Beruflich Dokumente
Kultur Dokumente
AbstractMassive multiple-input multiple-output (MIMO) has pilot contamination [6], [7]. As the number of antennas at
been proposed as a key technology for the future fth generation base station tends to innity, fast fading channels are approx-
(5G) cellular networks. In time division duplex (TDD) massive imately orthogonal, and the effects of uncorrelated noise and
MIMO systems, pilot contamination caused by channel estima-
tion error is crucial to the system performance. In this paper, fast fading vanish. The only remaining impairment is pilot
we propose a pilot sequences allocation strategy to mitigate the contamination which does not vanish with innite antennas
pilot contamination. In this strategy, the pilot sequences sets [1].
are identical for center users, but mutually orthogonal for edge Many previous literatures have addressed the issue of pilot
users in different cells. With mitigated pilot contamination, we contamination. AOA-based methods are proposed in [8][10],
analytically determine the approximate system capacity which is
accurate when the number of antennas at the base station tends which show that users with mutually non-overlapping AOA
to innite. The simulation results show that the proposed pilot PDFs hardly contaminate each other even if they use the
sequences allocation strategy achieves higher system capacity same pilot sequence. In [11], the eigenvalue-decomposition-
than the traditional pilot sequences allocation strategy whose based (EVD) estimation can estimate channel vectors using the
sequences reuse rate is one or three. There also exists an optimal statistics of the received data. A blind pilot decontamination
number of pilot sequences in different SNR to maximize the
system capacity. strategy is proposed in [12]. This strategy eliminates pilot
contamination by separating the interference subspace from
I. I NTRODUCTION the desired signal subspace. A time-shifted (asynchronous)
Massive multiple-input multiple-output (MIMO) systems protocol is proposed in [14] to mitigate the pilot contami-
where hundreds of antennas simultaneously serve a number nation. Only users from the same group create interference
of single-antenna users have attracted much attention in recent to each other. Another method to mitigate the effect of pilot
years. Massive MIMO systems can improve the spectral and contamination is through orthogonal frequency or orthogonal
energy efciency with orders of magnitude compared with pilot sequences [1]. Applying orthogonal frequency or or-
single-antenna systems [1], [2]. thogonal pilot sequences means fewer users can be served
Pilot sequences are used to training channel during the simultaneously. However, most of the proposed strategies
channel estimation phase. Channel estimation can be per- to eliminate pilot contamination suffer from the defect of
formed either in frequency division duplex (FDD) or time high computational complexities and the request of particular
division duplex (TDD) systems. In the FDD massive MIMO channel characteristics. When antennas tends to innity, the
system, channel estimation becomes more challenging because channel matrix becomes very large and the complexities of
traditional approaches incur a signicant overhead. Therefore, many operations about channel matrix cannot be tolerated,
most of the massive MIMO researches apply the assumption such as matrix inversion, eigenvalue decomposition.
of TDD systems with channel reciprocity. In the TDD system, In this paper, a low complexity and universal strategy
the orthogonal pilot sequences needed are independent of the of pilot sequences allocation is proposed for TDD massive
number of antennas at the base station and are proportional to MIMO systems. We allocate the same pilot sequences set for
the number of user antennas [3][5]. The base station is able the center users, while mutually orthogonal pilot sequences
to acquire the channel state information (CSI) which can be sets for the edge users in different cells to mitigate the
used for signal detecting for the uplink and precoding for the pilot contamination. With mitigated pilot contamination, we
downlink. analytically derive the approximate system capacity which is
Massive MIMO systems with a large number of users can accurate when the number of antennas at the base station
achieve better system performance at the expense of huge tends to innite. Finally, the simulation results show that the
complexity of channel estimation. As the number of pilot proposed pilot sequences allocation strategy achieves higher
sequences is limited by coherence time, pilot sequences are system capacity than the traditional pilot sequences allocation
mutually orthogonal in the same cell but reused in different strategies whose sequences reuse rate is one or three. The op-
cells, and perfect CSI can not be achieved in massive MIMO timal number of pilot sequences in different SNR to maximize
systems with a large amount of users [1]. The reuse of pilot the system capacity is also given.
sequences causes the inter-cell interference which is called The rest of paper is organized as follows. Section II in-
1489
2015 IEEE Wireless Communications and Networking Conference (WCNC): - Track 3: Mobile and Wireless Networks
3& 3( 3( 3(
N CM p is white complex Gaussian noise with zero mean
and unit variance.
We can get channel estimation
.F .H .H .H
1
1,k =
g Yp 1,k . (11)
Fig. 2. Pilot sequences sets
p p p
Pilot sequences for center users in different cells are reused
and l,kc = 1,kc . From (10), (11) and l,kc = 1,kc , the
channel estimation for the kc -th center user in cell 1 is derived
&HOOO as 1
1,kc = g1,kc +
g gl,kc + n1 . (12)
3& pp p
l=1
&HOOO 3(
3(
3(
For the ke -th edge user in cell 1, l,ke and 1,ke are
3&
mutually orthogonal. So the channel estimation is derived as
&HOOO
1
1,ke = g1,ke +
g n2 . (13)
p p p
3&
n1 and n2 are complex Gaussian noise vectors. The differ-
ence between (12) and (13) is caused by the use of orthogonal
Fig. 3. Pilot sequences strategy
pilot sequences among the edge users in different cells.
B. Data Transmission
During the data transmission phase, the received signal at
to Kc and Ke .
the base station in cell 1 is
PC = {1 , 2 , . . . , Kc },
yd = pu gl,k xdl,k + n (14)
PE,1 = {Kc +1 , Kc +2 , . . . , Kc +Ke }, l,k
PE,2 = {Kc +Ke +1 , Kc +Ke +2 , . . . , Kc +2Ke },
where xdl,k is the data symbol transmitted by the k-th user in
PE,3 = {Kc +2Ke +1 , Kc +2Ke +2 , . . . , p }. (9) cell l and n is white Gaussian noise.
The result of pilot sequences allocation is shown in Fig.3. The data symbol x d1,k detected by MF method is
The strategy of pilot sequences allocation is explained as x 1,k
d1,k = g H
yd . (15)
follows:
First, generating the mutually orthogonal pilot sequences
For the kc -th center user in cell 1, the detected data symbol
set P. is
Second, deriving Kc , Ke and Rc . x
d1,kc = pu g1,kc 2 xd1,kc + pu gl,kc 2 xdl,kc
Then, the center users are randomly allocated with the l=1
pilot sequences in PC . desired signal
pilot contamination
Finally, to make sure that pilot sequences of edge users
in adjacent cells are mutually orthogonal, the edge users + pu glH1 ,kc gl2 ,k xdl2 ,k + Wc (16)
in different cells are allocated with pilot sequences PE,1 , (l1 ,kc )=(l2 ,k)
noise
PE,2 or PE,3 , respectively. other interference
IV. where
PERFORMANCE ANALYSIS
pu H
We assume that the K users in each cell transmit pilot Wc = gl,k
H
n + n1 gl,k xdl,k . (17)
sequences simultaneously, which is the worst case [1]. Match
c
pp p
l l,k
lter (MF) detector is used at the base station. Then we analyze
The rst term in (16) is the desired signal, the second term
the capacity for center users and edge users, respectively.
is pilot contamination and the third term is interference from
A. Channel Estimation users with different pilot sequences. The last term is noise.
During the pilot transmission phase, the signals received by Similarly, for the ke -th edge user in cell 1,
the base station in cell 1 can be expressed as d1,ke =
x pu g1,ke 2 xd1,ke
Yp = p p gl,k H
l,k + N. (10) desired signal
l,k
+ pu g1,k
H
g xd + We (18)
M 1
e l,k l,k
l,k C , which is selected from pilot sequences set P, (l,k)=(1,ke )
noise
is the pilot sequence transmitted by the k-th user in cell l. other interference
1490
2015 IEEE Wireless Communications and Networking Conference (WCNC): - Track 3: Mobile and Wireless Networks
TABLE I
where
pu H
BASIC SIMULATION PARAMETERS
We = g1,ke n +
H
n2 gl,k xdl,k . (19)
p p p parameter value
l,k Number of cells L 7
Cell radius R 1000 m
The rst term in (18) is the desired signal, the second term is
Risk radius Rh 100 m
interference from users with different pilot sequences and the Normalized radius r0 100 m
last term is noise. (18) demonstrates that pilot contamination Decay exponent 3.5
of edge cells is eliminated. Shadow fading standard deviation shad 8 dB
Number of users in each cell K 20
Coherence time 196
C. System Capacity
From (16) we derive the capacity of the kc -th center user
and the ke -th edge user in cell 1. The results are shown in
1
0.9
(20) and (21), respectively.
pu g1,ke 4
0.6 Proposed strategy approximation
pu (l,k)=(1,ke ) |g1,k
Reuse rate = 3 approximation
e
l,k e 0.4
(21) 0.3
To simplify the results in (20) and (21), the following lemma 0.2
0.1
is used.
Lemma 1: Assuming p, q CM 1 are i.i.d standard com-
0
0 5 10 15 20 25
Capacity of per users in cell 1 (bits/s/Hz)
1491
2015 IEEE Wireless Communications and Networking Conference (WCNC): - Track 3: Mobile and Wireless Networks
p g 4
Ckc = E log2 1+ u 1,kc (20)
pu l=1 gl,kc 4 + pu (l1 ,kc )=(l2 ,k) |glH1 ,kc gl2 ,k |2 + var(Wc )
(a1) pu (M 2 + 2M )1,k 2
Ckc = E log2 1+ 2
c
pu (M 2 + 2M ) l=1 l,k c
+ pu M (l1 ,kc )=(l2 ,k) l1 ,kc l2 ,k + M l=1 l,kc + M pppup l,k l,k
(24)
(a1) pu (M 2 + 2M )1,k
2
Cke = E log2 1+ e
. (25)
pu M (l,k)=(1,ke ) 1,ke l2 ,k + M l=1 l,ke + M pppup l,k l,k
(a2) pu (M 2 + 2M )1,k 2
Ckc = E log2 1+ 2
c
pu (26)
pu (M 2 + 2M ) l=1 l,k c
+ M l = 1 l,k c + M pp p l,k l,k
capacity than the other two cases. As pilot contamination is the dominant factor at high SNR, the
strategy with pilot sequences mutually orthogonal in adjacent
80 cells is optimal.
70 VI. CONCLUSION
sum capacity of users in cell 1
Fig. 5. Sum capacity of users in cell 1 vs. the number of pilot sequences
A PPENDIX
A. Proof of Lemma 1
80
M
|pH q|2 = | pm qm |2
the optimal number of pilot sequences p
70
m=1
60
M
50
= |pm qm |2 + pm1 qm1 pm2 qm2
m=1 m1 =m2
40
= M E[|pm qm | ] + (M 2 M )E[pm1 qm1 pm2 qm2 ]
2
30 = M E[|pm |2 ]E[|qm |2 ]
20
0 10 20 30 40 50 60 70 80
+(M 2 M )E[pm1 ]E[qm1 ]E[pm2 ]E[qm2 ]
= M
Eu (dB)
(29)
Fig. 6. The optimal number of pilot sequences with different SNR
M
p4 = | |pm |2 |2
The optimal number of pilot sequences with different SNR m=1
is given in Fig.6. It is seen that the optimal number of
M
pilot sequences increases slowly with SNR when SNR is = |pm |4 + |gm1 gm2 |2
less than 60dB but increases rapidly to the maximum value m=1 m1 =m2
when SNR is over 60dB. With the increase of SNR, the = M E[|pm |4 ] + (M 2 M )E[|gm1 gm2 |2 ]
effect of pilot contamination becomes much more crucial on
= M E[|pm |4 ] + (M 2 M )E[|gm1 |2 ]E[|gm2 |2 ]
the performance of system, so the optimal number of pilot
sequences increases to reduce such effect. When SNR is over = 3M + (M 2 M )
62dB , our system degenerates to the reuse rate = 3 system. = M 2 + 2M (30)
1492
2015 IEEE Wireless Communications and Networking Conference (WCNC): - Track 3: Mobile and Wireless Networks
ACKNOWLEDGMENT
This work has been supported by the National Natural
Science Foundation of China (No.61171112) and National 863
High-Tech plan (No.2014AA01A704). The numerical calcu-
lations in this paper have been done on the supercomputing
system in the Supercomputing Center of University of Science
and Technology of China.
R EFERENCES
[1] T. L. Marzetta, Noncooperative cellular wireless with unlimited numbers
of base station antennas, IEEE Trans. Wireless Commun., vol. 9, pp.
3590-3600, 2010.
[2] H. Q. Ngo, E. G. Larsson, T. L. Marzetta, Energy and spectral efciency
of very large multiuser MIMO systems, IEEE Trans. on Commun., vol.
61, no. 4, pp. 1436-1449, 2013.
[3] F. Rusek, D. Persson, B. K. Lau, et al., Scaling up MIMO: Opportunities
and challenges with very large arrays, IEEE Sig. Proc. Mag., vol. 30,
no. 1, pp. 40-60, 2013.
[4] E. Larsson, O. Edfors, F. Tufvesson, et al., Massive MIMO for next
generation wireless systems, IEEE Commun. Mag., vol. 52, no. 2, pp.
186-195, 2014.
[5] L. Lu, G. Y. Li, A. L. Swindlehurst, et al., An Overview of Massive
MIMO: Benets and Challenges, IEEE J. Sel. Areas Commun Topics in
Signal Processing Journal, vol. 8, no. 5, pp. 742-758, 2014. 1137-1141,
2012.
[6] J. Jose, A. Ashikhmin, T. L. Marzetta, et al., Pilot Contamination and
Precoding in Multi-Cell TDD Systems, IEEE Trans. Wireless Commun.,
vol. 10, no. 8, pp. 2640-2651, 2011.
[7] H. Q. Ngo, T. L. Marzetta,E. G. Larsson, Analysis of the pilot contami-
nation effect in very large multicell multiuser MIMO systems for physical
channel models, in IEEE Int. Conf. on Acoustics, Speech and Sig. Proc.
(ICASSP), 2011, pp. 3464-3467.
[8] M. Filippou, D. Gesbert,Y. Haifan, Decontaminating pilots in cognitive
massive MIMO networks, in Int. Symposium on Wireless Commun.
Systems (ISWCS), 2012, pp. 816-820.
[9] Y. Haifan, D. Gesbert, M. C. Filippou, et al., Decontaminating pilots in
massive MIMO systems, in IEEE Int. Conf. on Commun. (ICC), 2013,
pp. 3170-3175.
[10] Y. Haifan, D. Gesbert, M. Filippou, et al., A Coordinated Approach to
Channel Estimation in Large-Scale Multiple-Antenna Systems, IEEE J.
Sel. Areas Commun., vol. 31, no. 2, pp. 264-273, 2013.
[11] N. Hien Quoc,E. G. Larsson, EVD-based channel estimation in multi-
cell multiuser MIMO systems with very large antenna arrays, in IEEE
Int. Conf. on Acoustics, Speech and Sig. Proc. (ICASSP), 2012, pp. 3249-
3252.
[12] R. R. Mueller, M. Vehkaperae,L. Cottatellucci, Blind Pilot Decontami-
nation, in 17th Int. ITG Workshop on Smart Antennas (WSA), 2013, pp.
1-6.
[13] Z. Qi,M. Xiaoli, Element-Based Lattice Reduction Algorithms for
Large MIMO Detection, IEEE J. Sel. Areas Commun., vol. 31, no. 2,
pp. 274-286, 2013.
[14] K. Appaiah, A. Ashikhmin,T. L. Marzetta, Pilot Contamination Reduc-
tion in Multi-User TDD Systems, in IEEE Int. Conf. on Commun. (ICC),
2010, pp. 1-5.
1493