Clustered Network MIMO and Fractional FrequencyReuse for the Downlink in LTE-A Systems
Ajay Thampi†, Simon Armour†, Zhong Fan‡, Dritan Kaleshi†
†Communication Systems and Networks Research Group, University of Bristol, UK‡Toshiba Research Europe, Telecommunications Lab, Bristol, UK
May 16, 2014
Ajay Thampi (University of Bristol) Clustered Network MIMO and FFR May 16, 2014 1 / 18
Thanks to...
The U.K. Research Council and Toshiba for jointly funding my PhD underthe Dorothy Hodgkin Postgraduate Awards.
Ajay Thampi (University of Bristol) Clustered Network MIMO and FFR May 16, 2014 2 / 18
The Problem
Worldwide data traffic to grow 7-fold in the next 3 yearsI 66% of that traffic will be video
Operators deploying 4G LTE NetworksI Reduced cell sizeI Aggressive frequency reuse (Reuse factor → 1)
Major Performance Bottleneck: Inter-Cell Interference
Ajay Thampi (University of Bristol) Clustered Network MIMO and FFR May 16, 2014 3 / 18
Possible Solutions
Network MIMO (aka CoMP)I Base stations pooled together to form a Virtual MIMO system
F Data and channel states are shared
I Interference channel becomes:F Broadcast channel on the downlinkF Multiple-access channel on the uplink
I Ideal solution if backhaul links have infinite capacityI Realistically, global coordination is unscalable
Fractional Frequency Reuse (FFR)I Split the spectrum into two bands:
F Band 1: Cell-Centre (Reuse factor = 1)F Band 2: Cell-Edge (Reuse factor > 1)
I Cancels interference entirely but inefficient use of spectrum
Ajay Thampi (University of Bristol) Clustered Network MIMO and FFR May 16, 2014 4 / 18
Clustered Network MIMO - System Model (1/2)
Scalable Network MIMO
C Clusters, each of size B (Here, C = 7 and B = 3)I Cluster 0: Home ClusterI Clusters 1 to C − 1: Neighbouring Clusters
R: Cell RadiusDc : Boundary between Cluster-Centre and Cluster-Edge
Ajay Thampi (University of Bristol) Clustered Network MIMO and FFR May 16, 2014 5 / 18
Cluster Network MIMO - System Model (2/2)NT : Number of transmit antennas (at the base station)
NR : Number of receive antennas (for each user in the cell)
K (c): Number of users in cluster cl(c)k : Length of data symbol for user k in cluster c
I Assumption is that l(c)k = NR ∀k , c
x(c)k : NR × 1 transmitted signal vector for user k in cluster c
y(c)k : NR × 1 received signal vector for user k in cluster c
y(c)k =
B∑b=1
H(c,b)k T
(c,b)k x
(c)k︸ ︷︷ ︸
desired signal
+B∑
b=1
H(c,b)k
K (c)∑i=1,i 6=k
T(c,b)i x
(c)i︸ ︷︷ ︸
intra-cluster interference
+C−1∑
c=0,c 6=c
B∑b=1
H(c,b)k
K (c)∑j=1
T(c,b)j x
(c)j︸ ︷︷ ︸
inter-cluster interference
+n(c)k
(1)
Ajay Thampi (University of Bristol) Clustered Network MIMO and FFR May 16, 2014 6 / 18
Existing Approach - Helper Clusters
Intra-Cluster InterferenceI Block diagonalisation (BD) precoding technique
F More practical than DPC and provides interference-free channels [1]
Inter-Cluster InterferenceI Get neighbouring clusters to help the edge users in the home cluster
F Not guaranteed to cancel interference
I Main idea is to increase the cluster size (B = 7 worked in [1])
Can we do better?
Ajay Thampi (University of Bristol) Clustered Network MIMO and FFR May 16, 2014 7 / 18
Proposed Approach - Network MIMO + FFR (1/3)
Set cluster size B = 3
FFR applied in cluster-scale to cancel inter-cluster interferenceI Bandwidth Partitioning is load-dependent
Ajay Thampi (University of Bristol) Clustered Network MIMO and FFR May 16, 2014 8 / 18
Proposed Approach - Network MIMO + FFR (2/3)
Bandwidth PartitioningI Let:
F M(i)c : Number of cluster-centre users in cluster i
F M(i): Total number of users in cluster iF W : Total available bandwidth
I Bandwidth allocated for cluster-centre users in cluster i :
W (i)c =
⌈(M
(i)c
M(i)
)W
⌉(2)
I Bandwidth allocated for cluster-edge users in cluster i :
W (i)e =
⌊(W −W
(i)c
3
)⌋(3)
I Choose largest W(i)e and corresponding W
(i)c for all clusters
Ajay Thampi (University of Bristol) Clustered Network MIMO and FFR May 16, 2014 9 / 18
Proposed Approach - Network MIMO + FFR (3/3)
Network MIMO with conventional FFR [2]
Proposed FFR scheme v/s Conventional FFRI 84% v/s 100% spectrum utilisationI 76.6% v/s 61.5% spectrum allocation for cluster-centre users (under
high load)
Location Classification: Use logistic regression approach in [3]I Employ Minimisation of Drive Test (MDT) reports specified in 3GPP
TS37.320
Ajay Thampi (University of Bristol) Clustered Network MIMO and FFR May 16, 2014 10 / 18
The Setup
Cell ParametersNumber of Cells 21
Cell Radius, R 1 km
Coordination Distance, Dc 350 m
MIMO ParametersNumber of Transmit Antennas, NT 4
Number of Receive Antennas, NR 2
Channel ModelCarrier Frequency 800 MHz
Fading Narrowband, Rayleigh
Power AllocationTotal Power Constraint, P 46 dBm
Algorithm Scaled Water Filling
SchedulingAlgorithm Proportional Fair
Window Size 100Ajay Thampi (University of Bristol) Clustered Network MIMO and FFR May 16, 2014 11 / 18
Results - Overall Sum Rate
Ajay Thampi (University of Bristol) Clustered Network MIMO and FFR May 16, 2014 12 / 18
Results - Cluster Edge Sum Rate
Ajay Thampi (University of Bristol) Clustered Network MIMO and FFR May 16, 2014 13 / 18
Results - Fairness
Ajay Thampi (University of Bristol) Clustered Network MIMO and FFR May 16, 2014 14 / 18
Results - Complexity
Channel State Information (CSI) ReductionI 71% reduction when compared to global coordinationI 14% reduction when compared to the helper approach
Execution time (in milliseconds)
B = 21(Global)
B = 7(Helper)
B = 3(FFR-Cell)
B = 3(FFR-Cluster)
239.75 10.43 3.37 2.01
Ajay Thampi (University of Bristol) Clustered Network MIMO and FFR May 16, 2014 15 / 18
Possible Future Directions
Performance study with imperfect CSI and better precodingtechniques
Clustered Network MIMO + FFR in a heterogeneous network
Ajay Thampi (University of Bristol) Clustered Network MIMO and FFR May 16, 2014 16 / 18
Thank you!
Q & A
http://ajaythampi.net
@thampiman
Ajay Thampi (University of Bristol) Clustered Network MIMO and FFR May 16, 2014 17 / 18
References
[1] J.Zhang; et al (2009)
Networked MIMO with Clustered Linear Precoding
IEEE Transactions on Wireless Communications, vol. 8, no. 4, pp. 1910-1921.
[2] L.C.Wang; et al (2011)
3-cell network MIMO architectures with sectorization and FFR
IEEE Journal on Selected Areas in Communications, vol. 29, no. 6, pp. 1185-1199.
[3] A.Thampi; et al (2013)
A Logistic Regression Approach to Location Classification in OFDMA-based FFRSystems
IEEE WoWMoM, pp. 1-9.
Ajay Thampi (University of Bristol) Clustered Network MIMO and FFR May 16, 2014 18 / 18