Determinant Based Multiuser MIMO Scheduling

with Reduced Pilot Overhead

Kyeongjun Ko and Jungwoo Lee
School of Electrical Engineering and Computer Sciences
Seoul National University, Seoul 151-744, Korea
Email: smuff@wspl.snu.ac.kr and junglee@snu.ac.kr

AbstractIn a multiuser MIMO cellular system where there

are many candidate users, it is critical to select a user group
which maximizes the overall throughput of the system. However,
the optimal scheduling strategy is computationally prohibitive
when the total number of users is large. In this paper, we propose
a determinant based user selection algorithm which reduces the
search complexity without much performance degradation. In
order to reduce pilot overhead by precoding matrix for multiuser
MIMO, we also propose a new pilot scheme with only one
set of pilot, and the new pilot scheme is combined with the
proposed scheduling algorithm. Simulation results show that the
proposed scheduling algorithm with a new pilot scheme reduces
computational complexity and pilot overhead with negligible
performance degradation compared to the exhaustive scheduling
with a conventional pilot scheme which uses two sets of pilots.

I. INTRODUCTION
Multiuser multiple-input multiple-output (MU-MIMO) systems have drawn much attention recently because of higher
capacity compared to single-user MIMO systems [1]. In multiuser single-input single-output (MU-SISO) systems, it was
shown that allocating entire power to the user who has the
best channel is the optimal strategy [2]. On the other hand, in
MU-MIMO systems, it has been known that allocating power
to multiple users simultaneously is better, which is different
from MU-SISO systems.
However, MU-MIMO systems have some practical issues. A
MU-MIMO system needs complicated user selection process
compared to a single-user MIMO (SU-MIMO) system. It
needs to find a user group which maximizes the throughput
while a single-user MIMO system finds only the best user. By
the way, it is not simple to find the best group of users because
a linear precoding matrix to remove inter-user interference
needs to be computed for each candidate group of users. Especially, the optimal user selection scheme is computationally
prohibitive when the total number of users is large in a given
system (cell). Many suboptimal user scheduling algorithms
have been proposed to reduce the computational complexity
of the optimal user selection scheme [3],[4],[9],[10].
Inter-user interference occurs when the resources are allocated to multiple users in MU-MIMO systems. The interuser interference must be eliminated since it limits the sumcapacity of the system. Dirty paper coding (DPC) is the
optimal non-linear precoding technique to avoid the interuser interference, but its complexity is prohibitive [5]. Zeroforcing beamforming (ZFBF) and block diagonalization (BD)

were proposed to reduce the complexity as linear precoding

techniques [6][8].
The linear precoding techniques require the channel state
information (CSI) at the transmitter. The channel estimation
has to be done at each receiver by embedding pilots in the
signal, and the estimated channel has to be fed back to the
transmitter. However, ZF and BD need two sets of pilots. One
is common pilot, which is used to estimate the raw channel.
The other is dedicated pilot, which is used to estimate the
effective channel which is a combined channel of the raw
channel and the precoding matrix. A single-user MIMO system
and a random beamforming system do not need dedicated
pilots since the beamforming matrix is available at the receiver.
But, in the ZF scheme and the BD scheme, the receiver does
not know the beamforming matrix since the beamforming
matrices depend on the selected users, and each receiver does
not know the users selected in transmitter. Therefore, ZF and
BD need dedicated pilots to estimate the effective channel.
Obviously, the data rate will suffer with large pilot overhead
due to the two sets of pilots, which may MU-MIMO less
attractive.
In this paper, we propose two new methods. One is the low
complexity MU-MIMO scheduling, and the other is a new
pilot scheme which uses only one set of (dedicated) pilots.
Simulation results show performance comparison between the
combination of two methods and the existing scheduling and
pilot schemes.
II. SYSTEM MODEL
We consider an MU-MIMO downlink channel with a single
base station (BS) which has M transmit antennas, and KT
users with N receive antennas. We assume that the receivers
estimate their channels perfectly and the BS knows the exact
channel state information of all the users. The broadcast
channel of MU-MIMO system is given by
yk = Hk x + nk ,

k = 1, . . . , KT

(1)

where Hk (k = 1,. . . ,KT ) is the N M channel matrix at

the kth user whose channel entries are independent identically
distributed (i.i.d.) complex Gaussian with zero mean and unit
variance, nk is the complex white Gaussian noise vector for
the kth user whose elements are i.i.d. complex Gaussian with
zero mean and unit variance, and yk is the received signal

978-1-4244-8331-0/11/$26.00 2011 IEEE

vector at the kth user. The transmitted signal, x, is expressed

as
K 

Pi Vi si
(2)
x=
i=1

where K is the number of the simultaneously selected users,

si is the ith symbol vector with E[si 2 ] = 1, Pi is the power
allocated for the ith selected user, and Vi is the precoding
matrix for the ith selected user. Then, (1) can be divided into
the desired signal, the interference signal, and the noise as
follows.
yk =

P k Hk Vk sk + Hk

K



Fig. 1.

Pi Vi si + nk

i=k,i=k

where we assume the user index k is re-indexed properly after

user selection. In the BD scheme, we can express (3) as

(4)
yk = P k Hk Vi si + nk .
III. D ETERMINANT BASED M ULTIUSER MIMO
S CHEDULING A LGORITHM
A. Power Allocation
An MU-MIMO system can be divided into K independent
parallel SU-MIMO in the BD scheme as in (4). Water-filling
based power allocation can be used since BS knows perfect
channel information of all the receivers. When we assume each
receiver receives N streams, there are total N K streams on
which we use water-filling. The capacity with water-filling [9],
[10] is written as


n

P i
i
log2 1 +
(5)
C(H(S), P ) =
M
i=1
where P is the total transmit power, n is the number of total
streams, S is the set of the selected users, H(S) is the set
of
of channels in S, i (i = 1, . . . , n) is the singular value

+

the ith channel of H(S). i satisfies i = PMi , and

n
i=1 i = M . The sum rate of the MU-MIMO system with
BD and water-filling among the selected users can then be
given by

R(S) =
C (H(S)V(S), P )
(6)
S

where V(S) is set of precoding matrix in S.

B. Low Complexity MU-MIMO Scheduling Algorithm
The optimal MU-MIMO scheduling with BD is performed
by
Ropt (S) =

max

S{1, ,KT },1|S|K

R(S)

Volume comparison by determinant

(3)

(7)

Therefore, the optimal scheduling is computationally proof all possible pairs

hibitive if KT  K sincethe number
K K T

.
The
most important
of the selected users is
i=1
i
consideration in selecting users in MU-MIMO systems with
ZFBF or BD is the orthogonality among the channels of the
selected users. If the orthogonality among the selected users is
not large, the beamforming direction of a user is going to be
misaligned from its own channel since the precoding matrix

becomes the null space of channels of the other selected users.

That is, the precoding matrix by ZFBF or BD eliminates not
only the undesired interference signal, but also the intended
signal, and the reduction of the desired signal is roughly
proportional to the distance between its own channel and the
precoding matrix. For example, if the channels among the
selected users are orthogonal to each other, the desired signal
does not suffer degradation because the precoding matrix is
aligned with the channel of the desired signal perfectly. In the
other extreme case, if the channels among the selected users
is in the same direction or plane, the desired signal is lost
completely since the precoding matrix is orthogonal to the
channel of the desired signal.
There exist many low complexity MU-MIMO scheduling
algorithms since the optimal scheduling scheme is too computationally complex to be used in the practical systems
[3],[4],[9],[10]. These schemes achieve performance close to
optimal scheduling algorithm but have high complexity still.
In this paper, determinant is used to estimate orthogonality
and channel power simultaneously in MU-MIMO user selection process with low complexity. The absolute value of
the determinant of a matrix corresponds to the volume of
the parallelepiped spanned by those vectors, which is roughly
proportional to orthogonality and channel power. Therefore,
the determinant based scheme can be regarded as an extended
scheme of Gram-Schmidt orthogonalization in [3]. For example, Figure 1 shows the relationship between orthogonality and
determinant, and computes the volume of an geometric object
spanned by vectors. Note that the volume of the cubic figure
in Fig. 1 (a) is larger than that in Fig. 1 (b).
The first step of the proposed algorithm is to find the user
which maximizes capacity among candidates in T . A next
user that maximizes determinant with the selected user set S is
chosen in a successive fashion without computing a precoding
matrix by SVD. The volume S of a figure composed by a
general m n complex matrix, B, is given by [11]

(8)
S = det(BH B).
The determinant between the channel H(S) of the selected
=
set and the channel of the nth candidate user Hn when H

H(S)
is expressed by
Hn




H(S)H(S)H H(S)HH
H
n

= det
det HH
. (9)
Hn H(S)H
Hn HH
n

TABLE I
D ETERMINANT BASED M ULTIUSER MIMO S CHEDULING A LGORITHM
Step I: Initialization
T = {1, 2, . . . , KT }; S = ;
H(S) = ; R = 0; X = IM ;

A. Optimal Scheduling Algorithm

The order of the optimal scheduling algorithm is given by
 

KT
3
K M
OP O
(12)
K
B. Suboptimal Scheduling Algorithm
We analyze computational complexity of the capacity based
algorithm, the Frobenius norm based algorithm [4], and the
chordal distance based algorithm [10]. The complexity of the
capacity based algorithm is expressed by

Step II: Loop

FOR i = 1 to K
si = arg max det(Hn XHH
n );
nT

Rtemp = R(Stemp ), where Stemp = S + {si };

IF (Rtemp < R)
break;
ELSE

CA < O(KT K 2 M 3 )

(13)

The complexity of the Frobenius norm based is given by

T
H(S) = [H(S)T , HT
si ] ;

F O(KT K 2 M 3 )

X = I H(S)H (H(S)H(S)H )1 H(S);

S = Stemp ; R = Rtemp ; T = T {si };

END-IF
END-FOR

(14)

The complexity of the chordal distance based scheduling

algorithm is
CO O(KT K M 3 ).

To lower computational complexity of (9), we use the following identity.



A B
det
= det(A) det(D CA1 B)
(10)
C D
From (9) and (10), we have


H(S)H(S)H H(S)HH
n
det
=
Hn H(S)H
Hn HH
n
det(H(S)H(S)H )

H
H 1
det(Hn HH
H(S)HH
n Hn H(S) (H(S)H(S) )
n)=

det(H(S)H(S)H )

(15)

C. Determinant based Scheduling Algorithm

As for the determinant based scheduling algorithm, the
complexity is given by

K 

4 3 3 2 13
2
2
DE <
8M N + 8M N + N N + N
3
2
6
i=2

(KT i + 1)


4
1
19
+ 8M N 2 + N 3 N 2 + N KT
3
2
6
O(8M 2 N KT K)
= O(KT M 3 ).

det(Hn (I H(S)H (H(S)H(S)H )1 H(S))HH

n ).
(11)
Note that det(H(S)H(S)H ) in (11) is a common value which
has nothing to do with Hn which is the channel of a candidate
user in T , and it can be skipped in the user scheduling process.
The detailed user scheduling process is described in Table I.
It should be noted that fewer than K users can be selected in
the user scheduling process. This may occur when a new user
which is close to the subspace of existing users reduces the
null subspace in the BD scheme.
IV. C OMPUTATIONAL C OMPLEXITY A NALYSIS
The key contribution of the proposed scheduling algorithm
is to achieve low complexity without sacrificing performance.
The complexity is counted as the number of flops, denoted as
. A flop is defined as a real floating point operation [12]. Each
of a real addition, a multiplication, and a division is counted
as one flop [4]. A complex addition and multiplication have
two flops and six flops, respectively. Although flop counting
cannot show complexity precisely in practical systems, it can
indicate a rough order of the computational complexity.
We assume KT  K, N M, and K =  M
N  in this
section. For an N M complex valued matrix H, the number
of typical matrix operation is representd in [4].

(16)
P
HHH which
The number of flops needed for computing I+ M
2
2
is a positive definite matrix is 8M N +N +N flops. We need
8M 2 N + 8M N 2 flops for Hn Xn HH
n which is also a positive
definite matrix. To reduce computation, we can use Cholesky
decomposition which decomposes a positive definite matrix
into a product of an lower triangular matrix and its hermitian.
With Cholesky decomposition, the determinant of a positive
definite matrix can be computed by
H
det(Hn XHH
n ) = det(LL ).

(17)

The Cholesky decomposition of an n n matrix needs n(n

1) real additions, n(n 1) real multiplications, n(n1)
real
2
divisions, 16 n(n1)(n2) complex additions, 16 n(n1)(n
2) complex multiplications, and n real square root calculations.
To compute the determinant, additional n real multiplications
are needed [9]. Therefore. the number of flops to compute
4 3
3 2
7
det(Hn XHH
n ) is 3 N 2 N + 6 N . The number of flops
for water-filling and the update of X is omitted since they are
dependent not on KT , but on K which is much smaller than
KT .
As shown in the above computational complexity analysis,
the determinant based scheduling algorithm has an advantage
in the systems with large K.

M = 8, N = 2

V. L OW OVERHEAD P ILOT D ESIGN FOR B LOCK

D IAGONALIZATION

50
45
Sum Capacity (bits/s/Hz)

Pilot signal is used for a receiver to estimate its channel

state information, and the pilot overhead decreases the data
throughput. In an open loop SU-MIMO or a closed loop SUMIMO with beamforming matrix, channel is estimated with a
single set of pilots which we call a common pilot.
In the ZFBF and BD schemes, the effective channel cannot
be estimated with common pilot only which is used to estimate
raw channel since receivers do not know the beamforming
matrix for the selected users determined at BS. Thus, another
pilot which we call dedicated pilot is needed to estimate the
effective channel which is obtained by combining the raw
channel with the beamforming matrix. This additional pilot
is an overhead which reduces the overall data throughput. The
requirement of the additional pilot overhead is a disadvtange of
ZFBF and BD although they have performance gain compared
to SU-MIMO or other MU-MIMO systems
We propose a reduced pilot scheme by matrix inversion for
ZF and BD in this section. In the proposed scheme, only the
dedicated pilot is used. We assume block fading channel and
zero feedback delay for simplicity. First, we define M M
unitary pilot matrix, , which is given by

40
35

SNR 20 dB

30

Optimal
Capacity based
Frobenius norm based
Chordal distance based
Determinant based

SNR 10 dB

25
20
15

SNR 0 dB
10
5
5

10

15

20
25
30
Number of total users

35

40

Fig. 2. Sum-capacity comparison of various scheduling algorithms when

M = 8, N = 2

A technical issue in (22) is whether V1 exists or not. If V1

does not exist, the raw channel in (22) may not be obtained.
In order for V1 to exist, V should be an M M square
(18) matrix, and it has to have full rank. As shown in Table I, the
= [T1 T2 TK ]T
determinant based scheduling algorithm does not always select
K users. However, to use this reduce pilot scheme, BS must
for
simplicity.
where i is N M pilot matrix and K = M
N
Here, i (i = 1, . . . , K) are orthonormal matrices each other. select K users for the existence of V1 .
Note that each row of i is a time-domain sequence of length
Our proposed scheduling algorithm selects the users who are
M which is orthogonal to each other. The received pilot signal near-orthogonal to each other, and thus the precoding matrices
at the kth user at time slot t is described by
also tend to near-orthogonal to each other. By well known
properties of an i.i.d. Rayleigh fading matrix channel, the
(t)
(t)
(t)
(t)
(t)
Hk V(t) = Hk [V1 1 + V1 1 + + VK K ]. (19) channel matrix is of full rank with probability one [13],[14]. In
our simulations, it was empirically observed that the inverse of
(t)
where Hk is the channel matrix of the kth user at time slot
V almost always exist, which may be due to the fact that the
(t)
t. Note that V at time slot t is given by
proposed algorithm select near-orthogonal users. In case the
(t)
(t)
V(t) = [V1 VK ] (for the selected users at time slot t 1)inverse does not exist, we can simply keep re-selecting users
until the corresponding V1 exist. This way, the existence
= IM
(otherwise)
of the inverse can be achieved with high probability. Note
(20)
that there is an extreme case where all the users have the
where IM is an M M identity matrix. The channel of the same channel, and V1 may not exist even if we re-schedule
users. In the proposed reduced pilot scheme, only one pilot
kth user at time slot t is estimated by
(dedicated pilot) is used so that the overall throughput is
(t)
(t)
(t)
(t)
Hk (V1 1 + V2 2 + + VK K )H
improved.
 (t) (t) (t) (t)
(t) (t) 
(21)
= Hk V1 Hk V2 Hk VK .
(t)

(t)

Each receiver estimate Hk V(t) instead of Hk by dedicated

pilot , which means that the raw channel is not estimated. In
fact, each receiver does not have to know its own raw channel
since effective channel is enough to decode the data. Note
that BS need to estimate the raw channel of each receiver to
construct the precoding matrices.
The base station estimates the raw channel of each receiver
by multiplying the inverse of the precoding matrix to (21).
 (t) (t) (t) (t)
(t) (t) 
Hk V1 Hk V2 Hk VK V1
= Hk V V1
= Hk

(22)

VI. S IMULATION R ESULTS

Fig. 2 compares sum-capacity of various scheduling algorithms with respect to KT and SNR when M = 8, N = 2.
It also shows an advantage of the determinant based scheduling algorithm. It achieves almost the same performance as
the capacity based algorithm and the Frobenius norm based
algorithm, and it is close to the optimal scheduling algorithm,
and has better performance than the chordal distance based
algorithm for all SNR range. Interestingly, our proposed
algorithm can be considered as a capacity upperbound in high
SNR since the determinant of a matrix is an approximation of
the channel capacity in high SNR.

M = 8, N = 2, Kt = 20

40

determinant based scheduling algorithm has the shortest run

time among the low complexity scheduling algorithms. It has
significant gain over the capacity based algorithm and the
Frobenius norm based algorithm in terms of run time, and it
is slightly better than the chordal distance based algorithm.
However, it achieves better sum capacity than the chordal
distance based algorithm as can be seen in Fig. 2 and Fig.
3.

30

VII. C ONCLUSIONS

Sum Capacity (bits/s/Hz)

80
70

Optimal
Optimal reduced pilot
determinant
determinant reduced pilot

60
50

20
10
0
0

10

15

20

25

30

SNR
Fig. 3. Effect of reduced pilot scheme in optimal scheduling algorithm and
determinant based scheduling algorithm when M = 8, N = 2, KT = 20
M = 8, N = 2, Run Time
35
30

Capacity based
Frobenius norm based
Chordal distance based
Determinant based

Run Time (ms)

25

In this paper, we propose a low complexity MU-MIMO

scheduling algorithm with block diagonalization and a reduced
pilot scheme. The low complexity scheduling algorithm finds
the best user set with the determinant (volume) of a matrix
composed of users channel matrices so that orthogonality
between the users is favored. Its performance is close to the
optimal scheduling algorithm, and has advantage over the
other low complexity scheduling algorithms in terms of both
sum-capacity and computational complexity. The new pilot
scheme reduces overhead, and it uses only one set of pilots
instead of two. In order to use the new pilot scheme, the
associated scheduling algorithm always has to select maximum
possible number of users. The simulations show that the new
single pilot scheme with the associated scheduling algorithm
performs as wells as the conventional two pilot scheme except
for slight degradation at low SNR.

20

R EFERENCES
15
10
5
0
5

10

15

20
25
30
Number of total users

35

40

Fig. 4. Run time comparison of various low complexity algorithms when

M = 8, N = 2

Fig. 3 shows the effect of the reduced pilot scheme on

the optimal scheduling algorithm and the determinant based
scheduling algorithm in terms of sum-capacity. It is observed
that the degradation due to the reduced pilot scheme is almost
negligible in terms of sum capacity. There is a slight gap in low
SNR between the two pilot scheme and the one pilot scheme,
but the two schemes have almost the same performance in high
SNR. The small degradation in the low SNR regime is due to
the fact that the number of the selected users is fewer than
K at low SNR in the scheduling algorithm with two pilots,
and it may be better to choose fewer than K users at low
SNR. In the reduced pilot scheme, K users always have to
be selected, which causes a minor degradation. On the other
hand, K users tend to be selected at high SNR even in the
scheduling algorithm with two pilots.
Fig. 4 compares run time of various low complexity algorithms when M = 8, N = 2. It is observed that the

[1] N. Jindal and A. Goldsmith, Dirty-paper coding versus TDMA for

MIMO broadcast channels, IEEE Trans. Inform. Theory, vol. 51, no.
5, pp. 1783-1794, 2005.
[2] R. Knopp and P. Humblet, Information capacity and power control in
single cell multiuser communications, in Porc. Of IEEE ICC, pp. 331335, 1995.
[3] T. Yoo and A. J. Goldsmith, Optimality of zero-forcing beamforming
with multiuser diversity, in Proc. IEEE Int. Cont. Communications, vol.
1, pp. 542-546, May 2005.
[4] Z. Shen, R. Chen, J. G. Andrews, J. R. W. Heath, and B. L. Evans, Low
complexity user selection algorithms for multiuser mimo systems with
block diagonalization, IEEE Trans. Signal Process., vol. 54, no. 9, pp.
3658-3663, 2006.
[5] M. Costa, Writing on dirty paper, IEEE Trans. Inform. Theory, pp.
439-441, May, 1983.
[6] C. B. Peel, B. M. Hochwald, and A. L. Seindlehurst, A vectorperturbation technique for near- capacity multiantenna multiuser communication - part I: channel inversion and regularization, IEEE Trans.
Inform. Theory, vol. 49, no. 7, pp. 1691-1706, Jul. 2003.
[7] L. U. Choi and R. D. Murch, A transmit preprocessing technique for
multiuser mimo systems using a decomposition approach, IEEE Trans.
on Wireless Commun., vol. 3, no. 1, pp. 20-24, 2004.
[8] Q. H. Spencer, A. L. Swindlehurst, and M. Haardt, Zero-forcing methods
for downlink spatial multiplexing in multiuser mimo channels, IEEE
Trans. on Signal Process., vol. 52, no. 2, pp. 461-471, 2004.
[9] X. Zhang and J. Lee, Low complexity multiuser MIMO scheduling with
channel decomposition using capacity upperbound, IEEE Trans. Comm.,
pp. 871-876, June. 2008.
[10] K. Ko and J. Lee, Low complexity multiuser MIMO scheduling with
chordal distance, in. Proc. CISS, pp. 80-84. Mar. 2008.
[11] http://en.wikipedia.org/wiki/Determinant, wikipedia.
[12] A. Ben-Israel and T. N. E. Greville, Generalized Inversed: Theory and
Applications. New York: Wiley, 1977.
[13] N. Jindal, Antenna combining for the MIMO downlink channel, IEEE
Trans. Wireless Comm., pp. 3834-3844, Oct., 2008.
[14] A. Tulino and S. Verdu, Random matrix theory and wireless communications, Foundations and Trends in Communications and Information
Theory, vol. 1, no. 1, 2004.