Вы находитесь на странице: 1из 7

Resource Allocation for Simultaneous Transmission

of Information ad RF energy in OFDM systems


Sumit Gautam and P. Ubaidulla
Signal Processing and Communication Research Center (SPCRC),
International Institute of Information Technology - Hyderabad (IIIT-H),
Hyderabad, INDIA.

AbstractIn this paper, we consider simultaneous transmission


of information and energy in a wireless network with channels
undergoing frequency-selective fading. The receivers are capable
of decoding the information as well as harvesting energy from the
same received signal. The network employs orthogonal frequency
division multiplexing (OFDM) for signal transmission. In this
context, we address two different problems and propose suitable
algorithms for transmit and receive processing. The first problem
is the optimal resource allocation for maximizing weighted sumrate of the users while guaranteeing a minimum harvested energy
at each receiver. In order to achieve this target, we propose
an iterative algorithm that allocates a set of subcarriers to
each user, allocates power to each subcarrier, and computes
the optimal ratio of the received signal power that goes to
the information decoder and the energy harvester. The second
problem is the optimal resource allocation for maximizing the
total harvested energy under constraints on the minimum rate
achieved by each user. We propose resource allocation and receive
processing algorithms as in the previous case. Numerical results
are provided, and the merits of the proposed design for various
operating conditions and parameters are illustrated.

I. I NTRODUCTION
Simultaneous transmission of information and energy in
wireless systems have received considerable attention recently
[1], [2], [3]. Such techniques enable the receivers to perform
both information decoding and energy harvesting from the
same electromagnetic wave. The received energy is either used
to operate the receiver circuit or stored for later use. Such
energy harvesting techniques hold great potential for energyconstrained networks with limited energy storage capability.
Over the years, the size of batteries used to drive the wireless
devices has decreased considerably. Rapid drainage of these
battery sources necessitate the devices to use battery optimally
by compromising on other important tasks [4]. To complement
the decreasing battery size, demand for efficient and compact
energy harvesters (EHs) have grown many folds [5]. In this
context, the wireless transmission of power along with information holds great promise.
Conventional receiver architectures have mainly focused
on increasing data rates. These optimal receivers used for
information decoding are incapable of performing both wireless information transfer and wireless energy transfer simultaneously. Therefore, this calls for a change in the receiver
architecture to support simultaneous wireless information and
energy transfer. Time switching (TS) and power splitting (PS)
are two practical receiver schemes proposed by researchers to

address this problem [4], [5]. In TS, the receiver is designed


in such a way that for a first few symbols, it harvests all the
energy from the signal power while for the remaining symbols,
all of the signal power is utilized for information decoding.
However, in case of PS, a fraction of the received signal power
is used for harvesting energy and the remaining power is used
for decoding information simultaneously.
Orthogonal frequency division multiplexing (OFDM) is a
spectrally efficient multi-carrier transmission technique and it
has been widely adopted in many recent standards. However,
the simultaneous information and energy transmission techniques in the context of OFDM has not been widely studied
[4]. In conventional OFDM, subcarriers are assigned to the
users based on two main resource allocation schemes: fixed [7]
and dynamic [8], [9]. When incorporated with wireless energy
transfer, OFDM not only preserves its existing advantages,
but it also allows multiple devices to charge simultaneously
using wireless recharging. A framework for OFDM-based
architecture along with power control algorithms to exploit
frequency diversity that improves information and energy
transfer efficiency is considered in [1].
In this paper, we consider simultaneous transmission of
information and energy in the downlink of a network that employs OFDM. We address the optimal resource allocation for
maximizing the weighted sum of rates and for maximizing the
total harvested energy. The receivers are assumed to employ
PS scheme to decode information and harvest energy simultaneously. Optimal resource allocation, which can significantly
enhance the performance of OFDM systems, is well-studied in
the conventional use of OFDM. Optimal subcarrier and power
allocation and receive processing is a challenging problem
when the additional requirement of energy transmission is
taken into account. To overcome this difficulty, we propose
an iterative algorithm for joint subcarrier and power allocation
at the transmitter and the optimal ratio of the received signal
that goes to the information decoder and energy harvester at
each receiver. The algorithms reported in [9] does not consider
allocation of subcarriers to each user, whereas the algorithm
proposed in this paper performs the subcarrier allocation as
well.
The paper is organized as follows. Section II presents a
description about the system model. Section III comprises
of solutions to the proposed problems involving both maximization of transmit rate and harvested energy. Section IV

Fig. 1: System model

Fig. 2: Receiver architecture for kth user

comprises of simulation results for the proposed algorithms.


Finally, section V concludes the paper.

PS plays a very important role in determining the efficiency


of the receiver system.
Let Rk denote the kth users rate expressed as:


X
Rk =
log2 1 + (1 k )pn k,n ,
(2)

II. S YSTEM M ODEL


The system model considered in this paper consists of a
transmitter that can transmit both information and energy to
K users. Fig. 1 depicts a simple block schema of the overall
model. It is considered that each user is capable of decoding
information and harvesting energy simultaneously.

nSk

where Sk is the 2set of subcarriers allocated to the kth user,


|h
|
and k,n = k,n
2 . We also define
n

Ek = k

A. Transmission and Channel Models


The transmitter considered in this paper consists of a single
antenna. Let xk (t) denotes the signal for kth user, k denotes
minimum harvested energy constraint known to the transmitter, Ek is the energy harvested by kth user at the receiver and
PT denotes the total transmit power. A frequency selective
fading channel is considered. Let Hk denote the channel gain
matrix for the kth user, hk,n be the channel gain for kth user in
the nth subcarrier, N0 be the power spectral density (PSD) of
additive white Gaussian noise (AWGN) and B be the overall
available bandwidth in MHz range. Let K denote the total
number of users where k = 1, 2,..., K. We assume that xk (t)
B
MHz, and
is a narrowband signal having a bandwidth of N
2
E[|xk (t)| ] = 1 where |.| denotes the absolute value or norm
and E[.] denotes the statistical expectation for the kth user.

Consider a receiver model with single receiving antenna as


shown in Fig. 2. Let the received signal to be denoted by yk (t)
and receiver noise for kth user be denoted as nk (t). Here, the
noise nk (t) can be modelled as a narrow-band Gaussian noise
B
with bandwith equals N
and centered around the frequency
fk . The complex equivalent of noise can be adopted from

2
j2fk t
2<{
}, where n
k (t)CN (0, 2k ),
[5] as: nk = 
nk (t)e
B
where k2 = No N
which implies that n
k (t) is a circulary
symmetric complex Gaussian (CSCG) random variable with
zero mean and variance k2 .
The signal received at the kth user on the nth subcarrier
can be expressed as
yk,n = hk,n xk,n + nk,n

(1)

The entire bandwidth is divided into N subcarriers and


the power allocated to the nth subcarrier is denoted be pn .
At each receiver, a power splitter (PS) is employed so that
an optimal fraction of the received signal is provided to the
information decoder and the remaining part is provided to the
energy harvester [5]. The PS ratio at the kth user is denoted
by k , where 0 < k 1. We can notice from Fig. 2 that

pn |hk,n |2 .

(3)

n=1

It should be noted that each user can harvest energy from all
the subcarriers.
III. M AXIMIZATION OF THE SUM - RATE
We consider the problem of maximizing the weighted sumrate of the users while ensuring that the energy harvested at
each user is above a given threshold and the total transmit
power does not exceed a given limit. To solve this problem, we require a mechanism that not only performs joint
subcarrier and power allocation, but also find an optimal PS
ratio, k , k = 1, 2, , K, at the receiver. The same k is
applicable for all the subcarriers received by the kth user. This
problem (P 1) can be expressed as
(P 1) :

B. Receiver Model

N
X

max

{k },{pk },{Sk }

subject to :

K
X

wk Rk

(4)

Ek k ,
N
X
p n PT ,

(5)

k=1

(6)

n=1

where wk is the weighting factor of the kth users rate.


It is evident that solution to problem (P 1) involves joint
optimization of subcarrier allocation and power. Now, since
this is a mixed integer problem, therefore the possible solutions
are not tractable. Hence, computing the optimal solution to
this problem is very challenging. Therefore, we propose an
algorithm to solve this problem iteratively.
The proposed algorithm (I) can be outlined as follows:
1) Initialize: U = {1, 2, , K}, Sk = , k [0 1],
k = 0, k = 1, 2, , K.
Iterate over steps 2 to 5 until convergence
2) V = {1, 2, , N }
3) for k = 1 to K,
a) find n such that |hk,n | > |hk,m |, m, n V
b) Sk = Sk {n}, k = k + PNT 
c) Rk = log2 1 + (1 k )k k
d) V = V {n}

4) While V 6= ,
a) find i such that wi Ri wk Rk , i, k U
b) for the i found in (4.a), find n such that
i,n i,l n, l V
c) Si = Si {n}, i =i + PNT
d) V = V {n}
P
e) Use water-filling algorithm with Ri =
nSi
P
log2 1 + 1 i i,n i,n and nSi pn = i
to compute pn , n Si
k
5) Compute PS ratio k = P |hk,n
|2 pn , k U.
n
It should be noted that if k = 0, then the receiver will not
be able to harvest any energy at all. However, if k = 1, all the
power from the transmit signal can be harvested at the receiver.
In step 5) of the algorithm, it is assumed that equality holds
for the constraint in (5). It should be noted that the algorithm
iteratively optimizes {pn } and {k }, until convergence. (6)
indicates that we can keep adding power until the equality is
obtained thereby maximizing the overall weighted sum rate.
The proposed algorithm (I) is motivated by the resource
allocation algorithm in [9], where energy transmission is not
considered. In general, the subcarriers and power allocation is
performed for K users involving (N +K) optimization parameters for an optimal solution. The proposed algorithm allocates
power using water-filling for (N K) times along with the
steps involved in [8]. After the subcarrier allocation to a user,
step 4(e) performs water-filling for optimal power allocation
whose complexity is O N log2 (N ) . Thus, the order of this
algorithm can be expressed as a non-linear function of N and
K for all the iterations. The proposed algorithm is evaluated
against pre-existing equal power allocation algorithm.
IV. M AXIMIZATION OF THE TOTAL HARVESTED ENERGY
In this section, we consider the problem of determining the
optimal resources allocation and the PS ratio for maximizing
the total energy harvested by the users under constraints
on the rates available to each user and the total transmit
power. Mathematically, this problem (P2) can be formulated
as follows:
K 
N

X
X
(P 2) :
max
k
pn |hk,n |2
(7)
{k },{pn },{Sk }

n=1

k=1

subject to :



log2 1 + j k , k, (8)

Therefore, we propose to solve this problem by alternatively


computing {pn } for fixed values of {k }, {Sk } and vice
versa. But, before we determine the power in each subcarrier
{pn }s and the power splitting ratio for each user{k }s, main
problem is to find out the exact subcarrier assignment that
maximizes the total harvested energy. Therefore, considering
fixed values of {k } and {pn } to find an optimal set of
subcarriers {Sk }, we can reduce the problem (P 2) as follows:
(P 3) :

max
{Sk }

subject to :

K 
X

N
X

pn |hk,n |2

(12)

n=1

k=1



log2 1 + j k , k, (13)

jSk
N
X

pi PT ,

(14)

i=1

pi 0,

i.

(15)

The optimum solution to subcarrier assignment problem can be


found by following greedy-like algorithmic approach. Clearly,
we can reduce the above optimization problem (P 3) into a
general form as given in [10] and [11]. When this generalized
non-convex optimization problem is considered, it can proved
that the duality gap is zero under certain conditions. The
overall problem can now be expressed as follows:
(P 4) :

max
{xk }

subject to :

K
X
k=1
K
X
k=1
K
X

fk (xk )

(16)

gk (xk )  ,

(17)

hk (xk )  P,

(18)

k=1

where xk RN are vectors of the optimization variable


({Sk })s, fk (.) are functions (RN R) that are not necessarily concave, and gk (.) and hk (.) are functions (RN RL and
RN R, respectively) that may not necessarily be convex.
is L-vectors of rate constraint and constant P is total power
constraint. Since we have assumed an OFDM scenario, it can
be assumed that L=K. The Lagrangian of (P 4) is defined as:

jSk
N
X

pi PT ,

(9)

i=1

pi 0,
0 k 1,
(1 )p |h

i,
k,

(10)
(11)

j
k
k,j
, k = 1, 2, , K, i =
where j =
j2
1, 2, , N , the constant k is the minimum rate required by
the kth user and the other variables have the same meanings as
defined previously. Again, the solutions to problem (P 2) are
not tractable as this is also a mixed integer problem involving
joint subcarrier and power allocation as in the previous case.

L({xk }; , ) =

K
X

K
h 
i
X
fk (xk ) + T
gk (xk )

k=1

k=1

+ T P

K
X


hk (xk )

(19)

k=1

where and are vectors of Lagrange dual variables. The


dual objective function r(, ) is defined as an unconstrained
maximization of the Lagrangian such that:
r(, ) = max{xk } L({xk }; , )

(20)

(P 5) :

min

r(, )

(21)

subject to :

 0,

(22)

 0.

(23)

The standard convex optimization results guarantee that both


(P 4) and (P 5) have same solution provided fk (xk )s are
concave, and gk (xk )s and hk (xk )s are both convex. In case
of non-convexity, the solution provided by (P 5) is an upper
bound to the solution of (P 4). The difference between the
upper bound and true optimum is called duality gap since
the upper bound is not always tight.
Definition: Let x?k and yk? represent the optimal soultions
obtained from (P 4) with = x and = y , and P = Px
and P = Py , respectively. To satisfy the time-sharing condition [10], for any 0 1, there exists an optimal solution
zk , such that:
K
X
k=1
K
X
k=1
K
X
k=1

"

#
pj
k = k
|hk,n | 1 +
(1 k ) + pj
n=1
jSk
! N
!
X
X
pj

pn |hk,n |2
(28)
(1 k ) + pj
n=1
N
X

Then, the dual optimization problem becomes:

jSk

where k = 1, 2, , K.
Considering fixed values of {k } and {Sk } to optimize
{pn }, we can reduce the problem (P 2) as:
(P 6) :

max
{pn }

subject to :

K 
X

N
X

pn |hk,n |2

(24)

hk (zk ) Px + (1 )Py

(25)

fk (zk )

K
X

fk (x?k ) + (1 )

k=1

K
X

fk (yk? )

(26)



log2 1 + j k , k, (30)

X
jSk
N
X

pi PT ,

n=1

n=1

(31)

i=1

i,

(32)

where k = 1, 2, , K, i = 1, 2, , N , and other variables


have same meanings as defined earlier. Note that the feasibile
region is bounded, so global maximum must exist as a continuous function. We can solve this problem using the KarushKuhn-Tucker (KKT) conditions. The objective function and
constraints can be written as:

k=1

From duality theory, r? f ? where f ? and r? are the primal


and dual optimal values, respectively. A globally optimal solution can be obtained by using Lagrange dual decomposition
in case the duality gap is zero. Though (P 4) is a non-convex
optimization problem, it can still be proved that the duality gap
is zero if either of the following two conditions is satisfied as
given below [11]:
Theorem 1: If x?k (, ) = arg max{xk } L({xk }; , ), as
a function of and , is continuous at ? and ? , the duality
gap is zero.
P
Theorem 2: Concavity of the optimal
k fk in and P
implies zero duality gap.
We find that the duality gap for the considered problems is
indeed zero [10]. If we now move our focus towards (19),
our aim is to determine the values of and . This can
be performed by iteratively solving (20) and (P 5) for
and . It should be noted that = [1 , 2 , , K ]T and
= [1 , 2 , , K ]T for the problem considered in this paper. In order to find the optimal set of subcarriers, we use (20)
to find the maximum Lagrangian, and then find the optimized
values of {pn }s and {k }s corresponding to the subcarrier
set obtained using conditions in (20). Mathematically, the
values k s and k s can be expressed as follows:
!
X
1 + (1 k )pj |hk,j |2
ln(2)
k =
(1 k )|hk,j |2 + pj |hk,j |2
jSk
!
N
N
X
X
2
2
k
|hk,n |
pn |hk,n |
(27)

(29)

n=1

k=1

pi 0,
gk (zk ) x + (1 )y

f (p) =

K 
X

hX

pn |hk,n |2

(33)

n=1

k=1

gk (p) =

N
X



i
log2 1 + j k 0, k,

(34)

jSk

r(p) =

N
X

p i PT

0,

(35)

i=1

Using (33), (34), and (35), the Lagrangian is given as:


L(p; k , ) = f (p) +

K
X

k gk (p) + r(p)

(36)

k=1

For optimality, p L(p; k , ) = 0. Thus, we can represent


the equation for satisfying optimality condition as:
K
X
k=1

N
X
n=1

|hk,n |2 +

K
X

k gk0 (p) + = 0

(37)

k=1



P
(1 )|h
|2
1
where gk0 (p) = jSk 2 +(1kk )pk,n
2
ln2 . The feasij |hk,n |
j
bility conditions can be represented as gk (p) and r(p) which
are expressed in (34) and (35). Complementary slackness
expressions can be represented as follows:
hX


i
k
log2 1 + j k = 0, k = 1, 2, , K, (38)
jSk
N
X


pi PT = 0, (39)
i=1

The conditions for non-negativity are: pj , k , 0. After


further analysis, it is found that if k 6= 0 and 6= 0, then
from (37) we can express pj s as:
pj = X

X
jSk

where X

j2
, j Sk ,
(1 k )|hk,j |2
k

and k

(40)

5) If 0 (i ) 0 and (i ) ), set ? = zoom(i , i1 )


and stop.
6) Choose i+1 such that i < i+1 < 1
7) Set i = i + 1.
8) Return to 2).
The second stage with zoom function definition is as follows:

(47)

1) Interpolate using bisection to find a trial step length j


between lo and hi .
2) Evaluate (j ) and (j ).
3) If [{(j ) > (0) + 1 j 0 (0)} and {(j ) >
}] or [(j ) (lo } and {(j ) }] , set
hi = j .
4) Else
a) Evaluate 0 (j ).
b) If [|0 (j )| 2 j 0 (0) and (j ) ] then
set ? = j and stop.
c) If 0 (j )(hi lo ) 0, set hi = lo .
d) lo = j .
Thus, in order to solve the proposed problem (P 2), following should be repeated till convergence: (i) perform subcarrier
allocation (Sk ) using iterative algorithm (II) proposed below,
(ii) solve (P 6) for fixed values of {k } and {Sk }, and (iii)
solve (P 7) for fixed values of {pn } and {Sk }. We reduce the
inequality constraints to equality constraints in (P 6), and then
the method of Lagrange multipliers has been applied which
generalizes the non-linear programming as a whole. Since we
get a closed-form solution when we derive analytically, the
proposed system of equations can be solved directly using
KKT approach. Then we propose an optimal solution for
problem (P 7) using the line search algorithm. Therefore, after
solving problem (P 2) under the proposed constraints, we can
see that the overall harvested energy is maximized at the
energy harvester.
To summarize the problem of maximizing harvested energy,
the proposed algorithm (II) can be outlined as:

j = j 2k,j , k = 1, 2, , K, and other variables


where
j
have same meanings as defined earlier. In order to solve this
problem, we can apply line search algorithm using strong
Wolfe conditions [12]. Define the univariate objective function:
() = f (xk +pk ) so that (0) = f (xk ) and define the
 nonP

linear constraint function () = jSk log2 1 + j such


that () . The solution to this problem is done in two
stages with the first stage outlining line search algorithm and
the second stage with definition of zoom function, which is
used in first stage. The first stage is as follows:
1) Set 0 = 0, choose 1 [0 1]. Set i = 1 and assume
0 < 1 < 2 < 1.
2) Evaluate (i ) and (i ).
3) If [{(i ) > (0) + 1 i 0 (0)} and {(i )
}]or [{(i ) > (i1 ) and i > 1} and {(i )
}] then, set ? = zoom(i1 , i ) and stop. Evaluate
0 (i ).
4) If |0 (i )| 2 0 (0) and (i ) , set ? = i and
stop.

1) Initialize: U = {1, 2, , K}; Sk = ; k [0 1]; V =


{1, 2,P
, N }; = {pn },n V; j = 0; {g} = {h}.
2) qi = i |gi,n |, i U and n V
3) Perform the following steps to obtain initial power in
each subcarrier using {q} obtained from 2):
P
T
a)
iV
i qi pi = q ,
P
T
b)  = P qi , i V
i
c) Find an optimal {} for subcarriers: = q
d) gk,n = 0, k = , k U and n V
Repeat over steps 4 to 8 until convergence
4) for i = 1 to N
a) find n such that gn,i gm,i , n,m U
b) j = j + 1, and set gn,i = gn,i and gn,i = 0
c) if j = then j = j + {n}
else if j 6= and n j , then find n again using
c(i) andc(ii) till n
/ j
d) gn,i = | 1 i gn,i |, i U
e) if log2 (1 + gn,i i ) j , then Sn = Sn + {i}
f) if j = K, then set j = 0 and k = , k U

PK

k=1

PN

n=1

|hk,n |2 + ln2

1, 2, , K. The values for k s, under certain feasible conditions, can be expressed using (37) as follows:

P
PN
K
2
n=1 |hk,n | ln2
k=1 k


k = P
(41)
(1k )|hk,n |2
jSk

j2 +(1k )|hk,n |2

The value of can also be obtained using (37).


Similarly, considering fixed values of {pn } and {Sk } to
optimize {k }, we can reduce the problem (P 2) as follows:
(P 7) :

K 
X

max
{k }

pn |hk,n |2

(42)

n=1

k=1

subject to :

N
X



log2 1 + j k , k, (43)

jSk

0 k 1,

k,

(44)

where k = 1, 2, , K. This problem can be understood as


maximization of K-independent terms each with maximized
{i }. Assuming k = 1 k , then problem (P 7) is further
reduced to K-independent minimization sub-problems as follows:
N
X
(P 8) :
max
(1 )
pn |hk,n |2 ,
(45)
{}

subject to :

n=1



j ,
log2 1 +

(46)

jSk

0 1,
p |h

Fig. 3: Convergence of Power Splitting (PS) ratio k for K=4


users with N=256 subcarriers and B=1 MHz.
5) Fix {k } and {S k } to maximize (7) over {pi } under the
constraints (8), (9) and (10) until convergence, k U
and i V. (Ref : (40))
6) Fix the optimized obtained, along with {S k }, to
maximize (7) over {i } under the constraints (8) and
(11) until convergence, i, k U. (Ref : (45))
7) Compute the values of k s and k s. (Ref : (27), (28)).
8) Find the Lagrangian using (19).
In the algorithm above, step 3) is used to initialize the power
of each subcarrier. It is noted that when the objective function
obtains its maxima, the equality in (9) always holds. However,
in case when the inequality constraint holds, it may be stated
that we can add some more power to obtain an optimal solution
until the objective function is maximized. From steps 5) and
6), we can infer that the solution to problem (P 2) can be
found using divide and conquer mechanism
by splitting (P 2)

into two sub-problems (P 6) and (P 7) corresponding to steps
5) and 6) respectively , and then applying the KKT conditions
as described previously. After the process of convergence is
completed, backtracking algorithmic approach is used to find
the maximum value of Lagrangian. Optimal set of subcarriers
{S k }, optimized power in each subcarrier {pn }, and the
estimated power splitting ratio {k } can then be obtained
using the maximum Lagrangian as computed earlier. The
algorithm is operational for (N+K+1) times over a single
iteration and therefore, the order of this algorithm can be
expressed as a linear function of N and K for all the iterations.
V. S IMULATION R ESULTS
In this section, we evaluate the performance of the two
proposed algorithms using simulations. We consider N = 256
OFDM subcarriers, total bandwidth B = 1 MHz, and N0 =90 dBW/Hz throughout the simulations where each user has
6-tap sample-spaced multipath channel with each tap experiencing independent Rayleigh fading. The weights wi , i =
1, 2, , K are all set to one. The lower limit of the harvested
energy () is assumed to be the same for all users.
Fig. 3 illustrates the convergence of PS ratio (k ) over the
indicated iterations. It can be observed from results in Fig. 3
that the PS ratio (k ), obtained from the proposed algorithm
(I) converges to the optimum value within around 20 iterations

Fig. 4: Spectral efficiency versus minimum harvested energy


requirement () for K=8 users with different PT values.

Fig. 5: Total Harvested Energy (in mW) versus number of


iterations for K=4 users with different PT values (in dBW).
for all considered scenarios. Fig. 4 depicts the performance of
proposed algorithm (I) in terms of spectral efficiency, where
it is compared with equal power allocation algorithm (EPA)
for K=8 users with different values of total transmit power
(PT ). The result indicates that maximum spectral efficiency is
obtained when minimum demanded harvested energy () is 0
dBm. However, it can be observed that the spectral efficiency
is found to decrease as the demand for harvested energy is
increased. It can also be noted that the spectral efficiency
improves considerably as PT increases.
In Fig. 5, performance of algorithm (II) is shown where
the convergence of total harvested energy (in mW) is shown
over different iterations for K=4 users and different values of
PT (in dBW). From the results, it can be observed that total
harvested energy (in mW) in proposed algorithm (II) converges
to the optimum value within 10 iterations for all the scenarios
considered. Again, the fast convergence of proposed algorithm
(II) confirms its practical implementations. Fig. 6 shows the
average maximized harvested energy {avg } by average rates
{avg } demanded by K=4 users for different PT values. The
maximized harvested energy (Ek ) is calculated using (3) for
all the K users. The
maximized harvested energy is
P average

k Ek
given as, avg =
, k = 1, 2, , K, i = 1, 2, , N .
K
For simplicity, uniform rates (avg =10 Mbits/s) were considered for all the users during simulations. From the results it
can be seen that, as the values of PT increases, avg increases

Fig. 6: Average maximized harvested energy (avg ) versus


average rate(avg ) for K=4 users with different PT values.

Fig. 7: Spectral Efficiency (in bits/seconds/Hertz) versus number of users for different PT values (in dBW).

appreciably. It can be observed from Fig. 6 that avg decreases


with increasing values avg . Fig. 7 depicts the plot for spectral
efficiency against the total number of users for different values
of PT (in dBW). The weighted sum rate maximization problem
at the transmitter can be verified using the results obtained at
receiver using this graph. It can be noted that the spectral
efficiency increases as the number of users increase while a
better system performance is observed for higher PT values.

[4] X Zhou, R Zhang, CK Ho, Wireless information and power transfer in


multiuser OFDM systems, in Global Communications Conference, 2013
IEEE.
[5] X Zhou, R Zhang, CK Ho, Wireless information and power transfer:
architecture design and rate-energy tradeoff, in Communications, IEEE
Transactions on Volume:61, Issue:11, Oct. 2013.
[6] D. Gunduz, K. Stamatiou, N. Michelusi, M. Zorzi, Designing intelligent
energy harvesting communication systems, in Communications Magazine, IEEE Volume:52, Issue:1, Jan 2014.
[7] E. Lawrey, Multiuser OFDM, in Proc. IEEE International Symposium
on Signal Processing and its Applications, vol. 2, pp. 761-764, Aug. 1999.
[8] Jang and K. B. Lee, Transmit power adaptation for multiuser OFDM
systems, IEEE J. Select. Areas Commun., vol. 21, pp. 171-178, Feb.
2003.
[9] W. Rhee and J. M. Cioffi, Increase in capacity of multiuser OFDM system using dynamic subchannel allocation, in Proc. 51st IEEE Vehicular
Technology Conference, vol. 2, pp. 1085-1089, Spring 2000.
[10] Y Wei, and R Lui, Dual methods for nonconvex spectrum optimization
of multicarrier systems, in IEEE Transactions on Communications, vol.
54 , issue: 7, 1310-1322, July 2006.
[11] K. Seong, M. Mohseni, and J.M. Cioffi, Optimal Resource Allocation
for OFDMA Downlink Systems, in International Symposium on Information Theory, IEEE, 2006.
[12] J. Nocedal and S. J. Wright, Numerical Optimization, chapter 3.
Springer, 2nd edition, 2006.

VI. C ONCLUSIONS
In this paper, multiuser OFDM environment is considered
for transmission. It is assumed that channel conditions are
known to both the transmitter and the receiver. Considering
the users frequency selective nature of the channels, we
performed joint subcarrier and power allocation along with
computation of a power splitting ratio, k . In this process,
transmit rate should be kept maximum, without compromising
the QoS. At the receiver, we use a power splitter (PS) for
splitting the received signal power for energy harvesting and
information decoding using the k value at the transmitter.
Then, the harvested energy is maximized for each user under
the constraints on minimum known rate and total transmit power. Proposed algorithm (I) is compared with equal
power allocation algorithm. Moreover, it is seen that the
harvested energy at energy harvester is maximized using our
approach from proposed algorithm (II). This work can further
be extended to many fascinating directions like addressing
the design of PS, exploiting the diversities originating from
time varying channels for integration of power control with
intellegent energy management policies at the wireless nodes,
and implementation of this framework in cooperative MIMO
multi-cell networks.
R EFERENCES
[1] K Huang, and E Larsson, Simultaneous information and power transfer
for broadband wireless systems, in IEEE Transactions on Signal Processing, 2013.
[2] T. Le, K. Mayaram, and T. Fiez, Efficient far-field radio frequency energy
harvesting for passively powered sensor networks, in IEEE Journal of
Solid-State Circuits, vol. 43, pp. 1287-1302, May 2008.
[3] W. C. Brown, The history of power transmission by radio wave, IEEE
Trans. on Microwave Theory and Techniques, vol. 32, pp. 1230-1242,
Sep. 1984.

Вам также может понравиться