannels can be characterized by three main aspects [12, 30, 88]: the hightransmission loss depending on signal frequencies, the time-varying mul-tipath propagation, and the low transmit speed of sound in water (about1500m/s) The fast time variations of UWA channels are mainly caused

by the relative movement [88], internal waves [32], and surface waves[16, 82] These features make UWA channels the most difficult commu-nication media in use today [88], and give rise to critical challenges forfurther developments

Acoustic Frequency

The frequency of underwater acoustics is in the range from 10 Hz to

1 MHz [92] When the bandwidth is between 10÷20 percent of the center

of signal, the communication system is called wide-band Although the

bandwidth of UWA communication systems is small, the signal frequency

is also small Thus, UWA communication systems are wide-band due tothe low relative center frequency in comparison with the bandwidth [88].Transmission Loss

The transmission loss of UWA propagation significantly depends onthe signal frequency The three factors that attenuate UWA signalsinclude spreading loss, absorption loss, and scattering loss The overallpath loss A (l, f ) is defined as [88]

in reference to some lr The symbol k is the path loss exponent, whichmodel the spreading loss and its value are usually between 1 and 2.The absorption coefficient α(f ), which increases rapidly with signal fre-quency, can be obtained using an empirical formula [20]

Noise

Noise in UWA channels consists of ambient noise and site-specificnoise Ambient noise is always present in the background of the sea,while site-specific noise is unique to certain places The first one isoften modeled as Gaussian and it is not white, while the latter one con-tains significant non-Gaussian components The power spectral density

of ambient noise decays at a rate of approximately 18 dB/decade Theattenuation growing with frequency whereas the noise decays with fre-quency, result in a signal-to-noise ratio (SNR) that varies over the signalbandwidth [88]

Propagation Delay

The speed of UWA waves increases with the salinity, temperature, andpressure of the water In shallow water environments, the temperatureand pressure are almost unchanging; thus, the speed of sound in shal-low water is considered to be a constant value (about 1500 m/s) Thepropagation delay τ can then be obtained as

τ = dcwhere d and c are the propagation distance (in meters) and the speed

of sound (in m/s), respectively Because of the low speed of sound, thepropagation delay τ = d/c is about tens milliseconds for transmissiondistances of longer than ten meters.

Multipath

In shallow water environments, the propagation of sound appears to

be a complicated multipath, which is mainly caused by reflections at thesurface and bottom The multipath interference in UWA communicationsystems is illustrated in Fig 1 Each path has its propagation delaydepending on its geometry The maximal propagation delay is called

as the delay spread of the UWA channel Because of the multipatheffect, the received signal is composed of various paths with differentamplitudes, propagation delays, and phase shifts

Figure 1: Multipath interference in UWA communication systems.

Doppler Effect

Another characteristic of UWA channels is time varying, which iscaused by two factors The first one is the result of the relative movementbetween the transmitter (Tx) and the receiver (Rx), while the latter one

is caused by inherent changes in the transmission medium such as thechanges in weather, surface wave, and storm, etc [9]

A relative movement between Tx/Rx or a moving medium results inthe change of frequency of the acoustic waves, which is called as Dopplershift An expression for the maximum Doppler frequency shift fD,max isgiven by [19]

fD,max = fcv

c,

where fc and c are the transmitted signal frequency (i.e carrier quency) and the sound speed, respectively; v stands for the speed of theobserver.

fre-The magnitude of Doppler effect is determined by the ratio a = v/cnamed as the relative Doppler shift, which is significant to the carrierfrequency due to the low speed of sound The non-negligible Dopplershift is a distinctive characteristic of UWA channels in comparison withthe radio channel

Moreover, even without intentional movements, the inherent changes

in transmission medium such as waves or drifts of transducers also lead

to the Doppler shift In shallow water environments, reflections from thesurface are the main reason of time-variant UWA channels The Dopplerspread presents the spectral width spreading of the received signal, whichdepends on the wave height, wind speed, reflections from the surface andbottom of the sea

3 UWA Channel Modeling Approaches and the State-of-the-Art

The characteristics of UWA channels are very complex due to Dopplereffects, high attenuations depending on signal frequencies, multipath ef-fects, and additive color noises Therefore, it is very difficult to modelexactly UWA channels, especially in shallow underwater environments,which have strong multipath effects on the signal UWA channel model-ing is not new research in underwater communication systems However,over the past few decades, although large variety of UWA channel modelshave been proposed, there is still no typical model that can be applied forall UWA channels because of differences in geographical areas, weatherconditions, and seasonal cycles [24, 70, 73, 88, 93, 96] Recent approaches

of designing UWA simulators in literatures are classified into two maincategories, which are the geometry and the measurement-based

The UWA geometry-based simulator has been designed by using thegeometrical channel model The well-known Bellhop code [69] is one ofpopular examples of this simulator The code built the UWA channelsimulator by using the ray theory for a given geometry, but it did not

consider the random channel variation [75] To deal with this issue,some studies run the Bellhop model in combination with environmentconditions, such as temperature and salinity [89], wind speeds [28], andsurface shapes [37] The simulated UWA channels obtained throughsuch Bellhop channel simulator showed the statistical properties thatare similar to those of the real UWA channels in some experimentalscenarios The difficulty of specifying the environment conditions is one

of the limitations of this simulator

Another kind of the UWA geometry-based simulator is developed bycombining the ray theory with statistical methods to describe the UWApropagation environment [13, 17, 27, 55, 73, 75, 103, 104] The statis-tical properties of the UWA channel were analyzed by using the prob-ability density function (PDF) of the angle-of-arrival (AOA), and theangle-of-departure (AOD) as key parameters The AOD is, however,

a derivative parameter of the AOA [55] In some research studies, thePDFs of the AOAs are assumed to be normally [103, 104] or uniformly[75] distributed Besides, in [13], the author approximated the PDF ofAOA with the half-circular Rice PDF The geometry-based simulatorcan describe the overall UWA channel with fewer estimated channel pa-rameters than the measurement-base one, and it is feasible to extendfrom one transmission environment to others without significant efforts.However, the geometrical modeling is not able to provide the statisticalcharacteristics of the simulated channel, which is close to those of thereal UWA channel This is because of the time and spatially varyingcharacteristics of the shallow UWA propagation environments

The UWA measurement-based channel modeling approach have beeninvestigated in [24, 74, 76, 85, 105] Almost all of these channel simu-lators are developed from given measurement data, which are obtainedfrom a specific underwater environment Based on analyzing the mea-surement data, the distribution of the propagation paths are specifiedsuch as Rayleigh [24, 85], Rician [76], K-distributed [105], and log-normal [74] Furthermore, in the replay-based simulators [58, 83, 95],the time variant channel impulse responses (TVCIRs) of the measuredUWA channel can be reproduced; or a new random TVCIR can be gener-

ated so that its statistical properties are similar to those of the measuredchannel The measurement-based simulator does not require physical in-put parameters, which may not be easy to set In addition, the simulatedchannels obtained by this simulator are extremely realistic based on ac-tual measurement data The disadvantage of the measurement-basedsimulator is that it can be only applied to the specific transmission en-vironment, where the channel is measured The best way to expand thediversity of this simulator is to collect a large amount of measurementdata at different time and locations [84] Moreover, for designing themeasurement-based channel simulator, a large number of channel pa-rameters, including path gains, Doppler frequencies, propagation delays,and phase shifts need to be estimated [56] There are some efficientcomputation algorithms to estimate these parameters, such as the rota-tional invariance techniques (ESPRIT) [34], the space-alternating gen-eralized expectation-maximization (SAGE) [33], the iterative nonlinearleast square approximation (INLSA) [31], the Lp-Norm Method (LPNM)[59] The measurements and computation efforts to estimate the largenumber of channel parameters make the measurement-based simulatormore complex than the geometry-based one.

In Vietnam, despite of a growing need of UWA communication plications in the military and commerce, there is not many researchpapers on UWA communication, especially in the field of channel mod-eling [2, 3, 6] Some characteristics of UWA propagations in Vietnam seahave been investigated in some earlier research [1, 4, 5, 7, 8]; however,the results of UWA channel modeling have not been given In [6], the au-thors have simulated the UWA propagation rays by solving the Eikonalequation for given environmental conditions As mentioned above, theseenvironmental parameters are hard to be specified due to the complexity

ap-of UWA propagation environments Besides, the simulated UWA agation rays is time-invariant that may not be able to describe the realUWA channel in most of cases

prop-4 Goals of the Dissertation

This dissertation aims at developing accurate and efficient approaches of

designing shallow UWA channel simulators based on the measurementdata of the real UWA channel in a specific shallow water environment.The proposed approaches should fulfill the following requirements:

• They should enable the accurate simulation of the shallow UWAchannel characterized by the measured channel impulse response(CIR), power delay profile (PDP), and/or Doppler power spectrum

• Determinations of the channel simulation model parameters should

be done in a simple and efficient manner

• The simulators designed by the proposed approaches should be able for the performance analysis of the UWA communication sys-tem based on Orthogonal Frequency Division Multiplexing (OFDM)technique

suit-To accomplish these goals, two simple and effective approaches areproposed for the design of UWA channel simulators for the two cases:(i) Fixed transmitter (Tx) and receiver (Rx) and (ii) Fixed Tx and

Rx moving The simulation results show that the proposal of design proaches emulate the statistical properties of the measured UWA chan-nels with high accuracy

ap-Furthermore, the statistical properties of UWA channels in terms ofDoppler power spectral densities (PSDs) is also the objective of this dis-sertation In this respect, we present a thorough analysis of Dopplereffects of shallow UWA channels having a time-variant surface motionand relative Tx/Rx movement As a result, the closed-form expression

of Doppler power spectrum is proposed and validated through the surement data

mea-Using the measurement-based UWA channel simulation model, a tailed analysis on the performance of UWA-OFDM communication sys-tems was presented; then, appropriate transmission parameters such asthe signal bandwidth, the number of sub-carriers, and the transmit powerwould be selected

de-5 Scope and Delimitations

The scope of the dissertation is for the approaches of designing low UWA channel simulators The design of measurement-based UWAchannel simulators, which are derived from the measurement data of thereal UWA channel in a specific shallow water environment, is mainlyfocused The aspects look into were the multi-path and Doppler effects

shal-of the measured shallow UWA channels The measurement results havebeen used for the input data of the simulators All parameters of theUWA channel simulation model are then derived from the measurementdata without considering the physical aspects of the acoustic wave prop-agation Therefore, the obtained UWA channel simulation model is justvalid for the specific transmission environment that the UWA channel ismeasured

The dissertation does not cover the analysis of effects of cal or environmental parameters (e.g., the water depth, the salinity, thetemperature, etc ) on the measured UWA channel The measurementdata itself reflects the influence of these parameters Furthermore, thestudy only concentrates on the shallow environments; thus, the channelsimulation model used in this dissertation does not capture the charac-teristics of UWA channels in the deep water

geometri-6 Motivations and Contributions of the Dissertation

In the previous section, the complexity of UWA channels has been cussed Furthermore, there is still no typical model that can be appliedfor all UWA channels because of the differences in geometrical and en-vironmental conditions This implies that we have to consider the UWAchannel characteristics specific to each different environments and lo-cations Hence, the studies of UWA channels and their characteristicssuch as path loss, delay and Doppler spread have been paid much at-tention for implementing UWA channel simulators as well as real UWAcommunication systems

dis-For the design of UWA channel simulators, we can implement the viously mentioned approaches, geometry-based and measurement-basedones Each approach has its own advantages and disadvantages Theperformance of each approach is analyzed by comparing the statistical

pre-properties of the simulated channels with those of the measured UWAchannel in a shallow water environment The simulation results showthat the measurement-based channel simulator provides the simulatedchannel which matches well with the measured UWA channel, but itrequires application of complex optimization computation methods toestimate a larger number of channel parameters The geometry-basedsimulator has a lower complexity than the measurement-based simula-tor However, the statistical properties obtained by this simulator do notfit with those of the measured UWA channel It motivates us to propose

an effective channel simulator, which is not only simple in computationbut also in good agreement with the measured UWA channel

Moreover, for the purpose of design and performance analysis of UWAcommunication systems, the statistical properties of UWA channels interms of correlation functions, Doppler power spectral densities (PSDs),and power delay profiles (PDPs) need to be analyzed [55, 56] As thematter of fact, many research papers have investigated the power delayprofile (PDP) of UWA channels, but the modeling of the Doppler spec-trum has been less well developed [94, 101] Despite the significant role

of the Doppler power spectrum in designing UWA channel simulatorsand in evaluating the UWA communication system performance, studies

on modeling it are still lacking in the literature In UWA communicationsystems, especially in shallow water environments, the Doppler effect is

a severe problem because of the fast-time variant process of the surface moving that results in unpredictable Doppler shifts [9, 21] This

sea-is our motivation to propose a Doppler power spectrum model for low UWA channels In this dissertation, the Doppler effects in shallowUWA communication systems is investigated in consideration of boththe Doppler components caused by the surface motion and by the trans-mitter/receiver movement Then, the Doppler power spectrum modelwas proposed and validated using the Doppler measurement results ofthe shallow UWA channel

shal-Another important task of underwater communication study is to alyze and verify performance of systems To mitigate inter-symbol in-terference (ISI) due to the large delay spread, Orthogonal Frequency

an-Division Multiplexing (OFDM) modulation has been widely applied toacoustic transmission [10, 22, 42, 68, 86] In OFDM systems, the trans-mission bandwidth is divided into many narrow sub-channel This makesthe symbol duration to be increased and then the inter-symbol interfer-ence (ISI) to be reduced but the inter-channel interference (ICI) caused

by the Doppler effect becomes more serious [47] In UWA-OFDM tems, the Doppler effect is much more severe because of following rea-sons: the sound speed is low and varying, the system is inherently wide-band [88], the Doppler frequencies are comparable to carrier frequencies,and the Doppler effects are unequal over sub-carriers [9] The frequency

sys-of the transmitted signal is significantly distorted by the Doppler fect [21] The motion-induced distortion has far-reaching implication forthe synchronization design and the channel estimation algorithms [88].Unlike other research studies that analyzed the ICI effect based on theassumption of Doppler spectrum such as classical, uniform or two-path[9, 22, 47, 86], we consider the ICI effect on the measurement-based UWAsimulator which is a wide-band shallow UWA channel model derived fromthe measurement data Moreover, the ambient noise from sources such

ef-as turbulence, waves, shipping, and rain is frequency dependent and notwhite noise [88] Most of the research studies have not evaluated noise

in combination with ICI effect or they considered noise as white noise[9, 22, 86] This calls for the needs of evaluating the UWA-OFDM sys-tem under the effects of both ICI and ambient noise In this dissertation,

we focus on both the ICI effect and ambient noise on UWA-OFDM tems Based on the measurement-based UWA channel, we present anexact analysis of the ICI of UWA-OFDM systems Consequently, impor-tant parameters of the system are evaluated: the transmit power, ICIpower, noise power and the resulting signal-to-interference ratio (SIR),signal to interference plus noise ratio (SINR), and channel capacity.This dissertation deals with the analysis of UWA channel modelingapproaches and UWA-OFDM system performance In this regard, themajor contributions of this dissertation are summarized as follows:

sys-1 The two typical approaches of designing shallow UWA channel

sim-ulators, the geometry-based and measurement-based ones, are vestigated The performance of each simulator is then analyzed bycomparing the statistical properties of the simulated channels withthose of the measured UWA channel in a shallow water environ-ments Thereafter, as presented in [J2], an effective approach fordesigning shallow UWA channel simulators, which is not only sim-ple in computation but also in good agreement with the measuredUWA channel, has been proposed Furthermore, the geometry-basedUWA channel model is used in [C1], while the measurement-basedone is addressed in [J1],[J2].

in-2 Theoretical background analysis of Doppler effects generated by thetransmitter/receiver (Tx/Rx) movement, and by the motion of sea-surface is presented Sequentially, a closed-form expression of Dopplerpower spectrum model for shallow UWA channels has been proposedand validated through curve fitting with the Doppler power spectrummeasurement results of real shallow UWA channels The detail ofthis contribution is presented in [J3]

3 The ICI plus ambient noise analysis for UWA-OFDM systems overthe measurement-based channel simulator for a real shallow UWAchannel is utilized The topic studied in depth includes the exact cal-culation of ICI power, ambient noise power and appropriate transmitpower, as well as their effects on performance of the UWA-OFDMsystem Signal to interference ratio (SIR), signal to interferenceplus noise ratio (SINR), and capacity performance are evaluated fordifferent parameters, including signal bandwidths, number of sub-carriers, and the transmit power These parameters should be cho-sen carefully in order to obtain the desired capacity and SINR withminimizing the ICI effect The results provide practical guidelinesfor choosing proper transmission parameters for UWA-OFDM sys-tems The paper [C1] deals with the SINR analysis, while differentperformance studies of UWA-OFDM systems are presented in [C2],[J1]

7 Organization of the Dissertation

This dissertation deals with the derivation and analysis of designing low UWA channel simulators The topics studied in depth include theapproaches of modeling shallow UWA channels, Doppler power spectrumbased on the measurement data Using the measurement-based UWAchannel simulator, the UWA-OFDM system performance is analyzed.Consequently, the system parameters can be optimized for the design ofUA-OFDM systems using the simulation results The structure of thisdissertation is as follows:

shal-• Chapter 1 presents the two approaches of designing shallow UWAchannel simulators, the geometry-based and measurement-based ones.This chapter is dedicated to review the main advantages and disad-vantages of each one Moreover, we also discuss simulation channelmodels, namely SOS (sum-of-sinusoids) and SOC (sum-of-cisoids),which can be used in modeling UWA channels Therein, the peculiarcharacteristics as well as advantages of employing the correspondingUWA channel models are also highlighted Specially, this chapterproposes an effective approach of designing shallow UWA channelsimulators that its implementation is simple in computation; how-ever, it outputs the simulated channel with statistical propertiesclose to the reality

• Chapter 2 The proposal for a closed-form expression of Dopplerpower spectrum model for shallow UWA channels has been pre-sented Using the geometry model for shallow UWA channels, thetheoretical background of Doppler effects generated by the trans-mitter/receiver (Tx/Rx) movement, or by the motion of sea-surfacehas been analyzed As a result, the Doppler power spectrum can bemodeled as a summation of the Spike-shape and the Gaussian-shape.The Spike-shape presents the Doppler component from the Tx/Rxmovement, while the Gaussian-shape presents the Doppler compo-nent from the sea-surface motion The proposed model is validatedthrough curve fitting with the Doppler power spectrum measurementresults of a real shallow UWA channel

• Chapter 3 This chapter utilizes the ICI plus noise analysis

UA-OFDM systems over the measurement-based channel models forshallow underwater acoustic channels We carry out the exact calcu-lation of ICI power, ambient noise power and appropriate transmitpower, as well as their effects on performance of the OFDM sys-tem Signal to interference ratio (SIR), signal to interference plusnoise ratio (SINR), and capacity performance are evaluated for dif-ferent signal bandwidths, number of sub-carriers, and the transmitpower Based on the simulation results, the system parameters can

be optimized for design of UA-OFDM systems

DESIGN OF SHALLOW UWA CHANNEL SIMULATORS

In this chapter, two typical approaches of designing underwater acoustic(UWA) channel simulators, the geometry-based and the measurement-based ones, are investigated Then, the performance of each one is an-alyzed by comparing the statistical properties of the simulated chan-nels with those of the real measured shallow UWA channel The re-sults show that the geometry-based simulator has a lower complexitythan the measurement-based simulator; however, the statistical proper-ties obtained by this simulator are in general not realistic and have alimited confirmation by measurements It is noteworthy that the de-signed channel models using this approach can be utilized for parameterstudies [55] This is because they can be extended without significanteffort when changing some parameters such as the water depth, thetransmission distance, the signal frequency, and so on On the contrary,the measurement-based channel simulator provides the simulated chan-nel which matches well with the real measured UWA channel, but itrequires application of complex optimization computation methods toestimate a larger number of channel parameters This is our motivation

to propose an effective approach of designing UWA channel simulators,which is not only simple in computation but also in good agreement withthe real UWA channel The parameters of the proposed simulator can bedirectly exploited from the measurement data without applying any op-timization computation method Moreover, the simulation results showthat the channel statistical properties obtained by the proposed simula-tor match well with those of the real measured UWA channel

15

1.1 Introduction

The shallow underwater acoustic channel can be characterized as a tipath fading channel Low propagation speed (1500 m/s) leads to UWAchannels with a long multipath delay spread that causes strong frequencyselectivity Time-varying multipath fading observed in the radio chan-nels is present also in the acoustic channels and is intensified in shallowwaters due to the strong surface and bottom reflections [13] The chan-nel typically has significant Doppler spread due to the low propagationspeed, even when the transmitter/receiver are not moving fast

mul-In order to design, analyze, and simulate underwater communicationsystems, efficient modeling of the acoustic channel is essential Thetwo main approaches of designing UWA simulators, the geometry andthe measurement-based ones, are depicted in Fig 1.1 and Fig 1.2, re-spectively It shoud be mentioned that, these figures only illustrate themethodology behind the design of shallow UWA channel simulators withthe simplified propagations, not a real shallow UWA channel

The methodology behind the geometry-based channel modeling is lustrated in Fig 1.1 The UWA geometry-based simulator is designed byusing the geometrical channel model In this case, the reference model isdeveloped by performing an ensemble averaging over an infinitely largenumber of scenarios [17, 55, 62] Parameters of the reference model such

il-as path loss, delay and Doppler spread can be derived from the etry model and physical laws of UWA propagation environments It isnoted that the reference is non-realizable because of an infinite number

geom-of scatterers Therefore, a simulation model can be derived by mating statistical properties of the reference model with a finite number

approxi-of scatterers The main advantage approxi-of this approach is the computationalsimplicity and it can be utilized for parameter studied [55] However,the simulation results show that its simulated channel is not realistic.That is because parameters of the model are hard to determine exactlyunder the real UWA propagation conditions

The methodology behind the measurement-based approach (see Fig 1.2)

is that the channel simulator are derived from the measurement data of

Figure 1.1: The methodology behind the geometry-based channel modelling [17, 55].

the particular scenario [24, 74, 76, 85, 105] By approximating the tical properties, parameters of the simulators such as path gains, delaysand Doppler frequencies are obtained The simulated channel of this sim-ulator is highly realistic but only valid for the specific scenario on whichthe measurement data have been taken Furthermore, the effort in fieldmeasurements and in computations of a large number of the simulationchannel parameters are major disadvantages of this approach

statis-Due to lack of standardized models that can meet all requirements

of UWA channel modeling, the choice of modeling approaches depends

on the application purposes Namely, for the design and comparisonpurposes, one should use the geometry-based simulator; whereas, themeasurement-based one can be a better choice for optimization and de-ployment of UWA communication systems

In this chapter, first of all, simulation models, namely SOS sinusoids) and SOC (sum-of-cisoids) methods [38, 102], can be used tomodel the UWA channel will be discussed The SOS method generatesthe channel simulation model by summing a finite number of sinusoids,whereas the SOC one models the channel by a finite sum of complex sinu-soids (cisoids) The SOS method generates the channel simulation model

(sum-of-by summing a finite number of sinusoids, whereas the SOC one modelsthe channel by a finite sum of complex sinusoids (cisoids) Then, both

Figure 1.2: The methodology behind the measurement-based channel modelling [31,

56].

approaches of designing UWA channel simulators, the geometry-basedand measurement-based ones, are evaluated The performance of eachapproach is verified by comparing its statistical properties with those

of the real UWA channel, which is measured in Halong bay, Vietnam.Furthermore, an effective approach to design UWA channel simulators isproposed for two different cases In the first case, the transmitter (Tx)and receiver (Rx) are fixed; whereas the Rx is moving in the latter one.All or a part of the channel parameters of the proposed simulators can bedirectly exploited from the measurement data of the real UWA channelwithout applying any parameter computation method; thus, its com-plexity is significantly reduced Moreover, the simulation results showthat our proposed simulator provides the statistical properties close tothose of the measured UWA channel

The rest of this chapter is organized as follows Section 1.2 presentsoverview of simulation models that can be used to model UWA channels.The design of the geometry-based UWA channel simulator is described

in Sect 1.3 In Sect 1.4, the steps to design the measurement-basedUWA channel simulator are presented using the measurement data ofthe real shallow UWA channel in Halong bay, Vietnam Sections 1.5 and1.6 describe the proposed approaches to design UWA channel simulatorsfor two different cases, fixed transmitter/receiver and moving receiver

Furthermore, their performance analysis in comparison with the-art is discussed Finally, Sect 1.7 highlights the specific findings ofthe chapter.

state-of-1.2 Overview of Simulation Models for UWA Channels

In this section, fading models, which are widely used in the cations community, will be discussed These fading models representthe characteristics of electromagnetic transmission over the air, both fornarrow-band and wide-band scenarios, and often successfully Therefore,they will be evaluated to model the distortion caused by the water inunderwater acoustic communications To capture the propagation char-acteristics of shallow UWA environments, parameters of these modelswill be determined by the physical laws of UWA propagation environ-ments or the measurement data of real shallow UWA channels

communi-1.2.1 Rayleigh and Rice channels

Rayleigh and Rice channels are the most popular channel models used

in wireless communications This is because of the fact that they areeasy and effective descriptions of fading channels for investigations usingcomputer simulations In multipath propagation environments, due todiffraction, reflection, and scattering, the received signal is composed ofmany different components, which are differences in amplitudes, phasesand delays A narrowband fading channel is often modeled by a complexGaussian process µ (t) = µ1(t) + jµ2(t), where µ1(t) and µ1(t) are real-valued Gaussian processes The Rice method is widely applied to modelthe real-valued Gaussian process µi(t) (i = 1, 2), which is expressed as

ci,ncos (2πfi,nt + θi,n) , (1.1)

where ci,n, fi,n, and θi,n are the gains, the frequencies, and the phases.For the case of having the LOS component at the receiver, a time-variant deterministic process m(t) will be introduced to represent theLOS component, i.e m (t) = ρej(2πf p t+θ p ), where ρ, fp, and θp denote theamplitude, the Doppler frequency and the phase of the LOS component,

respectively Consequently, the complex Gaussian process for the LOScase is expressed as µp(t) = µ(t) + m(t) By taking the absolute value of

µp(t), the Rice process is obtained as

summa-1.2.2 Deterministic SOS Channel Models

For simulation purposes, it is not necessary to take into account an nite number of components As proofed in [62, 64], the desired statisticalproperties can be described by a limit number of harmonic functions.This is called as a deterministic channel modeling principle Using thatprinciple, the real-valued Gaussian process is represented by a finite sum-of-sinusoids (SOS)

ci,ncos (2πfi,nt + θi,n) (1.4)

Sequentially, in the deterministic SOS model, the complex Gaussian cess is represented by

pro-˜

µ (t) = ˜µ1(t) + j ˜µ2(t) (1.5)Channel simulation models based on Rices SOS principle [14, 54] havewidely been in used in designing multipath radio channel simulators(e.g., see [15, 23, 43, 51, 81]) In the SOS channel simulation approach,the in-phase µ1(t) and quadrature µ2(t) (IQ) components of the channelscomplex envelope are assumed to be uncorrelated Under this considera-tion, it can only be used to simulate fading channels having symmetricalDoppler power spectral densities (DPSDs) In other words, the SOS-based approach is only applicable to model channels under isotropic

scattering conditions Hence, it can not be used to model the more alistic case of channels characterized by asymmetrical DPSDs, such asthe shallow UWA channels.

re-1.2.3 Deterministic SOC Channel Models

The SOC (sum-of-cisoids) model, where the Gaussian process is imated by a finite sum of complex sinusoids (cisoids) [78], is a solution toovercome the limitation of the SOS channel simulation approach In thedeterministic SOC model, the complex Gaussian process can be modeledand efficiently simulated as

to the physical wave propagation Therefore, it is suitable to use them

to develop the measurement-based channel simulators

1.3 The Geometry-based UWA Channel Simulator

The steps of implementing the geometry-based simulator are illustrated

in Fig 1.3 In the first step, a geometrical model for UWA channels iscreated The reference model is developed from the geometrical model

in the second steps Then, a simulation model for the UWA channelwill be selected in the third step The simulation model should be able

to describe closely the reference model, which is developed from thegeometrical channel model To estimate the parameters of the simulationmodel, a parameter computation method is applied in the fourth step.Finally, the channel parameters of the simulation model such as pathgains, propagation delays, and Doppler frequencies are determined

Figure 1.3: The scheme of designing the geometry-based channel simulator [17, 55].

1.3.1 Developing the Reference Model from the Geometrical Channel

Model

A shallow UWA geometrical channel model [55] is presented in Fig 1.4,where D is the distance between the transmitter (Tx) and receiver (Rx).The scatterers Si,n(n = 1, 2, , Ni and i = 1, 2) are assumed to be ran-domly distributed on the surface (i = 1) and the bottom (i = 2) of

a shallow-water environment The angle-of-departure (AOD) and theangle-of-arrival (AOA) of the nth path are denoted by βi,n and αi,n, re-spectively The receiver is moving with speed VR in the direction deter-mined by the angle-of-motion αR

v.Referring the geometrical model in Fig 1.4, the time variant channelimpulse response (TVCIR) h (τ, t) is composed of three components, andcan be expressed as

Figure 1.4: The geometrical model for shallow UWA channels with randomly

dis-tributed scatterers Si,n(•) on the surface (i = 1) and the bottom (i = 2) [55].

The LOS part h0(τ, t) is given by [55]

h0(τ, t) =

1 + cRAs(D0) Aα(D0) e

j(2πf 0 t+θ 0 )δ (τ − τ0) , (1.8)where cR denotes the Rice factor; the symbols τ0, f0, and θ0 stand forthe propagation delay, the Doppler frequency, and the phase shift of theLOS component, respectively The propagation delay τ0 = D0/cs, inwhich the speed of sound cs = 1500 m/s for shallow water environments,the total distance D0 between Tx and Rx is obtained by

in-computed as

α0 = π + arctan

yT

1 − yR 2

D



(1.11)The function As(D0) denotes the propagation loss coefficient due tospherical spreading, which is expressed as follows [19]

The second h1(τ, t) and the third part h2(τ, t) of the TVCIR h (τ, t)

in Eq 1.7 are given by

yR 2

2Dsin2(α), if π + arctan

yR 2

D



≤ α ≤ 3π

2 (1.17)

With reference to Fig 1.4, the AOD βi,n can be derived as a function ofthe AOA αi,n [55]

Di,n = y

T i

sin(βi,n) +

yR i

sin(αi,n), (1.19)for n = 1, 2, , Ni and i = 1, 2

The propagation delay τi,n of the nth path is obtained by

τi,n = Di,n

cs

where cs denotes the speed of sound, which is given by 1500 m/s

The propagation distance Di,n(αi,n) is a function of the AOA αi,n,which is obtained by substituting Eq 1.18 into Eq 1.19 Therefore, theparameters As(Di,n), Aa(Di,n), and τi,n can be also computed as functions

of the AOA αi,n These parameters are required for computation of thetime-frequency correlation function (T-FCF), which characterizes statis-tical properties of the reference model The T-FCF represents rapidity

of fading on the UWA channel in both time and frequency domain [72]

Therein, the time-correlation function (TCF) describes the time tion of the channel, and the frequency-correlation function (FCF) gives

varia-us criteria to measure the frequency-selectivity of the channel

For computation of T-FCF, the time-variant channel transfer function(TVCTF) needs to be derived By taking the Fourier transform of TV-CIR h(τ, t) with respect to the propagation delay τ , the TVCTF H(f, t)

As(Di(α)), the absorption loss coefficient Aα(Di(α)), and the tion delay τ (α) can be obtained

propaga-The TVCTF H(f, t) in Eq 1.21 is computed by using the tioned parameters; then, the normalized T-FCF of the reference model

aforemen-rHH (4f, 4t), where 4f and 4t denote the frequency and time tion variables, is obtained by [55]

separa-rHH(4f, 4t) = cR

1+c Rej2π[fmax cos(α 0 −αRv)4t−4f τ 0]+

to scattered power The symbol τ0 = D0/cs is the propagation delay

of the LOS component The maximum Doppler frequency is denoted

by fmax = VRfc/cs The second term of Eq 1.22 is the T-FCF of the

channel, which includes the scattered components from the surface andthe bottom.

1.3.2 The Simulation Model

In the reference model, the number of scatterers Ni are infinite; thus, it

is unable to implement In this section, a low-complexity sum-of-cisoids(SOC) simulation model is selected for the simulation of UWA channelsunder non-isotropic scattering conditions [55]

Applying the principle of deterministic channel modeling [62], the ulation model can be derived from the reference model by using only afinite number of scatterers number Ni Then, using Eq 1.22 with thefinite Ni, the T-FCF of the simulation model is given by [55]



,(1.23)where ci,n, fi,n, and τi,n are the path gain, the Doppler frequency, and thepropagation delay of the nth path received from the surface or bottom

of the ocean, respectively The path gain ci,n is computed as follows

ci,n = q As(Di,n)Aα(Di,n)

2PNin=1(As(Di,n)Aα(Di,n))2

whereas, the Doppler frequency fi,n is given by

fi,n = fmaxcos αi,n− αR

1.3.3 The Estimated Parameters of the Simulation Model

The parameters of the channel simulation model are derived by usingthe method of equally spaced scatterers (MESS) [55], which allows us tocompute the optimal positions xopti,n of the scatterers Si,n as

xopti,n = D

Ni



n − 12



for n = 1, 2, , Ni and i = 1, 2 Using xopti,n, the AOA αi,n and theAOD βi,n can be computed as in Eq 1.16 and Eq 1.18, respectively.Using these computed values of the AOA αi,n and the AOD βi,n, we cancalculate Di,n in Eq 1.19, then AS(Di,n) and Aα(Di,n) by using Eq 1.12and Eq 1.13, respectively The path gain ci,n is obtained by Eq 1.24.The Doppler frequency fi,n and propagation delay τi,n are computed by

Eq 1.25 and Eq 1.20, respectively Finally, the T-FCF of the simulationmodel is calculated by Eq 1.23

1.3.4 Simulation Results

In order to verify the performance of the channel simulator, the tistical properties of the simulation model are compared with those ofthe reference model, which are developed from the geometrical channelmodel

sta-For the case of the fixed Rx, i.e VR = 0, then the maximum Dopplerfrequency fmax = 0, and the Doppler frequencies fi,n in Eq 1.25 are thusequal to zeros Consequently, the T-FCF in Eq 1.23 becomes the FCFˆ

rHH (4f ), which is expressed as follows

a good fit between the reference model and the simulation model Thatmeans the simulation model with the computed parameters can describewell the geometrical model (i.e the reference model), not an measuredUWA channel in real conditions

1.4 The Measurement-based UWA Channel Simulator

The steps of designing a measurement-based channel simulator are ilar to those of the geometry-based simulator However, the reference

sim-Table 1.1: Parameters of the geometrical channel model

Δ f [Hz]

0 0.2 0.4 0.6 0.8 1

Simulation model Reference model

Figure 1.5: The comparison between the normalized FCF of the reference model and

that obtained by the geometry-based simulator.

model is achieved by experiments

1.4.1 The Reference Model from the Measurement Data

The reference model of the measurement-based approach is developedfrom the measurement data of a real UWA channel Experiments havebeen carried out in Halong bay, a shallow water environment, in Viet-nam, on 9th June 2015 The channel impulse responses (CIR) of the realUWA channel were measured for the distance of 150 m The testbedfor measuring the impulse response of a real-world underwater acous-tic channel is illustrated in Fig 1.6 The transmit side consists of acomputer, a power amplifier, and the transducer The ship (A) acts as

a transmitter which is equipped with all these components We

con-Figure 1.6: Illustration of the measurement setup in Halong bay.

sider an UWA channel in shallow water, where the distance between thetransmitter and the receiver boat is 150 m The receive side consists of ahydrophone, a low noise amplifier(LNA), and a receiver computer Themethod for signal processing to obtain the CIR of the measured UWA

is presented in [50] The correctness of the method has been proven bytheories, simulations, and experiments in [50] The brief review of thismethod is described in the following

At the transmitter side, the channel sounding signal as a Pseudo-Noisesequence P N (t) having the generator polynomial G(x) = x10 + x7 + 1

is generated After that, it is modulated by BPSK (Binary Phase ShiftKeying), amplified and transmitted via a transducer After traveling overthe real UWA channel, at the receiver side, the received sounding signal

is amplified by the low noise amplifier (LNA), then it is demodulated

by using the BPSK demodulator Sequentially, it leads to the channelparameter detector, which performs the correlation of the demodulatedsignal with the transmitted PN sequence P N (t) Choosing the peakvalues of the correlation results, the gain and the propagation delay

of each propagation path are determined Hence, the channel impulseresponse (CIR) ˆh (τl, tm) of the real UWA channel is obtained, where the

t [s]

100

50

0 0.02 0.015

τ [s]

0.01 0.005

0 0 0.5 1

Figure 1.7: The measured |ˆ h(τ, t)|2 for the transmission distance of 150 m.

observation time tm = m∆t for m = 1, 2, , M , and the propagationdelay τl = l∆τ for l = 0, 1, , L − 1 Therein, M denotes the number ofmeasured samples in observation time, L is the number of propagationpaths, and the path resolution is denoted by ∆τ

Figure 1.7 shows the results of the measured CIR ˆh (τl, tm) obtainedfrom 96 continuous measurements in Halong bay for the distance of

150 m The power delay profile (PDP) ˆρ(τl) of the measured UWA nel is computed by averaging the |ˆh(τl, tm)|2 in the observed time domainas

a function of the propagation delay τ [72]

Using the result of the PDP, we can derive the FCF, which izes the fading rapidity of the channel in the frequency domain The FCFˆ

character-RHH(∆f ) of the reference model, which is developed from the measuredUWA channel, is obtained by the Fourier transform of the measured

1.4.2 The Simulation Model

For describing the sophisticated channel, a wideband sum-of-cisoids correlated scattering (SOCUS) channel simulation model is widely ac-cepted [38, 56] The time-variant channel impulse response (TVCIR)

un-h (τ, t) of tun-he SOCUS model is expressed by [62]

θn,l

For further analysis of channel simulator performance, the time quency correlation function (T-FCF) of the simulation channel model

fre-need to be computed and then compared to that of the measured nel By taking the Fourier transform of the TVCIR h (τ, t) in Eq 1.30with respect to the propagation delay τ , the time-variant channel trans-fer function (TVCTF) H (f, t) is obtained as

RHH(∆f, ∆t) = E [H (f, t) H∗(f + ∆f, t + ∆t)] (1.32)Substituting Eq 1.31 to Eq 1.32, under the assumption that the UWAchannel simulator is wide-sense stationary, the T-FCF of the simulationmodel is formulated as follows [62]

1.4.3 Estimated Channel Parameters of the Simulation Model

The parameters of the simulation model, including the path gain cn,l,the Doppler frequency fn,l, the propagation delay τl, and the number ofpaths Nl, need to be determined It is noted that the number of thepropagation paths L is obtained directly from the measured PDP Forestimating the parameters of the simulation model, several computationmethods can be applied The Lp-Norm Method (LPNM) [59] is thewell-known method to compute these parameters The cost function is

RHH (∆t) − ˆRHH(∆t)

,

(1.35)

where RHH(∆t) and RHH(∆f ) are the TCF and the FCF of the nel given by the simulation model; ˆRHH(∆t) and ˆRHH(∆f ) stand forthe TCF and FCF of the reference model, respectively Notice that,ˆ

chan-RHH(∆t) and ˆRHH(∆f ) are derived from measurement data

For the case of static channels (i.e fixed Tx/Rx), the cost function in

1.4.4 Comparison of the Two Channel Simulators

To verify the performance of the two simulators, geometry-based andmeasurement-based, we compare the FCF obtained by these simulators

to that of the reference model, which is developed from the measurementdata of the real UWA channel The results illustrated in Fig 1.9 showthat the geometry-based simulator does not fit with the real UWA chan-nel (i.e the reference model) as does the measurement-based simulator.This simulator has a poor performance but a low complexity The laterone describes the real measured UWA channel; however, it is complex incomputation

Δ f [Hz]

0.2 0.4 0.6 0.8 1

Reference model Measurement-based simulation model Geometry-based simulation model

Figure 1.9: The comparison of the normalized FCF obtained by the two simulators

to that of the reference model.

It should be mentioned that the geometry-based simulator in Sect 1.3

is developed with the assumption of equally distributed scatterers Theassumption may not be valid for the measurement scenario; thus, thesimulation channel model obtained by the geometry-based approach isnot close to the measured UWA channel To make it fit well to the mea-sured channel, one can find the distribution of the PDF of the AOAsbased on the measurement data Using the distribution, other parame-ters of the simulation model can be derived using the geometry modeland UWA physical laws A combination between the geometrical modeland the measurement data may result in an effective simulator and also

be considered as an open research

1.5 The Proposed Approach for the Static UWA ChannelAfter analyzing the two approaches of designing UWA channel simula-tors, we can conclude that each one has advantages and disadvantages.The geometry-based one is simpler in computation because of the fewerestimated channel parameters but has a lack of closeness to reality Thetime and spatially varying characteristics of the shallow UWA propaga-tion environments lead to difficulties in fitting the geometrical model to

Figure 1.10: The flowchart of proposed approach to design the static UWA channel

simulator.

the real UWA propagation conditions Compared to this approach, themeasurement-based one is well-matched with the real measured UWAchannel, but it requires application of a complex computation method

to estimate the large number of the channel parameters Therefore, aneffective approach is proposed for the design of UWA channel simulatorthat its implementation is simple in computation; however, it outputsthe simulated channel with statistical properties close to the reality Themain advantage of this approach is that its channel parameters can bedirectly exploited from the measurement data

1.5.1 Descriptions

The steps of the proposed approach for the design of UWA static channelsimulator are illustrated in Fig 1.10, and described as follows

• Step 1: Based on the measurement data of the real UWA channel,the PDP ˆρ (τl), for l = 0, 1, , L − 1, is specified as described insection 1.4.1.

• Step 2: Our proposed simulator is also designed by using the surement data Therefore, the simulation model of the proposed sim-ulator is selected as the class of SOCUS model, which is suitable fordesigning measurement-based simulators as presented in Sect 1.4.2.The TVCIR of the simulation model is given in Eq 1.30 Subse-quently, the TVCTF H(f, t) and the T-FCF RHH(∆f, ∆t) of thesimulation model are computed as in Eq 1.31 and Eq 1.33, respec-tively In our considered case of the static channel, the Tx and

mea-Rx are fixed; thus, there is no Doppler shift Consequently, using

Eq 1.33 and setting fn,l = 0 for n = 1, 2, Nl and l = 0, 1, L − 1 ,the simulation FCF RHH (∆f ) can be expressed as

Referring to Eq.1.30, we can write the channel coefficient of the path

n=1c2n,l Therefore, in the proposed channelsimulator, the path gain cn,l can be obtained from the measured PDPˆ

∆ f [Hz]

0.2 0.4 0.6 0.8 1

Reference model Measurement-based simulation model Geometry-based simulation model The proposed simulation model

Figure 1.11: The comparison between the normalized FCF of the reference model

and that obtained by the measurement-based, the geometry-based, and the proposed simulators.

for n = 1, 2, , Nl The number of path Nl, for l = 0, 1, , L − 1,are obtained directly from the measured TVCIR ˆh (τl, t), that means

Nl = M , where M is the number of measured samples as presented

in Sect 1.4.1

Finally, all parameters of the proposed simulator are exploited fromthe measured PDP ˆρ (τl) without applying any complex computationmethod

1.5.2 Results and Discussions

For evaluating the proposed channel simulator performance, the tical characteristics of the simulated channels are compared to those ofthe measured real UWA channel (i.e the reference model)

statis-After obtaining the parameters {cn,l, τl} of the proposed channel lator, the time-variant frequency transfer function H (f, t) and the FCF

simu-RHH(4f ) of the channel simulation model are computed by Eq 1.31 and

Eq 1.33, respectively Then, the FCF of the simulated channel obtained

by each simulator in comparison with the measured real UWA channelare shown in Fig 1.11 As can be seen, the proposed channel simulator

Table 1.2: The performance comparisons

Approaches of

designing

The geometry-based channel simulator

The based channel simulator

measurement-The proposed channel simulator Required

The measured PDP of the real UWA channel

The measured PDP of real UWA channel

Number of the

estimated channel

parameters

Positions of scatters x opt

i,n

Parameters of propagation paths L×N l ×ncoptn,l, τlopto

None MSE 0.3824 1.2607 × 10−16 1.5010 × 10−16

is in good agreement with the reference model, which is derived from themeasurement data of the real UWA channel

To illustrate more detail of the accuracy of each simulator, we calculatethe Mean Square Error (MSE) of the FCF of the simulation channel asfollows

M SE = E



RHH(∆f ) − ˆRHH(∆f )

... varying characteristics of the shallow UWA propaga-tion environments lead to difficulties in fitting the geometrical model to