Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
I. I NTRODUCTION
The Long Term Evolution (LTE) mobile communications
system is being deployed around the Globe, providing significantly enhanced network capacity, user data rates and
latency compared to 3G systems. In the meantime the 3rd
Generation Partnership Project (3GPP) continues to make the
LTE technology evolve in order to cope with the sustained
mobile data traffic increase expected in the coming years.
Interference mitigation is key to further improve the network
capacity and the user data rates, especially at the cell edge.
To this end, interference mitigation techniques have formed
an integral part of the LTE design since the first version of
the LTE specifications, which correspond to the Release 8
of 3GPP. In this first LTE Release, Inter-Cell Interference
Coordination (ICIC) mechanisms allowed neighboring base
stations (called eNodeB in 3GPP) to coordinate semi-static
restrictions on the frequency domain power allocation to create
resources with reduced inter-cell interference levels [1]. In
Release 10, these mechanisms were improved to include a time
dimension to the resource allocation coordination, targetting
heterogeneous networks where low-power nodes (e.g. pico
eNodeBs) are deployed under the macro cell coverage to
densify the network. This technique, called eICIC, relies on
the transmission by the aggressor cell of an Almost Blank
Subframe (ABS), where the control and data channels are
not transmitted (or with a significantly reduced power). The
vicitim cell can then take advantage of the reduced interference
to schedule in ABSs its users mostly affected by the aggressor
interference [1]. Nevertheless, the ABSs still contain common reference signals, synchronisation signals and common
channels, which still can affect cell-edge users in the victim
cell. This is particularly the case if the network wishes to
maximise the offload of the macro cell towards the low-power
nodes, by biasing the User Equipments (UEs) measurement
3597
Fig. 1.
3598
Fig. 2.
Another alternative which does not suffer from this constraint is the Turbo SIC (see e.g. [5]), where the desired and
interfering signals are processed as outlined below (assuming
a single interferer for the sake of clarity):
1) soft estimation of UE1s information bits using a soft-in
soft-out channel decoder;
2) construction of soft estimates of UE1s symbols and
multiplication by the estimated channel response to
obtain a soft estimate of UE1s signal contribution.
3) subtraction of UE1s soft signal estimation from the
originally received signal;
4) do steps 1-2 for UE2 (interferer) instead of UE1, from
the signal obtained in step 3;
5) subtraction of UE2s soft signal estimation from the
originally received signal;
6) soft estimation of UE1s data, from the signal obtained
in step 5;
7) repeat steps 2-6 until successful decoding of UE1s data
(e.g. checked by means of CRC), or after a given number
of iterations.
As the estimate of the interference each transmitter creates for
a given UE is refined iteratively, the constraint on the MCS
of the interferers to enable a succesful interference decoding
is alleviated.
Note that IC can be employed for both inter-cell and intracell interference. In LTE, intra-cell interference originates from
the Multi-User (MU) Multiple Input Multiple Output (MIMO)
transmission [6], where two (or more, up to four) UEs can
be multiplexed on the same time frequency resources, being
separated in the spatial domain. This separation relies on the
MIMO precoding at the transmitter, which is sensitive to the
CSI accuracy and thus generally imperfect. Fig. 2 materializes
an example scenario where UE2 is an intra-cell interferer,
whereas UE3 is an inter-cell interferer.
III. C HALLENGES AND FEASIBILITY IN LTE
In contrast to LMMSE processing, which only needs an estimation of the covariance matrix of the interference (or received
3599
3600
TABLE I
S IMULATION PARAMETERS
Channel model
Penetration loss
UEs distribution
Traffic models
Antennas
MIMO scheme
Feedback
UE receiver
Interference modelling
Channel estimation
Scheduling
HARQ
Value
FDD downlink, 2 GHz
10 MHz
19 tri-sector macro sites with wrap-around,
500 m inter-site distance
ITU Uma, 3 km/h
20 dB for indoor UEs,
0 dB for outdoor UEs
Uniform 75 % UEs indoors,
10 UEs/cell for full buffer
Full buffer
FTP model 1 in [10] with 2 Mbytes packets
2 Tx X-pol, 2 Rx X-pol
SU-MIMO with adaptive number of layers
PUSCH 3-1 with 5 ms duty cycle
MMSE-SIC with IRC for
SU-MIMO w/wo inter-cell
interference cancellation up to 2 interferers
Fully modelled from 56 cells
Ideal
Proportional Fair, frequency selective
Explicit with 4 transmissions maximum
90
80
70
CDF
Parameter
Carrier
Bandwidth
Network layout
100
IC, N = 2
60
IC, N = 1
50
Baseline, no IC
40
30
20
10
0
0
10
100
90
C. Results
Using the system simulator described in the previous sections, we evaluate the performance gains brought by inter-cell
IC for a macro LTE network, first for the inter- & intraeNodeB case. This case typically represents a deployment
scenario where neighboring eNodeBs are inter-connected via
a low-latency (typically less than 1 ms) backhaul, in order
to enable a dynamic scheduling coordination between the
aggressor(s) and the victim cell.
For the full buffer traffic model, the considered performance
metrics are the system capacity, defined by the cell throughput,
and the cell-edge user throughput, defined as the 5th percentile
of the user throughput cumulative distribution function (CDF).
Fig. 3 shows the user throughput CDFs for the inter-eNodeB
IC, and Table II summarizes the associated capacity and
cell-edge user throughputs, for N = 1 and N = 2 processed
interferers.
As can be seen, the performance gains in the inter- & intraeNodeB case are huge, with roughly 30% and 65% increases in
both cell capacity and cell-edge user throughput, for N = 1 and
N = 2, respectively. Again, we need to keep in mind that these
results represent an upper bound of the achievable results, as
in practice the scheduling constraints imposed by the need
to ensure the aggressor MCSs decoding at the victim on the
3601
80
70
CDF
60
IC, N = 2
50
IC, N = 1
40
Baseline, no IC
30
20
10
0
0
10
TABLE II
F ULL BUFFER TRAFFIC , I NTER - & I NTRA - E N ODE B IC
No IC/IC
No IC
IC, N=1
IC, N=2
Capacity (Mbps)
15.5 (0 %)
20.3(+31 %)
26.4 (+70 %)
TABLE IV
FTP TRAFFIC , I NTER - & I NTRA - E N ODE B IC
No IC/IC
No IC
IC, N=1
IC, N=2
TABLE III
F ULL BUFFER TRAFFIC , I NTRA - E N ODE B IC
No IC/IC
No IC
IC, N=1
IC, N=2
Capacity (Mbps)
15.5 (0 %)
19.2 (+24 %)
23.2 (+50 %)
Capacity (Mbps)
9.04 (0 %)
11.3 (+25 %)
14.1 (+56 %)
TABLE V
FTP TRAFFIC , I NTRA - E N ODE B IC
No IC/IC
No IC
IC, N=1
IC, N=2
3602
Capacity (Mbps)
9.04 (0 %)
10.4 (+15 %)
11.8 (+31 %)