You are on page 1of 8

Available online at www.sciencedirect.

com

ScienceDirect
Available online at www.sciencedirect.com
Procedia Computer Science00 (2018) 000–000 www.elsevier.com/locate/procedia
ScienceDirect
Procedia Computer Science 131 (2018) 72–79

8th International Congress of Information and Communication Technology (ICICT-2018)

A Novel Pilot Decontamination Scheme for Uplink Massive MIMO


Systems
Weina Yuan一, Xinkai Yang, Rui Xu
School of Information Science and Engineering, East China University of Science and Technology. Shanghai, China

Abstract

The pilot contamination causing by the reuse in adjacent cells in massive MIMO for 5G is an important issue to deal with. In our
paper, a novel pilot decontamination scheme based on two existing schemes (soft pilot reuse and weighted-graphed-coloring-
based pilot decontamination) is proposed. All users are firstly separated into two categories: cell centered users that reuse the
same pilot sequences and cell edged users that use other orthogonal pilot subgroups to get rid of the severe contamination. But
the slight contamination among the cell centered users still exists. Then, in order to improve the decontamination of cell centered
users, a weighted-graph-based method is applied. With such combination, the proposed scheme is able to mitigate the
contamination and improve the quality of communication significantly, which is verified by simulation results.
© 2018 The Authors. Published by Elsevier Ltd.
© 2018 The Authors. Published by Elsevier B.V.
This is an open access article under the CC BY-NC-ND license (https://creativecommons.org/licenses/by-nc-nd/4.0/)
Peer-review
Selection andunder responsibility
peer-review of organizing committee
under responsibility of thecommittee
of the scientific 8th International Congress
of the 8th of Information
International Congressand Communication
of Information and
Technology (ICICT-2018).
Communication Technology.
Keywords:Massive MIMO, Pilot contamination, Soft pilot reuse, Graph coloring problem.

1. Introduction

With the development of communication technologies, the upcoming 5th Generation (5G) mobile communication
systems are excepted to fulfill the requirements of mobile services in the future. In order to fulfill the demand of
capacity and reliability in 5G, the massive multi-input-multi-output (MIMO) system has been investigated1,2,3, in
which every base station (BS) is loaded with a great quantity of antennas to serve a few users. It offers several
advantages such as less energy consumption of BSs, less intracell interference and noise, more robustness and low

* Corresponding author. Tel.: +86-137-6126-9049


E-mail address:wnyuan@ecust.edu.cn

2018 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license
https://creativecommons.org/licenses/by-nc-nd/4.0/)
Selection and peer-review under responsibility of the scientific committee of the 8th International Congress of Information and Communication
Technology

1877-0509 © 2018 The Authors. Published by Elsevier Ltd.


This is an open access article under the CC BY-NC-ND license (https://creativecommons.org/licenses/by-nc-nd/4.0/)
Selection and peer-review under responsibility of the scientific committee of the 8th International Congress of Information and
Communication Technology
10.1016/j.procs.2018.04.187
Weina Yuan et al. / Procedia Computer Science 131 (2018) 72–79 73
2 W. YUAN / Procedia Computer Science00 (2018) 000–000

complexity precoding, but it suffers severe pilot contamination (PC) problem because of the reuse of the limited
pilots among the adjacent cells4.
Plenty of efforts have been made to solve this challenging problem. The performance of channel estimation (CE)
and the condition of pilot contamination in massive MIMO system are analyzed5,6. A simple practical channel
estimator and a data-aided MMSE estimation are presented, both of them address the performance under
moderate/serious PC without previous knowledge7,8. But they all haven’t taken pilot reuse scheme into consideration.
Since CE requires pilot resources while the resources are limited but should be reused in multicell scenarios which
results in PC, a modified pilot reuse scheme to mitigate PC is also necessary. Pilot sequences are designed to be
orthogonal to each other and assigned for both single cell and neighbor cell9, so that the interference only occurs
between cells. Allocation schemes based on coloring graph and weighted coloring graph (WGC-PD) are proposed
respectively10,11. Those methods not only can reduce PC, but also achieve large uplink rate. In addition, a kind of
pilot reuse scheme so called Soft Pilot Reuse Scheme (SPRS) is proposed, in which the users are separated into two
categories: cell centered users and cell edged users. The cell centered users (likely subjected to slight PC) can use
the same pilot resource while the cell edged users (likely subjected to serious PC) in adjacent cells apply other cell-
edge pilot subgroups12, which probably enables to promote the pilot decontamination due to the orthogonality.
A novel pilot decontamination scheme based on SPRS and WGC-PD is proposed in this paper. In the proposed
scheme, SPRS is firstly applied to divide users and cells generally, then the focus is on the pilot decontamination
among the cell centered users, and WGC-PD is applied to solve this problem more thoroughly than the previous
solution so that the performance is better in uplink massive MIMO.
The remainder of this article is structured as follows: Section2 presents the system model and explains the new
pilot decontamination scheme based on SPRS and WGC-PD. In Section3, simulation results and the relating
analyses are given. And the conclusion of this paper is presented in Section4.

2. Experimental and Computational Details

2.1. System Model

In this article, an uplink multicell MIMO system is considered. The system includes J cells. Each cell has a BS
which is loaded with N antennas providing service to K (K<<N) single antenna users. The channel impulse response
vector connecting the kth user in the jth cell to the BS of the ith cell can be written as
hi ,k , j  bi ,k , j  i ,k , j (1)
where b i , k , j and  i , k , j represent the small-scale fading coefficient vector and the large-scale fading coefficient
respectively. b i , k , j obeys the distribution of  (0, I N ) and  i , k , j is equal among different antennas in the same BS.

Fig. 1. Example of multicell massive MIMO


74 Weina Yuan et al. / Procedia Computer Science 131 (2018) 72–79
W. YUAN / Procedia Computer Science00 (2018) 000–000 3

The BSs’ channel information can be acquired by pilot-aided CE. But, pilot resources are limited, the same set 
of orthogonal pilot sequences with length l ought to be reused and that no such pilots are reused in one single cell
should be guaranteed. Hence, the available Q (Q≥K) pilots’ matrix is Φ  1 ,  2 ,...,  Q   Ql whose column
vectors are orthonormal to each other so that Φ H Φ =I . To show the pilot assignment, a variable a j , k  1, 2,..., Q is
introduced to denote that the pilot sequence  a is assigned to the user uk , j , Then, A  [a j , k ] .
j ,k

Thus, the uplink communication system can be written as 12


YiP   p  j 1 H i , j Φ  N
J
(2)
where Y is the signal of pilot sequences received at the ith BS, H i , j is the channel matrix composed of (1), N
i
P

represents the noise matrix and  p is the transmission power. Due to the reuse of the same pilot, the estimation Hˆ i , j
includes the channels of the other cells’ users, which is so called PC. Thus, if a pilot assignment matrix A is
assumed, the uplink SINR of uk , j can be formulated by :
4
h Hj , k , j
SINR= 4
 j ', k '  j , k
h Hj , k , j + 2j , k /  2
(3)
 2j , k , j

 j ', k '  j , k
 2j ', k ' , j

where j ', k '   j,k


represents the users in different cells using the same pilot as uk , j ,  2j , k denotes the uncorrelated
noise and interference, and  j ', k '  j , k
 2j ', k ' , j is the PC. Hence the related average achievable rate R of uplink uk , j is
formulated as:
R  (1  0 ) E{log 2 (1  SINR)} (4)
where μ0 is the fraction of spectral efficiency loss allocated to channel estimation.
It is clear that our mission is to maximize the total communication rate R. So, the optimization problem P1 is:
max
ak , j
 j,k
log 2 1  SINR   (5)

Fortunately, P1 can be rewritten as P2 approximately11 :


   2j , k , j 

max  log 2 1   (6)
ak , j j,k
   2

  j ', k '  j , k j ', k ' , j

The exhaustive search method of P2 is as follows: the number of different kinds of pilot allocations in the jth cell
for K users and Q pilots is PQK = Q ! (Q  K )! , but considering all the cells, the assignment can be as huge as ( PQK ) J 1 ,
which is infeasible in reality11. So, an optimized scheme is desired.

2.2. The Proposed Scheme

In the ith cell, the Ki users are firstly divide into two groups: the cell centered users and the cell edged users based
on the coefficient  i , k ,i , which is12
?  Yes  center users
 i2,k ,i  i   (7)
 No  edge users
and the threshold value i can be calculated as
 K 2
i   i ,k ,i
K k 1
(8)

where  is adjustable according to the real configuration12. Thus, a simple example of the division is illustrated in
Fig.2.
Weina Yuan et al. / Procedia Computer Science 131 (2018) 72–79 75
4 W. YUAN / Procedia Computer Science00 (2018) 000–000

Fig. 2. An illustration of the users’ division.

Therefore, we introduce the SPRS in which the whole orthogonal pilot resource set is divided into two subsets,
assigning to the cell edged users and cell centered users respectively. Furthermore, the latter one is reused by cell
centered users in all cells. According to (7),
K i  K i ,c  K i ,e (9)
where Ki,c and Ki,e represent the number of cell centered users and cell edged users in the ith cell. Define
K ct  max K i ,c , i  1, 2,..., J  and K ed   i 1 K i ,e . Thus, in this proposed scheme, the total number of the required
J

orthogonal pilot sequences is


K all  K ct  K ed (10)
Then the available Φ can be assigned as
T
Φ all  ΦTct ΦTed  (11)
in which Φ ct   K l is the pilots for cell centered users and Φ ed   K l is the pilots for cell edged ones. In addition,
ct ed

the matrix Φ ed can be separated into J row partitions, corresponding to the Ki,e cell edged users. Thus, the
orthogonality of the pilots of all the cell edged users is guaranteed, so that the quality of the communication service
for them can be improved significantly.
So, since the pilot sequences in the single cell are orthogonal to each other and therefore no PC exists, and the
cell edged users are free from very severe PC due to SPRS, the achievable decontamination mainly comes from
eliminating the interference among different cell centered users reusing the same pilot resource. We firstly introduce
a variable  j , k j ', k ' which is used to measure the intensity of the PC between any two users, which is :
 j , k j ', k '   2j ', k ' , j /  2j , k , j   2j , k , j ' /  2j ', k ' , j ' (12)
The larger τ is, the more serious PC will take place between these two users when the same pilot resource is reused
by them.
And mathematically, the MIMO interference can be translated as a graph theory problem: WG  U ,W  , where U
represents the total users and W denotes the potential PC among them which is measured by τ11. In our proposed
scheme, the U and W only adapt to the cell centered users. To achieve the reduction of PC and the high rate of
uplink transmission, the WGC-PD is introduced. The main idea of the WGC-PD is based on greedy algorithm in
which the larger the weight W is, the prior these two users are assigned with pilot sequences to achieve the minimum
PC. The concrete algorithm is shown in Table.1.

Table 1.The algorithm of WGC-PD


Input: Parameters: K, J, and Q
WG  U ,W 

Output: Pilot allocation: a j , k  


1  j  J ,1  k  K
Initialization:
1:  j1 , k1 , j2 , k2   arg max  j , k , j ', k ', j  j '  j,k j ', k '
76 Weina Yuan et al. / Procedia Computer Science 131 (2018) 72–79
W. YUAN / Procedia Computer Science00 (2018) 000–000 5

 
2: a j , k  0, a j1 , k1  1, a j2 , k 2  2

3:    j1 , k1 , j2 , k2 
While a j , k  0 do:

4:  j , k   j ', k '  , j  j '


 j,k j ', k '

5: j0 , k0  arg max j,k  j,k


: j, k   

6:   q : k , a j0 , k  q,1  q  Q 
7:  q   j , k  , a j ,k  s
 j0 , k0 j ,k

8: a j0 , k0  arg min q q : q    
9:      j0 , k0 
End and return a j , k  
Step 1-3: the initialization is executed where the two users with the largest weight are selected and they are
merged into the assigned set Ω. After that, the other users are assigned with pilots until a j , k  0 . Step 4-5: the users
are selected in the order of  j , k .  j , k denotes the sum of the weighted-edges connecting uk , j with the other cells’
users within Ω. The uk , j with the largest weight sum will be preferentially considered in the next stage. Step 6-9:
0 0

selecting the pilot allocation that causes the minimum potential PC of uk , j (which is measured by  q ) from the 0 0

available pilot resources  , which represents the unused pilots in the j0th cell. After all, j0 , k0 is added into Ω.
Finally, return a j , k  at the end of the loop.
So, combine these two schemes, the novel proposed pilot decontamination scheme is able to mitigate PC
significantly and that can be supported by the simulation results in the next chapter.

3. Results and Discussion

Hence, the performance of our proposed scheme is simulated through Monte Carlo methods. The parameters in
simulation are listed in Table.2.

Table 2.Parameters in simulation


Number of users in one cell Ki 10
Number of cells J 19
Number of pilot sequences Q K≤Q≤KJ
15 if fixed
Loss of spectral efficiency μ0 0.05
Antenna’s number in a BS N 32~2048
512 if fixed
The threshold parameter θ 0.05≤θ≤1
0.1 if fixed
Transmission power  p 5~30 dB
15 if fixed
The shadowing fading  8 dB
6 W. YUAN / Procedia Computer Science00 (2018) 000–000
Weina Yuan et al. / Procedia Computer Science 131 (2018) 72–79 77

Fig. 3. CDF of uplink SINR for different schemes

Fig.3 shows the cumulative distribution function (CDF) of uplink SINR for different schemes. It is obvious that
the proposed scheme so called SPRS+WGC-PD, attains a fairly higher SINR for uplink users than both of SPRS12
and WGC-PD11 scheme respectively. The Random pilot scheme performs worst without doubt. In other words, with
the same SINR value SINR0, the order of the values of CDF is Random>SPRS>WGC-PD> SPRS+WGC-PD, which
means the proposed SPRS+WGC-PD scheme has the least users whose SINR are less than SINR0.

Fig. 4. CDF of uplink achievable rate for different schemes

Fig.4 shows the relationship between CDF and rate R. It can be concluded that the proposed scheme is superior to
the existing schemes. The meaning of this is the same as explained above. And when R comes to about 6 bps/Hz, the
CDF curves are all nearly reaching to 1.
W. YUAN / Procedia Computer Science00 (2018) 000–000 7
78 Weina Yuan et al. / Procedia Computer Science 131 (2018) 72–79

Fig. 5. Average uplink achievable rate R against N for different schemes

Fig.5 shows the curves of rate R against N. It is obvious that with the increase of N, the rates of these four
schemes all rise. But the proposed scheme’s curve rises fastest. The next two schemes are WGC-PD and SPRS. And
in comparison, the Random one doesn’t perform ideally.

Fig. 6. Average uplink achievable rate R against δ for different schemes

Fig.6 shows the curves of the rate R against the transmission power δ. All the considered schemes can improve
the rate R when δ increases. At the same time, obviously, the effect of the SPRS+WCG-PD scheme is the best.

4.Conclusions

In this paper, we have proposed a novel pilot decontamination scheme which combines the two existing schemes:
SPRS and WGC-PD scheme. The uplink massive MIMO users are separated into cell centered users and cell edged
users, and their pilot assignments are generally based on SPRS. Furthermore, WGC-PD is introduced to improve the
decontamination of the cell centered users. Then, according to the simulation results, this combination outperforms
than those existing schemes in many aspects. Thus, the proposed scheme has a good prospect for application in the
massive MIMO system.
Weina Yuan et al. / Procedia Computer Science 131 (2018) 72–79 79
8 W. YUAN / Procedia Computer Science00 (2018) 000–000

Acknowledgements

This work is supported by the National Natural Science Foundation of China (Grant No.61501187), and the
Fundamental Research Funds for the Central Universities.

References

1. Alliance, N. G. M. N., "5G white paper. " Next generation mobile networks, white paper (2015)
2. A. Gupta and R. K. Jha, "A Survey of 5G Network: Architecture and Emerging Technologies," in IEEE Access, vol. 3, pp. 1206-1232, 2015.
doi: 10.1109/ACCESS.2015.2461602
3. E. Björnson, L. Sanguinetti, J. Hoydis and M. Debbah, "Optimal Design of Energy-Efficient Multi-User MIMO Systems: Is Massive MIMO
the Answer?," in IEEE Transactions on Wireless Communications, vol. 14, no. 6, pp. 3059-3075, June 2015. doi:
10.1109/TWC.2015.2400437
4. F. Rusek et al., "Scaling Up MIMO: Opportunities and Challenges with Very Large Arrays," in IEEE Signal Processing Magazine, vol. 30, no.
1, pp. 40-60, Jan. 2013. doi: 10.1109/MSP.2011.2178495
5. H. Q. Ngo, T. L. Marzetta and E. G. Larsson, "Analysis of the pilot contamination effect in very large multicell multiuser MIMO systems for
physical channel models," 2011 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), Prague, Czech
Republic, 2011, pp. 3464-3467. doi: 10.1109/ICASSP.2011.5947131
6. A. Zaib, M. Masood, A. Ali, W. Xu and T. Y. Al-Naffouri, "Distributed Channel Estimation and Pilot Contamination Analysis for Massive
MIMO-OFDM Systems," in IEEE Transactions on Communications, vol. 64, no. 11, pp. 4607-4621, Nov. 2016. doi:
10.1109/TCOMM.2016.2593924
7. F. A. P. De Figueiredo, F. A. C. M. Cardoso, I. Moerman and G. Fraidenraich, "Channel Estimation for Massive MIMO TDD Systems
Assuming Pilot Contamination and Frequency Selective Fading," in IEEE Access, vol. 5, pp. 17733-17741, 2017. doi:
10.1109/ACCESS.2017.2749602
8. M. Ju, L. Xu, L. Jin and D. Defeng Huang, "Data aided channel estimation for massive MIMO with pilot contamination," 2017 IEEE
International Conference on Communications (ICC), Paris, 2017, pp. 1-6. doi: 10.1109/ICC.2017.7996889
9. M. AlKhaled and E. Alsusa, "Impact of pilot sequence contamination in massive MIMO systems," in IET Communications, vol. 11, no. 13, pp.
2005-2011, 9 7 2017. doi: 10.1049/iet-com.2017.0062
10. X. Zhu, L. Dai and Z. Wang, "Graph Coloring Based Pilot Allocation to Mitigate Pilot Contamination for Multi-Cell Massive MIMO
Systems," in IEEE Communications Letters, vol. 19, no. 10, pp. 1842-1845, Oct. 2015. doi: 10.1109/LCOMM.2015.2471304
11. X. Zhu, L. Dai, Z. Wang and X. Wang, "Weighted-Graph-Coloring-Based Pilot Decontamination for Multicell Massive MIMO Systems,"
in IEEE Transactions on Vehicular Technology, vol. 66, no. 3, pp. 2829-2834, March 2017. doi: 10.1109/TVT.2016.2572203
12. X. Zhu et al., "Soft Pilot Reuse and Multicell Block Diagonalization Precoding for Massive MIMO Systems," in IEEE Transactions on
Vehicular Technology, vol. 65, no. 5, pp. 3285-3298, May 2016. doi: 10.1109/TVT.2015.2445795

You might also like