Analysis and Optimization of
Transmission Strategies for Two Hop
Networks with Multiple Antennas
MUHAMMAD NAEEM ADIL
Master’s Degree Project
Stockholm, Sweden September 2013
XR-EE-KT 2013:007
Analysis and Optimization of
Transmission Strategies for Two Hop
Networks with Multiple Antennas
MUHAMMAD NAEEM ADIL
<mnadil@kth.se>
September 6, 2013
Master of Science Thesis Project performed at TU, Berlin, Germany
School of Electrical Engineering Communication Theory KTH Royal Institute of Technology
Stockholm, Sweden
Supervisors : Dipl.-Math. Jan Schreck MSc. Michal Kaliszan
iii
Acknowledgements
I would like to express my deepest gratitude to Dipl.-Math. Jan Schreck and MSc. Michal Kaliszan for their wonderful guidance, contin-uous encouragement and feedbacks throughout my thesis work. With-out there supervision this thesis project could not have been realized.
Also, I would like to extend special thanks to all the people at In-stitut fur Telekommunikationssysteme in TU Berlin for their assistance in every matter during my stay in Berlin.
Last but not the least, I would like to thank my parents for the unconditional support and love.
iv
Abstract
Two hop relay based networks consist of three network nodes: source, relay station, and destination in which relay station assists the source to communicate reliably and efficiently with the destination. More-over, these networks provide cost efficient solution for achieving high data rate via cooperative communication between relays with single antennas.
In two hop relay based networks, communication from a source to destination takes place over two phases, i.e , in first phase from source to relay station and in second phase from relay station to the desti-nation. Therefore, it is essential to formulate transmission strategies, i.e, TDMA, SDMA, Hybrid TDMA-SDMA and multicast in terms of resource allocation, beamforming over two phases so that interference is taken into account and high data rates are achieved. In this the-sis, some relay selection methods have been proposed to optimize the network performance. Different proposed transmission strategies are compared in different scenario settings in order to analyse and decide the best strategy in each setting.
Based upon simulation results it is recommended to use adaptive time split ratio between the two phases. Brute force relay selection gives the optimal relay assignment but Hungarian assignment algorithm also performs pretty close to brute force performance. SDMA with cooper-ative relays connection with multiple antennas at the relays performs much better than the other transmission strategies. However, multicast strategy performs much better if second phase channel knowledge is not available at the base station.
Keywords : Two hop relay based networks, Transmission strate-gies, Resource allocation, Beamforming, Relay selection
Contents
Contents v
List of Figures vii
List of Acronyms and Abbreviations ix
1 Introduction 1
1.1 Motivation . . . 1
1.2 Objective . . . 3
1.3 Outline of the thesis . . . 4
1.4 Contributions . . . 4
2 Two Hop Relay Networks 5 2.1 Relay forwarding strategies . . . 6
2.2 Channel and rate model . . . 6
2.2.1 Block fading channel model . . . 8
2.2.2 Resource allocation . . . 8
2.2.3 Phase I . . . 9
2.2.4 Phase II . . . 10
2.2.5 End-to-end rate . . . 11
2.2.6 Time Split ratio between two phases . . . 12
3 Transmission Strategies for Two-Hop Relay Networks 15 3.1 Baseline 1: Single relay . . . 15
3.1.1 TDMA . . . 16
3.1.2 SDMA with fixed relay assignment . . . 18
3.1.3 SDMA with relay selection . . . 20
3.1.4 SDMA-TDMA (Hybrid) with fixed relay assignment . . 25
3.2 Baseline 2: Cooperative relays . . . 28
3.2.1 TDMA . . . 28
3.2.2 SDMA with fixed relay assignment . . . 29
vi CONTENTS
3.2.3 SDMA with relay selection . . . 29
3.2.4 Hybrid SDMA-TDMA with with fixed relay assignment 30 3.3 Baseline 3: Multicast . . . 30
3.3.1 MF Multicast beamforming, single relay selection . . . 31
3.3.2 Max-min beamforming, single relay selection . . . 32
3.3.3 MF Multicast beamforming, cooperative transmission . 33 3.3.4 Max-min beamforming, cooperative transmission . . . 33
4 Simulation Scenarios with Results and Analysis 35 5 Conclusion and reflections 39 5.1 Conclusion . . . 39 5.2 Reflections . . . 40 5.2.1 Economic impact . . . 40 5.2.2 Environmental impact . . . 40 5.2.3 Social aspect . . . 40 Bibliography 41
List of Figures
1.1 Conventional cellular network . . . 2
1.2 Two hop relay networks with direct link . . . 3
1.3 Two hop relay networks without direct link . . . 3
2.1 Schematic figure for system and channel model . . . 7
2.2 Split ratio comparison for TDMA transmission strategy . . . 13
3.1 TDMA . . . 17
3.2 TDMA sequential transmission for 3 connections . . . 17
3.3 Comparison between rate maximization relay selection and fixed assignment . . . 18
3.4 SDMA with single fixed relay . . . 19
3.5 SDMA with relay selection . . . 20
3.6 Comparison between relay selection methods for SDMA with fixed number of connections . . . 24
3.7 Comparison between relay selection methods for SDMA with vary-ing connections . . . 25
3.8 Hybrid strategy . . . 26
3.9 Sum rate over SNR variation for single relay strategies . . . 27
3.10 Sum rate over relay variation for single relay strategies . . . 27
3.11 Cooperative relay strategies . . . 31
3.12 Multicast relay strategies comparison . . . 34
4.1 TDMA with variation in number of users . . . 36
4.2 SDMA in single relay configuration with variation in number of users 36 4.3 SDMA single relay connection with multiple antennas at the users 37 4.4 Multiple antennas at relays vs multiple antennas at the users . . . 37
4.5 Phase 1 Multicast comparison with TDMA and SDMA . . . 38
4.6 Multicast comparison with TDMA and SDMA . . . 38
List of acronyms and abbreviations
Bs Base station
Rs Relay station
Ue User equipment
TDMA Time Division Mutliple Access
FDMA Frequency Division Mutliple Access
SDMA Space Division Multiple Access
AF Amplify and Forward
DF Decode and Foward
CF Compute and forward
IBC Interfering Broadband Channel
IFC Interference Channel
MRT Maximuim Ratio Transmission
MRC Maximuim Ratio Combining
ZF Zero Forcing
SVD Singular Value Decomposition
SIMO Single Input Multiple Output
MIMO Multiple Input Multiple Output
CSI Channel Side Information
i.i.d independent and identically distributed
SINR Signal to Interference plus Noise Ratio
MF Match Filter
Chapter 1
Introduction
This chapter gives a brief overview of this thesis report. It describes the motivation and main objectives of this thesis work. It also gives an outline of the thesis report and my contributions in the project.
1.1
Motivation
Two hop relay based networks have been an area of great interest. They provide cost and energy efficient solution for increasing cellular network ca-pacity so that network operator can meet the ever increasing demands for higher data rates for different data services [1]. In the conventional cellular network, there is a direct connection between base station and the user as shown in Figure 1.1. But we may have coverage issues due to shadow fading problem in large cell radius with mostly non-line of sight conditions. The demand for high data rates has been increasing due to different application and services that are being used on different user equipments. Further, it is also predicted that the traffic data usage with high data rate demands will increase exponentially in the future as well [2]. Therefore increasing capacity of the existing network in cost and energy efficient manner is an interesting area of research. In conventional single hop network, demands for higher user rate can be fulfilled by deployment of more and more high power base station in order to increase the cell capacity. But this requires bigger infrastructure, more energy and high operational and maintenance cost. [3, 4].
The challenges mentioned above can be met with the deployment of low power and low-cost nodes (relay stations) along with the existing conventional network with base stations, instead of deploying new base stations with high power and cost. Two hop relay networks overcome the coverage issues by minimizing the distance between the user and relay stations. It also minimizes
2 CHAPTER 1. INTRODUCTION
Figure 1.1: Conventional cellular network
the interference as relay stations are low power nodes.
A relay based network also reduces infrastructure cost mainly due to ra-dio frequency backhaul link between base stations and relays. Either same frequency (inband) band is used in both phases with separation in time be-tween them or different frequency bands are used in two phases (outband) [5]. Furthermore, relay stations can be environmentally powered by using wind or solar energy, thus reducing operational costs. High data rates gains in two hop relay networks can be achieved through cooperative communication, in which relays with single antennas cooperates to form a relay with multiple antennas scenario.
In two hop relay based networks, generally we can have two different com-munication links between base station and user. Firstly, there is direct link between base station and user. The other link, consists of two phases or hops. Phase one is between base station and relay station and phase two between relay station and user as depicted in Figure 1.2.
In this thesis report, we only consider the two phase connections between base station and user via relay station without any direct links as shown in Figure 1.3.
A two hop relay based network consists of three nodes, i.e, source, relay and destination. The source transmits the information via the relay to the destination. The relay receives the information message from a source and performs some processing depending upon the relay forwarding scheme de-scribed in Section 2.1 and retransmits the information to the user. Relays can also cooperate with each other to form a relay with multiple antennas and offer gains in terms of capacity and reliability to the network.
1.2. OBJECTIVE 3
Figure 1.2: Two hop relay networks with direct link
Figure 1.3: Two hop relay networks without direct link
1.2
Objective
The main objective of this thesis is to formulate different transmissions strategies like SDMA, SDMA-TDMA (Hybrid) and multicast with single and cooperative relays connection scenarios. Furthermore, different transmission strategies for two hop relay based networks in different scenario settings are
4 CHAPTER 1. INTRODUCTION
analysed. Moreover, network performance, in terms of combined user rates, is improved by optimizing resource allocation between the two phases and minimizing interference with different precoders.
1.3
Outline of the thesis
Thesis report is organized as follows: Chapter 2 describes main features of two hop relay networks, relay forwarding principles, channel and rate models and resource block allocations for two phases in the system. Chapter 3 de-scribes the transmission strategies used for two hop relay networks in terms of beam forming and resources allocation with different relay connections. Chap-ter 4 presents the results and analysis of the transmission strategies used in different scenarios settings. Finally, concluding statements about main results are summarized in Chapter 5.
1.4
Contributions
My main contributions in the project are as follows:
• Formulation of transmission strategies and propose relay selection strate-gies.
• Resource allocation between two phases.
• Comparison of different transmission strategies.
• Comparison of different system setups.
• Run simulations for different scenario settings, analysing the results and deriving conclusion based upon the results.
Chapter 2
Two Hop Relay Networks
Classical two hop relay based networks consist of three nodes: base sta-tion, relay stasta-tion, and user. In the present work, we consider a generalized setup consisting of multiple base stations, relay stations and end users, with several (logical) connections between base stations and users. We assume that communication from base station to user or from user to base station always takes place in two phases. In the downlink, phase I transmission takes place from base station to relay station (Bs → Rs) and phase II transmission takes place from relay station to user (Rs → U e). Similarly uplink transmission also takes place in two phases. D/L and U/L transmissions as well as two phase transmissions in them take place in separate time slots. In this thesis report, we only focus on downlink transmission.
As one complete transmission from base station to user takes place in two phase via relay stations, therefore additional resources are needed to maintain orthogonality between the two phases. In frequency domain, FDMA is used in order to have different frequency bands in both phases to communicate with relays. In time domain, TDMA is used where two phase transmissions are separated in time by allocating different time slots for two phases but using the same frequency in both phases. In one time slot, only phase I transmission takes place from base station to relay and in second time slot, only phase II transmission from relay to user by applying one of the forwarding principles described in the Section 2.1.
Relays cannot receive and transmit simultaneously. We focus on time division multiple access approach between two phases for our project with the half duplex constraint assumption but it also hold true for other strategies. We also assume that there is no direct transmission from the base station to the user.
6 CHAPTER 2. TWO HOP RELAY NETWORKS
2.1
Relay forwarding strategies
Relay forwarding strategies define different types of processing done at re-lays on the information received at them from source before it is retransmitted to the destination [6]. Some of the commonly used relay forwarding strategies are as follows :
• Amplify and forward (AF)
In AF strategy, relays amplify the received signal from source and re-transmit the scaled version of the received signal to the destination. Relay acts like a repeater which amplifies the received signal from the source. No decoding is done at the relay stations. All processing is done at the destination on the information transmitted from the source. In AF, noise as well as interference get amplified from the previous phase [7] [6].
• Decode and forward (DF)
In DF strategy, relay stations perform processing on received signal at relay and extract the message information transmitted by source. Relay encodes it again and forwards it to the destination. By decoding the message, a relay gets rid of the interference and noise and retransmits much cleaner version of the information to the destination. One draw-back of DF is that as the number of transmitters increases, interference increases so it is hard to decode the transmitted message reliably [6].
• Compute and forward (CF)
In CF relay forwarding strategy, relay partially decodes information mes-sage received from source. Afterwards, relay retransmit some informa-tion according to some funcinforma-tion related to the informainforma-tion message re-ceived at the relay, which helps in decoding the actual information mes-sage at the destination. In CF, direct link between source and destina-tion is essential in order to assist user to decode message with addidestina-tional information from relay [6].
2.2
Channel and rate model
We consider a cellular network with B base stations, R relays and U users. Base station b is equipped with Nb transmit antennas and performs some
linear processing, each relay r is equipped with a single antenna and user u is equipped with Nu receive antennas. As two hop relay network consists of
2.2. CHANNEL AND RATE MODEL 7
Figure 2.1: Schematic figure for system and channel model
three network nodes users, relay stations, and base stations. All the users, relays, base stations in the system are collected in following index sets.
u ∈ U = {1, . . . , U }. r ∈ R = {1, . . . , R}. b ∈ B = {1, . . . , B}.
We consider k connections between base stations and users.
k ∈ K = {1, . . . , K}.
Connections are defined in terms of logical and physical connections. For every connection k, we define logical connection(b,u) between base station b and user u. Physical layer connection k ∈ K is defined as
ck= (b(k), R(k), u(k))
where, b(k) is the base station involved in connection k. R(k) ⊆ R as the set of relays used by connection k. u(k) is the user active in connection k.
8 CHAPTER 2. TWO HOP RELAY NETWORKS
In general, there may be multiple physical connections that corresponds to a logical connection (b,u).
2.2.1
Block fading channel model
We assume frequency flat channels experiencing block fading so that the channels are random but remain constant for entire duration of super frame. Super frame designates the duration of time for which the current resource allocation decision is made and for which the channels are assumed to be con-stant. A super frame is further divided into multiple time slots corresponding to resource allocation blocks assigned for various transmissions (e.g. Phase I and II transmissions for different connections). Unless it is stated otherwise, the channels are drawn from circular-symmetric unit variance gaussian distri-bution. Furthermore, channel realizations in two different super frames are statistically independent.
The channel from base station b to relay r is denoted as hHr,b ∈ C1×Nb.
Similarly, channel from relay r to user u is denoted as gu,r∈ CNu×1.
2.2.2
Resource allocation
As mentioned above, we consider time separation between the two phases of transmission. Therefore, a super frame is subdivided into T time slots to have time separation between the two phase transmissions for all active connections. In our assumption phase I and phase II transmission cannot take place simultaneously. A super frame is subdivided into at least two time slots (T ≥ 2). The exact super frame division depends on the transmission strategy (e.g. TDMA, SDMA or Hybrid ) is being used. They are described in more detail in Chapter 3.
Let τ (t) denotes the relative duration of time slots t, where t ∈ T = {1, . . . , T } and
T
X
t=1
τ (t) = 1, (2.1)
i.e, we normalize the duration of a super frame to 1. For each connection k, we have time slot tk that is divided into two times slots.
tk = t
(1)
k + t
(2)
k (2.2)
t(1)k is the duration of the time slot where phase I transmission takes place for connection k. Similarly, t(2)k is the duration of the time slot where phase II transmission takes place for connection k.
2.2. CHANNEL AND RATE MODEL 9
All connections that are active during a super frame are denoted by A. Furthermore, the connections that are active during time slot t are denoted as At.
We hve only analysed downlink transmission therefore we define the two phases with respect to downlink channels in the next two subsections.
2.2.3
Phase I
Phase I between base stations and relay stations (Bs → Rs). Phase I channel is modelled as an interfering broadcast channel (IBC) [8]. There can be two different relay connection configurations. First as the single relay in which only one relay is used in a connection and second as cooperative relays in which a subset of relays are involved in transmission.
In single relay case, the received signal at a relay r is given by
yr(1) = B X b=1 hHr,bx(1)b + n(1)r Where n(1) r ∼ CN (0, σr2).
In case of cooperative relays a subset of relays R(k) cooperates with each other to improve the network performance. The signal received at relays R(k) in the system is:
y(1)R(k)= B X b=1 HR(k),bx(1)b + n(1)R(k) (2.3) and we define • HR(k),b = h hHr,bi
r∈R(k) is the channel from the base station b to relays
R(k) (vertical concatenation of row vectors).
• The signal transmitted by base station b as x(1)b ∈ CNb × 1.
• n(1)R(k) is the additive white Gaussian noise where n(1)R(k) =hn(1)
r
i
r∈R(k).
The signal x(1)b transmitted by base station b is formed by linear precoding in a given time slot t,
x(1)b = X k∈At vks (1) k √ pk where, we define
10 CHAPTER 2. TWO HOP RELAY NETWORKS
• The precoding vectors vk ∈ CNb× 1, with kvkk2 = 1 is used by the base
station.
• s(1)k is the data information symbol intended for user u in connection k in phase I .
The transmitted signal must satisfy a power constraint kx(1)b k2
2 ≤ Pmax for
all b = 1, . . . , B.
The signal power is the mean square value of the received signal. The desired signal power in the phase I is
E[|HR(k),bvk √ pks (1) k | 2] = p kHR(k),bvkE h |s(1)k |2ivH kH H R(k),b= pk|HR(k),bvk|2 (2.4) where Eh|s(1)k |2i= 1.
Depending upon connection of relays in cooperation, we may perform some joint linear operation on received signals using a receive filter vector u(1)k at relays to obtain
ZR(k)(1) = u(1)Hk y(1)R(k)
We only consider DF relay forwarding therefore the bit stream is fully decod-able at the relays R(k) in connection k.
In case of single relay transmission is with R(k)= {r(k)} in every connec-tion k ∈ K , then the received SINR at relay staconnec-tion in phase one is given by SINR(1)k = pk|h H R(k),b(k)vk|2 X l∈K l6=k pl|hHr(k),b(l)vl|2+ σr(k)2 . (2.5)
Similarly, in cooperative relays case with R(k) as the group of relays in every connection k ∈ K which perform some joint processing, then the received SINR at relays R(k) in phase one is given by
SINR(1)k = pk|u H kH H R(k),b(k)vk|2 X l∈K l6=k pl|uHkH H R(k),b(l)vl|2+ σ2R(k) . (2.6)
2.2.4
Phase II
Phase II is from relay stations to user equipments (Rs → U e). Phase II is modelled as an interference channel. The signal received by user u(k) from relays R(k) in connection k is given by yu(k)(2) .
yu(k)(2) = X k∈At Gu(k),R(k)x (2) R(k)+ n (2) u ,
2.2. CHANNEL AND RATE MODEL 11
where we define :
• The channel from relays R(k) to a user u as Gu,R(k)=
h
gHu,ri
r∈R(k) for a
connection k .
• The signal transmitted by relays R(k) as
x(2)R(k) = wks
(2)
k
√
qk
where s(2)k are the data information symbols in second phase with |s(2)k |2 =
1.
• The precoder used by relays R(k) as wk ∈ CR(k)×1.
• The transmit power used by relays in a connection k is denoted by qk.
• n(2)
u denotes additive white Gaussian noise with zero mean and unit
variance.
The transmit power used by the relays is power constrained such that
qk≤ Qmax. The receive filter used by user u is tk∈ CNu×1.
The estimate of the desired signal at user u in phase II is given by
ˆ
sk = tHky
(2)
u(k).
The SINR of user u in phase two is defined by
SINR(2)k = qk|t H kGu(k),R(k)wk|2 X l∈K l6=k ql t H kGu(k),R(l)wl 2 + σu(k)2 . (2.7)
2.2.5
End-to-end rate
Effective data rate R(1)k of phase I in a connection k from base station b(k) to relays R(k) is calculated by equation (2.8),
R(1)k = τ (t(1)k ) log(1 + SINR(1)k ), (2.8) Similarly, effective data rate R(2)k of phase II in a connection k from relays
R(k) to user u is calculated in equation (2.9) below,
12 CHAPTER 2. TWO HOP RELAY NETWORKS
where SIN Rk(1) and SIN R(2)k are calculated for time slots t(1)k and t(2)k respec-tively.
The end to end rate Re2e(k) is defined as the end rate that a user u gets
in a connection k. Re2e(k) for a connection k is the minimum of the effective
data rates achieved in phase I and phase II. It is given in equation(2.10)
Re2e(k) = min(R
(1)
k , R
(2)
k ). (2.10)
2.2.6
Time Split ratio between two phases
Due to different transmission environments in two transmission phases, different rates are achieved for each phase. Secondly, we also assume only decode and forward relaying without additional buffering at relay stations. Therefore, the bits received in phase I time slot at relay stations have to be transmitted in phase II in next time slot otherwise data is lost. So in order to overcome this problem we introduce a transmission time split between two phases to optimize the end to end rate. Split ratio between two phases can be symmetrical or adaptive.
• Symmetrical : In symmetrical case, we set by default activation factor to 0.5 of the total transmission time. So both the phases phase I and phase II get equal time to transmit.
• Adaptive : In case of adaptive split ratio we adjust activation factor such that both the phases give equal rate. For example if phase I achiev-able rate is higher than the phase II rate. The lower rate achieved in second phase can be the bottleneck because entire data received at relay in phase I cannot be transmitted in phase II in given time slot. This problem is addressed by the adaptive split ratio. By allocating more time for transmission to lower performance phase, it balances the bits transmitted in both phase. Denoting:
˜
R(1)k = log(1 + SINR(1)k ), (2.11) ˜
R(2)k = log(1 + SINR(2)k ). (2.12) We calculate the optimal slot lengths in TDMA mode for connection k as
2.2. CHANNEL AND RATE MODEL 13
t(2)k = α(2)k × tk, (2.14)
where, α(1)k and α(2)k are adaptive time split factors for connection k calculated as α(1)k = R˜ (2) k ˜ R(1)k + ˜Rk(2), α(2)k = 1 − α(1)k .
Symmetrical can be a case of adaptive split but its hard to say because of the different channel conditions, number of antennas etc. in both the phases.
TDMA strategy with three connection with one base station and three users, Fig 2.2 shows that adaptive split ratio performs in terms of sum end to end rate over SNR variations which is much better than the symmetrical time split ratio. For rest of the transmission strategies,the same improvement was observed. −50 0 5 10 15 20 25 1 2 3 4 5 6 SNR [dB]
mean network spectral efficiency [bits/sec/Hz]
TDMA with fixed relay assignment TDMA with rate maximization
Chapter 3
Transmission Strategies for Two-Hop
Relay Networks
In this section, different set of transmission strategies are defined in dif-ferent relay connection settings. Following assumptions are made for all the transmission strategies as stated below:
1. Only downlink transmission is analysed. Phase I from Bs → Rs and Phase II from Rs → U e.
2. Half duplex transmission constraint is assumed i.e, relay stations cannot transmit and receive at the same time.
3. Each base station has Nb antennas.
4. Each relay station has a single antenna (Relay with multiple antenna are simulated though cooperative relays).
5. DF relay forwarding is used.
6. No additional buffering at relay stations is assumed.
3.1
Baseline 1: Single relay
In single relay connections, the base station selects one relay (i.e.|R(k)| = 1) in a connection k based upon the relay selection method to transmit one stream per user. Single relay transmission strategies are summarized in Table 3.1 and described in more detail below.
16 CHAPTER 3. TRANSMISSION STRATEGIES FOR TWO-HOP RELAY NETWORKS
Table 3.1: Single Relay
ID Description 1a TDMA;
Phase I: MRT; full power
Phase II: point–to–point SIMO; full power; MRC 1b SDMA fixed assignment
Phase I: ZF; uniform power
Phase II: SIMO interference channel; full power; MRC 1c SDMA with relay selection based upon perfect CSI
Phase I: ZF; uniform power
Phase II: SIMO interference channel; full power; MRC 1d Hybrid SDMA-TDMA with with fixed relay assignment
Phase I: ZF; uniform power
Phase II: TDMA, full power; MRC
3.1.1
TDMA
In TDMA transmission strategy, only one network entity transmits in one time slot so transmissions take place on orthogonal resources and in sequential manner. Super frame time is divided into times slots. In TDMA transmission strategy, the total number of times slots in which a super frame is divided into, is equal to twice the number of connections. Two phases in a connection
k have differently assigned time slots for transmission. Phase I transmissions
take place in time slot t(1)k ,
t(1)k = 2k − 1.
Similarly, phase II transmissions take place in time slot t(2)k ,
t(2)k = 2k.
In TDMA strategy, the active base station and the relay station transmits with full power. For example in the case of a system with one base station, three relays and three users as shown in Figure 3.1. The transmission will take place in six phases in a sequence shown in Figure 3.2.
The relay assignment in case of TDMA strategies are either fixed or rate maximization.
Fixed relay assignment : In fixed relay assignment, relays are preassigned
according fixed connection order to every connection without taking channel realization into consideration. For example we have three active
3.1. BASELINE 1: SINGLE RELAY 17
Figure 3.1: TDMA
Figure 3.2: TDMA sequential transmission for 3 connections
connections and we assign relays having identities [1, 2, 3] to connection number [1, 2, 3] respectively.
Rate maximization based relay selection : In this relay selection method,
relay R∗(k) is selected based upon maximization of end to end rate for each connection k,
R∗(k) = {r∗(k)} = arg max
r∈R
(Re2e(k)) (3.1)
For TDMA strategy with three connections, one base station and three users, Figure 3.3 shows that relay selection by rate maximization performs much better than fixed relay assignment in terms of sum end to end rate.
Phase I Precoding Vector
In single relay scenario, the maximum effective channel gain can be achieved by maximum ratio transmission precoder. It is calculated by
18 CHAPTER 3. TRANSMISSION STRATEGIES FOR TWO-HOP RELAY NETWORKS −50 0 5 10 15 20 25 1 2 3 4 5 SNR [dB]
mean network spectral efficiency [bits/sec/Hz]
TDMA with fixed relay assignment
TDMA with rate maximization based relay selection
Figure 3.3: Comparison between rate maximization relay selection and fixed assignment
signal divided by its norm. The MRT precoding vector is given by
vk =
hr(k),b
khr(k),bk2
. (3.2)
Base station transmits with maximum power pk = Pmax.
Phase II Precoding Vector
If user u has Nu > 1 receive antennas the receive filter tk is given by
maximum ratio combining (MRC)
tk =
gu(k),r(k)
kgu(k),r(k)k2
, (3.3)
where Gu(k),R(k) is the phase II channel gain between relays R(k) and
user u(k).
wR(k) is the precoder used by a group of relays. In case of fixed relays
we have wR(k) = 1.
The relays R(k) forwards the data to user u(k) using full power qR(k)=
Qmax.
3.1.2
SDMA with fixed relay assignment
In SDMA transmission strategy, the super frame is always divided into two time slots irrespective of the number of active users. In the first time slot,
τ (1), phase I transmission from all active base stations to all active relays takes
place and in second time slot, τ (2), relays transmit to users. In fixed relay assignment, relays are preassigned to every connection according to a fixed
3.1. BASELINE 1: SINGLE RELAY 19
connection order without taking channel realization into consideration. For example we have four active connections and we assign relays having identities [1, 2, 3, 4] to connection number [1, 2, 3, 4] respectively.
Figure 3.4: SDMA with single fixed relay
Phase I Precoding Vector
In phase I zero forcing beamforming is used to minimize the interference in phase I. ZF precoder are selected such that they are orthogonal to channel realizations other than the channel between active base station and the assigned relay station. For example for the user u(k) in con-nection k, the ZF precoder vector is chosen at base station b(k) such that
vk⊥uHl HR(l),b, for all l ∈ K with l 6= k, (3.4)
Phase II Precoding Vector
If user u has Nu > 1 receive antennas the receive filter tk is given by
20 CHAPTER 3. TRANSMISSION STRATEGIES FOR TWO-HOP RELAY NETWORKS
Figure 3.5: SDMA with relay selection
3.1.3
SDMA with relay selection
Relay selection methods
There are a number of relay available in two hop network to assist the base station by retransmitting the information either replicating in a better way or by processing it.One of the important factors that effects the rate performance of two hop relay networks is the relay selection. Therefore, relay selection methods are essential in order to achieve maximum data rate gains. [9].
All relay selection methods that are described in this subsection depend upon the channels state information feedback except fixed relay selection method. In our simulations we assumed perfect channel state feedback.As the transmissions consist of two phases so channels of both phase HR(k),b(k)
and Gu(k),R(k) affects the performance of the system [10]. In the described
re-lay selection methods, we take channel state of both phases into consideration unless it is stated otherwise.
In this section we propose different relay selection method based upon different selection criteria for SDMA transmission strategy. We try to optimize relay assignment such that either the sum rate is maximized or the interference
3.1. BASELINE 1: SINGLE RELAY 21
is minimized.
Channel norm In this approach, the relays selection criteria is based upon
maximization of the minimum of channel norms between the phase I and phase II. The channel norm based relay selection method works in the following steps :
Algorithm 1 Relay assignment based upon channel norm
For a given number of connections K.
Objective: To to find r∗(k) for each k ∈ K so that R∗(k) = {r∗(k)} based upon channel norms.
Initialization : Ra = R = {1, . . . , R}
for k ∈ K ( all active connection A in a super frame) do
for r(k) ∈ Ra (all available relay stations in the system) do
r(k)∗ = arg max r(k)∈Ra minnkhr,bk22, kgu(k),rk 2 2 o (3.5) end for
Relay r∗(k) is assigned to the connection k and it is made unavailable for the next connection,
Ra= Ra− {r∗(k)}
end for
First short coming of channel norm based relay selection is that it does not take into consideration interference conditions of the phases. Sec-ondly, the initial active connection is allocated the best possible relay and it is unavailable for further connections.
Brute force approach In brute approach, we check for every possible relay
and user combination which can give best sum of end to end data rates
Re2e(k) over all active connections K over all possible connection
com-binations. Brute force relay selection algorithm works in the following steps:
In brute force approach, the number of active connections can be fixed or varied depending upon the data rate criteria. Active connection number is chosen based upon how many connections give the maximum data rate. Brute force gives the best possible solution as it checks all the possible combination but it comes at the cost of extensive relays selection
22 CHAPTER 3. TRANSMISSION STRATEGIES FOR TWO-HOP RELAY NETWORKS
Algorithm 2 Brute Force Approach with all possible combinations
Initialization: Rates = 0
Calculate all possible relay combinations without repetition RCkfor also all
possible user combination such that user u(k) is active out of total users U .
for k = 1, . . . , K (all possible active connection) do for number of possible user combination for k do
for RCk do
Calculate sum rate (Rates) as a sum over end to end data rates for all
active connections k for corresponding possible relay combination and user combination.
if current Rates> previous maximum Rates then
The active connection number k, relay and user combination that give maximum Rates is selected for that transmission.
r(k)∗ = arg max AllRsandU ecombinations (Rates) end if end for end for end for
search with R choose k without repetition for all possible relay and user combinations.
Complexity (c) is in terms so iterations search that brute force relay selection goes though for relay selection.
For varying number of active connections A ≤ K complexity (c) is given by c = A Y k=1 ( A! (A − k)!k!) ∗ ( R! (R − k)!) (3.6)
where A denotes the maximum active connections. R number of relays are available in the system. First part of the equation 3.6 A!
(A−k)!k! is
for checking number of active connections 1 ˙A that gives the best sum
end-to-end rate. Second part, (R−k)!R! is for choosing k relays out of total number of relays R. Where k = A For any fixed number of active
3.1. BASELINE 1: SINGLE RELAY 23
connections complexity is given by
c = ( A!
(A − k)!k!) ∗ (
R!
(R − k)!) (3.7)
Greedy approach Greedy relay selection approach is similar to brute force.
In greedy approach, we increase the number of active connection in stepwise manner.
Algorithm 3 Brute Force Approach with all possible combinations for k = 1, . . . , K (all possible active connection) do
for r = 1, . . . .R do
Check which available relay gives the maximum Ratesfor a given connection
k. The relay which gives the maximum rate for a connection k is selected
and is made unavailable for further connections.
end for
if currentRates > previousmaximumRates then
number of connections, connection number and assigned relay is stored.
end if end for
In greedy approach, we select the relay for every connection based upon sum rate criteria and check how many active users can give us the max-imum rate. The complexity for greedy approach relay selection is given by c = K+1 X i=0 (u − i) ∗ (R − i) (3.8)
where u is the number of user.
Hungarian approach In Hungarian relay selection approach relays
combi-nation is selected such that interference is minimized in the all the active connections based upon the Hungarian algorithm. Hungarian method assign A relays out of R for U users such that interference matrix(cost matrix) is minimizes. How Hungarian assignment works and it was implemented according to [11] [12]. Interference matrix is based upon Phase II that between relay and user equipment without taking phase I into consideration. Interference matrix is calculations are based upon
24 CHAPTER 3. TRANSMISSION STRATEGIES FOR TWO-HOP RELAY NETWORKS
interference caused to a user u by other users. Main advantage of Hun-garian method is that its complexity (c) of search is always
c = R4 (3.9)
Hungarian assignment algorithm have one challenging problem, how to incooperate phase I along with phase II for relay selection decision.
Figure 3.6 shows if we have fixed number of active connections equal to 3, brute force method performs the best because it checks all possible combinations. But we observe Hungarian method is also pretty close to brute force in sum rate performance and it has low complexity also for greater number of relays and number of active connections. Figure 3.6 shows if we set varying number of active connections, greedy or brute force perform better because less number of user are active. Because if next newly added connection they have a condition to check whether newly added connection increases the system performance or not.
Figure 3.7 shows the comparison between all above mentioned relay se-lection methods for varying number of connection from 1 to 3. Only in brute force and greedy approach, we can have varying number of connections so still brute force perform better than the rest.
−50 0 5 10 15 20 25 1 2 3 4 5 SNR [dB]
mean network spectral efficiency [bits/sec/Hz]
nBsAnt=8, nRs=8, nRsAnt=1, nConnections=3, nMs=3 relaySelectionSDMA=fixed;
relaySelectionSDMA=channel norm; relaySelectionSDMA=Optimal Brute Force;
relaySelectionSDMA=Optimal with Hungarian algorithm;
Figure 3.6: Comparison between relay selection methods for SDMA with fixed number of connections
3.1. BASELINE 1: SINGLE RELAY 25 −50 0 5 10 15 20 25 1 2 3 4 5 SNR [dB]
mean network spectral efficiency [bits/sec/Hz]
SDMA relaySelection =fixed; SDMA relaySelection=channel norm; SDMA relaySelection=greedy;
SDMA relaySelection=Optimal Brute Force; SDMA relaySelection= Hungarian algorithm;
Figure 3.7: Comparison between relay selection methods for SDMA with vary-ing connections
3.1.4
SDMA-TDMA (Hybrid) with fixed relay assignment
In SDMA and TDMA transmission strategies we have some issues that effect the system performance. In SDMA strategy in phase II user experience strong interference from other active relays due to single antennas at the relays. In case of TDMA we need to subdivide the super frame time into twice the number of active user time slots and available transmission time for every connection is quite short. So in this SDMA-TDMA strategy, we try to combine the benefits of both strategies as by using SDMA with ZF precoder at Bs with multiple antennas in phase one which almost nullifies interference. In second phase interference is avoided by using TDMA strategy. Here in Hybrid SDMA-TDMA case, a super frame is divided into k + 1 time slots for k connections. Phase I is always assigned the time slot number 1,
t(1)k = 1. (3.10)
In phase 2 we have a sequential slot allocation,
26 CHAPTER 3. TRANSMISSION STRATEGIES FOR TWO-HOP RELAY NETWORKS
Figure 3.8: Hybrid strategy
Phase I Precoding Vector
Uses ZF beamforming as described in phase I of Section 3.1.2.
Phase II Precoding Vector
Users are sequentially processed as described in Section 3.1.1 .
Figure 3.9 depicts a comparison between transmission strategies for sin-gle relay: sum rate over SNR variation for fix number of relays nRs= 8 and only 2 are assigned according to relay selection strategies. We can observe that SDMA gives much better sum rate than others in low SNR to average SNRs because SDMA is not in interference limited region upto 15 dB. How-ever, TDMA performs much better in high SNR regions because of the high interference in the SDMA case. Figure 3.10 shows that as number of relays is increased all strategies sum rate increases but SDMA gives a much higher gain in sum rate than TDMA. Hybrid strategy with SDMA in first phase and TDMA in second perform almost the same as TDMA, which indicates that second phase is the limiting factor.
3.1. BASELINE 1: SINGLE RELAY 27 −50 0 5 10 15 20 25 1 2 3 4 5 6 SNR [dB]
mean network spectral efficiency [bits/sec/Hz]
Hybrid with fixed relay assignment
TDMA with rate maximization based relay selection SDMA with brute force relay selection
Hybrid with fixed relay assignment
Figure 3.9: Sum rate over SNR variation for single relay strategies
3 3.5 4 4.5 5 5.5 6 6.5 7 7.5 8 1.5 2 2.5 3 3.5 4 number of relays
mean network spectral efficiency [bits/sec/Hz]
TDMA with fixed relay assignment
TDMA with rate maximization relay selection SDMA with brute force relay selection Hybrid with fixed relay assignment
28 CHAPTER 3. TRANSMISSION STRATEGIES FOR TWO-HOP RELAY NETWORKS
3.2
Baseline 2: Cooperative relays
In cooperative relay connections, we assume a subset of relays are con-nected via error and delay free backhaul network. Each base station transmits one data stream per user to a subset of relays. Scenario for multiple antennas is simulated thorough cooperative relaying because of the limitation of number of antennas deployment at relay as the number of user increases. Secondly, we get uncorrelated propagation path to provide high diversity gains. Table 3.2 summarizes the transmit strategies which are described in more details below.
Table 3.2: Cooperating Relays
ID Description 2a TDMA/FDMA;
Phase I: point–to–point MIMO; SVD precoding; full power Phase II: point–to–point MIMO; SVD precoding; full power 2b SDMA with fixed relay assignment
Phase I: ZF; uniform power
Phase II: MIMO interference channel; full power; SVD precoding 2c SDMA with relay selection based on perfect CSIT
Phase I: ZF; uniform power
Phase II: MIMO interference channel; full power; SVD precoding 2d Hybrid SDMA-TDMA with fixed relay assignment
Phase I: ZF; uniform power
Phase II: TDMA, full power; SVD precoding
3.2.1
TDMA
The TDMA strategy is similar to single relay but relays are assumed to cooperate. Transmission to each user via a relay in two phase takes place on orthogonal resources. All active users are sequentially processed. The relay selection is based upon the channel norm as given in (3.5)
Phase I Precoding Vector
Point–to–point MIMO with SVD precoding/filtering,
HR(k),b(k) = U(1)k Σ
(1)
k V H(1)
k (3.12)
Precoder or the transmit filter used by base station b(k) for user u are
uk = [U
(1)
3.2. BASELINE 2: COOPERATIVE RELAYS 29
i.e, the first column of U(1)k is selected. The receive filter used by relay R(k)
vk= [V
(1)
k ]1. (3.14)
Hence, the effective channel gain between base station b(k) and relays R(k) is |uH
R(k)HR(k),bvk|2 = |σ1|2, where σ1 is the largest singular value
of the channel matrix HR(k),b.
Phase II Precoding Vector
Point–to–point MIMO with SVD precoding/filtering. That is, the re-ceive filter tk and precoder wk used by user u and relay group R(k) in
connection k. Gu,R(k)= U (2) k Σ (2) k V (2)H k tk= [U (2) k ]1 (3.15) wk= [V (2) k ]1. (3.16)
3.2.2
SDMA with fixed relay assignment
The relays to connections are assigned independently of the channel re-alization, i.e., the mapping R(k) is fixed a priori for all k = 1, . . . , K. The receive filters at the groups of cooperating relays are given by unit vectors. That is, uR(k) = ei for all R(k) and some integer 1 ≤ i ≤ |Rr|, where ei is
the vector with all zeros and a single 1 at the i–th position. The receive filters are fixed a priori and chosen such that kuH
kHR(k),b(k)k is maximized.
Phase I Precoding Vector
Each base station uses zero forcing beamforming according to (3.4). The available power at the base stations is distributed uniformly, i.e.,
pk= P|Umax
b|.
Phase II Precoding Vector
The receive filter tkand precoder wkused by user u and relay group R(k)
are given by SVD precoding/filtering according to (3.16) and (3.15). The active relays R(k) retransmits the data, received from base station b(k) to user u after some processing using full power qR(k) = Qmax for all
R(k).
3.2.3
SDMA with relay selection
Here the relay group R(k) for a user k are selected for active connection based upon one of relay selection methods in Section 3.1.3 such that either sum
30 CHAPTER 3. TRANSMISSION STRATEGIES FOR TWO-HOP RELAY NETWORKS
rate is maximized or interference is minimized. In cooperative case, multiple relays are selected to process and forward the received information but in real world we have problems of transmission synchronization from multiple relays. Here, we assume delay free backhaul connection between relays.
Phase I Precoding Vector
Each base station uses zero forcing beamforming according to (3.4). The available power at the base stations is distributed uniformly, i.e.,
pk= P|Umax
b|.
Phase II Precoding Vector
The active relays R forward the data to user u using full power qR(k)=
Qmax for all R(k). The receive filter tk and precoder wk used by user
u and relay group R(k) are given by SVD precoding/filtering according
to (3.16) and (3.15).
3.2.4
Hybrid SDMA-TDMA with with fixed relay
assignment
It is a combination of of SDMA in phase I and TDMA in phase II with fixed relay assignments.
Phase I Precoding Vector
Uses ZF beamforming as described in phase I of Section. 3.1.2.
Phase II Precoding Vector
It uses SVD precoding with group of relays transmitting with full power.
Figure 3.11 depicts the sum rate performance for cooperating relay con-nections. We observe that the SDMA strategy performs better than all others because multiple cooperating relay to form a multiple antenna scenario, which allows to cancel the interference in phase two.
3.3
Baseline 3: Multicast
With multicast strategies, we try to see the effect when relays are not con-nected though backhaul network. Instead base station multicast the data of user u(k) to all the relays R(k). Multicast is expected to be useful in scenar-ios when base station have no channel information about phase II channels
3.3. BASELINE 3: MULTICAST 31 −50 0 5 10 15 20 25 1 2 3 4 5 6 7 8 9 SNR [dB]
mean network spectral efficiency [bits/sec/Hz]
TDMA with fix relay assignment
TDMA with rate maximization based relay selection SDMA with brute force based relay selection Hybrid with fixed relay assignment
Figure 3.11: Cooperative relay strategies
between relays and users. The transmission rate in the first phase is thus determined by the weakest connection
Ck(1) = τ (t(1)k ) min
r(k)∈R(k) log(1 + SINR
(1)
k,r)
Where SINR(1)k,r designates the SINR in phase I of a connection k at relay
r(k) and is given by SINR(1)u(k),r(k)= pk|h H r(k),b(k)vk|2 X l∈K l6=k pl|hHr(k),b(l)vl|2+ σ2r(k) (3.17)
Table 3.3 summarizes the transmit strategies which are described in more details below.
3.3.1
MF Multicast beamforming, single relay selection
In this baseline scheme, the BS multicasts the data for user u(k) to all relays R(k). In the second phase, all relays have the same data and the best one is chosen to transmit to user u.
• Phase II transmissions to each user u is performed on orthogonal re-sources just like TDMA. The UEs are processed sequentially.
• In each transmission from base station b to user u, all available relays are selected as R(k) = R.
32 CHAPTER 3. TRANSMISSION STRATEGIES FOR TWO-HOP RELAY NETWORKS
Table 3.3: Multicast
ID Description
3a Matched filter; Best single relay
Phase I: MF multicast beamforming; full power Phase II: Single relay; MRC; full power
3b Max-min; Best single relay
Phase I: Max-min multicast Tx beamforming; full power Phase II: Single relay; MRC; full power
3c Matched filter; Cooperative transmission
Phase I: MF multicast beamforming; full power
Phase II: Virtual Tx MIMO; full power; SVD precoding 3d Max-min; Cooperative transmission
Phase I: Max-min multicast Tx beamforming; full power Phase II: Virtual Tx MIMO; full power; SVD precoding
• In each transmission the active base station and the active relays trans-mit with full power.
Phase I Precoding Vector
Base station b(k) transmits the data for user u(k) to relays R(k) using the matched filter (MF) approach. Specifically, the transmit filter vk is
chosen as the eigenvector corresponding the greatest eigenvalue of matrix
HHR(k),b(k)HR(k),b(k).
Phase II Precoding Vector
In the second phase, the best available relay (i.e. with the strongest channel) is selected: r(k)∗ = arg max
r=1,...R
kgu(k),R(k)k2
2 and R(k)
∗ = {r(k)∗}.
3.3.2
Max-min beamforming, single relay selection
In this strategy, the multicast precoder in the first phase is chosen accord-ing the max-min SNR approach. Here we try maximize the minimum received SNR at users by finding optimal precoder in phase I, which is however com-putationally more intensive.
Phase I Precoding Vector
Base station b(k) transmits the data for user u(k) to relays R(k) using the max-min problem formulation:
3.3. BASELINE 3: MULTICAST 33 vk= arg max vk min r∈R(k)SN IR (1) r,k = arg max vk min r∈R(k)h H r,b(k)vkvHkhr,b(k)
We try to selected beam forming such that it maximize the minimum pre-scribed SNR over all relays, which is bounded by the maximum transmit power at the base stations. It is a convex optimization problem solved by semi definite programming. In our matlab implementation max min problem was solved by cvx toolbox [13].
Phase II Precoding Vector
If user u(k) has Nu > 1 receive antennas the receive filter tk is given by
maximum ratio combining (MRC)
tk=
Gu,R(k)wR(k)
kGu,R(k)wR(k)k2
, (3.18)
where Gu,R(k) is the phase II channel between relays R(k) and user u.
wR(k)is the precoder used by a group of relays and in case of fixed relays
wR(k) = 1.
The relays R(k) forwards the data to user u using full power qR(k) =
Qmax.
3.3.3
MF Multicast beamforming, cooperative
transmission
This strategy assumes that the relays are capable of forming a virtual transmit array by coordination between relays by the corresponding user to which they transmit.
Phase I Precoding Vector
The transmission in Phase I is performed as described in Section 3.3.1.
Phase II Precoding Vector
The relays form a virtual transmit MIMO array. On the resulting point-to point MIMO channel from the relay array point-to the UE, SVD precoding according to (3.15) and (3.16) is performed.
3.3.4
Max-min beamforming, cooperative transmission
This strategy combines the elements of Section 3.3.2 (Phase I, Max-min) and Section 3.3.3 (Phase II,Virtual MIMO).
Figure 3.12 shows that max-min beamforming perform much better than matched filter based approach. Furthermore, as expected multiple relays in
34 CHAPTER 3. TRANSMISSION STRATEGIES FOR TWO-HOP RELAY NETWORKS
virtual transmit array perform better than single best relay. Therefore the max-min approach with cooperative transmission gives the best performance.
−50 0 5 10 15 20 25 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 SNR [dB]
mean network spectral efficiency [bits/sec/Hz]
Multicast; phase1TxBeamforming=MaxMin; rsVirtualArrayMode=cooperative; Multicast; phase1TxBeamforming=MaxMin; rsVirtualArrayMode=best relay; Multicast; phase1TxBeamforming=Matched Filter; rsVirtualArrayMode=cooperative; Multicast; phase1TxBeamforming=Matched Filter; rsVirtualArrayMode=best relay;
Chapter 4
Simulation Scenarios with Results and
Analysis
In this chapter, we present the results of different simulations that are performed using NetSim tool. This tool was developed by Jan Schreck and Michal Kaliszan and subsequently adapted and refined during the present master thesis project.
In different simulations, we try to investigate relay based network with single base station with eight antennas and eight relay stations available with single antenna per relay and relay configurations in terms of single relay or cooperative relays, number of users and number of antennas at the users for different strategies or changing some parameters. Performance of different system settings is evaluated by sum end to end rate Rates. In these
simula-tion results, split ratio was set to symmetrical in the system in which we try to observe the same strategy with different parameters.As split ratio affects the strategy in the same manner. But in case where different strategies are compared adaptive split is used.
Figure 4.1 depicts performance of a TDMA strategy with SNR variation as number of users are increased from 1 to 4. We observe that sum end to end rate remains almost the same as the number of users is increased. Figure 4.2 depicts that SDMA with single relay strategy with SNR variation as number of users is increased from 1 to 4. We observe that Rates increases in the low to
average SNR region as the number of users is increased. In high SNR region
Rates decreases as it is in the interference limited region.
This performance limiting factor due to interference can be overcome by either deploying multiple antennas at the relay stations or at the users. Figure 4.3 shows that by increasing number of antennas at the user end the perfor-mance in the high SNR region improves significantly. This is because SDMA system switches the interference limited system to be not in the interference
36 CHAPTER 4. SIMULATION SCENARIOS WITH RESULTS AND ANALYSIS −5 0 5 10 15 20 25 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 SNR [dB]
mean network spectral efficiency [bits/sec/Hz]
TDMA with symmetric split and 8 Rs
1 Ue; 2 Ue 4 Ue
Figure 4.1: TDMA with variation in number of users
−50 0 5 10 15 20 25 1 2 3 4 5 SNR [dB]
mean network spectral efficiency [bits/sec/Hz]
Multiple Bs Ants=8; Single antenna at Rs and Ue SDMA DF; Ue=1;nRs=8
SDMA DF; Ue=2; nRs= 8 SDMA DF;Ue=4;nRs=8
Figure 4.2: SDMA in single relay configuration with variation in number of users
limited region.
Figure 4.4 shows that having the same number of multiple antennas trans-mit antennas at relay station give the same results as having the same number of multiple receiving antennas at the user.
Figure ?? depicts that if phase two channel information is not available then multicast outperforms the rest of the transmission strategies.
37 −5 0 5 10 15 20 25 0 1 2 3 4 5 6 7 8 SNR [dB]
mean network spectral efficiency [bits/sec/Hz]
SDMA DF; nUe=1; nMsAnts=1 SDMA DF; nUe=2; nMsAnts=1 SDMA DF; nUe=2; nMsAnt=2
Figure 4.3: SDMA single relay connection with multiple antennas at the users
−50 0 5 10 15 20 25 1 2 3 4 5 6 7 8 SNR [dB]
mean network spectral efficiency [bits/sec/Hz]
nConnections=2
SDMA;Multiple antennas at Rs SDMA; Multiple antenna at users
Figure 4.4: Multiple antennas at relays vs multiple antennas at the users
Figure 4.6 shows that in low and average SNR region TDMA, multicast and SDMA with single relay perform almost the same and better than SDMA with cooperative relays. In high SNR region SDMA with cooperative relay connection perform much better than the rest.
38 CHAPTER 4. SIMULATION SCENARIOS WITH RESULTS AND ANALYSIS −50 0 5 10 15 20 25 1 2 3 4 5 6 SNR [dB]
mean network spectral efficiency [bits/sec/Hz]
TDMA with relay selection based upon channel 1 SDMA with relay slection based upon phase 1 channel Multicast with max min and best relay
Figure 4.5: Phase 1 Multicast comparison with TDMA and SDMA
−50 0 5 10 15 20 25 1 2 3 4 5 6 7 8 9 SNR [dB]
mean network spectral efficiency [bits/sec/Hz]
SDMA with single relay TDMA with adaptive split COoperative SDMA
Multicast with max min and best relay
Chapter 5
Conclusion and reflections
5.1
Conclusion
In this thesis, we have formulated various transmission strategies, i.e, TDMA, SDMA, Hybrid for two hop relay based networks for different re-lay configurations like single rere-lay, cooperative and multicast. We proposed to improve system performance in terms of sum end to end rate by optimizing resource allocation between two phases, i.e, phase 1 and phase 2, and applying beamforming vectors at transmitters as well as at receivers with interference taken into account properly or to maximize the received signal at the receiver. Based upon our simulations results, it is concluded that performance of the system can be improved by applying adaptive split ratio between the two phases. Secondly, in SDMA transmission strategy, brute force relay selection method assigns the optimal relay combination after checking all the possible combinations. Relay assignment based on Hungarian algorithm also perform pretty close to brute force method.
It is also observed that as the number of relays increases, the system per-formance for all the strategies is improved. Also, for a system using SDMA, with single relay station and with single antenna at user end, which operates in interference limited region (at high SNR) so it is better to use cooperative relays with multiple antennas. If phase 2 channel information is not available at base stations then multicast performs better then other strategies. In the case that both channel information are available then TDMA, multicast, and SDMA strategies with single relay configuration give better performance at low/average SNR region but cooperative relay configuration performs much better at high SNR.
40 CHAPTER 5. CONCLUSION AND REFLECTIONS
5.2
Reflections
This master thesis was part of FundNet project at the Technical University of Berlin. It was mainly focussed on two hop relay based networks in which we tried to formulate and optimize transmission strategies for the two phases.
5.2.1
Economic impact
One of the main motivation of deploying two hop relay based network is to reduce network operational and maintenance cost by deploying relay stations which are connected via wireless channel to the base stations instead of physical backhaul of optical fibre or microwave link. Therefore, they are of great significance from the network operator/vendor point of view. Users do not get direct economic benefits as they are already paying low prices for data services.
5.2.2
Environmental impact
From environmental point of view, two hop relay based network use low power nodes (relay station). Also they are small in size, so it does not distort landscape of the city with big towers everywhere. One of the major benefits of relays is that they can be self-powered, for example, they can operate on solar panels as they consume less energy.
5.2.3
Social aspect
Eventually the society benefits from the high data rates achieved via two hop relay based networks. This can enable high speed applications on different devices used in different aspects of our daily lives, from basic high quality communication to business, entertainment and all we can think of that can affect our life style.
Bibliography
[1] J. Hoydis, M. Kobayashi, and M. Debbah, “Green small-cell networks,”
Vehicular Technology Magazine, IEEE, vol. 6, no. 1, pp. 37–43, 2011.
[2] “Cisco visual networking index: Global mobile data traffic forecast up-date,” 2012.
[3] R. Pabst, B. H. Walke, D. Schultz, P. Herhold, H. Yanikomeroglu, S. Mukherjee, H. Viswanathan, M. Lott, W. Zirwas, M. Dohler, H. Agh-vami, D. Falconer, and G. Fettweis, “Relay-based deployment concepts for wireless and mobile broadband radio,” Communications Magazine,
IEEE, vol. 42, no. 9, pp. 80–89, Sept.
[4] B. Stiller, T. Bocek, F. Hecht, G. Machado, P. Racz, and M. Waldburger, “Heterogeneous networks-Increasing Cellular Capacity,” tech. rep., 2011.
[5] K. Loa, C.-C. Wu, S.-T. Sheu, Y. Yuan, M. Chion, D. Huo, and L. Xu, “Imt-advanced relay standards [wimax/lte update],” Communications
Magazine, IEEE, vol. 48, no. 8, pp. 40–48, 2010.
[6] J. Zhao, “Analysis and design of communication techniques in spectrally efficient wireless relaying systems,” Tech. Rep. Band 12, 2010.
[7] B. Nazer and M. Gastpar, “Compute-and-forward: Harnessing interfer-ence through structured codes,” Information Theory, IEEE Transactions
on, vol. 57, no. 10, pp. 6463–6486, Oct.
[8] C. Suh and D. Tse, “Interference alignment for cellular networks,” in
Communication, Control, and Computing, 2008 46th Annual Allerton Conference on, pp. 1037–1044, 2008.
[9] B. Wang, Z. Han, and K. J. R. Liu, “Distributed relay selection and power control for multiuser cooperative communication networks using buyer/seller game,” in INFOCOM 2007. 26th IEEE International
Con-ference on Computer Communications. IEEE, pp. 544–552, 2007.