15 - A NOMA-PSO Based Cooperative Transmission Method in SatCom

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 6

A NOMA-PSO Based Cooperative Transmission

Method in Satellite Communication Systems


Ruimin Wan, Lina Zhu, Tian Li, Lin Bai
School of Electronic and Information Engineering, Beihang University, Beijing,100191, China
Email: ruimin wan@buaa.edu.cn; zhulina@buaa.edu.cn; tian.li@buaa.edu.cn; l.bai@buaa.edu.cn

Abstract—Due to the limited frequency and energy resources, access (FDMA), time division multiple access (TDMA), code
cooperative satellite communication (SATCOM) systems have division multiple access (CDMA), space division multiple
been widely studied. To improve the spectrum efficiency further, access (SDMA), and orthogonal FDMA (OFDMA). According
we propose to combine non-orthogonal multiple access (NOMA)
with cooperative SATCOM systems in this paper. Under the con- to [9] and [10], due to the excellent spectrum efficiency of
dition of fixed total transmit power, an optimization problem is orthogonal frequency division multiplexing (OFDM) systems,
formulated to maximize the sum-rate with the variables of power the OFDM-based satellite data transmission systems has been
allocation factors. In order to solve this non-convex optimization proposed as one promising solution for supporting high-rate
problem, the particle swarm optimization (PSO) algorithm is data traffic. However, according to [11], in order to eliminate
applied with an optimal power allocation strategy. Simulation
results show that the proposed NOMA-PSO scheme outperforms the inter symbol interference (ISI), it is necessary to insert
existing orthogonal multiple access (OMA) approach. the cyclic prefix in an OFDM symbol. In traditional OFDM
Index Terms—Non-orthogonal multiple access (NOMA), coop- systems, cyclic prefix occupies about twenty-five percent of
erative satellite communication systems, PSO algorithm, power the resources, which results in a waste of spectrum. Thus,
allocation. its spectrum can be further improved. Recently, because of
superposition coding and no cyclic prefix, non-orthogonal
I. I NTRODUCTION multiple access (NOMA) is attracting much attention as an
Suffering from the scarce spectrum and orbit resources of improvement of OMA. In [12], we can obtain that applying
satellite communication (SATCOM) systems, different system NOMA to OFDM-based systems can further enhance the
architectures and techniques have been considered to improve spectrum efficiency as well as accommodate more users than
the capacity and performance of systems based on the existing traditional systems.
satellite platform. Among them, the one using the multi- In light of the above discussion, NOMA is applied to a
satellite cooperative transmission method, which is considered multi-satellite cooperative transmission system which consists
as a possible structure of future SATCOM systems, is carried of GEO satellites and LEO satellites in this paper. On one
out to increase the capacity without acquiring extra power and hand, from [13] and [14], we can conclude that the greater
spectrum resources. In [1] and [2], cooperative transmission the channel gain gap, the better the performance of NOMA
methods have been proposed to improve the performance system under a certain circumstance. On the other hand, we
of SATCOM systems while [3]-[5] developed schemes that assume the GEO satellite and the LEO satellite communicate
enable single antenna mobiles in a multi-user environment simultaneously to multiple ground users who are locating in
to share their antennas for cooperative transmission. For the overlapped coverage area of the satellites. In this case, the
geostationary earth orbit (GEO) satellite collocation systems, channel gains from a user to GEO and LEO satellites will be
cooperative transmission schemes that aim to increase the significantly different, which makes it feasible and reasonable
transmission rate and improve the spectrum efficiency are to apply NOMA to the cooperative system. Theoretical analy-
investigated in [6]-[8]. Although different techniques have sis and simulation results show that NOMA scheme can serve
been considered by using multiple satellites transmit signals a better performance than that of OMA in SATCOM systems.
cooperatively to increase the spectrum/energy efficiency, the In addition, the power allocation of each satellite to maximize
satellite diversity (i.e., a cooperative satellite system that the system sum-rate is investigated under the circumstances
consists of different types satellites) is not exploited. They that the total transmit powers of the GEO and LEO satel-
all assume that one kind of satellites among GEO, the highly lite are fixed respectively. Since the proposed optimization
elliptical orbit (HEO), low earth orbit (LEO) satellites, are problem is non-convex, particle swarm optimization (PSO)
employed at a time. Thus, in this paper, we mainly focus on a algorithm is considered to solve the problem. As it is expected,
cooperative transmission scheme by using GEO satellites and the proposed NOMA-PSO cooperative transmission method
LEO satellites. in SATCOM systems can improve the system performance
Furthermore, to meet the overwhelming demand for high- further.
rate data traffic in SATCOM systems, several orthogonal The rest of this paper is organized as follows. Section
multiple access (OMA) schemes have been applied to im- II presents the system model and the optimization problem
prove the performance, such as frequency division multiple formulation. Then, the PSO algorithm is applied to solve

978-1-5386-2062-5/17/$31.00 ©2017 IEEE


the proposed optimization problem in Section III. The per- where K represents the Rican K-factor. In this paper, we
formance comparisons between NOMA scheme and OMA consider the LOS signal path as the dominating wave prop-
scheme are presented and analyzed in Section IV and the paper agation. Our assumption is reasonable because in SATCOM
is concluded in Section V. the dominating radio wave propagation is mainly formed by
the LOS signal part. According to [17], the channel can be
II. S YSTEM MODEL AND PROBLEM FORMULATION assumed to be frequency-flat. Assuming that fc,n represents
A. System Model the subcarrier frequency of the n-th user and f << fc,n ,
the frequency dependence can be ignored. Then, (3) can be
In cooperative satellite collocation systems, we suppose that
rewritten as
there are κ satellites located on GEO and LEO. To use the
satellite resources efficiently, we consider a κ = 2 system that hk,n (f ) = hk,n (fc,n ) = [hLOS ](k,n) (fc,n ) (4)
consists of a GEO satellite and an LEO satellite as shown in
Fig. 1. where [hLOS ](k,n) denotes the LOS channel signal part from
the n-th user to the k-th satellite. According to the free-space
propagation model, we can obtain
rk,n
[hLOS ](k,n) = ak,n exp(−j2πfc,n ) (5)
c0
*(2
where c0 is the speed of light in the free space. In addition,
ak,n represents the complex envelope, rk,n is the distance from
KQ
the n-th user to the k-th satellite. Using ϕ0 to denote the carrier
phase angle taken as zero at the time of observation, ak,n can
/(2 be calculated by
KQ
RYHUODSSHG c0 ejϕ0
ak,n = . (6)
DUHD 4πfc,n rk,n
It is obviously that the length of ak,n is a constant which
WKHQWK can be described as
XVHU c0
|ak,n | = . (7)
Fig. 1. A simplified system model of cooperative transmission 4πfc,n rk,n
Then, in order to obtain the distance rk,n , we need to
Assume that the number of ground users which are equipped describe the location of each satellite and the ground user at
with single antennas is N and the OFDM scheme is considered first. It is noteworthy that geographical longitude and latitude
throughout the paper. Thus, different users are allocated with are used to determine the coordinates of positions. Generally
orthogonal subcarrier frequencies. Just as uplink NOMA, in speaking, in geographical coordinates, the position of the n-th
each subcarrier, the user receives useful signals from different ground user is denoted by the longitude θn and the latitude φn .
satellites. For user n, we suppose the useful coded signal Then, the position vector of the n-th ground user in Cartesian
transmitted by the k-th satellite xk,n is coordinates is given by

xk,n = pk,n sk,n (1) αn = [xn , yn , zn ]T (8)
where sk,n and pk,n represent the useful data signal and where xn = Re cos(φn ) cos(θn ), yn = Re cos(φn ) sin(θn ),
the transmit power from the k-th satellite to the n-th user and zn = Re sin(φn ). Here Re = 6378km is the mean earth
respectively. Let E[|sk,n |2 ] = 1. For user n, the received signal radius. Consider that the heights of the GEO satellite and the
yn is expressed as LEO satellite are presented by H1 and H2 respectively. Then,

κ the position vectors of the two satellites are expressed as

yn = hk,n pk,n sk,n + wn (2)
k=1
β s,1 = [d1 cos θs,1 , d1 sin θs,1 , 0]T (9)
where hk,n is the baseband channel coefficient from the k-th
satellite to the n-th user , wn is the additive white Gaussian and
noise, whose power density is denoted by Wn . β s,2 = [xs,2 , ys,2 , zs,2 ]T (10)
According to [15]-[16], the frequency selective satellite
radio channels contain a line-of-sight (LOS) signal part and a where β s,1 and β s,2 represent the position vector of the GEO
Non-line-of-sight (NLOS) signal part is given by satellite and the LEO satellite, respectively. Here, d1 = Re +
  H1 , d2 = Re + H2 , and xs,2 = d2 cos(φs,2 ) cos(θs,2 ), ys,2 =
K 1 d2 cos(φs,2 ) sin(θs,2 ), zs,2 = d2 sin(φs,2 ). Denote by φs,k and
h(f ) = hLOS (f ) + hN LOS (f ) (3)
K +1 K +1 θs,k the latitude and longitude of the k-th satellite. Since the
latitude of the GEO satellite is 0◦ , its position is determined
p
the n-th ground user, we can obtain that t = [ Pk,n
k
]κ×N . It is
by the longitudes θs,1 . obviously that tk,n ∈ (0, 1). Then, the optimization problem
Then, we can calculate the distance of k-th satellite and can be expressed with tk,n as
n-th ground user by
N κ
tk,n Pk |hk,n |2
OP 1 : max log2 (1 + k=1 ) (16a)
t W0
rk,n = β s,k − αn  (11) n=1

N
where k = 1, 2 and n = 1, 2, . . . , N , respectively. s.t. t1,n = 1, ∀n (16b)
n=1
B. Problem Formulation N

In this section, we let k = 1 represents the GEO satellite t2,n = 1, ∀n (16c)


whose total transmit power is P1 and k = 2 denotes the LEO n=1
satellite whose total transmit power is P2 . tk,n > 0, ∀k, n (16d)
At the n-th ground user terminal, successive interference where k ∈ {1, 2}, n ∈ {1, 2, . . . , N }, and κ = 2. It is
cancellation (SIC) in [18] is employed to separate the received apparently that the optimization problem in this paper is non-
signals. Generally speaking, according to the principles of convex and the optimal value is closely related to the power
NOMA, the SIC process at the n-th user is related to the allocation factor matrix t. Hence, PSO algorithm is considered
magnitude of channel gain. However, due to the large fading to solve this optimization problem because of twofold: i)
of the SATCOM system, the channel gain of the GEO and the object function of this optimal problem is single; ii) the
the LEO satellites is very low. Thus, it is hard to achieve solution of the optimization problem is in a certain range.
SIC in our satellite system just rely on channel state although
|h1,n |2 < |h2,n |2 . In this case, in order to apply SIC, III. PSO ALGORITHM
power allocation strategy should be considered. Since we have PSO resembles a school of flying birds which is proposed
E[|sk,n |2 ] = 1 for k = 1, 2 and n = 1, . . . , N , in this paper, by Eberhart and Kennedy in [19]. In PSO, the individuals
the SIC process on the n-th user mainly based on the power are “evolved” by cooperation and generations. Generally, each
allocation and channel gain pk,n |hk,n |2 . Moreover, since the individual represents a potential solution to an optimization
fading of the GEO satellite is larger than that of the LEO problem. Moreover, we consider each individual as a “parti-
satellite, the total transmit power of GEO is larger than LEO, cle”, and each particle is treated as a point in a D-dimensional
that is to say, P1 > P2 . In summary, at the n-th user, the signal space. According to the principles of PSO algorithm, each
from the GEO satellite s1,n can be detected at first owing to particle adjusts its flying not only according to its own flying
p1,n |h1,n |2 > p2,n |h2,n |2 . Consequently, at the n-th user, the experience but also its companions’ flying experience.
transmission rate of the k-th satellite in a unit bandwidth can In PSO, since each “particle” represents a potential so-
be derived as lution and the solution of the optimization problem t =
p1,n |h1,n |2 [tk,n ]κ×N , (κ = 2) has 2N elements, i.e. 2N variables, we
R1,n = log2 (1 + ). (12) can assume that the dimension of each particle D can be taken
p2,n |h2,n |2 + W0
as 2N . Then, the i-th particle can be represented as
In SIC, the transmission rate of the LEO satellite at the n-th
user as X I = [xi1 , xi2 , . . . , xi(2N ) ]. (17)
2
p2,n |h2,n | In order to solve the proposed optimization problem OP 1,
R2,n = log2 (1 + ). (13)
W0 we let
Then, the rate of the n-th user is given by [xi1 , xi2 , . . . , xiN ] = [t1,1 , t1,2 , . . . , t1,N ] (18)
Rn,noma = R1,n + R2,n and
p1,n |h1,n |2 + p2,n |h2,n |2 (14)
= log2 (1 + ). [xi(N +1) , xi(N +2) , . . . , xi(2N ) ] = [t2,1 , t2,2 , . . . , t2,N ] (19)
W0
where t1,n , n ∈ {1, 2, . . . , N } is the power allocation factors
Thus, the sum-rate of the NOMA scheme system is expressed
of the GEO satellite, and t2,n , n ∈ {1, 2, . . . , N } denotes the
as
κ power allocation factors of the LEO satellite.
N
pk,n |hk,n |2 Given by the best fitness value, the best previous position
Rnoma = log2 (1 + k=1 ). (15) of any particle is recorded and represented as
n=1
W0
In this paper, we aim to maximize the sum-rate of the N P I = [pi1 , pi2 , . . . , pi(2N ) ]. (20)
ground users with the variables of power allocation factors. Suppose the rate of the position change (velocity) for the i-th
For simplicity, we use p = [pk,n ]κ×N to denote the power particle is
p
allocation matrix. Then, letting tk,n = Pk,n , (k = 1, 2)
k
represent the power allocation factor for the k-th satellite to V I = [vi1 , vi2 , . . . , vi(2N ) ]. (21)
In [20], the particles are manipulated according to the Algorithm 1: PSO algorithm
following equations: Input: number of particle M , number of iterations
vid = ω ∗ vid +c1 ∗ rand() ∗ (pid − xid ) tmax ,xmin ,xmax ,vmin ,vmax
(22a) Output: the best fitness value zg and the corresponding
+ c2 ∗ Rand() ∗ (pgd − xid ) solution P G
xid = xid + vid (22b) 1 for each particle i ≤ M do
2 initialize X I according to equation (17);
where c1 and c2 are two positive constants, rand() and Rand()
3 initialize V I according to equation(21);
are two random functions in the range [0, 1]. pgd represents the
4 end
group’s best experience and the subscript symbol g is the index
5 for each iterations t ≤ tmax do
of the best particle among all the particles in the population. ω
6 for each particle i ≤ M do
is inertia weight and plays an important role of balancing the
7 calculate the fitness value z according to equation
global search and local search. Generally, ω can be a positive
(16a);
constant or a positive linear or nonlinear function of time. In
8 if the fitness value z is better than the best fitness
this paper, w can be calculated by
value zp in history then
t 9 set current value as the new zp ;
ω = ωmax − (ωmax − ωmin ) (23)
tmax 10 set current position as the new best position
where ωmax and ωmin are the maximum and minimum P I;
11 end
inertia weights, respectively. Here, tmax denotes the maximum
12 end
number of iterations (Evolving algebra) while t represents the
13 choose the particle with the best fitness value zg of
number of present iterations.
all the particles as P G ;
The equation (22a) is usually utilized to calculate the
14 for each particle i ≤ M do
particle’s new velocity in terms of its previous velocity and
15 calculate particle velocity according V I to
the distances of its current position from its own best position
equation (22a);
and the group’s best position. In this paper, the minimum
16 if vid > vmax then
velocity vmin is set to 0 and the maximum velocity vmax
17 vid > vmax ;
is 1. Then, the particle flies toward a new position based on
18 else if vid < vmin then
equation (22b). Moreover, the performance of each particle is
19 vid = vmin ;
measured according to a pre-defined fitness function, which
20 end
is related to the problem to be solved as shown in (16). The
21 end
PSO algorithm is presented in Algorithm 1.
22 update particle position X I according to equation
It is noteworthy that, different from other optimization prob- (22b);
lems with PSO algorithm, since tk,n denotes the power alloca- 23 if xid > xmax  xid < xmin then
tion factor, xid ∈ (0, 1), that is to say, xmin = 0, xmax = 1 in 24 xid = rand(1, 1) ∗ (xmax − xmin ) + xmin ;
Algorithm 1. Moreover,
N according to2N(16b), (16c), (18), and 25 end
(19), we have d=1 xid = 1 and d=N +1 xid = 1 in our 26 end
optimization problem. Also, we keep vid ∈ [vmin , vmax ] in 27 end
Algorithm 1. 28 return the best fitness value zg and P G ;

IV. S IMULATIONS
A. Comparison with OMA
Therefore, the rate of the n-th user with OMA scheme can be
In this section, we compare the performance of the coop- expressed as
erative satellite system based on NOMA and OMA at first.
According to [21], at the n-th user, assume the bandwidth of Rn,oma = R1,n + R2,n . (26)
the GEO satellite is ε. Since there is no SIC of OMA scheme in
Then, we also use PSO algorithm to calculate the maximum
receiver, we can obtain the transmission rate of GEO satellite
of sum rate with OMA scheme and the best power allocation
with OMA(OFDM) as
strategy. Except for changing the fitness function, the other
t1,n P1 |h1,n |2 constraints are the same as in OP 1. The process for OMA
R1,n = ε log2 (1 + ). (24)
εW0 scheme optimal problem is also shown in Algorithm 1. In
summary, under the OMA circumstances, the fitness function
At the n-th user, the transmission rate of the LEO satellite is
is given by
given by

N
t2,n P2 |h2,n |2 Roma = Rn,oma . (27)
R2,n = (1 − ε) log2 (1 + ). (25)
(1 − ε)W0 n=1
TABLE I
PARAMETERS OF THE PSO ALGORITHM the OMA scheme when satellites transmit signals to the n-th
user.
Parameters Values
Number of Particle M 20
Maximum Inertia Weight ωmax 0.95 73

Minimum Inertia Weight ωmin 0.4


Number of Iterations tmax 500 72
Positive Constant c1 ,c2 2

71

fitness value
Rnoma
B. Simulation Parameters 70 Roma
Runi
In our simulations, the number of particle M is 20. Accord-
69
ing to [18], the PSO with the inertia weight ω in the range
[0.9, 1.2] on average will have a better performance; that is,
68
the probability to find the global optimum within reasonable
number of iterations is high. Therefore, we set ωmax = 0.95
67
and ωmin = 0.4. In addition, the numbers of the iterations 0 10 20 30
the index of iterations (t)
40 50

tmax is 500, and the two positive constant c1 and c2 in (22a)


are set to 2. For convenience, these above parameters of PSO
Fig. 2. Sum rate using by PSO algorithm
algorithm can be summarized as in Table 1.
In this paper, we assume that the heights of GEO H1 and
LEO H2 are 36000km and 1000km respectively, the number Affected by increasing carrier frequency, the channel gain of
of ground users N is 10, and the subcarrier frequency of the the n-th ground user is decreasing in terms of (7). Correspond-
n-th user fc,n is from 15GHz to 20GHz. For the sake of ing to the optimal fitness value in Fig. 2, Fig. 3 depicts the
simplicity, we choose the LEO satellite whose orbit is also in optimal power allocation strategy calculated by the Algorithm
the equatorial plane, i.e. the latitude of it φs,2 is 0◦ . Suppose 1. From Fig. 3, according to the maximized fitness value in
the longitude of GEO satellite θs,1 is 28◦ , then, the value of NOMA scheme, we can conclude that the power of the GEO
LEO satellite θs,2 is 30◦ . The latitude of the n-th ground user satellite is almost uniformly allocated among the N ground
φn and the longitude of the user θn are all in the range of the users while the LEO satellite assign most of its power to the
overlapped area of the two satellites. Moreover, let σk = W Pk ground user with best channel gain. As shown in Fig. 3, when
0
represent the signal-noise ratio, due to long distance, the fading N = 10, the power allocation factor of GEO satellite for each
of satellite system is larger than that of the terrestrial system. user is about 0.1, where more than 70% power of the LEO
Therefore, in order to obtain good performance, the total satellite is allocated to the first user. Since P1 > P2 (σ1 > σ2 ),
power constraint of the GEO satellite P1 and LEO satellite P2 the power from the GEO satellite contribute more to a better
are larger than terrestrial system, respectively. Thus, σ1 > σ2 . performance for the n-th user. Moreover, since the channel
Using these parameters we can obtain the channel state to state of the first user is best in our simulations, the more power
calculate the fitness value as shown in (16a) and (27). assign to it, the better performance can be expected under the
circumstance that the other power allocation factors of LEO
t2,n , n ∈ {2, . . . , N } are greater than zero (i.e. t2,n > 0).
C. Results of Simulation
The sum-rate of NOMA scheme Rnoma and OMA scheme 0.8
Roma calculated by PSO algorithm are shown in Fig. 2. It t
1,n
0.7
is apparently that the performance of NOMA scheme surpass t
2,n
OMA scheme. For comparison purpose, with NOMA scheme,
power allocation fators

0.6

we consider the sum-rate of uniform power allocation (i.e. 0.5

NOMA-uniform) both of the GEO satellite and the LEO 0.4


satellite as shown by the green line. In this case, the power
allocation factor can be expressed as tk,n = N1 . Obviously, the
0.3

NOMA-PSO scheme outperforms NOMA-uniform scheme. In 0.2

Fig. 2, as the fitness value converges fast, we only shows the 0.1

curve of different fitness value within 50 iterations although 0


1 2 3 4 5 6 7 8 9 10
the number of iterations is 500 in our simulations. According the index of ground user (n)
to the figure, we can also obtain that the fitness value of
NOMA scheme Rnoma is converged at 71.9 approximately. It Fig. 3. Optimal power allocation strategy
is advantage about 5% over the OMA scheme Roma , while
Roma is converged at 68.3 approximately. This is mainly In Fig. 3, the optimal power allocation strategy has already
because the bandwidth of NOMA scheme is wider than that of been given by PSO algorithm. For the sake of analysis, Fig.
4 gives the rate of each ground user according to the power [2] S. Zhang, Q. Guo, and Y. Wei, “Performance of cooperative satellite
allocation strategy as shown in Fig. 3. From Fig. 4, for each communication based on space-time trellis code,” IEEE 26th Annual
International Symposium on Personal, Indoor, and Mobile Radio Com-
ground user, we can observe that the performance of NOMA munications, pp. 1814-1818, 2015
scheme with PSO algorithm can improve slightly than that [3] A. Nosratinia, T. E. Hunter and A. Hedayat, “Cooperative communication
with uniform allocation scheme. Moreover, the rate of each in wireless networks,” IEEE Communications Magazine, vol. 42, no. 10,
pp. 74-80, Oct. 2004.
ground user with OMA scheme is shown by the red line in Fig. [4] M. Janani, A. Hedayat, and T. E. Hunter , “Coded cooperative in wireless
4. Apparently, from Fig. 3 and Fig. 4, we can conclude that communications: space-time transmission and iterative decoding,” IEEE
NOMA-PSO scheme not only provides larger sum-rate than Transactions on Signal Processing, vol. 52, no. 2, pp. 362-371, Feb. 2004.
[5] K. Azarian, H. E. Gamal and P. Schniter, “On the achievable diversity
OMA scheme but also can improve the rate of each ground multiplexing tradeoff in half-duplex cooperative channels,” IEEE Trans-
user. actions on Information Theory, vol. 51, pp. 4152-4172, Dec. 2005.
[6] S. Dou, L. Bai, J. Xie, and Z. Xiao,“Cooperative transmission for
geostationary orbiting satellite collocation system,” 2014 IEEE Global
7.6 Communications Conference, pp. 2880-2885, Dec. 2014.
Rn,noma [7] L.Bai, LN. Zhu, J. Choi, F. Liu and Q. Yu, “Cooperative transmission
7.4
7.4
Rn,uni over rician fading channels for geostationary orbiting satellite collocation
Rn,oma system,” IET Communications, vol. 11, no. 4, pp.538-547, Mar. 2017
7.2 7.3 [8] B. Paillassa, B. Escrig, R. Dhaou, M.-L. Boucheret and C. Bes,
“Improving satellite services with cooperative communications,” In-
7.2
7 ternational Journal of Satellite Communications and Networking,
R [bit/s/Hz]

DOI:10.1002/sat.989, Oct. 2011.


6.8 3 4
[9] J. Li, Y. Zhang, and Y. Zhang, “ Fast tracking doppler compensation
for OFDM-Based LEO Satellite Data,” 2016 2nd IEEE International
n

6.6 Conference on Computer and Communications, pp. 1814-1817, 2016


[10] A. Ginesi, F. Potevin: “OFDM Digital Transmission Techniques for
6.4
Broadband Satellites,” International Communications Satellite System
Conference, pp. 1049-105, Feb. 2006.
6.2
[11] X. Liu, H. Chen, B. Lyu, and W. Meng, “Symbol cyclic shift equal-
ization PAM-OFDM - a low complexity CP-free OFDM scheme,” IEEE
6
1 2 3 4 5 6 7 8 9 10 Transactions on Vehicular Technology, pp. 1-14, 2017.
the index of ground user (n)
[12] P. Parida and S. S. Das, “Power allocation in OFDM based NOMA
systems: A DC programming approach,” IEEE Globecom Workshops
(GC Wkshps), pp. 1026-1031, Dec. 2014.
Fig. 4. Rate of the n-th ground user Rn
[13] L. Dai, B. Wang, and Y. Yuan, “Non-orthogonal Multiple Access for
5G: Solutions, Challenges, Opportunities, and Future Research Trends,”
IEEE Communications Magazine, vol. 53, no. 9, pp. 76-83, 2015.
[14] A. Benjebbour, A. Li, and K. Saito, “NOMA: From Concept to Stan-
V. C ONCLUSION dardization,” IEEE Conference on standards for Communications and
Networking, pp. 18-23, 2015.
In this paper, we propose to apply NOMA-PSO scheme to [15] A. Knopp, R. T. Schwarz, D. Ogermann, C. A. Hofmann and
the OFDM-based cooperative SATCOM systems. In order to B.Lankl,“Satellite system design examples for maximum MIMO spectral
avoid wasting resources, we consider a two satellites system efficiency in LOS channels,” IEEE Global Telecommunications Confer-
ence, pp. 1-6, Nov. 2008.
that consists of a GEO satellite and an LEO satellite. Then, [16] LN. Zhu, L. Bai, JD. Xie, and T. Li, “The ranging error influence on
PSO algorithm was used to maximize the sum-rate of the channel capacity in geostationary orbiting satellite collocation systems,”
NOMA scheme system. For comparison purpose, we also 2015 International Conference on Wireless Communications & Signal
Processing, October 2015.
studied the OMA scheme system. The simulation results show [17] L. Bai and J. Choi, Low Complexity MIMO Detection, Springer, New
that the sum-rate of NOMA scheme is better than that of OMA York, 2012
scheme. Through PSO algorithm, the optimal power allocation [18] G. Maral and M. Bousquet, Satellite Communications Systems, Wi-
ley&Sons, 2010
strategy can be described as: the power of GEO satellite is [19] Eberhart, Russ C., and James Kennedy, “A new optimizer using particle
almost uniformly allocated while the LEO satellite assigns swarm theory,” In Proceedings of the Sixth International Symposium on
most of its power to the ground user with the highest channel Micro Machine and Human Science, vol. 1, pp. 39-43, 1995.
[20] Y. Shi, and R. Eberhart, “A modified particle swarm optimizer,” IEEE
gain. Moreover, we can conclude that NOMA-PSO scheme International Conference on Evolutionary Computation Proceedings, pp.
can offer a better performance than that of NOMA-uniform 69-73, 1998.
scheme. [21] Y. Saito, Y. Kishiyama, A. Benjebbour, and T. Nakamura, “Nonorthog-
onal multiple access (NOMA) for cellular future radio access,” IEEE
VTC Spring 2013, pp. 1-5, Jun. 2013.
VI. ACKNOWLEDGMENTS
This work was supported by the National Natural Sci-
ence Foundation of China (NSFC, Grant Nos. 61231011 and
61231013).

R EFERENCES
[1] X. Jie, S. Feng, and W.Li, “Research on cooperative diversity in mobile
satellite communication system,” 2010 Second International Workshop on
Education Technology and Computer Science, vol. 1, pp. 304-307, 2010.

You might also like