Beruflich Dokumente
Kultur Dokumente
K (1)
, 2
1
,
M
k m
m
M k
=
K (2)
Let { }
, ,
0,1
k m n
be the RB allocation indicator, where
, ,
1
k m n
= represents that the n
th
RB in the m
th
CC is
allocated to the k
th
user, and otherwise,
, ,
0
k m n
= . In order
to avoid co-channel interference, each RB should be
assigned to at most one user in each frame. Thus,
, ,
( )
1, ,
k m n
k m
m n
K
M N (3)
where K(m) is the feasible set of users for the m
th
CC. Due to
the various carrier aggregation capabilities of users, the
feasible sets of users for different CCs are possibly distinct
from each other.
III. PROBLEM FORMULATION
A utility function can be used for balancing the efficiency
and fairness. It maps the RBs that a user utilizes into a real
number in order to evaluate the satisfaction of users. It is
well-known that if the system utility is defined as the sum of
the logarithmic user throughput, then Proportional Fair (PF)
algorithm is the optimal algorithm for the maximization of
system utility. Here, the utility function of the k
th
user in the
t
th
frame is defined as
[ ] [ ]
, ,
( ) ( , )
k k m n
m k n k m
U t U t
=
M R
(4)
Spectrum
Time
CC 1
CC 2
CC 3
CC 4
NB user1
NB userK1
.
.
.
BB user1
.
.
.
BB userK2
CC
allocation
Control phase
Frame
RB assigned to NB user1
RB assigned to BB user1
RB
allocation
197 197 197 197
where [ ]
, , k m n
U t represents the utility of the k
th
user on the n
th
RB of the m
th
CC in the t
th
frame, M(k) is the set of CCs
assigned to the k
th
user, and R(k,m) is the set of RBs in the
m
th
CC allocated to the k
th
user. Since scheduling is at the
beginning of each frame, we omit the time index t for
notation simplicity.
We formulate the CC and RB joint allocation problem as
a constrained optimization model, denoted by P1, with the
objective of maximizing the system utility. P1 is a non-linear
integer programming problem. An exhaustive search over all
combinations is computationally impossible when the value
of K, M and N are non-trivial. The objective of our work
aims at a suboptimal resource allocation algorithm that can
obtain the near-optimal solution of P1.
{ }
{ }
, , ,
, , , , ,
1 1 1
{ }{ }
, 1
1
, 2
1
, ,
( )
,
, ,
P1: max
. .
1,
,
1, ,
0,1 , ,
0,1 , ,
k m k m n
K M N
k m n k m k m n
k m n
M
k m
m
M
k m
m
k m n
k m
k m
k m n
U
s t
k
M k
m n
k m
k
= = =
=
=
K
K
K
M N
K M
K , m n M N
(5)
IV. MSUL ALGORITHM
To reduce the computational complexity of the optimal
solution of P1, we propose a suboptimal resource algorithm,
referred to as the MSUL (Minimizing System Utility Loss)
algorithm. In the MSUL algorithm, a relaxed optimization
problem is first solved to obtain the upper bound of the
optimum of P1. Then, based on the optimal solutions of the
relaxed optimization problem, a iterative adjustment
algorithm is performed, until all the NB users have met the
carrier aggregation capability constraint. In each adjustment
iteration, if a NB user is selected, then it should adopt one
CC for the maximum utility and releases the RBs on the
other CC. The released RBs will be reassigned among the
other feasible users, which may decrease the utility of these
RBs. The decreasing utility is called as the lost of system
utility. We need to minimize the loss of system utility
incurred by the adjustment process.
It is supposed that the NB users can utilize all the CCs.
Then, ( ) ,
B
m m = K K M , with
1 2 B
= K K K * .
Accordingly, P1 can be reduced to the following relaxed
optimization problem, denoted by P2.
{ }
, ,
, , ,
1 1 1
{ }
, ,
( )
, ,
P2: max
. .
1, ,
0,1 , , ,
k m n
K M N
k m n k m
k m n
k m n
k m
k m n
U
s t
m n
k m n
= = =
K
M N
K M N
(6)
Obviously, P2 can be divided into M subproblems for
each CC, where each CC optimizes its RB allocation
independently. For the m
th
CC, the n
th
RB is allocated to the
user k
*
which achieves the maximum utility on this resource
block, that is
*
, ,
( )
( , ) arg max
k m n
k m
k m n U
=
K
(7)
It is worth mentioning that the optimum of P2 is the upper
bound of that of P1, since the constraint that each NB user
can be assigned on only one CC is relaxed in P2.
Based on the optimal solutions of the relaxed
optimization problem, iterative adjustment process is
performed with the constraints on the carrier aggregation
capabilities of NB users. The iterative process runs at most
K
1
turns. In every iteration, among the remaining unqualified
NB users that do not satisfy the carrier aggregation capability
constraint, the user k
A
with the minimum loss of the system
utility
min
k
U is selected to adjust its resource. This
procedure will iterate until all the NB users have met the
carrier aggregation capability constraint.
min
arg min
A
A k
k
k U
=
K
(8)
where
min
,
( )
min
k k m
m k
U U
M
(9)
and
, , ,
( )\ ( , )
, ,
( ) \ ( , )
( ) \
max
k m k l n
l k m n k l
p l n
l k m n R k l
p l k
U U
U
M R
M
K
(10)
The
, k m
U represents the resulted loss of the system
utility, when the k
th
NB user chooses the m
th
CC as its
working spectrum whereas reallocates the RBs
in ( , ), ( ) \ k l l k m R M to other users. The RBs reallocation
decision is made as follows
, ,
( ) \
( , ) arg max , ( ) \ , ( , )
k l n
k l k
k l n U l k m n k l
=
K
M R (11)
On the other hand, among the CCs in M(k), the CC m(k)
with minimum
, k m
U is selected as the working spectrum of
the k
th
NB user, thus
,
( )
( ) arg min
k m
m k
m k U
=
M
(12)
Consequently, the steps of MSUL are described in
TABLEI.
TABLE I. MSUL ALGORITHM
1. Initialization:
1 2 B
= K K K * , and ( ) ,
B
m m = K K M;
2. Solve the P2 according to (7) to
obtain
*
, k m
and
*
, , k m n
;
* *
( , ), , , ,
1, 0, ( , )
k m n m n k m n
k k m n = =
* *
( , ), ,
1, 0, ( , )
k m n m k m
k k m n = =
3. Update the following sets based on
*
, k m
and
*
, , k m n
:
{ }
*
,
( ) 1,
k m
k m m = = M M
{ }
*
, ,
( , ) 1,
k m n
k m n n = = R N
198 198 198 198
{ }
*
, 1 2
1
1,
M
B k m
m
k k
=
= >
K K K *
4. Update the set of NB users requiring resource
adjustment { }
1
|
A B
k k K K K , and if
A
K , then go to Step 5; otherwise, the
algorithm exits;
5. Based on (8), (9) and (10), select a NB user
A
k for
resource adjustment, and choose its working CC
according to (12);
6. According to (11), reallocate the resource released
by the selected NB user
A
k to other users and
update
*
, k m
and
*
, , k m n
;
'
*
,
'
1 ( , )
( ) \ ( ), ( , )
0 ( , )
p l A A A
p k l n
l k m k n k l
p k l n
=
=
M R
'
*
, ,
'
1 ( , )
( ) \ ( ), ( , )
0 ( , )
p l n A A A
p k l n
l k m k n k l
p k l n
=
=
M R
7. Update the feasible set of users for CCs, such that
{ } ( ) ( ) \ , ( )
A A
m m k m m k = K K , and go to Step 3.
V. PERFORMANCE EVALUATION
To verify the performance of the proposed algorithms,
simulation experiments of downlink multi-carrier are
conducted via MATLAB. A cell scenario with the radius 1
km is considered. One BS locates at the center of the cell.
The users are uniformly dropped within the cell coverage.
Both NB and BB users are full-buffered. There are 4 CCs to
be aggregated in the BS. The bandwidth of each CC is set to
be 25MHz. Let each CC contain 64 RBs. The wireless
channel is modeled including distance-dependent path loss,
shadowing fading and the small scale fading. The simulation
parameters are listed in TABLE II. The performance of the
MSUL algorithms is evaluated in terms of the network
utility, the average throughput of users, the fairness of users,
and the efficiency of RB usage.
TABLE II. SIMULATION PARAMETERS
Parammeter Setting / description
CC aggregation pattern 4 CCs with 25MHz per CC
Number of RBs per CC 64
Path loss model Okumura-Hata model
Shadowing standard deviation 7 dB
Thermal noise spectral density -174 dB/Hz
Small scale fading distribution Rayleigh distribution
Moreover, the utility of the k
th
user on the n
th
RB of the
m
th
CC in the t
th
frame is defined by
[ ]
, , , ,
( ) ( )
k m n k m n k
U t r t r t = (13)
In (13),
, ,
( )
k m n
r t is the instantaneous throughput of user k on
the n
th
RB in the m
th
CC at the t
th
frame. ( )
k
r t is the average
aggregated throughput of user k up to time t, which is
updated as follow
, ,
( ) ( , )
( ) (1 1 ) ( 1) 1 ( )
k k k m n
m k n k m
r t W r t W r t
= +
M R
(14)
It is shown in [10] that with sufficient large value of t,
( )
k
r t weakly converges to the stationary throughput of user k.
Applying the utility function (13), the RB allocation rule
(7) can be rewritten by
*
, ,
( )
( , ) arg max ( ) ( )
k m n k
k m
k m n r t r t
=
K
(15)
Equation (15) is the classic Proportional Fair (PF)
criterion which is proposed by Kelly [9][11] take the balance
between maximizing the system throughput and maintaining
fairness among all users. In addition, it has been
demonstrated in [12] that (15) converges to the optimal
solution of NUM (Network Utility Maximization) problem
where the network utility is defined as the sum of the
logarithmic of user throughput.
We compare the performance of MSUL algorithm with
cross-CC-PF algorithm, which improves the performance of
PF criterion in multi-carrier system. Reference [8] proves
that cross-CC-PF maximizes network utility if the set of
working CCs for each user is predetermined. Both MSUL
and cross-CC-PF can be considered to belong to PF
algorithm.
The comparison of the network utility versus the number
of BB users between MSUL and Cross-CC-PF is illustrated
in Fig.2. During the simulation, the total number of users of
20 is fixed, but the ratio of the BB users varies from 0 to
100%. Fig.2 shows that Cross-CC-PF suffers lower network
utility than MSUL with a mixture of both NB and BB users.
The network utility increment is achieved by MSUL,
because there are more NB users to be selected in each CC,
and hence more multi-user diversity gains to be exploited.
Furthermore, from Fig.2 we can find that the network utility
increases with the ratio of BB users. The reason is that BB
user operating on all CCs can achieve more frequency
diversity gain than NB user dose. When the ratio equal 1,
there are only BB users in the network. That means every
CC could be assigned to any user in the network. In this case,
MSUL attains the same network utility as Cross-CC-PF,
which has been proved to be the optimal scheduler once the
CC assignment of each user is determined.
100 90 80 70 60 50 40 30 20 10 0
250
255
260
265
270
275
280
285
290
295
300
Ratio of BB Users (%)
N
e
t
w
o
r
k
U
t
ilit
y
(
S
u
m
o
f
ln
(
R
a
t
e
)
)
MSUL
Across-CC-PF
Figure 2. Network utility with the different ratios of BB users
Fig. 3 shows the CDF (cumulative distribution function)
of the average throughput for different categories of users. In
199 199 199 199
this scenario, 50% of users are NB users. It clearly shows
that the CDF curves of MSUL are closer than those of Cross-
CC-PF, and the average throughput of NB users of MSUL is
higher than that of Cross-CC-PF. This indicates that the
MSUL algorithm not only improves the fairness between NB
and BB users, but also significant increases the average
throughput of NB users. This is due to the fact that MSUL
expands the feasible set of user for each CC, so that it could
assign each NB user to the good CC with the higher utility,
compared with Cross-CC-PF.
0 0.84 1.89 2.94 3.99
0
0.2
0.4
0.6
0.8
1
User Throughput(Mbps)
C
u
m
m
u
la
t
iv
e
D
is
t
r
ib
u
t
io
n
Across-CC-PF: BB user
Across-CC-PF: NB user
MSUL: NB user
MSUL: BB user
Figure 3. CDF for NB and BB users with different algorithms
In order to evaluate the efficiency of a CC to be used, we
examine the throughput ratio of CC, which is defined as the
ratio of the actual achieved throughput over the supremum
throughput of the CC. With the feasible set of users for each
CC, the supremum throughput of a CC is obtained according
to (7). The ratio can be defined as
_ ,max
( )
m m
CC th n n
n n
R m R R
(16)
Where
m
n
R is the actual attained throughput of the n
th
RB in
the m
th
CC by the algorithm, and
,max
m
n
R is the supremum
throughput of the n
th
RB in the m
th
CC.
Fig. 4 illustrates the CDF of the average CC throughput
ratio using the two algorithms with the equal number of NB
and BB users. As shown in Fig. 4, MSUL exploits each CC
more efficiently than Cross-CC-PF does. This is because
that, MSUL tends to schedule a user with better channel
quality on a CC to achieve higher throughput by jointly
scheduling CCs and RBs.
0 0.2 0.4 0.6 0.8 1
0
0.2
0.4
0.6
0.8
1
Average Throughput Ratio of CC
C
u
m
m
u
la
t
iv
e
D
is
t
r
ib
u
t
io
n
Across-CC-PF
MSUL
Figure 4. CDF of the average CC throughput ratio in the two algorithms
VI. CONCLUSION
In this paper, we investigate the joint resource allocation
for pervasive wireless networks with carrier aggregation. A
maximum network utility model is conducted to formulate
the resource allocation problem. By relaxing the carrier
capability constraints, the assignment rule is derived to
obtain the optimal solution of the model. Furthermore, based
on the obtained optimal solution of the relaxed model, an
iterative algorithm is proposed to adjust resource allocation
to meet the carrier capability requirement of each user.
Numerical results show that MSUL offers better
performance than the existing algorithm which schedules
CCs and RBs separately. Consequently, we can conclude
that the MSUL is an efficient resource allocation scheme for
the pervasive wireless with carrier aggregation due to its
simplicity and high performance.
ACKNOWLEDGMENT
This work is partly supported by the National S&T Major
Project of China under Grant No.2010ZX03005-001, the
National Natural Science Foundation of China under Grant
No. 60802024, and the Program for New Century Excellent
Talents in University (NCET-10-0294), China.
REFERENCES
[1] ITU-R, Framework and overall objectives of the future
development of IMT-2000 and systems beyond IMT-2000,
M. 1645, Jun 2003.
[2] Taewon Hwang, Chenyang Yang, Gang Wu, Shaoqian Li,
Geoffrey Ye Li, OFDM and Its Wireless Applications: A
Survey, IEEE Trans. Veh. Technol., VOL. 58, pp. 1673-
1694, May 2009.
[3] 3GPP TR 36.814 v. 1.2.1, Further advancements for
EUTRA: physical layer aspects, Tech. Spec. Group Radio
Access Network, Rel. 9, June 2009.
[4] Y. Wang, K. Pedersen, T. Srensen, and P. Mogensen,
Carrier load balancing and packet scheduling for multi-
carrier systems, IEEE Trans.Wireless Commun., vol. 9, pp.
17801789, May 2010.
[5] L. Zhang, K. Zheng, W. Wang, L. Huang, Performance
analysis on carrier scheduling schemes in the long-term
200 200 200 200
evolution-advanced system with carrier aggregation, IET
Commun., Vol. 5, pp. 612-619, 2011.
[6] Liu Liu, Mingju Li, Juejia Zhou, Xiaoming She, et. al.,
Component Carrier Management for Carrier Aggregation in
LTE-Advanced System, Proc. IEEE VTC Spring 2011.
[7] Yao-Liang Chung, Lih-Jong Jang, Zsehong Tsai, An
Efficient Downlink Packet Scheduling Algorithm in LTE-
Advanced Systems with Carrier Aggregation, IEEE
Consumer Communications and Networking Conference
(CCNC2011), 2011, pp. 632-636.
[8] Yuanye Wang, Klaus I. Pedersen, Troels B. Srensen, Preben
E. Mogensen, Utility Maximization in LTE-Advanced
Systems with Carrier Aggregation, Proc. IEEE VTC Spring
2011.
[9] F. Kelly, Charging and rate control for elastic traffic,
European Transactions on Telecommunications, vol. 8, pp.
3337, 1997.
[10] A. Jalali, R. Padovani, and R. Pankaj, Data throughput of
CDMA-HDR a high efficiency-high data rate personal
communication wireless system, Proc. IEEE VTC, May
2000, pp. 18541858.
[11] F. Kelly, A. Maulloo, D. Tan, Rate control for
communication networks: shadowprices, proportional fairness
and stability, Journal of the Operational Research Society,
Vol. 49, pp. 237-252, 1998.
[12] J. Holtzman, Asymptotic analysis of proportional fair
algorithm, Proc. IEEE PIMRC, Sep. 2001, pp. 33-37.
201 201 201 201