systems and computer engineering - carleton university
TRANSCRIPT
Dynamic ICIC
M. Rahman
Introduction
Motivation
Contributions
Literature
Framework
Two-Level Partial
Overview
Central Algo.
Simulation Para.
Results
Distributed
Overview
Inter-eNB
Inter-eNB
Simulation Para.
Results
Clustered Central
Overview
Simulation Para.
Results
Complexity
Conclusions
Future Work
Publications
Dynamic Inter-Cell Interference Coordinationin Cellular OFDMA Networks
Mahmudur Rahman
Thesis Supervisor
Prof. Halim Yanikomeroglu
Department of Systems and Computer EngineeringCarleton University
Ottawa, Canada
July 29, 2011
1
Dynamic ICIC
M. Rahman
Introduction
Motivation
Contributions
Literature
Framework
Two-Level Partial
Overview
Central Algo.
Simulation Para.
Results
Distributed
Overview
Inter-eNB
Inter-eNB
Simulation Para.
Results
Clustered Central
Overview
Simulation Para.
Results
Complexity
Conclusions
Future Work
Publications
1 Introduction
2 Optimization Framework for Dynamic InterferenceCoordination
3 Proposed Two-Level Algorithm with Partial CentralProcessing
4 Proposed Distributed Algorithm with Neighboring CellCoordination
5 Proposed Cluster-based Centralized Scheme
6 Conclusions
7 Future Work
8 List of Publications2
Dynamic ICIC
M. Rahman
Introduction
Motivation
Contributions
Literature
Framework
Two-Level Partial
Overview
Central Algo.
Simulation Para.
Results
Distributed
Overview
Inter-eNB
Inter-eNB
Simulation Para.
Results
Clustered Central
Overview
Simulation Para.
Results
Complexity
Conclusions
Future Work
Publications
Motivation
• 4G and beyond systems adopted Orthogonal FrequencyDivision Multiple Access (OFDMA) air-interfacetechnology
• Dense spectrum reuse is obvious in all cellular systems
• Consequently, high inter-cell interference and itsmitigation are concerns
• Mitigation techniques:• Inter-cell interference cancellation• Inter-cell interference randomization• Inter-cell interference avoidance
• Literature focuses mainly on static or semi-staticschemes
• Our approach: dynamic interference avoidance usingnetwork level coordination
3
Dynamic ICIC
M. Rahman
Introduction
Motivation
Contributions
Literature
Framework
Two-Level Partial
Overview
Central Algo.
Simulation Para.
Results
Distributed
Overview
Inter-eNB
Inter-eNB
Simulation Para.
Results
Clustered Central
Overview
Simulation Para.
Results
Complexity
Conclusions
Future Work
Publications
Motivation (Cont.)
• WINNER (Phase II, 2006-2007) has been an importantmotivating factor
• One of the pioneers to propose dynamic ICIC
4
Dynamic ICIC
M. Rahman
Introduction
Motivation
Contributions
Literature
Framework
Two-Level Partial
Overview
Central Algo.
Simulation Para.
Results
Distributed
Overview
Inter-eNB
Inter-eNB
Simulation Para.
Results
Clustered Central
Overview
Simulation Para.
Results
Complexity
Conclusions
Future Work
Publications
Contributions
• Optimization framework for dynamic interferencecoordination
• Different suboptimal solutions-• Proposed two-level algorithm with partial central
processing• Proposed distributed algorithm with neighboring cell
coordination• Proposed cluster-based centralized scheme
• Evaluation of the proposed schemes using extensivesystem simulations
• The proposed schemes significantly outperform thereference schemes
5
Dynamic ICIC
M. Rahman
Introduction
Motivation
Contributions
Literature
Framework
Two-Level Partial
Overview
Central Algo.
Simulation Para.
Results
Distributed
Overview
Inter-eNB
Inter-eNB
Simulation Para.
Results
Clustered Central
Overview
Simulation Para.
Results
Complexity
Conclusions
Future Work
Publications
Some Interference Coordination in Literature
Reuse 1
• all resources available
• interferers are closer to each other
• worst interference condition and bestreuse opportunity
Reuse 3
• 1/3 resources available
• interferers are further apart
• proactive static interferencecoordination
Fractional Frequency Reuse (FFR)
• part of the resources in the cell centerwith reuse 1
• compromise between the two
• different variants are available
Our proposed schemes are dynamic and opportunistic6
Dynamic ICIC
M. Rahman
Introduction
Motivation
Contributions
Literature
Framework
Two-Level Partial
Overview
Central Algo.
Simulation Para.
Results
Distributed
Overview
Inter-eNB
Inter-eNB
Simulation Para.
Results
Clustered Central
Overview
Simulation Para.
Results
Complexity
Conclusions
Future Work
Publications
Some Interference Coordination in Literature(cont.)
S1
S2
S3
Pow
er
FrequencyS1
S2
S3
c) SFR
S1
S2
S3
Po
we
r
Frequency
S1
S2
S3
d) PFR
S1
Po
wer
Frequency
S2
S3
S1
S2
S3
b) Reuse 3
S1
S2
S3
Pow
er
Frequency
S1
S2
S3
a) Reuse 1
7
Dynamic ICIC
M. Rahman
Introduction
Motivation
Contributions
Literature
Framework
Two-Level Partial
Overview
Central Algo.
Simulation Para.
Results
Distributed
Overview
Inter-eNB
Inter-eNB
Simulation Para.
Results
Clustered Central
Overview
Simulation Para.
Results
Complexity
Conclusions
Future Work
Publications
Optimization Framework
Resource unit → n and user → m
Maximize
∑i
[M∑
m=1
N∑n=1
u(m,n)i ρ
(m,n)i
], (1)
Sector
i
kth
Interfering
sector
jth
Interfering
sector
subject to
ρ(m,n)i ∈ {0, 1}; ∀m and ∀n, (2)
I(n)i =
M∑m=1
ρ(m,n) =
{0; resource n is restricted in i1; otherwise.
(3)
8
Dynamic ICIC
M. Rahman
Introduction
Motivation
Contributions
Literature
Framework
Two-Level Partial
Overview
Central Algo.
Simulation Para.
Results
Distributed
Overview
Inter-eNB
Inter-eNB
Simulation Para.
Results
Clustered Central
Overview
Simulation Para.
Results
Complexity
Conclusions
Future Work
Publications
Optimization Framework (contd.)
γ(m,n)i =
PcH(m,n)i ,i
Pc
K∑k=1
H(m,n)i ,k · I (n)
k + Pc
J∑j=1H(m,n)
i ,j · I (n)j + PTN
,
(4)
Find I(n)k = 0 & I
(n)j = 0; ∀k , ∀j , ∀n that maximizes (1)
• r(m,n)i
BER,AMC← γ(m,n)i and
u(m,n)i = f
[r
(m,n)i , d
(m)i
]• Demand factor:d
(m)i = R̄i/(R
(m)i + δ), where
R(m)i → mth user’s throughput
and R̄i =
(M∑
m=1R
(m)i
)/M →
avg. user throughput
Sector
i
kth
Interfering
sector
jth
Interfering
sector
9
Dynamic ICIC
M. Rahman
Introduction
Motivation
Contributions
Literature
Framework
Two-Level Partial
Overview
Central Algo.
Simulation Para.
Results
Distributed
Overview
Inter-eNB
Inter-eNB
Simulation Para.
Results
Clustered Central
Overview
Simulation Para.
Results
Complexity
Conclusions
Future Work
Publications
Optimization Framework (contd.)
Challenges
• Problem is similar to 3AP which is NP-Complete
• Optimal solution requires centralized algorithm
• Large scale combinatorial optimization
Propose suboptimal solutions considering the following
• Fact: only neighboring interferers are of concern– we
assume, I(n)j = 1
• 4G provides cell-specific orthogonal reference signals;
i.e., determine γ(m,n)
i |I (n)k =0
; ∀k & ∀n
• Cell-edge (rate deprived) users are more prone inter-cellinterference
10
Dynamic ICIC
M. Rahman
Introduction
Motivation
Contributions
Literature
Framework
Two-Level Partial
Overview
Central Algo.
Simulation Para.
Results
Distributed
Overview
Inter-eNB
Inter-eNB
Simulation Para.
Results
Clustered Central
Overview
Simulation Para.
Results
Complexity
Conclusions
Future Work
Publications
Sub-optimal Solutions
1 The proposed two-level algorithm with partial centralprocessing
2 The proposed distributed algorithm with neighboringcell coordination
3 The proposed cluster-based centralized algorithm
11
Dynamic ICIC
M. Rahman
Introduction
Motivation
Contributions
Literature
Framework
Two-Level Partial
Overview
Central Algo.
Simulation Para.
Results
Distributed
Overview
Inter-eNB
Inter-eNB
Simulation Para.
Results
Clustered Central
Overview
Simulation Para.
Results
Complexity
Conclusions
Future Work
Publications
Overview of the Proposed Two-Level Algorithm
• sector i selfishly finds I(n)k = 0; ∀k and ∀n such that[
M∑m=1
N∑n=1
u(m,n)i ρ
(m,n)i
]is maximized
• The above is a 2D assignment problem (solved usingHungarian algorithm)
• prepare utility matrix UM×N =[u(m,n)
]heuristically
• Matrix UM×N reflects heuristically found I(n)k = 0 based
on users’ service status and interference avoidance gains• Hungarian algorithm is applied to UM×N to see if
chosen entries have I(n)k = 0
• Central-level: the 3rd dimension is handled at a centralcontroller
12
Dynamic ICIC
M. Rahman
Introduction
Motivation
Contributions
Literature
Framework
Two-Level Partial
Overview
Central Algo.
Simulation Para.
Results
Distributed
Overview
Inter-eNB
Inter-eNB
Simulation Para.
Results
Clustered Central
Overview
Simulation Para.
Results
Complexity
Conclusions
Future Work
Publications
Central Processing of the Proposed Two-LevelAlgorithm
For chunk n, at the central controllermaximize
Z =∑
i ,j∈Φ
u(mi ,n)i (1− x
(n)i ) + u
(mj ,n)j (1− x
(n)j );
(5)subject to
x(n)i + x
(n)j ≤ 1; (6)
A
B C
• Φ→ sectors that either request or are requested forrestriction
• mi , mj → candidate UTs in sector i and j , respectively
• x(n)i , x
(n)j → binary variables for restrictions
13
Dynamic ICIC
M. Rahman
Introduction
Motivation
Contributions
Literature
Framework
Two-Level Partial
Overview
Central Algo.
Simulation Para.
Results
Distributed
Overview
Inter-eNB
Inter-eNB
Simulation Para.
Results
Clustered Central
Overview
Simulation Para.
Results
Complexity
Conclusions
Future Work
Publications
System and Simulation Parameters
Parameter Assumption
Cellular Layout Hexagonal grid, 19 sites, 3 sectors/siteInter-site Distance 1000 mPath-loss exponent 3.57
Shadowing Independent Log-normal std. 8 dB
Antenna Pattern A (θ) = -min
[12
(θ
θ3dB
)2
, 20
], θ3dB = 70o
Bandwidth@Carrier 45 [email protected] GHzChannel model 20-Tap WINNER Channel
UE speeds of interest 20 km/hrSector TX power 39.81 Watts
Minimum UE Distance 50 mAMC modes BPSK, QPSK, 16- & 64-QAM
with rates 1/2, 2/3 & 3/4
14
Dynamic ICIC
M. Rahman
Introduction
Motivation
Contributions
Literature
Framework
Two-Level Partial
Overview
Central Algo.
Simulation Para.
Results
Distributed
Overview
Inter-eNB
Inter-eNB
Simulation Para.
Results
Clustered Central
Overview
Simulation Para.
Results
Complexity
Conclusions
Future Work
Publications
Some Summarized Results: Two-Level Algorithm
0 100 200 300 400 500 600 700 80040
45
50
55
60
65
70
75
80
85
90
Cell Edge Throughput (kbps)
Sec
tor
Thr
ough
put (
Mbp
s)
Cell−Edge Vs. Sector Throughput: Full−Buffer, Hungarian Scheduler
Prop. 2 with r2TH
CE: ~16X, S: −0.7%
Prop. 1 with r2TH
CE: ~26X, S: −17.5%
Prop. 1 with r1TH
CE: ~17X, S: −12.2%
Prop. 2 with r1TH
CE: ~11X, S: 0%
Reuse 1CE: 29.3 kbps, S: 88.6 Mbps
PFRCE: ~12X, S: −27.4%
Reuse 3CE: ~27X, S: −52.6%
15
Dynamic ICIC
M. Rahman
Introduction
Motivation
Contributions
Literature
Framework
Two-Level Partial
Overview
Central Algo.
Simulation Para.
Results
Distributed
Overview
Inter-eNB
Inter-eNB
Simulation Para.
Results
Clustered Central
Overview
Simulation Para.
Results
Complexity
Conclusions
Future Work
Publications
Summarized Results: Two-Level Algorithm(cont.)
0 100 200 300 400 500 600 700 80040
50
60
70
80
90
100
110
Cell Edge Throughput (kbps)
Sec
tor
Thr
ough
put (
Mbp
s)
Cell−Edge Vs. Sector Throughput: Full−Buffer, PF Scheduler
PFRCE: ~12X, S: −27.6%
Reuse 3CE: ~20X, S: −54.1%
Reuse 1CE: 33.1 kbps, S: 100.2 Mbps
Prop. 1 with r1TH
CE: ~18X, S: −15.5%
Prop. 2 with r1TH
CE: ~12X, S: −1.2%
Prop. 1 with r2TH
CE: ~23X, S: −20.7%
Prop. 2 with r2TH
CE: ~16X, S: −1.8%
16
Dynamic ICIC
M. Rahman
Introduction
Motivation
Contributions
Literature
Framework
Two-Level Partial
Overview
Central Algo.
Simulation Para.
Results
Distributed
Overview
Inter-eNB
Inter-eNB
Simulation Para.
Results
Clustered Central
Overview
Simulation Para.
Results
Complexity
Conclusions
Future Work
Publications
Overview of the Distributed Algorithm
• Developed LTE systems;central processing discouraged
• Centralized processing isreplaced by pair-wise utilitycomparison
• Inter-cell interference iscategorized as intra-eNB andinter-eNB interference
S1
S2
S3
Tri-sector BS antenna
Inter-eNB interference
Intra-eNB interference
• Intra-eNB interference avoidance is done using a novelHungarian method applied to multi-cellular environment
• Inter-eNB interference avoidance is similar to theprevious scheme
17
Dynamic ICIC
M. Rahman
Introduction
Motivation
Contributions
Literature
Framework
Two-Level Partial
Overview
Central Algo.
Simulation Para.
Results
Distributed
Overview
Inter-eNB
Inter-eNB
Simulation Para.
Results
Clustered Central
Overview
Simulation Para.
Results
Complexity
Conclusions
Future Work
Publications
Intra-eNB Avoidance: Multi-Cellular Hungarian
Uintra =
Sectors 1,2,3 Tx︷ ︸︸ ︷
U1|{} U2|{} U3|{}
Sectors 1,2 Tx︷ ︸︸ ︷U1|{3} U2|{3}
Sectors 1,3 Tx︷ ︸︸ ︷U1|{2} U3|{2}
Sector 2,3 Tx︷ ︸︸ ︷U2|{1} U3|{1}
.(7)
UM×N =
u(1,1) u(1,2) · · · u(1,N)
u(2,1) u(2,2) · · · u(2,N)
u(3,1) u(3,2) · · · u(3,N)
...... · · ·
...
u(M,1) u(M,2) · · · u(M,N)
(8)
u(m,n) =
{r (m,n) · d (m); all 3 transmit,
(r (m,n) − rp) · d (m); 1 among 3 is restricted.(9)
18
Dynamic ICIC
M. Rahman
Introduction
Motivation
Contributions
Literature
Framework
Two-Level Partial
Overview
Central Algo.
Simulation Para.
Results
Distributed
Overview
Inter-eNB
Inter-eNB
Simulation Para.
Results
Clustered Central
Overview
Simulation Para.
Results
Complexity
Conclusions
Future Work
Publications
Inter-eNB Inter-Cell Interference Avoidance
• Preparation of utility matrix, Uinter
• Similar to earlier scheme: selfishly prepares UM×N
• Heuristically find potential I(n)k = 0, ∀k in the inter-eNB
sectors
• Applying Hungarian algorithm to Uinter
• If a chosen entry has I(n)k = 0 marked, n is noted as to
be masked in k• Prepare lists of the resource restrictions based on users’
status
• Inter-eNB negotiations for resource restrictions• Each eNB negotiates the potential resource restrictions
by pairwise comparison• Restriction is imposed to the sector in which utility is
lower
• Resource units resulting from intra- and inter-eNBavoidance are restricted in each sector
19
Dynamic ICIC
M. Rahman
Introduction
Motivation
Contributions
Literature
Framework
Two-Level Partial
Overview
Central Algo.
Simulation Para.
Results
Distributed
Overview
Inter-eNB
Inter-eNB
Simulation Para.
Results
Clustered Central
Overview
Simulation Para.
Results
Complexity
Conclusions
Future Work
Publications
System and Simulation Parameters
Parameter Assumption
Cellular Layout Hexagonal grid, 19 sites, 3 sectors/siteInter-site Distance 500 m
Path-loss LD = 128.1 + 37.6 log10(D), D in kmShadowing Independent Log-normal std. 8 dB
Penetration Loss 10 dB
Antenna Pattern A (θ) = -min
[12
(θ
θ3dB
)2
, 20
], θ3dB = 70o
Bandwidth@Carrier 20 MHz (100 PRBs) @2.0 GHzChannel model 6-Tap SCME
AMC modes Attenuated Shannon boundfor QPSK, 16- and 64-QAM
with varying ratesUE speeds of interest 30 km/hr
Sector TX power 46 dBmMinimum UE Distance ≥ 35 m
20
Dynamic ICIC
M. Rahman
Introduction
Motivation
Contributions
Literature
Framework
Two-Level Partial
Overview
Central Algo.
Simulation Para.
Results
Distributed
Overview
Inter-eNB
Inter-eNB
Simulation Para.
Results
Clustered Central
Overview
Simulation Para.
Results
Complexity
Conclusions
Future Work
Publications
Power Allocation to Physical Resource Blocks
Scheme Allocated Power
Reuse 1 Pt/NReuse 3 3Pt/NPFR 1.3 outer resource: 2.25Pt/N; inner resource: 1.13Pt/NSFR 1.5 outer resource: 1.5Pt/N; inner resource: 0.75Pt/NSFR 2.0 outer resource: 2.0Pt/N; inner resource: 0.5Pt/NSFR 2.5 outer resource: 2.5Pt/N; inner resource: 0.25Pt/NSFR 2.75 outer resource: 2.75Pt/N; inner resource: 0.13Pt/NProposed 1 eligible resource: Pt/(N − Nr); restricted resource: 0Proposed 2 eligible resource: 10Pt/(10Ne + Nr);
restricted resource: Pt/(10Ne + Nr)
21
Dynamic ICIC
M. Rahman
Introduction
Motivation
Contributions
Literature
Framework
Two-Level Partial
Overview
Central Algo.
Simulation Para.
Results
Distributed
Overview
Inter-eNB
Inter-eNB
Simulation Para.
Results
Clustered Central
Overview
Simulation Para.
Results
Complexity
Conclusions
Future Work
Publications
Summarized Results: Distributed Algorithm
50 100 150 200 250 300 350 40016
18
20
22
24
26
28
30
32
34
36
Cell Edge Throughput (Kbps)
Ave
rag
e S
ecto
r T
hro
ug
hp
ut (M
bp
s)
Reuse 1CE: 99 Kbps,S:32.8 Mbps SFR eff.reuse 2.0
CE:+21.4%,S:-4.3% Proposed 2(P
res=0.1XP
elig)
CE:+171.3%,S:+7.2%
Proposed 1 (P
res =0)
CE:+266.1%,S:+3.9%
PFR eff reuse 1.3 CE:+149.7%,S:-26%
Reuse 3CE:+168.4%,S:-49.2%
SFR eff.reuse 2.75CE:+114.7%,S:-26.1%
SFR eff.reuse 2.5CE:+69.4%,S:-8.6%
SFR eff.reuse 1.5CE:-13.2%,S:-1.4%
22
Dynamic ICIC
M. Rahman
Introduction
Motivation
Contributions
Literature
Framework
Two-Level Partial
Overview
Central Algo.
Simulation Para.
Results
Distributed
Overview
Inter-eNB
Inter-eNB
Simulation Para.
Results
Clustered Central
Overview
Simulation Para.
Results
Complexity
Conclusions
Future Work
Publications
Cluster-based Centralized Scheme: Size 3
• UE forms cluster with two most dominantinterferers- ψ1 & ψ2
• UEs’ rates are computed with combinations(Π) of these interferers restrictions
maximize
∑i
∑Π
[M∑
m=1
N∑n=1
u(m,n)i|Π ρ
(m,n)i|Π
], (10)
subject to
ρ(m,n)i|Π ∈ {0, 1};∀m and ∀n, (11)
Sector i
User 1
User 2
ψ1ψ2
I(n)i =
∑Π
M∑m=1
ρ(m,n)i|Π =
{0; resource n is restricted in i1; otherwise.
(12)
23
Dynamic ICIC
M. Rahman
Introduction
Motivation
Contributions
Literature
Framework
Two-Level Partial
Overview
Central Algo.
Simulation Para.
Results
Distributed
Overview
Inter-eNB
Inter-eNB
Simulation Para.
Results
Clustered Central
Overview
Simulation Para.
Results
Complexity
Conclusions
Future Work
Publications
Cluster-based Centralized Scheme: Size 3(contd.)
γ(m,n)i =
PcH(m,n)i ,i
Pc∑x 6=i
x /∈G(m)i
H(m,n)i ,x + Pc
∑ψ∈G(m)
i
H(m,n)i ,ψ · I (n)
ψ + PTN , (13)
• γ(m,n)i |ψ1
= γ(m,n)i given, I
(n)ψ1
= 0
• r(m,n)i |ψ1
← γ(m,n)i |ψ1
• u(m,n)i |ψ1
= f(r
(m,n)i |ψ1
, d(m)i
) Sector i
User 1
User 2
ψ1ψ2
24
Dynamic ICIC
M. Rahman
Introduction
Motivation
Contributions
Literature
Framework
Two-Level Partial
Overview
Central Algo.
Simulation Para.
Results
Distributed
Overview
Inter-eNB
Inter-eNB
Simulation Para.
Results
Clustered Central
Overview
Simulation Para.
Results
Complexity
Conclusions
Future Work
Publications
Cluster-based Centralized Scheme: Size 3(contd.)
Constraints for inter-cluster relations:
ρ(m,n)i |{ψ1} + I
(n)ψ1
= 0 or 1,
ρ(m,n)i |{ψ2} + I
(n)ψ2
= 0 or 1,
ρ(m,n)i |{ψ1,ψ2} + I
(n)ψ1
= 0 or 1,
ρ(m,n)i |{ψ1,ψ2} + I
(n)ψ2
= 0 or 1. (14)
Scheduling constraints:∑Π
∑n
ρ(m,n)i |Π ≤ 2;∀i , ∀m. (15)
25
Dynamic ICIC
M. Rahman
Introduction
Motivation
Contributions
Literature
Framework
Two-Level Partial
Overview
Central Algo.
Simulation Para.
Results
Distributed
Overview
Inter-eNB
Inter-eNB
Simulation Para.
Results
Clustered Central
Overview
Simulation Para.
Results
Complexity
Conclusions
Future Work
Publications
Systems and Simulation Parameters
Parameter Assumption
Cellular layout Hex grid, 19 cell sites, 3 sectors/site,Inter-site distance 500 mBandwidth@Carrier 10 MHz (50 PRBs) @ 2.0 GHzPath-loss exponent 3.76Shadowing std. 8 dB (independent)UE speeds 30 km/hrPenetration loss 10 dBAntenna configuration Single-input single-outputeNB antenna gain 14 dBiUE antenna gain 0 dBiAMC modes Attenuated Shannon bound
for QPSK, 16- and 64-QAMwith varying rates
Channel model 6-Tap SCMETotal sector TX power 46 dBmUE close-in distance 35 mTraffic model Full buffer
26
Dynamic ICIC
M. Rahman
Introduction
Motivation
Contributions
Literature
Framework
Two-Level Partial
Overview
Central Algo.
Simulation Para.
Results
Distributed
Overview
Inter-eNB
Inter-eNB
Simulation Para.
Results
Clustered Central
Overview
Simulation Para.
Results
Complexity
Conclusions
Future Work
Publications
Summarized Results: Cluster-based CentralizedScheme
10 20 30 40 50 60 70 80 90 100 110 12010
10.5
11
11.5
12
12.5
13
13.5
14
Cell Edge Throughput (kbps)
Sec
tor
Thr
ough
put (
Mbp
s)
Cell−Edge Vs. Sector Throughput: Different Utility, Cluster Size, and Coordination
No Coordination: u=rd2
CE: 20.9 kbps, S: 11.8 Mbps
Coordination Cluser Size 2: u=rdCE: 76.3 kbps, S: 12.8 Mbps Coordination Cluser Size 3: u=rd
CE: 99.3 kbps, S: 12.6 Mbps
Coordination Cluser Size 2: u=rd2
CE: 82.5 kbps, S: 11.4 Mbps
Coordination Cluser Size : u=rd2
CE: 109 kbps, S: 11.0 Mbps
No Coordination: u=rdCE: 23.2 kbps, S: 12.1 Mbps
27
Dynamic ICIC
M. Rahman
Introduction
Motivation
Contributions
Literature
Framework
Two-Level Partial
Overview
Central Algo.
Simulation Para.
Results
Distributed
Overview
Inter-eNB
Inter-eNB
Simulation Para.
Results
Clustered Central
Overview
Simulation Para.
Results
Complexity
Conclusions
Future Work
Publications
Complexity Comparison
Complexity Comparison among Studied Schemes1
Studied No. of YALMIP Time CPLEX Solver TimeScenario Variables (sec./iteration) (sec./iteration)
No ICIC 1155 0.189 0.016ICIC with cluster size 2 2310 0.520 0.076ICIC with cluster size 3 4620 1.378 0.300
1Computed in a personal computer with AMD Athlon X2 6400+ CPU, 4 GB main memory, and
Windows XP 64-bit OS.
28
Dynamic ICIC
M. Rahman
Introduction
Motivation
Contributions
Literature
Framework
Two-Level Partial
Overview
Central Algo.
Simulation Para.
Results
Distributed
Overview
Inter-eNB
Inter-eNB
Simulation Para.
Results
Clustered Central
Overview
Simulation Para.
Results
Complexity
Conclusions
Future Work
Publications
Conclusions
• Dynamic interference coordination is formulated usingsum-utility maximization problem
• Three sub-optimal approaches have been pursued
• A unique, two-level algorithm is investigated. The cell-edgeperformance reaches close to that of Reuse 3 whilemaintaining Reuse 1 cell throughput.
• A distributed algorithm with neighboring cell coordination ispresented. The intra-eNB interference is coordinated using anovel method based on Hungarian algorithm in amulti-cellular context. The proposed schemes attains cellthroughput comparable to that in Reuse 1 while achieving acomparable to or slightly better reuse 3 cell-edge throughput.
• A cluster-based centralized scheme with different cluster sizesand utility functions is studied. The presented schemesachieve significant gain in cell-edge throughput withoutimpact on cell throughput.
29
Dynamic ICIC
M. Rahman
Introduction
Motivation
Contributions
Literature
Framework
Two-Level Partial
Overview
Central Algo.
Simulation Para.
Results
Distributed
Overview
Inter-eNB
Inter-eNB
Simulation Para.
Results
Clustered Central
Overview
Simulation Para.
Results
Complexity
Conclusions
Future Work
Publications
Future Work
• Utility functions
• Power control
• Inclusion of MIMO
• Integration with other mitigation techniques
• Combining with PHY-layer CoMP
30
Dynamic ICIC
M. Rahman
Introduction
Motivation
Contributions
Literature
Framework
Two-Level Partial
Overview
Central Algo.
Simulation Para.
Results
Distributed
Overview
Inter-eNB
Inter-eNB
Simulation Para.
Results
Clustered Central
Overview
Simulation Para.
Results
Complexity
Conclusions
Future Work
Publications
List of PublicationsWINNER Deliverables
1 Description of Identified New Relay Based Radio Network Deployment Concepts and FirstAssessment by Comparison Against Benchmarks of Well- Known Deployment Concepts UsingEnhanced Radio Interface Technologies, WINNER I Deliverable D3.1, Nov. 2004. Availableonline at http://www.ist-winner.org/DeliverableDocuments/D3.1v1.1.pdf.
2 Concept and Criteria for Coordination Across Base Stations to Improve the Mutual InterferenceSituation, WINNER I Deliverable D3.3, Jun. 2005.
3 Description of Identified New Relay Based Radio Network Deployment Concepts and FirstAssessment by Comparison Against Benchmarks of Well-Known Deployment Concepts UsingEnhanced Radio Interface Technologies, WINNER I Deliverable D3.2, Feb. 2005. Availableonline at http://www.ist-winner.org/DeliverableDocuments/D3.2v1.1.pdf.
4 Proposal of the Best Suited Deployment Concepts for the Identified Scenarios and Related RANProtocols, WINNER I Deliverable D3.5, Jan. 2006. Available online athttp://www.ist-winner.org/DeliverableDocuments/D3.5.pdf.
5 Interference Avoidance Concepts, WINNER II Deliverable D4.7.2, Jun. 2007. Available online athttp://www.ist-winner.org/WINNER2-Deliverables/D4.7.2.pdf.
PhD Work Included in the Thesis
6 M. Rahman and H. Yanikomeroglu, “QoS provisioning in the absence of ARQ in cellular fixedrelay networks through intercell coordination,” in Proc. IEEE Global CommunicationsConference (GLOBECOM2006), San Francisco, California, USA, Nov. 2006.
7 M. Rahman and H. Yanikomeroglu, “Multicell downlink OFDM subchannel allocations usingdynamic intercell coordination,” in Proc. IEEE Global Communications Conference(GLOBECOM2007), Washington DC, USA, Nov. 2007.
31
Dynamic ICIC
M. Rahman
Introduction
Motivation
Contributions
Literature
Framework
Two-Level Partial
Overview
Central Algo.
Simulation Para.
Results
Distributed
Overview
Inter-eNB
Inter-eNB
Simulation Para.
Results
Clustered Central
Overview
Simulation Para.
Results
Complexity
Conclusions
Future Work
Publications
List of Publications (contd.)8 M. Rahman and H. Yanikomeroglu, “Interference avoidance through dynamic downlink OFDMA
subchannel allocation using intercell coordination,” in Proc. IEEE Vehicular TechnologyConference (VTC2008-Spring), May 2008, pp. 1630–1635.
9 M. Rahman, H. Yanikomeroglu, and W. Wong, “Interference avoidance with dynamic inter-cellcoordination for downlink LTE system,” in Proc. IEEE Wireless Communications andNetworking Conference (WCNC2009), Budapest, Hungary, Apr. 2009.
10 M. Rahman and H. Yanikomeroglu, “Enhancing cell-edge performance: A downlink dynamicinterference avoidance scheme with inter-cell coordination,” IEEE Transaction on WirelessCommunications, vol. 9, April 2010, pp. 1414-1425.
11 M. Rahman and H. Yanikomeroglu, “Inter-cell interference coordination in OFDMA networks: anovel approach based on integer programming,” Proc. IEEE Vehicular Technology Conference(VTC2010-Spring), Taipei, Taiwan, May 2010.
12 M. Rahman and H. Yanikomeroglu, “A cluster-based integer linear programming approach indynamic interference coordination,” submitted to IEEE Transactions on Vehicular Technology,June 2011.
13 M. Rahman and H. Yanikomeroglu, “A distributed ICIC algorithm with neighbor coordination”,manuscript in-preparation for a possible submission to an IEEE magazine.
PhD Work Not Included in the Thesis
14 M. Salem, A. Adinoyi, M. Rahman, H. Yanikomeroglu, D. Falconer, and Y.-D. Kim, “Apparatusand method for allocating subchannels and controlling interference in OFDMA systems,” Patentfiled by Samsung Korea, Korea patent application no: P2008-0054726 (application date: 11June 2008); US patent application no: 12/341,933 (application date: 22 December 2008);international patent application no: PCT/KR2009/002119 (filing date: 23 April 2009).
15 M. Salem, A. Adinoyi, M. Rahman, H. Yanikomeroglu, D. Falconer, Y.-D. Kim, and E. Kim,“Fairness-aware joint routing and scheduling in OFDMA-based multi-cellular fixed relaynetworks,” in Proc. IEEE International Conference on Communications (ICC2009), Dresden,Germany, Jun. 2009.
32
Dynamic ICIC
M. Rahman
Introduction
Motivation
Contributions
Literature
Framework
Two-Level Partial
Overview
Central Algo.
Simulation Para.
Results
Distributed
Overview
Inter-eNB
Inter-eNB
Simulation Para.
Results
Clustered Central
Overview
Simulation Para.
Results
Complexity
Conclusions
Future Work
Publications
List of Publications (contd.)
16 M. Salem, A. Adinoyi, M. Rahman, H. Yanikomeroglu, D. Falconer, and Y.-D. Kim,“Fairness-aware radio resource management in OFDMA-based cellular fixed relay networks,”submitted to IEEE Transaction on Wireless Communications, submitted Nov. 2008, revised Jul.2009.
17 M. Salem, A. Adinoyi, M. Rahman, H. Yanikomeroglu, D. Falconer, Y.-D. Kim, E. Kim, andY.-C. Cheong, “An overview of radio resource management in relay-enhanced OFDMA-basednetworks,” to appear in IEEE Communications Surveys and Tutorials, 2010.
18 F. A. Bokhari, H. Yanikomeroglu, W. K. Wong, and M. Rahman, “Fairness assessment of theadaptive token bank fair queuing scheduling algorithm,” in Proc. IEEE Vehicular TechnologyConference (VTC2008-Fall), Calgary, Alberta, Canada, Sep. 2008.
19 F. A. Bokhari, H. Yanikomeroglu, W. K. Wong, and M. Rahman, “Cross-layer resourcescheduling for multimedia traffic in the downlink of 4G wireless multicarrier networks,”EURASIP Journal on Wireless Communications and Networking, Special Issue on Fairness inRadio Resource Management for Wireless Networks, 2009.
20 P. Djukic, M. Rahman, H. Yanikomeroglu, and J. Zhang, “Advanced radio access networks forLTE and beyond,” in Evolved Cellular Network Planning and Optimization for UMTS and LTE,L. Song and S. Jia, Eds. Auerbach Publications, CRC Press, Taylor & Francis Group, 2010.
33
Dynamic ICIC
M. Rahman
Introduction
Motivation
Contributions
Literature
Framework
Two-Level Partial
Overview
Central Algo.
Simulation Para.
Results
Distributed
Overview
Inter-eNB
Inter-eNB
Simulation Para.
Results
Clustered Central
Overview
Simulation Para.
Results
Complexity
Conclusions
Future Work
Publications
Thank You!
34