Sunteți pe pagina 1din 8

6052 IEEE INTERNET OF THINGS JOURNAL, VOL. 6, NO.

4, AUGUST 2019

Interbeam Interference Constrained Resource


Allocation for Shared Spectrum Multibeam
Satellite Communication Systems
Min Jia , Senior Member, IEEE, Ximu Zhang, Xuemai Gu, Member, IEEE,
Qing Guo , Member, IEEE, Yaqiu Li, and Ping Lin

Abstract—Future Internet of Things should contain space seg- satellite network and terrestrial network coordinate with each
ment and terrestrial segment. In addition, the multibeam satellite other. Satellite system has evolved to meet the high quality
communication systems, especially working in S shared band, and low-cost requirements, equipped with on-board multibeam
have gained more attention, which plays a significant role in
providing direct-to-user satellite mobile services. Besides, due transmitting antennas. In addition, the multibeam satellite
to the limited on-board resources, it is increasingly urgent to system can meet the increasing demands for large commu-
improve resource utilization. Taking the interbeam interference, nications business through resource allocation technology.
channel conditions, delay factor, capacity, bandwidth utilization The multibeam satellite antenna technology with the advan-
variance into consideration, a novel joint resource allocation tages of beam space isolation and frequency reuse has become
algorithm is proposed in this paper. Interbeam interference
coefficient matrix derived from frequency reuse is established an effective way to alleviate the resource shortage. The satel-
to measure the level of co-channel interference. Moreover, lite spectrum sharing scheme, spectrum sensing scheme, and
the proposed algorithm can allocate resources flexibly accord- intelligent spectrum control strategy are proposed in [4]–[6],
ing to specific traffic requirements and channel conditions. which improved spectral efficiency and throughput signifi-
A novel joint power and bandwidth allocation algorithm is pro- cantly. However, the interbeam users suffer from intensive co-
posed by optimizing throughput and approximation problem of
actual requirement. The optimal solution to this optimization frequency interference, particularly for the beam-edge users.
problem can be obtained by golden section theory and sub- Ignoring the asymmetrical beam traffic, fixed resources are
gradient iteration. The evaluation results demonstrate that the allocated to each beam in most of current satellite [7], [8].
proposed algorithm can maximize the capacity, minimize the As on-board resources become scarcer, it is inevitable that the
bandwidth utilization variance, and it can also allocate resource fixed allocation scheme will result in the mismatch between
intelligently adapting to the user requirements and channel
conditions. traffic demand and allocated capacity. It means that the lower
limit of traffic demand cannot be satisfied due to the waste of
Index Terms—Bandwidth utilization variance, capacity, inter- resources for some beams. Therefore, the smart resource allo-
beam interference, multibeam, shared spectrum, time delay.
cation for multibeam satellite communication systems gains
more attention. In [8]–[11], different capacity allocation mod-
I. I NTRODUCTION els have been established to adjust power or bandwidth to
achieve an optimal resource allocation in on-board processor.
O ENSURE the requirements of large coverage, satellites
T play a vital role in the future Internet of Things (IoT)
system. Researchers in various fields paid close attention to
The research in [8] and [9] measured the capacity based on
the Shannon equation. Allocation models with two criteria are
proposed in [10]. One balanced the power and spectrum, the
satellite-based IoT as soon as it was proposed. However, other optimized the bandwidth and symbol rate. Different
the characteristics of large connection and low power con- from [10], the capacity computing model based on satellite
sumption of the IoT have brought great challenges to the link budget equation is established in [11]. However, it only
traditional satellite communication system [1]. In view of the optimizes the power resource, without considering the cou-
characteristics of the work of satellite-based IoT, informa- pling between power and bandwidth, which will cause the
tion cascades are important dynamical processes in the future waste of bandwidth resource.
IoT society [2], [3]. The on-board resources are limited in Moreover, due to requirements of massive user access,
the multibeam satellite communication system, with which it will cause huge interbeam interference. The effective
Manuscript received August 7, 2018; revised August 24, 2018 and and reasonable resource allocation will bring significant
September 7, 2018; accepted September 11, 2018. Date of publication benefits [12], [13]. Therefore, interbeam interference should
September 18, 2018; date of current version July 31, 2019. This work was be considered, which is caused by frequency reuse in multi-
supported by the National Natural Science Foundation of China under Grant
61671183, Grant 61771163, and Grant 91438205. (Corresponding author: beam satellite communication systems. A dynamic resource
Min Jia.) allocation algorithm of the multibeam satellite system has been
The authors are with the Communication Research Center, School of proposed in [14]. Unlike static resource allocation in the satel-
Electronics and Information Engineering, Harbin Institute of Technology,
Harbin 150001, China (e-mail: jiamin@hit.edu.cn). lite system, a dynamic Stackelberg game model is designed
Digital Object Identifier 10.1109/JIOT.2018.2870878 to maximize cost-efficiency which describes the relationship
2327-4662 c 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
JIA et al.: INTERBEAM INTERFERENCE CONSTRAINED RESOURCE ALLOCATION 6053

between the satellite system’s profit, interference pricing,


and user’s power allocation [15]. Moreover, the essential
discussions and proofs for the pricing rationality are pro-
vided in [16]. However, time delay and bandwidth utilization
variance are not considered in time-varying scenarios.
As frequency resources become scarcer, joint resource allo-
cation with high capacity and effective bandwidth utilization
has been introduced in [16]. A joint power and bandwidth allo-
cation scheme is proposed to improve the spectral efficiency
and throughout [17]. However, to the best of the author’s
knowledge, bandwidth utilization variance is neglected in the
satellite systems. It is hard to guarantee the fairness among
beams without joint resource allocation, especially in the poor
channel conditions. Moreover, the greater interference has
been brought to the satellite system. All of these studies did not
consider the interbeam interference among co-channel beam, Fig. 1. Multibeam satellite communication system.
which should not be neglected.
The main contributions are summarized as follows.
1) Taking the impacts of interbeam interference and traf-
fic demand into account, the multibeam satellite system
model is established. Moreover, the interbeam interfer-
ence coefficient matrix is used to measure the interfer-
ence in the system.
2) Minimum mean squared error principle and the mini-
max principle are adopted to reduce the gap between
the required bandwidth and actual assigned bandwidth.
3) The factor of time delay is considered in this paper, and
the idea that the delay constraint can be transformed into
traffic demand constraint is adopted.
The rest of this paper is organized as follows. Section II
shows the multibeam model over satellite downlink. Moreover,
the problem of resource allocation is taken into the math-
ematical optimization model, and the proposed algorithm is
introduced in Section III. In Section IV, the simulation results
are presented and discussed with the analysis of interbeam Fig. 2. Co-channel interference in satellite downlink.
interference, channel conditions, and delay metrics. Finally,
interbeam interference power. Taking the angle between the
Section V concludes this paper.
user Um in beam Bi and satellite as an example, the calculation
method of axis-off angle is explained as follow.
II. S YSTEM M ODEL
Due to the large transmission distance between satellite
A. Model of Multibeam Satellite Communication System and user, the problem of inclined projection can be ignored.
The multibeam satellite communication system is shown in Therefore, the co-channel interference scenario in satellite
Fig. 1. The system consists of a geostationary earth orbit satel- downlink of beam is subsatellite point as shown in Fig. 2.
lite, multiple spot beams, and satellite users. Assuming that Assuming that the locations of satellite and user have
there are M spot beams and the ith beam can be expressed been known, the length of the arc between subsatellite point
>
as Bi . The real-time business requirements of satellite users, and user can be obtained easily, which is expressed as lui .
which lead to the total demand of relevant beam, are highly According to the calculation formula of arc length, the central
dynamic [17]. angle αui can be obtained
>
B. Calculation Method of Axis-Off Angle 180lui
αu =
i
. (1)
By using multibeam antenna and frequency reusing scheme, πR
multiple signals are transmitted in co-frequency channels Then the distance between satellite and user terminal dui can
simultaneously. In this case, the side lobe gain of the beam be calculated based on cosine law
  
will cause severe interference with the other co-channel beam dui = (h + R) h + R − 2R cos αui + R2 . (2)
as shown in Fig. 2.
Assuming that the gain of multiple satellite antennas com- The axis-off angle ui can be obtained by the sine law
plies with the definition of ITU-R S.672-4 [18], the off-axis as shown in (2). It means that the satellite transmitting
angle is the main factor to be considered in calculating antenna gain can be calculated depending on ITU-R S.672-4.
6054 IEEE INTERNET OF THINGS JOURNAL, VOL. 6, NO. 4, AUGUST 2019

C. Capable Allocation Model With Interbeam Interference where N0 is noise power spectral density, αi is the integrated
It is assumed that all the receivers use the antenna of the attenuation factor of the transmission channel in Bi , which can
same size. Such as a satellite terminal in the center of beam be influenced by signal frequency band, free space loss, atmo-
Bm communicates with satellite in downlink. There are many spheric absorption loss, antenna gain, and so on. The value of
co-interference paths in this system. Due to the large distance αi can be expressed as
between the satellite and users, we can consider that the path  
1 4π d 2
of useful signal is similar to the path of interference signal. αi = . (7)
γ Gmax Gr λ
The distance of this path is expressed as d. Therefore, the
free space loss (4π d/λ)2 and mobile channel fading factor γ It can be seen from (7) that increasing the resource allocated
are explicit. The required power Pir and interference power Iri to Bi can improve the channel capacity. However, once the
received by satellite users can be expressed as allocated channel capacity exceeds the traffic requirements of
  the beams, it will not only waste the limited resources but also
λ 2 enlarge the interference power.
Pr = γ Pt Gmax Gr
i i
(3)
4π d
M  
λ 2 III. P ROPOSED O PTIMAL J OINT P OWER
Iri = γ Pm G
t m,i rG AND BANDWIDTH A LLOCATION
4π d
m=1
  Different spot beams have various traffic requirements.
M
Gm,i λ 2 Some traffic imposes strict delay requirements. According
= γ Pm
t Gmax G r
Gmax 4π d to [18], the constraint of system capacity impacted on delay
m=1
M   is shown as
λ 2
= γ Pt hm,i Gmax Gr
m
(4) Ti
4π d Ci ≥ (8)
m=1 (1 − ei )Di
where Pm
t is the transmitted power allocated to Bm by the satel- where Di is the average delay limit of the Bi , ei is the packet
lite, the received antenna gain of satellite users is represented error rate over the downlink of Bi , and (1 − ei )Di > 1. The
by Gr , Gmax is the maximal antenna gain of satellite, and the delay constraint is determined the minimal traffic demand of
antenna gain of Bk in the direction of Bi is represented by Gk,i . each beam.
The antenna gain is normalized, which is defined as the To evaluate the system performance, it is vital to select
interference coefficient matrix H ∈ CM×M and M is the a reasonable metric. The optimal objective minimizes the
number of satellite beams. Interference coefficient matrix can gap between traffic demand and allocated capacity. Based
be used to measure the interference level among co-channel on above constraints, this paper adopts the second order
beams and can be expressed as objective function to minimize the gap. To quantify the gap
⎡ ⎤ between bandwidth requirement and actual assigned band-
G1,1 G1,2 · · · G1,M width, the bandwidth requirement variance function is defined
1 ⎢ ⎢ G2,1 G2,2 · · · G2,M ⎥ ⎥ as (Ti − Ci )2 , where Ti is bandwidth requirement, and Ci is
H= ⎢ .. .. .. .. ⎥ (5)
Gmax ⎣ . . . . ⎦ actual assigned bandwidth.
GM,1 GM,2 · · · GM,M Optimal approximation solution includes two principles as
follows.
where Gi,j represents the antenna gain of Bi in the direction of 1) Minimum bandwidth requirement variance.
Bj , and the value of Gi,j depends on frequency reuse scheme. 2) Maximum error minimum method (minimax principle).
In this paper, the whole spectrum of the satellite integrated In order to meet the communication requirements of each
network is divided into seven parts, 7 bands are a cluster, and point beam, the capacity allocated to point beam should be
the frequency reuse factor is 7 [20]. If different frequency is close to its requirements. Therefore, the mathematical model
reused in Bi and Bj , Gi,j = 0, Gj,i = 0 can be satisfied. If the of optimization problem can be expressed as
inner-beam interference is ignored, Gi,i = 0 is always satisfied.
Assuming that the traffic demand of Bi is Ti , the power 
M
min f (P, W) = min (Ti − Ci )2 (9)
allocated to Bi is Pi , and the bandwidth allocated to Bi is {Pi ,Wi }
i=1
Wi . Shannon bounded capacity Ci allocated to Bi can be s.t. Ci ≤ Ti (10)
expressed as Ti
  Ci ≥ (11)
Pir (1 − ei )Di
Ci = Wi log2 1 + M
Wi N0 + Iri
  2  Pi ≤ Ptotal (12)
γ Pit Gmax Gr 4πλ d i=1
= Wi log2 1 +   λ 2
Wi N0 + M M
m=1 γ Pt hm,i Gmax Gr 4π d
m
  Wi ≤ Wtotal (13)
Pit i=1
= Wi log2 1 +  (6)
αi Wi N0 + M m
m=1 Pt hm,i (Ti − Ci )2 ≤ ε. (14)
JIA et al.: INTERBEAM INTERFERENCE CONSTRAINED RESOURCE ALLOCATION 6055

Equation (10) denotes that the actual bandwidth distribution Algorithm 1 Proposed Optimal Algorithm
of beam should be less than the traffic demand of each beam, Step 1: Set appropriate initial values for the dual variables ρ, λ, μ, ν,
which shows the upper limit of allocated. According to (8), the iteration accuracy ε, the maximum number of iteration Niter . The
bandwidth of each beam is set to Wi = Wtotal /M, i ∈ (1, 2, . . . , M),
lower limit of allocated capacity is relevant to delay in (11). and the indicator variable p = 1.
Equations (12) and (13) indicate the total allocated resources Step 2: Substitute the above relevant values into (17) and calculate
opt
cannot exceed total power and bandwidth resources, respec- the optimized power Pi , i ∈ (1, 2, . . . , M).
opt
tively. Equation (14) indicates that the difference between Step 3: Substitute the values of Pi and ρ, λ, μ, ν into (18), and
actual bandwidth distribution of each satellite beam and traffic opt
then calculate the optimized bandwidth Wi , i ∈ (1, 2, . . . , M).
demand cannot be too wide. opt opt
Step 4: Substitute the values of Pi and Wi , i ∈ (1, 2, . . . , M)
By introducing the non-negative Lagrange multipliers ρ, λ, into (20) to update the dual variables. 
  i+1 
ν, and μ, the Lagrange function of the optimization problem Step 5: If the conditions of Ptotal − M Popt  < ε,
 M  M i+1
i=1
can be obtained  i+1 
Wtotal − i=1 Wopt  < ε, Ptotal ≥ i=1 Popt and Wtotal ≥
M i+1
i=1 Wopt are satisfied simultaneously, or i = Niter is satisfied,
L(P, W, ρ, λ, μ, ν) then jump to Step 6. Otherwise, jump to Step 2.
M 
M   Step 6: the resources optimal allocation is completed.
Ti
= (Ti − Ci )2 + ρi − Ci
(1 − ei )Di
i=1 i=1

M 
M
+λ (Pi − Ptotal ) + μ (Wi − Wtotal ) Specifically, the solution processes can be divided into the
i=1 i=1 following three subproblems.
M 
  Subproblem 1 (Power Allocation): The dual variables ρ, λ,
+ν (Ti − Ci )2 − ε (15) μ, and ν are given. The partial derivative is taken for each
i=1 Pi (i = 1, 2, . . . , M) in (15) based on Karush–Kuhn–Tucker
condition [22]. The approximate solution P∗i satisfies (16) as
where P = {P1 , P2 , . . . , PM }, W = {W1 , W2 , . . . , WM }, ρ = follows:
{ρ1 , ρ2 , . . . , ρM }.  

Due to the fact that the consideration of interbeam interfer- ρi αi λN0 ln 2 P∗i + M k=1 Pk hi,k
ence is taken in resource allocation model, this optimization Ti − Ci + = 1+ . (17)
2 2 αi Wi N0
problem does not belong to convex optimization. Therefore,
we solve this problem from the perspective of dual problem.
The comparison of relationship between the optimal solu- Assuming the condition Ci > Ti is justifiable, it is easy to
tion and dual variable, the primary problem can be expressed find that the condition (11) is satisfied due to (1 − ei )Di > 1.
as [19] From (17), we can infer the value of λ is negative, which
is conflicting with the premise of λ ≥ 0. Therefore, the con-
g(ρ, λ, μ) = min L(P, W, ρ, λ, μ, ν) straint (10) can be ignored while the value of λ is non-negative.
P,W In this paper, the golden section method is adopted to obtain
⎛   ⎞
P∗i , then we get the optimal power Pi = max(0, P∗i ).
opt
f (P, W) + M i=1 ρ i
Ti
(1−ei )Di − C i opt
⎜  ⎟ Subproblem 2 (Bandwidth Allocation): Pi is brought
⎜+ λ M (P − P ) ⎟
= min⎜ i=1 i total ⎟ into (16), then the partial derivative is taken for each Wi (i =
P,W ⎝ + μ M
(Wi − Wtotal )  ⎠
 i=1 1, 2, . . . , M). The approximate solution Wi∗ satisfies
+ν M i=1 (T i − C i ) 2
− ε
M      
 ∗ ∗ Ti ∗
 ρi 
opt
Pi
≤ f P ,W + ρi − Ci 2 Ti − Ci + × H log2 1 + ∗ opt
− αi Wi Pi N0
(1 − ei )Di 2 αi Wi∗ N0
i=1

M
 ∗  
M
 ∗  − Eμ = 0 (18)
+λ Pi − Ptotal + μ Wi − Wtotal
i=1 i=1
M  where the value of E can be represented as
 
+ν (Ti − Ci )2 − ε ⎡ 2
 i=1
 
M
≤ f P∗ , W ∗ = p∗ (16) E=⎣ αi Wi∗ N0 + Pk hi,k
k=1
  
where d∗ and p∗ are the optimal solution of dual problem 
M
αi Wi∗ N0
opt
and primary problem, respectively. d∗ is a lower bound of p∗ , + + Pk hi,k Pi ln 2. (19)
and the dual gap |p∗ − d∗ | is not zero. The optimal solution k=1
d∗ can be obtained based on the Lagrange duality theory and
subgradient iteration. The proposed algorithm is described as By adopting the golden section method to obtain Wi∗ , the
optimal bandwidth Wi = max{0, Wi∗ } is obtained.
opt
follows.
6056 IEEE INTERNET OF THINGS JOURNAL, VOL. 6, NO. 4, AUGUST 2019

TABLE I
PARAMETERS IN M ULTIBEAM S ATELLITE S YSTEM

Fig. 3. Capacity allocation to beams at the same channel conditions.

Subproblem 3 (Dual Variables Update): The dual variables


can be obtained
 opt opt opt opt  OPUB, which optimally distributes power among the
ρ ,λ ,μ ,ν = arg max , g(ρ, λ, μ, ν) satellite and terminals. The bandwidth allocated to each
ρ,λ,μ,ν
  beam is Wi = Wtotal /M, i ∈ {1, 2, . . . , M}.
= arg max min L Popt , W opt , ρ, λ, μ, ν . 3) The optimal bandwidth allocation algorithm (UPOB).
ρ,λ,μ
(20) The convex optimization method is used to implement
UPOB, which optimally distributes bandwidth among
A subgradient method is applied to update the duality the satellite and terminals. The power allocated to each
variables beam is Pi = Ptotal /M, i ∈ {1, 2, . . . , M}.
  + 4) The uniform resource allocation algorithm (UPUB).
opt Ti
ρin+1 = ρin −
nρi Ci − Total system resources are averagely assigned to
(1 − ei )Di
  + each beam.
M
opt 5) The OPOB without delay constraint (OPOBND). The
λ n+1
= λ −
λ Ptotal −
n n
Pi
delay constraint can transform into traffic demand con-
i=1
  + straint, which can be expressed as (8).

M
opt The feasibility and effectiveness of the proposed algorithm
μn+1 = μn −
nμ Wtotal − Wi will be proved in terms of the following four aspects.
i=1
  +
ν n+1
= ν n −
nν ε − (Ti − Ci )2 (21)
A. Impacts of Delay Constraints on Resource Allocation
where [x]+ = max{0, x}, n is iteration number, and
n is iter-
The following reasonable assumptions are made to explore
ation step sequence. As long as the value of
n is reasonable,
the performance of OPOB algorithm under two channel
the subgradient method can ensure the convergence of dual
conditions.
variables. The optimal solution of dual problem satisfies the
When the channel conditions of all beams are the same,
lower bound of primary problem [23]. However, the duality
we set the traffic demand of the ten beams as 120–255 Mb/s.
gap is close to zero infinitely when the number of iterations is
The spot beam with higher traffic demand will be allocated
large enough [24]. The duality gap is zero, therefore the local
more resources to minimize the objective function. As shown
optimal solution is also the global optimal solution [23].
in Fig. 3, compared OPOBND with OPOB, the performance of
OPOBND is slightly better than that of OPOB. However, the
IV. N UMERICAL A NALYSIS AND S IMULATION capacity allocated to beam 1–3 by OPOBND is less than cor-
The S band multibeam satellite communication sys- respondingly minimal traffic demand, while OPOB can meet
tem model is established in the simulations. According the minimal traffic demand of each beam at least.
to [25] and [26], network parameters provided for simulation When the channel conditions of all beam are inconsistent,
in this paper are shown in Table I. The performance of the fol- we set the same demand for some beams which have different
lowing five algorithms are compared to prove the effectiveness channel conditions to indicate the impact of the traffic demand
of the proposed algorithm. on resource allocation. The normalized noise power spectral
1) The optimal joint power and bandwidth allocation algo- density parameters of ten beams are set to [0.2, 0.25, 0.3, 0.35,
rithm (OPOB). 0.4, 0.2, 0.2, 0.2, 0.2, 0.2], respectively. Each traffic demand
2) The optimal power allocation algorithm (OPUB). The of ten beams is [120, 150, 150, 150, 150, 180, 200, 220, 240,
convex optimization method is used to implement 260] Mb/s.
JIA et al.: INTERBEAM INTERFERENCE CONSTRAINED RESOURCE ALLOCATION 6057

Fig. 6. Bandwidth utilization variance versus the beam number.

Fig. 4. Capacity allocation to beams at different channel conditions.

Fig. 5. Capacity distributions of spot beams comparisons among OPOB,


OPUB, UPOB, and UPUB. Fig. 7. Bandwidth utilization variance versus activity ratio of users.

Fig. 3 illustrates the capacity allocated to each beam when is Wi = Wtotal /M, i ∈ {1, 2, . . . , M}. However, the optimiza-
the channel conditions are the same. The comparison of total tion method is used to implement the OPOB algorithm, which
capacity among different channel conditions is shown in Fig. 4. optimally distributes power and bandwidth among the satel-
In beam 4 and 5, the OPOBND fails to satisfy their minimal lite and terminals. Therefore, OPOB algorithm allocates more
traffic demand. In contrast, the OPOB can meet the mini- capacity to higher demand beams and allocates less capac-
mal demand of each beam at least. However, OPOB slightly ity to lower demand beams. Therefore, OPOB algorithm can
reduces the capacity beside beam 4 and 5, the proposed provide greater flexibility in resource allocation to meet the
algorithm has a better fairness among all beams. different demand of each beam, and significantly enhance the
system capacity. It means that the proposed method is more
B. Capacity suitable to the traffic demands.
Assuming that the channel conditions for each beam are the
same, the power and bandwidth allocated to each beam of four C. Bandwidth Utilization Variance
algorithms are shown in Fig. 5, respectively. Compared with Fig. 6 shows the bandwidth utilization variance between
UPUB, OPUB, and UPOB, the beams with higher demand are the traffic demand and the capacity in each spot-beam. The
allocated with more power and bandwidth, and the beams with bandwidth utilization variance of OPOB is slightly higher than
lower demand are allocated with fewer resources through the that of other algorithms in the beam 1–5. However, OBOP
joint allocation algorithm. Therefore, OPOB could approach keeps on a steady bandwidth utilization variance value from
the users’ demand, to a large extent. beam 6 to 10 while other algorithms tend to grow rapidly.
Fig. 5 shows the capacity distributions of each spot-beam. In It means that the proposed method has better performance in
terms of capacity allocated to each beam, traditional UPUB interbeam fairness.
algorithm allocates the same power and bandwidth to each As shown in Fig. 7, the convex optimization method and
beam without considering the traffic requirements. In UPOB, the minimax method are used to implement the OPOB, which
the power allocated to each beam is Pi = Ptotal /M, i ∈ optimally reduce bandwidth utilization variance. Hence, the
{1, 2, . . . , M}. In OPUB, the bandwidth allocated to each beam proposed method not only fits the traffic demands better but
6058 IEEE INTERNET OF THINGS JOURNAL, VOL. 6, NO. 4, AUGUST 2019

capacity, minimize the gap between capacity and demand, and


will not introduce too much interference.

V. C ONCLUSION
Due to the shortage of on-board resources in satellite com-
munication system, based on golden section theory and subgra-
dient iteration, a joint power and bandwidth resource allocation
scheme called OPOB has been proposed. Moreover, the capac-
ity allocation model in satellites downlink is established, which
takes the interbeam interference, channel conditions, and delay
factors into consideration. Furthermore, a joint power and
bandwidth allocation algorithm is proposed, and the optimal
solution can be obtained through the Lagrange duality theory
Fig. 8. Total system capacity comparisons among OPOB, OPUB, UPOB,
and subgradient iteration. Especially, compared with OPUB,
and UPUB. UPOB, and UPUB, the proposed algorithm can improve the
total system capacity and minimize the gap between the band-
width requirement and actual assigned bandwidth without
also can reduce the bandwidth utilization variance at any introducing too much interference based on minimax princi-
traffic density. Compared with OBUP, UBOP, and UBUP, ple and golden section theory. The evaluation results indicate
the proposed algorithm can minimize the gap between the that the proposed algorithm has advantages against the eval-
traffic demand and the capacity at different traffic densities, uated methods in time delay, bandwidth utilization variance,
which meet the business requirements of each beam to a large capacity, and fairness, which is robust in practical scenarios.
extent. Therefore, under different traffic densities, the proposed
scheme OPOB provides the best performance.
R EFERENCES
D. Impact of Interbeam Interference on Resource Allocation [1] Z. Qu, G. Zhang, H. Cao, and J. Xie, “LEO satellite constellation for
Internet of Things,” IEEE Access, vol. 5, pp. 18391–18401, 2017.
The total system capacity and interference power of four [2] D. Helbing et al., “Saving human lives: What complexity science and
algorithms are illustrated in Fig. 8. Ignoring the interbeam information systems can contribute,” J. Stat. Phys., vol. 158, no. 3,
interference, UPUB algorithm can achieve maximum system pp. 735–781, Feb. 2015.
[3] M. Jalili and M. Perc, “Information cascades in complex networks,”
capacity. Taking interbeam interference into consideration, the Complex. Netw. J., vol. 5, no. 5, pp. 665–693, Jul. 2017.
total capacity of each algorithm has a certain level of decline. [4] M. Jia, X. Gu, Q. Guo, W. Xiang, and N. Zhang, “Broadband
In this case, the total system capacity provided by OPOB algo- hybrid satellite-terrestrial communication systems based on cognitive
rithm is superior to the other algorithms. Although the level radio toward 5G,” IEEE Trans. Commun., vol. 23, no. 6, pp. 96–106,
Dec. 2016.
of interference of OPOB algorithm is slightly larger than oth- [5] D. Wang, N. Zhang, Z. Li, F. Gao, and X. Shen, “Leveraging high
ers, it is worth to sacrifice in interference compared with total order cumulants for spectrum sensing and power recognition in cogni-
system capacity. tive radio networks,” IEEE Trans. Wireless Commun., vol. 17, no. 2,
pp. 1298–1310, Feb. 2018.
When the interbeam interference is ignored, the channel [6] Z. Li, L. Guan, C. Li, and A. Radwan, “A secure intelligent spectrum
conditions of the spot beams are the same. Hence, the UPUB control strategy for future THz mobile heterogeneous networks,” IEEE
algorithm can be seen as a special case of the water filling Commun. Mag., vol. 56, no. 6, pp. 116–123, Jun. 2018.
[7] U. Park, H. W. Kim, D. S. Oh, and D.-L. Chang, “Performance analysis
algorithm. When the power is equally allocated among the of dynamic resource allocation for interference mitigation in integrated
spot beams, the maximum channel capacity can be reached satellite and terrestrial systems,” in Proc. ICNGMA, Cambridge, U.K.,
in UPUB. In addition, the UPOB algorithm divides the power Sep. 2015, pp. 217–221.
[8] F. Qi, L. Guangxia, F. Shaodong, and G. Qian, “Optimum power allo-
equally into multiple spot beams, but the allocated scheme of cation based on traffic demand for multi-beam satellite communication
the bandwidth resources affects the total capacity of the sys- systems,” in Proc. ICCT, Jinan, China, Sep. 2011, pp. 873–876.
tem. The capacity of the OPOB algorithm is slightly less than [9] R. Dhaou et al., “Optimized handover and resource management: An
802.21-based scheme to optimize handover and resource management in
UPUB algorithm, which is close to the theoretical maximum hybrid satellite-terrestrial networks,” Satellite Commun. Netw. J., vol. 32,
channel capacity. no. 1, pp. 1–23, 2014.
When the interbeam interference is considered, the capacity [10] G. Ding, Q. Wu, and J. Wang, “Sensing confidence level-based joint
spectrum and power allocation in cognitive radio networks,” Wireless
of the UPUB algorithm is weakened, however, the pro- Pers. Commun., vol. 72, no. 1, pp. 283–298, Sep. 2013.
posed joint power bandwidth optimization allocation algorithm [11] H. Wang, A. Liu, X. Pan, and J. Yang, “Optimization of power alloca-
achieves a better capacity. Considering the co-frequency inter- tion for multiusers in multi-spot-beam satellite communication systems,”
ference in the system, the interference size of the OPOB Math. Prob. Eng., vol. 3, pp. 1–10, Feb. 2014.
[12] X. Chen, R. Q. Hu, and Y. Qian, “Distributed resource and power alloca-
algorithm is slightly higher than that of the others. It is tion for device-to-device communications underlaying cellular network,”
worthwhile to improve the system capacity with a small inter- in Proc. GCC, Austin, TX, USA, Dec. 2014, pp. 4947–4952.
ference increase. Therefore, compared with the single resource [13] X. Hu, S. Liu, R. Chen, W. Wang, and C. Wang, “A deep reinforcement
learning-based framework for dynamic resource allocation in multibeam
allocation algorithm and the UPUB algorithm, the proposed satellite systems,” IEEE Commun. Lett., vol. 22, no. 8, pp. 1612–1615,
algorithm can improve resource utilization, maximize system Aug. 2018.
JIA et al.: INTERBEAM INTERFERENCE CONSTRAINED RESOURCE ALLOCATION 6059

[14] J. Lei and M. A. Vazquez-Castro, “Joint power and carrier allocation Ximu Zhang received the M.Sc. degree from the
for the multibeam satellite downlink with individual SINR constraints,” Harbin Institute of Technology, Harbin, China, in
in Proc. ICC, Cape Town, South Africa, May 2010, pp. 1–5. 2017, where she is currently pursuing the Ph.D.
[15] F. Li et al., “Joint pricing and power allocation for multibeam satellite degree at the Communication Research Center,
systems with dynamic game model,” IEEE Trans. Veh. Technol., vol. 67, School of Electronics and Information Engineering.
no. 3, pp. 2398–2408, Mar. 2018. Her current research interests include broadband
[16] F. Li et al., “Spectrum trading for satellite communication systems with satellite communications and heterogeneous cloud
dynamic bargaining,” IEEE Trans. Commun., to be published. radio access network.
[17] H. Wang, A. Liu, and X. Pan, “Optimization of joint power and band-
width allocation in multi-spot-beam satellite communication systems,”
Math. Prob. Eng., vol. 1, pp. 1–9, Dec. 2014.
[18] T. M. Cover and J. A. Thomas, Elements of Information Theory.
New York, NY, USA: Wiley, 1991. Xuemai Gu (M’16) received the M.Sc. and Ph.D.
[19] “Satellite antenna radiation pattern for use as a design objec- degrees from the Department of Information and
tive in the fixed-satellite service employing geostationary satellites,” Communication Engineering, Harbin Institute of
Int. Telecommun. Union, Geneva, Switzerland, ITU-Recommendation Technology (HIT), Harbin, China, in 1985 and 1991,
ITU-R S.672-4, 1997. respectively.
[20] M. Jia, X. Zhang, X. Gu, and Q. Guo, “Energy efficient cognitive spec- He is currently a Professor and the President of the
trum sharing scheme based on inter-cell fairness for integrated satellite- Graduate School, HIT. His current research interests
terrestrial communication systems,” in Proc. VTC, Porto, Portugal, include integrated and hybrid satellite and terrestrial
Jun. 2018, pp. 1–6. communications and broadband multimedia commu-
[21] S. Boyd and L. Vandenberghe, Convex Optimization. Cambridge, U.K.: nication technique.
Cambridge Univ. Press, 2004.
[22] Z. Hou, H. Chen, Y. Li, and B. Vucetic, “Incentive mechanism design
for wireless energy harvesting-based Internet of Things,” IEEE Internet
Things J., vol. 5, no. 4, pp. 2620–2632, Aug. 2018.
[23] N. Hu, B. Duan, and H. Cao, “Robust optimization with convex model Qing Guo (A’11–M’16) received the M.Sc.
considering bounded constraints on performance variation,” Struct. degree from the Beijing University of Posts and
Multidiscipl. Optim., vol. 56, no. 1, pp. 59–69, Nov. 2017. Telecommunications, Beijing, China, in 1985, and
[24] G. Tychogiorgos and K. K. Leung, “Optimization-based resource allo- the Ph.D. degree from the Harbin Institute of
cation in communication networks,” Comput. Netw., vol. 66, pp. 32–45, Technology (HIT), Harbin, China, in 1998.
Jun. 2014. He is currently a Professor and the President of the
[25] K. Kang et al., “Analysis of interference and availability between School of Electronics and Information Engineering,
satellite and ground components in an integrated mobile-satellite ser- HIT. His current research interests include satellite
vice system,” Satellite Commun. Netw. J., vol. 33, no. 4, pp. 351–366, communications and broadband multimedia commu-
Apr. 2015. nication techniques.
[26] “Technical specification group radio access network, evolved universal
terrestrial radio access (E-UTRA), radio frequency (RF) system scenar-
ios (release 10), V10.3.0,” 3GPP, Sophia Antipolis, France, Rep. TR
36.942, pp. 14–15, 2012.
Yaqiu Li received the M.Sc. degree from the
China Academy of Space Technology, Beijing,
China.
She is currently a Deputy Chief Engineer with
the China Academy of Space Technology. Her cur-
rent research interests include integrated satellite-
terrestrial network and communication mechanism
of UHF mobile communication satellite system.

Min Jia (M’13–SM’17) received the M.Sc. degree


in information and communication engineering
from the Harbin Institute of Technology (HIT),
Ping Lin received the M.Sc. degree from the
Harbin, China, in 2006, and the Ph.D. degree from
Harbin Institute of Technology, Harbin, China,
Sungkyunkwan University, Seoul, South Korea, and
in 2017.
HIT in 2010.
She is currently with Hisense, Qingdao,
She is currently an Associate Professor and a
China. Her current research interests include
Ph.D. Supervisor with the Communication Research
integrated satellite-terrestrial network and satellite
Center, School of Electronics and Information
communication.
Engineering, HIT. Her current research interests
include advanced mobile communication technology
and nonorthogonal transmission scheme for 5G, cognitive radio, digital signal
processing, machine learning, and broadband satellite communications.

S-ar putea să vă placă și