You are on page 1of 14

Received: 2 May 2019 Revised: 8 July 2019 Accepted: 10 August 2019

DOI: 10.1002/ett.3748

RESEARCH ARTICLE

Multiobjective cooperative swarm intelligence algorithm


for uplink resource allocation in LTE-A networks

M. Leeban Moses1 B. Kaarthick2

1
Bannari Amman Institute of Technology,
Sathyamangalam, India Abstract
2
Coimbatore Institute of Engineering and Nowadays, Long-Term Evolution Advanced (LTE-A) network is the primary
Technology, Coimbatore, India
innovation in 4G networks. The LTE-A networks convey exceptional data rates
Correspondence and low latency for few sorts of application. Sometimes, in LTE-A, the mul-
M. Leeban Moses, Bannari Amman tiobjective uplink resource allocation is the perplexing optimization problem,
Institute of Technology,
Sathyamangalam-638401, India. and this problem is considered as 0-1 multiobjective knapsack trouble. In
Email: mleebanmoses1003@gmail.com order to overcome this trouble, a multiobjective cooperative swarm intelligence
algorithm for resource allocation is proposed in this paper. In our proposed
work, hybrid firefly algorithm (FFA) and particle swarm optimization (PSO)
algorithm are utilized to solve 0-1 multiobjective knapsack problem. Initially, a
priority and urgency factor (urgency of packets)–based user ranking and quan-
tifying scheme is designed for scheduling process. After the scheduling process,
the resource allocations employ three objective functions, such as maximization
of resource utilization, maximization of quality of service (QoS), and interfer-
ence minimization. The optimization problem is overcome by using the hybrid
FFA and PSO algorithm. The experimental outcomes demonstrate that it has the
best QoS and less interference in the resource allocation in LTE-A network than
state-of-art methods in the proposed strategy.

1 I N T RO DU CT ION

In Long-Term Evolution (LTE) and LTE Advanced (LTE-A), the resource allocation and link adaptation are discussed
with an intention on the location; and about the pertinent reference and control signal's formatting; and, in addition, to
that they enable the decisions.1,2 The communication between individuals is done by utilizing wireless communication.
The wireless connectivity between machines has been extended by the human-to-human (H2H) communications and
machine-to-machine (M2M) communications in cellular networks (CNs).3 By the third-generation partnership project,
the machine-type communication (MTC) is also called M2M communications as a key standardization body.4,5 MTC is
generally utilized for transportation systems, healthcare, smart metering, and monitoring systems.6,7
The LTE/LTE-A technology plays a major role in back filling of MTC applications.8 However, because of the diverse
characteristics of quality of service (QoS) requirement, power efficiency, transmission periodicity, and payload size, MTC
poses huge challenges to LTE/LTE-A technologies.9,10 Devaluation of user equipment (UE) cost and radio access network
(RAN) overload control issue is mainly focused by the LTE-A standards.11 LTE-A networks are planned fundamentally
for H2H communications; therefore, typically, the amount of uplink (UL) traffic is lower than the downlink (DL) traffic.12
Specifically, a standout among the most basic difficulties is efficient management of massive random access. The pream-
ble collision's probability increases extremely on account of massive random access attempts, and in this manner, the
LTE/LTE-A random access decreases forcefully while executing.13

Trans Emerging Tel Tech. 2019;e3748. wileyonlinelibrary.com/journal/ett © 2019 John Wiley & Sons, Ltd. 1 of 14
https://doi.org/10.1002/ett.3748
2 of 14 MOSES AND KAARTHICK

By embracing LTE/LTE-A system for supporting M2M communication, numerous issues can happen. For instance, the
quantity of M2M devices in a normal LTE cell is considerably higher than the quantity of H2H terminals that might be
normal in that same LTE cell.14 As the result, the excessive delay and congestion are likely to occur; moreover, the packet
arrival rates for M2M devices are much lower than H2H devices. M2M devices that take over the conventional random
access methodology, which would first require an arrangement, then send their information after receiving the connec-
tion, and finally give it up rapidly until the next packet.15 Various possible answers are present to ease congestion and
enhance the performance of random access. Some of them are class barring scheme, back-off technique, slotted access
method, pull-based scheme, dynamic physical random access channel (PRACH) resource allocation method, and PRACH
resource separation scheme.8,16-19 However, these techniques have some advantages and disadvantages. In this context,
the method adopted a cooperative swarm intelligence algorithm in the LTE-A network for the UL resource allocation.
Some existing performance such as delay, cost, overhead, and complexity are minimized and the throughput, fairness,
resource utilization is increased by the proposed schemes. At the time of allocating the resources in the LTE-A network,
some optimization problems arise; this optimization problem is considered as the multiobjective 0-1 knapsack problem.
To overcome the optimization problem, the metaheuristics technique is used. In our work, the problem is avoided by
adopting the firefly algorithm and particle swarm optimization (FFA-PSO) algorithm in the proposed method. Our pro-
posed algorithm can well contribute to maximize the throughput and provide an excellent balance between analysis and
corruption of the search space. The FFA searches optima by investigating the hunt space, and PSO is used for refreshing
FFA to locate the global optimum arrangement. Thus, the advantages of both nature-inspired algorithms are combined.
User ranking based on the urgency of the packet is known as the urgency factor before the resource allocation. The exhibit
of the proposed method can expand the system limit, and QoS also shows the complexity and the overhead reduction
when compared and the existing optimization techniques.
The remaining sections of this paper are depicted as follows. Sections 1 includes the introduction part. Sections 2
includes related works. Sections 3 includes the proposed method. The allocation of UL resource in LTE-A network is
discussed in Sections 3.1. In Sections 3.2, the multiobjective 0-1 knapsack problem in optimal UL resource allocation is
depicted. The optimal FFA-PSO algorithm is depicted in Sections 3.3. The results and the conclusion sections are included
in Sections 4 and 5.

2 RELATED WORKS

To achieve the desired features simultaneously, Lee et al20 designed a multiobjective resource allocation scheme. Three
objective functions are first formulated by the author to minimize interference, guarantee a high degree of fairness, and
maximize resource utilization efficiency. To form a single multiobjective optimization problem, the objective functions
are combined using a weighted sum approach. To this problem, the Pareto-optimal solution was found by utilizing an ant
colony optimization algorithm.
Shukla and Bhatia21 presented a packet scheduling algorithm (PSA) in the LTE/LTE-A in CNs for assuring the video
in real-time and data-centric traffic become effective in the DR wireless technologies. Because of high traffic inflow in
the LTE-based CNs (LCNs) at the time of an emergency or natural calamity, the LCN will encounter high congestion in
the remote/optical connections, bringing about high packet loss of basic data from the influenced circle. In their examine
work, a capable PSA was situated on the class of service, prioritization, information rates utilization, and area of traffic
origin in the CN. Their PSA will lessen the packet loss in the LCN and, hence, guarantee the consolidation of QoS.
In RAN and CN of the LTE network, Alavikia and Ghasemi22 have proposed a technique named as an adaptive mecha-
nism to manage an expansive number of MTC devices. To regulate the MTC traffic, they have utilized access class barring
(ACB) scheme, as indicated by the congestions level in the CN and RAN. The author considered the MTC device based
on the class on two priority schemes such as RAN resources and contending. Initially, the RAN formulated the overload
issue to locate the quantity of allowable contending MTC devices. At that point, an active load management policy was
proposed in the view of additive increase multiplicative decrease rule to control the incoming load to the CN from mul-
tiple cells. In their method, each advanced node B updates the ACB factor upon overburden identification in the RAN or
CN in an adaptive way, as far as possible the quantity of MTC devices in both CN and RAN.
Ragaleux et al23 have concentrated on the UL schedule imposed by the LTE/LTE-A standard with the practical con-
straints. The authors considered the multimedia traffics for generating the mobile user for heterogeneous requirements in
help of QoS. They have executed an adaptive and potential aware scheduling scheme (APASS) to handle the subsequent
MOSES AND KAARTHICK 3 of 14

time and frequency issue. APASS was standard agreeable and covers an extensive variety of scheduling goals. Three
algorithms are made by scheduling plan, which works jointly to give a productive solution.
To address the expanding requests of mobile traffic, BenMimoune et al24 have discussed about LTE/LTE-A networks.
Relay selection and DL resource allocation are engaged in their work; a user might be associated with a multihop relay
in a base station (BS) and have some relay stations from which to pick in his range. A relay selection and dynamic joint
resource allocation scheme were introduced by the authors to beat the extra difficulties presented by multihop relay nodes.
An adaptive LTE-A cross-layer packet scheduling (ALS) was proposed by Chang et al,25 the real-time high-speed packet
service is guaranteed for LTE-A. ALS consists of two mechanisms, ie, adaptive reward priority scheduling and dynamic
resource allocation algorithm. The average packet delay, packet dropping probability, system capacity, etc, are compared
in the result.
While satisfying QoS requirements for M2M devices, Ghavimi et al26 have proposed a group-based M2M commu-
nications for proficiently assign resource blocks (RBs), in which M2M devices are clustered based on their wireless
transmission protocols, QoS requirements, and their characteristics. Based on LTE-A networks to perform power alloca-
tion and joint RB in single carrier frequency division multiple access (SC-FDMA), they have figured out a sum-throughput
maximization issue in M2M devices while regarding every one of the imperatives related with SC-FDMA method and
QoS requirements. In LTE-A networks, the requirements in UL SC-FDMA air interface increased the resource alloca-
tion issue. First, this issue is constructed as a binary integer programming; then, the authors have tackled the resource
allocation issue; afterward utilizing the Lagrange duality theory, a dual issue is formulated.
Sun et al27 have presented a device-to-device (D2D) resource allocation in LTE-A network using hybrid PSO genetic
algorithm (GA). For maximizing the throughput of the system, this approach shares the same frequency resource with
one cellular UE (CUE), which allows two D2D pairs only. D2D pair should employ the frequency resource of the CUE
after the mapping process of the PSO algorithm by utilizing a mode selection scheme. Their approach could upgrade
the diversity of the particles to avoid the local optima. Their presented scheme also upgrades the spectral efficiency and
mitigates interference between the D2D pairs and the CUEs.
Hybrid genetic simulated annealing (GSA) energy efficient resource allocation was presented by Xu et al28 for orthogo-
nal frequency division multiple access system. Hybrid GSA and support vector machine (SVM) are integrated to enhance
the performance. To compute the outage probability constraint, SVM is employed in their system. In their approach, a
power allocation objective function was utilized to minimize the power consumption over subcarrier.
Coordinated multipoint (CoMP) is a technique that coordinates the clusters using radio controller coordinator (RCC)
to enhance the network throughput. The RCC's position is updated as per the arrange qualities, for example, the topo-
graphical size, cells dispersion and thickness, accessible transmission capacity, and so forth. When picking the ideal RCC
position, additionally, the determination of legitimate coordination methods (eg, coordinated scheduling (CS) versus
joint transmission) assume a significant job as they give generously unique throughput addition relying upon bunch size
and idleness.29 Marotta et al analyzed the convergence time to the transfer CoMP information among the coordinated
cluster of enhanced node Bs (eNBs) with various instantiation deployments of CS functionality over network function
virtualization infrastructure.30
In distributed CoMP, the eNBs transfer load information messages from one to another cell through the X2 interface.
The X2 topology outcomes from the instantiation of neighborhood relations among eNBs, which influences the quantity
of traded messages among eNBs. In C-CoMP, a centralized radio resource coordinator (CRRC) is embedded in the system
to receive the messages from eNBs. The CRRC is responsible for settling on planning the choices. It is critical to make
reference when eNBs utilize another characterized XN interface rather than standard X2 interface.31

3 P RO P O S ED MET H OD

Resource allocation in LTE-A networks played a major role in an extraordinary research movement. Compared with the
DL resource allocation, the UL resource allocation has less consideration. Here, the UL scheduling challenges are man-
aged by the proposed strategy. In this paper, by considering the QoS requirements, the outline of an optimal scheduling
algorithm for LTE-A UL, optimal resource utilization, and a decreased measure of interference determined for every appli-
cation are addressed. We plan to clarify how the request of the bearers in the application architecture should be suited by
relegating the available radio resources to such an extent that the application requirements and resource constraints in
the network are satisfied without giving up the system throughput. The procedure relies on the idea of optimization issue
in which the resource allocation issue is planned as a multiobjective knapsack issue. This takes into consideration brisk
4 of 14 MOSES AND KAARTHICK

User Equipment (UE)


eNB1- enhanced Node B
BS -Base Station
Interference
BS1
BS1
eNB1

eNB1

BS1

BS1

BS1

FIGURE 1 System model

and found the exact solutions of optimal allocation decisions. The principle commitments of this paper are condensed as
takes after.
1. A priority-based user ranking and quantifying scheme is designed based on priority by which the charge of requesting
and filtering users and urgency of packets.
2. Resource utilization maximization, QoS maximization, and interference minimization are the three resource alloca-
tion objective function.
3. A resource allocation algorithm in view of a multiobjective cooperative swarm intelligence algorithm,32 centralizing
to take care of the optimization problem by utilizing the FFA and PSO, is utilized.
Figure 1 demonstrates the system model of an UL resource allocation in the LTE-A network. By a set of cells, the
system model is composed and network nodes (including eNB, UE, and BS) are distributed among cells. Here, the UE
is communicating with the network by means of the BS, which is wirelessly associated with the eNB. The associations
between the BS and the UE are known as the direct link (Figure 1). Here, some interference is happened between the
BS and the eNBs at the time of resource allocation to the user because of minimum QoS, minimum resource utilization.
This issue is considered as the 0-1 multiobjective knapsack problem (MOKP) in the resource allocation. To dodge such
optimization problems, the proposed method used the scheduling schemes and the optimal resource allocation schemes
and these are portrayed as takes after.

3.1 Resource allocation in UL LTE-A


In a wireless network, to enhance the network execution, a segment of bandwidth or power is assigned to the distinctive
users that are known as the resource allocation. In the resource allocation process, the scheduling process is the most
imperative one. In the meantime, there are two essential attributes that are incorporated into the scheduling process. The
initial one is to fulfill the QoS requirements of the user and the second one is to build the efficiency of resources allocated
to the users. In the proposed technique, the scheduling process merges numerous phases, which permits encouraging the
arrangement of QoS too. Here, the scheduling problems are dodged in view of the priority of the user. The priority of the
user is registered in the segment underneath.
Phase 1
In the UL resource allocation scheme, the users are ranked in light of the priority and the urgency of packets, which are
also called the urgency factor. In the UL resource allocation scheme, for each QoS class identifier (QCI), the packet delay
budget is processed. The application that coincides with the LTE-A network has diverse QoS and this is isolated into the
MOSES AND KAARTHICK 5 of 14

distinctive bearer. Here, one QCI is dispensed to every bearer to distinguish the specific service for a fitting application.
The urgency factor is processed in light of this packet delay budgets and it is figured as takes after:
w
UFi = , (1)
d
where, in the buffer of user i, the urgency of the head of line (HoL) packet is represented as UF, the time interval that the
HoL packet waiting in the buffer is denoted as w, and the packet delay budget is represented as d. Here, the priority factor
for each user is computed to find the QoS mechanism and it is given as follows:
{
2 if app = GBR
PFi = (2)
1 if app = Non GBR.

The greater priority for applications with guaranteed bit rate (GBR) is obtained by the scheduling algorithm indicated by
the QCI definition. In the aforementioned equation, the urgency factor UFi is utilized to compute the packet delay budget
with a specific end goal to separate the each QCI. From Equations (1) and (2), it is believable to characterize the metric
that controls the ordering of users in phase, one as follows:

𝛿Fi = UFi .PFi . (3)

The aforementioned equation 𝛿Fi is utilized to ordering the N active users. The priority of the user and the differences of
priority among users are defined by using the factor 𝛿Fi . After that, the user ranking list is given to the next phase of the
scheduling process.
Phase 2
There are two errands incorporated into the second phase: (1) if the latency of the packet is near the packet delay budget
(ie, UFi ≥ 0.9), the resources are straight forwardly allocated to the users; (2) from the stage one, discovering which users
of the ranking list is passed to the phase 2. On the off chance that the value of UFi ≥ 0.9, either the RBs are accessible or not
is recognized in phase 2. Meanwhile, if any RBs are now allotted by the hybrid automatic repeat request entity or the value
of QCI is equivalent to 0, the specific RB is absent. In this place, the biggest interval of contiguous RBs is characterized
by phase 2 and, furthermore, processes the demand Di for every RB for every user to designate the contiguous RB in light
of the buffer status report and modulation coding scheme. In the event that the value of UFi < 0.9, the user is selected for
phase 3.
Phase 3
In phase 3, the rest of the resources are dispensed to the users chosen from phase 2 and it has appeared in Figure 2. This
resource allocation depends on a cooperative swarm intelligence algorithm combining FFA and PSO and is utilized to
take care of the optimization problem. Figure 2 demonstrates the feasible scheduling process of the LTE-A network.

3.2 Multiobjective 0-1 knapsack problem in optimal UL resource allocation


In the multiobjective resource allocation, the packet sending treatment can be taken care of by each QCI in the system as
indicated by the predefined specifics of the QCI. To boost the framework throughput and fulfill the QoS constraints, the
QCI must take after a typical rate and planning approach. In the time of resource allocation, there are some interference

RB1 RB2 RB3 RB4 RB5 RB6

UE1

UE2

UE3 U3 U2 U4 U1 U4 U5

UE4
FIGURE 2 Feasible scheduling
UE5 process. RB, resource block; UE,
user equipment
6 of 14 MOSES AND KAARTHICK

that is happening between BS and eNBs because of minimum QoS, minimum resource utilization. Therefore, there is a
need to create an objective function to solve these problems. In the proposed technique, there are three objective func-
tions, for example, (1) resource utilization maximization, (2) QoS maximization, and (3) interference minimization, ie,
incorporated into the resource allocation as individually communicated by Equations (4)-(6) and these objective functions
are figured as takes after. The resource utilization is computed as follows:

⎧ [ ( n )]
⎪mink 𝜀k ∑ ρwk .qwk
⎪ w=1

Max Ruti =⎨ ∑
K
(4)
⎪subject to k=1 qwk = 𝓁w ; w = 1, … … , W

⎪ qwk ≥ 0; k = 1, 2, … , K.

In the aforementioned equation, the quantity of resource is represented as qwk ; here, the resource w allocated to the
activity k. The available number of resource is represented as lw , the effectiveness of the resource w allocated to activity k
is represented as 𝜌wk , and the return function for activity k is represented as 𝜀k . The second objective function is maximum
QoS and it is computed as follows:
⎧ ∑n
⎪min ci x
⎪ i=1
Max QoS = ⎨ (5)
subject to Bx ≤ xm ,

⎪ Beq x = 1K .

In the aforementioned notation, the allocation vector is represented as x, which consolidates all feasible solutions;
the binary inequality matrix is represented as B, which belongs to the binary set {0, 1}; the binary equality matrix is
represented as Beq ; and the cost vector is represented as c with m users. The allocation vector x = [x1 , x2 , … , xM ]T and
xm = [xm,1 , , … , xm,a ]T , where a denotes the number of possible allocation pattern. The cost vectors are extracted from the
feasible solution. Here, the cost function is computed for all feasible solutions. Among the all pattern, the best pattern is
selected based on the minimum value. The third objective function is the interference and it is computed as follows:


n
Int = min li + di . (6)
i=1

In the aforementioned equation, packet delay rate and the packet error loss rate are represented as di ,li . In the resource
allocation process, the utility and merits are evaluated for each task. Here, the allocation problem is considered as a variant
of the 0-1 MOKP to optimize the overall utility subject to multiple resource constraints. The utility of the resources is
evaluated based on the variety of workloads of each task. The knapsack problem derives its name from the thief's problem
of choosing which of a variety of treasures to steal in such a way that the value of his spoils is maximized and that he
does not exceed the weight carrying capacity of his knapsack. This problem is referred to as a 0-1 knapsack problem, as
opposed to a fractional knapsack problem because each treasure is indivisible. The multiobjective knapsack issue is simple
knapsack issue where numbers of goals are maximized simultaneously. The 0-1 MOKP can be calculated as follows:

⎧ 𝑗 ∑
n
𝑗
⎪max X (z) = i=1 ai zi 𝑗=1,. … q

MOKP ⎨subject to ∑ 𝜂 k z ≤ b
n
(7)
i i k k=1,. … ,r
⎪ i=1

⎩ zi 𝜀 {0, 1} i = 1, … s.

In the aforementioned Equation (7), the number of object is denoted as s, the decision variable is represented as zi ,
the numbers of objectives are represented as q, and the number of constraints of the problem is represented as r. In the
multiobjective function x, the jth component is represented as x j . Nowadays, the real trouble is the resource allocation
trouble and these problems can be solved by applying the multiobjective 0-1 knapsack problem. To solve the MOKP, the
MOSES AND KAARTHICK 7 of 14

researchers used several metaheuristic and cooperative swarm intelligent algorithms in the literature. Here, the proposed
MCSIA-RA method employed a hybrid FFA and PSO to solve the knapsack optimization problem.

3.3 Multiobjective cooperative swarm intelligence algorithm for resource allocation


In the proposed multiobjective cooperative swarm intelligence algorithm for resource allocation (MCSIA-RA) system, the
enhanced version of the FFA-PSO32 is introduced to solve the (0-1) knapsack problem. To find the best optimal solution,
FFA is used to search the path for optima and PSO is utilized to update FFA. Our algorithm's main aim is to diminish the
resource allocation time, the computational cost of the network.

3.3.1 Firefly algorithm


The procedure of the FFA is given as follows for the optimal allocation of UL resource in the LTE-A network.
1. The fireflies are attracted each other in light of their sex.
2. The fireflies are attracted by different fireflies in light of the splendor. At the same time, the fireflies are moved
randomly, if there is no brighter one.
3. The brightness of the firefly is selected by the landscape of the objective function. By using the FFA, the fitness function
is computed for every solution. Here, the brighter firefly is selected based on the best fitness function and this better
fitness is updated.
Step (i): Initialization
In the initialization step, the population of the fireflies is randomly initialized. In the LTE-A resource allocation, there are
m number of users and n number of RBs that are present. In the FFA, each resource is considered as the firefly. The value
of a specific user can be taken by each firefly; it shows the RBs that have been allowed to that particular specific user.
The randomly generated resources are depicted as (r1, r2, r3 … rn ). The total number of resources in the LTE-A network is
represented as R.
Step (ii): Objective function creation
The resource is allocated based on the fitness function. Here, the resource is allocated based on the resource utilization
maximization, QoS maximization, and interference minimization, ie,

⎧minimize ∶ Int

F (ri ) = ⎨maximize ∶ Ruti (8)

⎩Maximize QoS.

The distance of two fireflies ri and rj is computed by using the Euclidean distance and it is given as follows. Here, ri,k is
the k component of the spatial coordinating of the firefly ri and h is denoted by the number of dimension, ie,

√ h
√∑ ( )2
dij = √ ri,k − r𝑗,k . (9)
k=0

Step (iii): Updation


The firefly ri is attracted by another brightest firefly rj and it is determined by the following equation. The solutions are
updated using the FFA and the solutions are updated utilizing the following equation:
2 ( )
𝛽ik+1 = 𝛽ik − 𝛾0uk e−𝛽dij 𝛽𝑗k − 𝛽ik + Ωk 𝜂ik , (10)

where the new update solution is represented as 𝛽ik+1 , the present ith solution is represented as 𝛽ik , and the jth solution is
represented as 𝛽𝑗k . Here, the arbitrary parameter is represented as Ωk , at time t, the random distribution is represented as
𝜂ik and the constants are denoted as 𝛾0𝑢𝑘 and 𝛽.

3.3.2 PSO algorithm


Here, the PSO algorithm is utilized to update the objective function of the FFA. Here, the updating process is done based
on the PSO to improve the performance of the FFA. Here, the QoS and the resource utilization are maximized and the
8 of 14 MOSES AND KAARTHICK

interference is minimized. In this paper, the aforementioned objectives are satisfied and the optimal results are obtained
with the help of the PSO algorithm.
Step (iv): PSO for updating
The latest position of the ri is calculated by using the following equation:
( ) ( )
𝜒it+1 = u.𝜒it + c1 .x1 . pti − rit + c2 .x2 . gt − rit (11)

rit+1 = rit + 𝜒it+1 , (12)

where the velocity of the resources is represented as 𝜒it , the iteration is represented as t, in the iteration t, the ri th current
position is represented as rit , the best solution of the ith resource is represented as pti , the best position of the particle pop-
ulation gt , and the inertia weight of the each resource is represented as u. Here, the cognitive parameters are represented
as c1 and c2 and the random numbers between 0 and 1 are represented as x1 and x2 .
Step (v): Termination
Save the best solution achieved so far and minimize the energy consumption and total task completion time of the real
and reactive task variations. Then, save the corresponding parameters, verify the range of iteration, if the iteration has
not attained the maximum range, then the iteration count increases as k = k+1, or else terminate the process.

The optimal performance of the proposed 0-1 MOKP can be minimized by using the aforementioned FFA-PSO
algorithm. The optimal resources are allocated to the users selected in phase 3. Here, the selecting and the scheduling
process for users' selection to allocate the resources are clearly depicted in Section 3.2.

4 R E S U LT S AN D D ISCU SSION S

In this area, the optimization property of LTE-A resource sharing is exploited to address the knapsack scheduling issue
and the total performance of the network is optimized. The challenging issues in the planning procedure are deal with
the QoS and the throughput aware ranking function. In order to obtain the performance results for the proposed work, a
Vienna LTE-A simulator in MATLAB is used.33
MOSES AND KAARTHICK 9 of 14

Parameter Values TABLE 1 Simulation parameters


Carrier frequency 2.15 GHz
Subcarrier spacing 15 kHz
eNB TX power 46 dBM
UE TX power 23 dBM
Bandwidth 10 MHz
Noise figure 9 dB at eNB, 5 dB at UE
Path loss (dB) 128.1 + 37.6 log10 (d) with d in km
Path loss model COST 231
Fading model Jakes
Confidence interval 95%
Simulation time 50 ms

In the proposed MCSIA-RA method, for the optimal UL resource allocation, the efficient FFA, and the PSO algorithm
is implemented and efficient optimal selection method is also utilized to solve the 0-1 MOKP. The user ranking is the
important task before the resource allocation and this user ranking process is done based on the urgency factor and
the priority factor. These two factors are computed by the packet delay budget. Due to this efficient proposed method,
the overall network performance is also maximized. Table 1 summarized the parameters and their values are used in the
simulation.

4.1 Performance comparison


This section shows the comparison performance of throughput, peak signal-to-noise ratio (PSNR), packet loss ratio (PLR),
delay, fairness index, and execution time values, which were obtained for proposal and other methods. In our method, for
resource allocation, the FFA-PSO algorithm is utilized. In the first section, FFA-PSO is compared with other optimization
algorithms such as FF GA, FF differential evolution, and GA-PSO with different users (20, 40, 60, 80, and 100). Figure 3
shows the throughput and packet delay comparison.
Throughput values for FFA-PSO are 0.56 Mbps for 20 users, 1.47 Mbps for 40 users, 2.28 Mbps for 60 users, 3 Mbps for
80 users, and 3.78 Mbps for 100 users. For increasing number of users, throughput value also increased. Analyzing all
users, FFA-PSO is 12.08% higher than FF-GA, 12.62% higher than FF-DE, and 16.68% higher than GA-PSO. PSNR values
for FFA-PSO are 42.8 for 20 users, 42.9 for 40 users, 43.2 for 60 users, 44.223 for 80 users, and 43 for 100 users. For high
users, PSNR value is also in high level. However, comparing with other methods, FFA-PSO is 2.85% less than FF-GA,
7.08% less than FF-DE, and 7.32% less than GA-PSO.
Figure 4 depicts the comparison of loss of packet's ratio and fairness index of our model with other optimizations. PLR
for our method is 0.13 for 20 users, 0.27 for 40 users, 0.42 for 60 users, 0.61 for 80 users, and 0.77 for 100 users. The overall
ratio for FFA-PSO is 3.16% less than FF-GA, 42.23% less than FF-DE, and 31.54% less than GA-PSO. For increasing the
users, fairness index is also increased. Fairness index for FFA-PSO is 1.05 for 20 users, 0.98 for 40 users, 0.9 for 60 users,
0.82 for 80 users, and 0.75 for 100 users. For increasing number of users, fairness index value also increased. Analyzing
all users, FFA-PSO is 32.8% higher than FF-GA, 2.22% higher than FF-DE, and 17.3% higher than GA-PSO.

FIGURE 3 Comparison of throughput and peak signal-to-noise ratio. DE, differential evolution; FFA, firefly algorithm; GA, genetic
algorithm; PSO, particle swarm optimization
10 of 14 MOSES AND KAARTHICK

FIGURE 4 Comparison of packet loss ratio and fairness index

FIGURE 5 Comparison of delay and time consumption. DE, differential evolution; FFA, firefly algorithm; GA, genetic algorithm; PSO,
particle swarm optimization

Figure 5 shows the comparison of delay and time consumption. Delay for FFA-PSO are 0.25 for 20 users, 0.26 for 40
users, 0.27 for 60 users, 0.28 for 80 users, and 0.292 for 100 users. For increasing the number of user, delay of FFA-PSO
is slightly increased. Comparing to other methods, overall delay for our model is 33.21% less than FF-GA, 14.35% less
than FF-DE, and 38.65% less than GA-PSO. Time consumption for FFA-PSO is 0.04 ms for 20 users, 0.06 ms for 40 users,
0.06 ms for 60 users, 0.12 ms for 80 users, and 0.18 ms for 100 users. Total execution time for our method is 19.56% higher
than FF-GA, 46.23% higher than FF-DE, and 48.67% higher than GA-PSO.
In the second section, the system performance is analyzed with the aid of comparison of the proposed method with
existing techniques. Our model MCSIA-RA for the resource allocation is contrasted with the existing resource allo-
cation method such as ACO-RA scheme,20 PSA,21 ALS,25 three-step GA (TSGA),34 round robin (RR) approach,35 and
proportional fair (PF) resource allocation.36
Figure 6 shows the average throughput found in the experimentation. Our proposed model gives the equal opportunity
to all the users to allocate the resources. Here, a number of users exploited the different types of application and the
FFA-PSO algorithm provides the higher throughput when compared with the existing method. For 20 users, throughput
of our system is 1.9% superior to ACO-PA, 2.5% superior to PSA, 3.5% superior to ALS, 3.6% superior to TSGA, 3.84%
superior to RR, and 5.7% superior to PF. For 40 users, our system is 2.7% superior to ACO-PA, 2.73% superior to PSA, 3.5%
superior to ALS, 2.78% superior to TSGA, 3.2% superior to RR, and 3.51% superior to PF. For 60 users, our system is 2.6%
superior to ACO-PA, 2.83% superior to PSA, 3.65% superior to ALS, 4.2% superior to TSGA, 5.2% superior to RR, and 5.9%
superior to PF. For 80 users, our system is 2.23% superior to ACO-PA, 2.5% superior to PSA, 10.5% superior to ALS, 13.6%
superior to TSGA, 14.7% superior to RR, and 26% superior to PF. For 100 users, our system is 3.2% superior to ACO-PA,
3.8% superior to PSA, 3.9% superior to ALS, 14.56% superior to TSGA, 16.2% superior to RR, and 24.5% superior to PF.
The packet delay described for the resource application is shown in Figure 7. A graph is plotted for the packet delay
against the number of users and the packet delay budget. From these values, until reaching the threshold values, the
schedules were adjusted to discard the packets. When allocating the application to all users, the schedules met the delay
MOSES AND KAARTHICK 11 of 14

MCSIA-RA ACO-RA [21] PSA [22] ALS [26]


TGSA [35] RR [36] PF [37]
4
3.5
Throughput (Mbps)

3
2.5
2
1.5
1
0.5 FIGURE 6 Throughput comparison. ALS, adaptive LTE-Advanced
0
cross-layer packet scheduling; MCSIA-RA, multiobjective cooperative
20 40 60 80 100 swarm intelligence algorithm for resource allocation; PF,
Users proportional fair; PSA, packet scheduling algorithm; RR, round robin

MCSIA-RA ACO-RA [21] PSA [22] ALS [26]


TGSA [35] RR [36] PF [37]
0.5

0.4
Delay (s)

0.3

0.2

0.1 FIGURE 7 Packet delay comparison. ALS, adaptive LTE-Advanced


cross-layer packet scheduling; MCSIA-RA, multiobjective cooperative
0
20 40 60 80 100 swarm intelligence algorithm for resource allocation; PF,
Number of users proportional fair; PSA, packet scheduling algorithm; RR, round robin

and this delay is noted by using this proposed method, which computes the PLR and is necessary to analyze the packet
delay budget.
The proposed method minimized the packet delay budget efficiently when compared with existing methods. For 20
users, the delay of our system is 3.6% less than ACO-PA, 3.8% less than PSA, 4.92% less than ALS, 10.25% less than TSGA,
11.2% less than RR, and 13.2% less than PF. For 40 users, the delay of our system is 0.79% less than ACO-PA, 1.25% less
than PSA, 2.36% less than ALS, 3.25% less than TSGA, 4.5% less than RR, and 10.7% less than PF. For 60 users, the delay
of our system is 0.38% less than ACO-PA, 1.62% less than PSA, 2.92% less than ALS, 3.25% less than TSGA, 4.76% less than
RR, and 12.3% less than PF. For 80 users, the delay of our system is 1.4% less than ACO-PA, 2.8% less than PSA, 3.92% less
than ALS, 5.25% less than TSGA, 9.7% less than RR, and 13.5% less than PF. For 100 users, the delay of our system is 6.6%
less than ACO-PA, 8.8% less than PSA, 9.92% less than ALS, 13.25% less than TSGA, 18.32% less than RR, and 22.35% less
than PF.
The PLR of the proposed system is shown in Figure 8. The PLR is computed based on the number of packets loss at
which the time the packet delay reaches the budget. Here, PLR is plotted against the number of users and the number
of packet loss. Here, the number of users varied from 20, 40, 60, 80, and 100, respectively. Overall, PLR of our method is
3.96% less than ACO-PA, 5.98% less than PSA, 6.78% less than ALS, 8.92% less than TGSA, 13.1% less than RR, and 17.13%
less than PF.
The notable fairness index appears in Figure 9. In the graph, the fairness index further plotted against the number
of users and the class fairness. Here, there are two sorts of fairness index that are incorporated; they are interclass and
intraclass fairness index. The objective of the interclass fairness index is to gauge for what number of clients the network
could guarantee the required bit rate. To process the normal throughput fairness index, the fairness between the various
QoS classes is essential.
Here, the proposed MCSIA-RA technique, for example, FFA-PSO, gives the better fairness record when contrasted and
the current strategies are used for the resource allocation. Here, the effect of non-GBR application is lessened by utilizing
this algorithm. Total fairness index of our method is 12% superior to ACO-PA, 12.68% superior to PSA, 13.2% superior to
ALS, 14.3% superior to TGSA, 18.57% superior to RR, and 22.38% superior to PF.
12 of 14 MOSES AND KAARTHICK

MCSIA-RA ACO-RA [21] PSA [22] ALS [26]


TGSA [35] RR [36] PF [37]
1.2

Packet loss Ratio


0.8

0.6

0.4
FIGURE 8 Packet loss ratio comparison. ALS, adaptive
LTE-Advanced cross-layer packet scheduling; MCSIA-RA, 0.2
multiobjective cooperative swarm intelligence algorithm for resource 0
allocation; PF, proportional fair; PSA, packet scheduling algorithm; 20 40 60 80 100
RR, round robin Number of users

MCSIA-RA ACO-RA [21] PSA [22] ALS [26]


TGSA [35] RR [36] PF [37]
1.2
1

Fairness index
0.8
0.6

FIGURE 9 Fairness index comparison. ALS, adaptive 0.4


LTE-Advanced cross-layer packet scheduling; MCSIA-RA, 0.2
multiobjective cooperative swarm intelligence algorithm for resource
0
allocation; PF, proportional fair; PSA, packet scheduling algorithm; 20 40 60 80 100
RR, round robin Number of users

MCSIA-RA ACO-RA [21] PSA [22] ALS [26]


TGSA [35] RR [36] PF [37]
50

40
PSNR (db)

30

20
FIGURE 10 Peak signal-to-noise ratio comparison. ALS, adaptive
LTE-Advanced cross-layer packet scheduling; MCSIA-RA, 10
multiobjective cooperative swarm intelligence algorithm for
0
resource allocation; PF, proportional fair; PSA, packet scheduling 20 40 60 80 100
algorithm; RR, round robin Users

The proposed method of PSNR is shown in Figure 10. To calculate the network quality, the PSNR metric is the most
pivotal objective function. Here, the PSNR is almost constantly protected by the proposed MCSIA-RA when the number
of users increased. Compared with the other method, the proposed method has better PSNR ratio. In the existing TSGA
method, the PSNR range is good but does not clearly provide the better quality packet delivery. In PSNR, MCSIA-RA is
3.3% superior to ACO-PA, 3.65% superior to PSA, 4.6% superior to ALS, 6.9% superior to TGSA, 7.23% superior to RR, and
13.8% superior to PF.
The time consumption of our method is shown in Figure 11. From the graph, we know that our model has less con-
sumption time than existing systems. In resource allocation, the TSGA based method performs very well; however, the GA
requires more consumption time for the resource allocation because it requires so many parameters for the optimization.
The average waiting time is very high; this is the main drawback in the RR scheduling algorithm. The PF share method
increases the system performance; however, the response time is very high. Time consumption is 15% less than the
ACO-PA, 19% less than PSA, 23% less than ALS, 29.1% less than TGSA, 32.7% less than RR, and 46.5% less than PF.
MOSES AND KAARTHICK 13 of 14

MCSIA-RA ACO-RA [21] PSA [22] ALS [26]


TGSA [35] RR [36] PF [37]
0.4
Time Consumption (ms)

0.4

0.2
FIGURE 11 Time consumption comparison. ALS, adaptive
0.1
LTE-Advanced cross-layer packet scheduling; MCSIA-RA,
multiobjective cooperative swarm intelligence algorithm for resource
0
20 40 60 80 100 allocation; PF, proportional fair; PSA, packet scheduling algorithm;
Users RR, round robin

5 CO N C LUSION S

Here, we inspected MCSIA-RA to solve the LTE resource allocation issue in LTE-A network. Initially, based on the ser-
vice requirement and urgency factor, the candidate users ought to be chosen in the scheduling process. Then, in LTE-A
UL resource allocation, we showed the applicability of FFA-PSO to tackle the optimization issue. Finally, the proposed
approach gave the finest solution based on the fractional knapsack optimization issue. For various classes of QoS with
separable constraints, while lessening the delay and loss, flexible resource allocation system is acquired by the proposed
algorithm and this result is shown in the simulation process. Additionally, the proposed optimization algorithms are
compared with the existing resource optimization schemes to compute the throughput and the efficiency of the LTE-A
network. The comparison shows that the proposed strategy has better QoS and less interference in the resource allocation
in LTE-A network than state-of-art methods. The future work of our proposed method is to reduce the complexity of the
method and to hybrid another optimization method also and, in simulation result, the throughput and efficiency will be
improved more in the future.

ORCID
M. Leeban Moses https://orcid.org/0000-0002-8946-2863

REFERENCES
1. Hasan M, Hossain E, Kim D. Resource allocation under channel uncertainties for relay-aided device-to-device communication underlay-
ing LTE-A cellular networks. IEEE Trans Wirel Commun. 2014;13(4):2322-2338.
2. Xiang X, Lin C, Chen X, Shen X. Toward optimal admission control and resource allocation for LTE-A femtocell uplink. IEEE Trans Veh
Technol. 2015;64(7):3247-3261.
3. Gu J, Yoon H, Lee J, Bae S, Chung M. A resource allocation scheme for device-to-device communications using LTE-A uplink resources.
Pervasive Mob Comput. 2015;18:104-117.
4. Zhou Z, Dong M, Ota K, Wang G, Yang L. Energy-efficient resource allocation for D2D communications underlaying cloud-RAN-based
LTE-A networks. IEEE Internet Things J. 2016;3(3):428-438.
5. Pateromichelakis E, Shariat M, ul Quddus A, Tafazolli R. On the evolution of multi-cell scheduling in 3GPP LTE/LTE-A. IEEE Commun
Surv Tutor. 2013;15(2):701-717.
6. Sheng Y, Peng M, Wang W. QoS-aware resource allocation algorithm for broadcast/multicast OFDM systems. J Electron Inf Technol.
2011;30:2943-2946.
7. Liao H, Chen P, Chen W. An efficient downlink radio resource allocation with carrier aggregation in LTE-Advanced networks. IEEE Trans
Mob Comput. 2014;13(10):2229-2239.
8. Ali M, Hossain E, Kim D. LTE/LTE-A random access for massive machine-type communications in smart cities. IEEE Commun Mag.
2017;55(1):76-83.
9. de Andrade T, Astudillo C, da Fonseca N. Allocation of control resources for machine-to-machine and human-to-human communications
over LTE/LTE-A networks. IEEE Internet Things J. 2016;3(3):366-377.
10. Mehaseb M, Gadallah Y, Elhamy A, Elhennawy H. Classification of LTE uplink scheduling techniques: an M2M perspective. IEEE
Commun Surv Tutor. 2016;18(2):1310-1335.
11. Misic J, Misic V, Khan N. Sharing it my way: efficient M2M access in LTE/LTE-A networks. IEEE Trans Veh Technol. 2016;66(1):696-709.
12. Misic J, Misic V. Adapting LTE/LTE-A to M2M and D2D communications. IEEE Network. 2017;31(3):63-69.
13. Nardini G, Stea G, Virdis A, Sabella D, Caretti M. Resource allocation for network-controlled device-to-device communications in
LTE-Advanced. Wireless Networks. 2016;23(3):787-804.
14 of 14 MOSES AND KAARTHICK

14. Li N, Cao C, Wang C. Dynamic resource allocation and access class barring scheme for delay-sensitive devices in machine to machine
(M2M) communications. Sensors. 2017;17(6):1407.
15. Zhang X, Wang Y, Wang R, Hou Y, Wang D. Group-based joint signaling and data resource allocation in MTC-underlaid cellular networks.
Sci China Inf Sci. 2017;60(10).
16. Morvari F, Ghasemi A. Two-stage resource allocation for random access M2M communications in LTE network. IEEE Commun Lett.
2016;20(5):982-985.
17. Hasan M, Hossain E, Niyato D. Random access for machine-to-machine communication in LTE-Advanced networks: issues and
approaches. IEEE Commun Mag. 2013;51(6):86-93.
18. Condoluci M, Araniti G, Dohler M, Iera A, Molinaro A. Virtual code resource allocation for energy-aware MTC access over 5G systems.
Ad Hoc Netw. 2016;43:3-15.
19. Zheng K, Ou S, Alonso-Zarate J, Dohler M, Liu F, Zhu H. Challenges of massive access in highly dense LTE-Advanced networks with
machine-to-machine communications. IEEE Wirel Commun. 2014;21(3):12-18.
20. Lee Y, Loo J, Chuah T, El-Saleh A. Multi-objective resource allocation for LTE/LTE-A femtocell/HeNB networks using ant colony
optimization. Wirel Pers Commun. 2016;92(2):565-586.
21. Shukla S, Bhatia V. Packet scheduling algorithm in LTE/LTE-Advanced-based cellular networks. IETE Tech Rev. 2017;35(6):551-561.
https://doi.org/10.1080/02564602.2017.1342573
22. Alavikia Z, Ghasemi A. Overload control in the network domain of LTE/LTE-A based machine type communications. Wireless Networks.
2016;24(1):1-16.
23. Ragaleux A, Baey S, Karaca M. Standard-compliant LTE-A uplink scheduling scheme with quality of service. IEEE Trans Veh Technol.
2017;66(8):7207-7222.
24. BenMimoune A, Khasawneh F, Rong B, Kadoch M. Dynamic joint resource allocation and relay selection for 5G multi-hop relay systems.
Telecommunication Systems. 2017;66(2):283-294.
25. Chang B, Liang Y, Chang P. Adaptive cross-layer-based packet scheduling and dynamic resource allocation for LTE-Advanced relaying
cellular communications. Wirel Pers Commun. 2017;96(1):939-960.
26. Ghavimi F, Lu Y, Chen H. Uplink scheduling and power allocation for M2M communications in SC-FDMA-based LTE-A networks with
QoS guarantees. IEEE Trans Veh Technol. 2017;66(7):6160-6170.
27. Sun S, Kim K, Shin O, Shin Y. Device-to-device resource allocation in LTE-Advanced networks by hybrid particle swarm optimization and
genetic algorithm. Peer-to-Peer Netw Appl. 2015;9(5):945-954.
28. Xu L, Zhou X, Li Q, Zhang X. Energy-efficient resource allocation for multiuser OFDMA system based on hybrid genetic simulated
annealing. Soft Computing. 2016;21(14):3969-3976.
29. Musumeci F, De Silva E, Tornatore M. Enhancing RAN throughput by optimized CoMP controller placement in optical metro networks.
IEEE J Sel Areas Commun. 2018;36(11):2561-2569.
30. Marotta A, Kondepu K, Giannone F, et al. Impact of CoMP VNF placement on 5G coordinated scheduling performance. Paper presented
at: European Conference on Networks and Communications; 2017; Oulu, Finland.
31. Marotta A, Kondepu K, Giannone F, et al. Performance evaluation of CoMP coordinated scheduling over different backhaul infrastruc-
tures: a real use case scenario. Paper presented at: IEEE International Conference on the Science of Electrical Engineering (ICSEE); 2016;
Eilat, Israel.
32. Zouache D, Moussaoui A, Ben Abdelaziz F. A cooperative swarm intelligence algorithm for multi-objective discrete optimization with
application to the knapsack problem. Eur J Oper Res. 2018;264(1):74-88.
33. Vienna LTE-A Simulators. http://www.nt.tuwien.ac.at/ltesimulator/
34. da Mata S, Guardieiro P. Resource allocation for the LTE uplink based on genetic algorithms in mixed traffic environments. Computer
Communications. 2017;107:125-137.
35. Aijaz A, Tshangini M, Nakhai M, Chu X, Aghvami A. Energy-efficient uplink resource allocation in LTE networks with M2M/H2H
co-existence under statistical QoS guarantees. IEEE Trans Commun. 2014;62(7):2353-2365.
36. Mosleh S, Liu L, Zhang J. Proportional-fair resource allocation for coordinated multi-point transmission in LTE-Advanced. IEEE Trans
Wirel Commun. 2016;15(8):5355-5367.

How to cite this article: Moses ML, Kaarthick B. Multiobjective cooperative swarm intelligence
algorithm for uplink resource allocation in LTE-A networks. Trans Emerging Tel Tech. 2019;e3748.
https://doi.org/10.1002/ett.3748

You might also like