Applied Intelligence

https://doi.org/10.1007/s10489-020-01834-w

hPSO-SA: hybrid particle swarm optimization-simulated annealing

algorithm for relay node selection in wireless body area networks
Naveen Bilandi 1,2 & Harsh K. Verma 1 & Renu Dhir 1

# Springer Science+Business Media, LLC, part of Springer Nature 2020

Abstract
In the modern world, wireless body area networks (WBANs) play an essential role in psychological and biomedical applications.
The use of WBANs in medical applications is limited due to various issues related to the sensors, viz., irregularity in data
production, replacement and recharging of their batteries and the energy consumed by the networks. This manuscript addresses
how these problems can be solved along with optimization of the energy consumption through efficient design of the system by
applying routing protocols and heuristic-based optimization algorithms. In this paper, the particle swarm optimization (PSO)
algorithm is a heuristic search algorithm that relies on an upgrade mechanism of the velocity and position of swarms. Although
PSO has excellent exploration capability in global search, it becomes quickly stuck in local minima. To enhance the local search
function of the current PSO algorithm, a simulated annealing (SA) algorithm has been incorporated in the exploitation phase. The
newly developed hybrid PSO-SA (hPSO-SA) algorithm is validated with other state-of-the-art nature-inspired algorithms on
eighteen benchmarks and five real engineering design problems. The statistical results of the proposed hPSO-SA algorithm are
promising and indicate very good efficiency. The paper also aims at the application of the proposed algorithm to the WBAN
design problem for minimization of the energy consumption through better selection of the relay node. The proposed hPSO-SA
algorithm outperforms twelve other metaheuristic algorithms, taking hybrid variants for comparison.

Keywords Wireless body area networks . Particle swarm optimization . Simulated annealing . Energy efficiency . Relay sensor
node

1 Introduction The early diagnosis of diseases based on the initial symp-

toms, may prevent many of the fatal effects of diseases. A
Over the years, the manifold growth in industrialization and WBAN can be an effective technology for detecting the
urbanization has led to the degradation of natural resources onset of such disorders at an early stage and has enhanced
(air, water, land) and has resulted in life-threatening diseases, the capabilities of modern-day medical facilities. A WBAN
such as cardiovascular diseases, and various other chronic allows placement of minuscule biosensors on or in the pa-
fatal diseases, and in millions of people losing their lives. tient’s body to observe important bodily signals. With the
mounting need for pervasive data transmission and present
developments in lower energy consumption wireless tech-
nology, there is much interest in the medical prognosis field
* Naveen Bilandi for developing wireless systems for patients. The motiva-
naveenbilandi77@gmail.com
tion behind using a WBAN is to improve the correctness
and effectiveness of sensors in detecting the initial stages
Harsh K. Verma
vermah@nitj.ac.in of many diseases. A WBAN makes use of sensors in, on, or
around the body, which are designed for specific applica-
Renu Dhir
dhirr@nitj.ac.in
tions, to measure the biological conditions of the human
body. For continuous monitoring of the human body, a
1
Department of Computer Science and Engineering, National Institute WBAN makes use of sensor devices that continuously send
of Technology, Jalandhar, India health-related information to personal digital assistants
2
Department of Computer Science and Engineering, DAV University, (PDAs), which leads to loss of energy of the sensors. To
Jalandhar, India address this issue, various existing studies have suggested
 N. Bilandi et al.

the use of a relay node. Relay node selection can be consid- problems. Dhadwal et al. [10] utilized a hybrid genetic
ered an optimization problem. In this regard, this paper pre- algorithm (GA) technique combined with PSO to increase
sents a hybrid PSO-SA approach for the relay node selection the search proficiency and for selection of the finest route,
problem to optimize energy utilization in WBANs. among others. Palvinder et al. [26] proposed an energy-
This subsequently results in effective use of the energy of efficient bee clustering protocol using artificial bee colony
these sensors [1, 2]. Researchers worldwide have contributed to detect optimal CHs and enhance the energy efficiency
greatly to improving routing algorithms and enhancing the for WSNs. Nitin et al. [27] proposed a moth flame
lifespan of WBANs. optimization-powered threshold-sensitive energy-efficient
In WBANs, routing techniques can be categorized from clustering protocol (TECP) for enhancing the network sta-
an energy perspective as mobility-, thermal-, link-, or bility period.
distance-aware techniques and many more. Energy- This is the motivation behind this study. The PSO algo-
efficient routing protocols, such as those in topology and rithm has been modified by incorporating a simulated anneal-
multihop-based WBANs [3–5], medium-access-control- ing algorithm (hPSO-SA) to avoid being trapped in local min-
based WBANs [6, 7], and priority-based WBANs [8], ima, and this algorithm has been applied for constrained en-
and various optimization algorithms have also been ana- gineering optimization problems and validated with other
lyzed [9–14]. However, these algorithms do not solve the state-of-the-art nature-inspired algorithms.
problems associated with WBANs since a single algo- As per the no free lunch (NFL) [28] theorem, no single
rithm cannot be effective for all problems. algorithm is suitable for solving every kind of optimization
Recently, various nature-inspired metaheuristic approaches problem, so there is always room for improvement. In this
have been proposed, such as the coyote optimization algo- context, this paper presents a hybrid variant of PSO-SA,
rithm [15], black widow optimization algorithm [16], new where SA is used as a local search mechanism to enhance
ant colony optimization algorithm [17], hybrid PSO-GA the exploitation capability of the PSO algorithm. The per-
[18], and Harris hawks optimization [19]. formance of the proposed algorithm is compared with that
Some of the work on solving relay node selection prob- of various metaheuristic algorithms, viz., PSO [29], GSA
lems using metaheuristics can be summarized as follows: [30], FEP [31], ALO [32], SMS [33], BA [34], FPA [35],
PSO [20], genetic ant colony [11], artificial bee colony CS [36], FA [37], GA [38], MFO [39] and MVO [40], on
[21], hybrid PSO-GA [22], PSOBAN [23], cuckoo search various benchmark functions, on engineering design prob-
[24], the opportunistic energy-efficient routing with load lems and in solving the relay node selection problem in
balancing (OE2-LB) algorithm [25] and the optimized WBANs. The scenario of the proposed research is shown
cost-effective and energy-efficient routing protocol in Fig. 1. The techniques are summarized in Table 1 in
(OCER) [14] have been explored to solve WBAN-based terms of their advantages and disadvantages.

Fig. 1 Application of the proposed algorithm for a health monitoring system in a WBAN
Table 1 Summary of various techniques

Authors Year Algorithm used Advantages Disadvantages Ref.

D’Andreagiovanni 2015 Ant colony optimization and approximate Robust model for finding solutions of better quality Lower computational efficiency [12]
nondeterministic tree search
Mirjalili S 2015 Ant lion optimization Solves a real world problem with an unknown search space Inefficient compared to Levy flight [32]
Yang X-S 2010 Bat algorithm Utilizes the best combination with the enhanced feature of Low convergence rate [34]
echolocation
Hayyolalam V 2020 Black widow optimization Fast convergence in obtaining the global optimum solution Takes a long processing time [16]
for nonlinear optimization problems
Qais MH 2019 Coyote optimization algorithm Minimizes the fitness function and root mean squared Shortage of photovoltaic manufacturer data [15]
current error function
Samal 2019 Cuckoo search-based algorithm Proposed fitness function for placement of relay nodes Lack of a synchronization mechanism [24]
Yang X 2012 Firefly algorithm Can easily balance exploration and exploitation Higher computation time [36]
X-s. Y 2012 Flower pollination algorithm Provides better solutions in image compression, graph Unable to solve combinatorial optimization problem [35]
coloring and multiobjective optimization and extensive simulations
Kaur N 2017 Genetic algorithm Lower energy consumption for WBANs Considers less complex network scenarios [14]
Rashedi E 2009 Gravitational search algorithm Enhancement in the solution compared to the central force Lack of computational efficiency for high dimensional [30]
optimization algorithm functions
Mirjalili 2014 Gray wolf optimization Better exploitation capability Higher computation time [41]
Asghar A 2019 Harris hawks optimization Higher potential for finding the optimal solution Premature convergence in solving different problems [19]
Garg H 2016 Hybrid PSO-GA for constrained optimization Optimal solution for constrained nonlinear optimization Not validated on benchmark functions [18]
problems problems
Singh P 2019 Improved artificial bee colony Better selection of cluster head and improved energy Computational time is longer [26]
efficiency
Yao X 1999 Improved fast evolutionary programming using Shows for the first time the relationship between the Does not provide the solution of multimodal [31]
Cauchy mutation distance to the global optimum and the search step size optimization problems
Xin-She Yang 2010 Levy flight firefly algorithm More efficient in determining the global optimum solution Low convergence [37]
Mittal N 2018 Moth flame optimization Acquires the optimal solution for balancing the load Low data transmission rate [27]
Raj 2020 Opportunistic energy-efficient routing with load Improved performance compared to other existing Only assumes a mesh topology [25]
hPSO-SA: hybrid particle swarm optimization-simulated annealing algorithm for relay node selection in...

balancing algorithm protocols

Wu T 2015 PSO Optimum relay placement position for delivery of Low packet transmission rate [20]
information
Kennedy J 1995 PSO Efficient in solving various engineering design problems Less exploitation capability [29]
Jamili 2011 PSO-SA Better for periodic job shop scheduling problem Not validated on benchmark functions and constrained [42]
problems
Wang 2004 PSO-SA Higher potential for finding the optimal solution Not validated on constrained and WABN problems [43]
Bilandi N 2019 PSOBAN Better selection of relay node for WBAN problems Not validated on benchmark functions and constrained [23]
problems
Agnihotri 2019 PSO-GA Routing scheme raises lifetime by 12% to 23% compared Energy harvesting not done properly [22]
to the shortest path approach
Cuevas 2013 States of matter search Improves the balance between exploration and exploitation Higher computation time [33]
 N. Bilandi et al.

Fig. 2 Pseudocode for the hybrid PSO-SA algorithm

2 Materials and methods Step 1: Population generation and iteration

2.1 Particle swarm optimization The population is randomly generated using Eq. (1), and
the initial velocity is assigned through Eq. (2). Based on the
PSO, inspired by birds’ social behavior in searching for food velocity, the next iterations of the particles are decided.
and sharing their findings with other birds, was created by J.
Kennedy and R. Eberhart [29, 44]. The birds use their own Pos0i ¼ LB þ randðÞ*½UB−LB ð1Þ
best experience (Pbest) and the swarm’s best experience (Gbest) vel 0i ¼ zerosðNP; N Þ ð2Þ
for adjustment of their routes to obtain food. The mathemati-
cal formulation of the PSO algorithm is given below: where
hPSO-SA: hybrid particle swarm optimization-simulated annealing algorithm for relay node selection in...

Fig. 3 Proposed model for the WBAN problem

Pos0i Initial allocation or position of particle i    

vel 0i Initial velocity of particle i Velikþ1 ¼ wVeli;k þ c1 r1 PosLocal
i;k −Posi;k
Current
þ c2 r2 PosGlobal
g;k −PosCurrent
i;k

UB Upper bound for the examined variables ð4Þ

LB Lower bound for the examined variables
i;kþ1 ¼ Posi;k
PosCurrent þ wVeli;k ð5Þ
Current
rand() Random number within (0,1)
NP Size of the population (the population size of
where
particles is chosen as 100)
N Total number of dimensions in the search space Veli, k Velocity of the i-th particle at the k-th iteration
* The number of iterations taken is 100. PosCurrent
i;k Current position of the i-th particle at the k-th
iteration
PosLocal
i;k Local best position of the i-th particle at the k-
(Pbest) th iteration
Step 2: Fitness function PosGlobal Global best position of the i-th particle at the k-
g;k
(Gbest) th iteration
The fitness function for every particle is calculated using Eq. c1, c2 Acceleration coefficients for the PSO
(3): algorithm
f fit : Entrans ¼ EnT  p þ Enamp  n  p  d 2 ð3Þ r1,r2 Uniformly distributed random numbers within
elec
(0,1)


where ffit is the fitness function, f fit Locki at the kth iteration.
To achieve better results, c1 and c2 [20] have been set to
the value of 2.0. We used a value of 1.0 for c1 and c2 and
Step 3: Searching
achieved better results. The parameter ‘w’ is the inertia weight
and can be calculated using Eq. (6):
A new position is searched for, and the best ffit for the


position is found; subsequently, the particles store the new weight max −weight min iter
position. weight ¼ weight −max
ð6Þ
itermax

Step 4: Position-velocity updated where weightmin and weightmax are the lowest and highest
values of the inertia weights. For the proposed hPSO-SA,
Each particle updates both the velocity and position of Pbest the lowest and highest inertia weights weightmin and weight-
max
and Gbest using Eqs. (4) and (5): have been taken as 0.2 and 0.9, respectively [20].
 N. Bilandi et al.

Table 2 Deployment of sensor nodes on the human body

Sensor node S1 S2 S3 S4 S5 S6 S7 S8 SINK

Location (X, Y) (0.55, 1) (0.25, 1) (0.28, 0.2) (0.48, 0.25) (0.3, 0.5) (0.5, 0.5) (0.45, 0.13) (0.35, 0.9) (0.4, 1.1)

consequently, the local best position (PosLocal i;k Þ is updated if

Step 5: Local and global best updated
and only if ffitk + 1 is superior to ffitk using Eq. (7) and formu-
First, a comparison between the local best ffitk (current lating it using Eq. (7) for global best.
iteration) and f f i t k + 1 (next iteration) is made, and

Fig. 4 Flowchart of proposed

hybrid PSO-SA for WBAN
problem
hPSO-SA: hybrid particle swarm optimization-simulated annealing algorithm for relay node selection in...

8     Table 3 Parameter settings

< PosCurrent
i;kþ1 ; if f fit Posi;kþ1
Current
≤ f fit PosLocal
i;k
PosLocal ¼     ð7Þ Parameter Description of parameter Value
i:kþ1
: PosLocal ; if f PosCurrent > f Pos Local
i;k fit i;kþ1 fit i;k
weightmin Minimum value of the inertia weight 0.2
In the global best update, PosGlobal
g;kþ1 is calculated by weightmax Maximum value of the inertia weight 0.9
updating the variables in Eq. (7) as PosCurrent
i;k to PosLocal
i;k and i Number of particles 30
PosLocal
i;k to PosGlobal
k . NP Number of iterations 100
where rand() Random number (0,1)
c1, c2 Acceleration coefficients 1.0
PosLocal
i;k Local-best position of the ith particle at the T Temperature 0.93
kth iteration
PosGlobal
k Global-best position at the kth iteration
PosGlobal
g;kþ1 Global-best position at the (k+1)th iteration
PosLocal
i:kþ1 Local-best position of the ith particle at the The hybridization of PSO and SA has been performed by
(k+1)th iteration adopting a heuristic procedure, and the simulated annealing
algorithm is operated sequentially after PSO [46]. PSO is a
Figure 14 in Appendix presents the pseudocode for the
population-based algorithm, and SA belongs to the category
PSO algorithm.
of single solution-based algorithms. PSO after execution gen-
erates a single best solution, and SA takes that single solution
as input and exploits the search space around the solution to
2.2 Simulated annealing (SA) algorithm
find the best solution.
The best tuning of the proposed algorithm is performed by
Simulated annealing is a metaheuristic search algorithm that
selecting the initial temperature designated in Eq. (9).
works based on the concept of the hill-climbing principle. The
problem of optimal neighborhood shrinkage is overshadowed T initial ¼ 2* jN j ð9Þ
by the particular probability of recognizing a worse structure
[45]. where Tinitial is Initial temperature of the proposed algorithm,
The initial solutions are generated, which are random so- |N| is the number of attributes.
lution vectors, and a new neighbor solution is created by The cooling schedule for SA is calculated using Eq. (10), as
taking into account the predefined neighborhood structure adopted in [46].
evaluated by the objective function [46]. The move is always
T ¼ 0:93*T initial ð10Þ
accepted if the new neighbor result is better than the original
result. In contrast, a worse neighbor solution is adopted with 2.4 Wireless body area network problem formulation
a probability determined using the Boltzmann probability,
Eq. (8): In a WBAN, the main concerns are energy conservation
and network stability enhancement because the greater
−θ
P ¼ eT ð8Þ the amount of energy of the WBAN, the more stable the

where θ is the fitness difference of the generated neighbor Table 4 Simulation parameters for the WBAN problem
result and best result, and T is the temperature. SA systemat-
Symbol Parameter Description of parameter Value Unit
ically reduces the temperature as the algorithm proceeds,
which reduces the chances of the algorithm converging to a Entrans Energy Initial energy 0.6 J
local minimum. Figure 15 in Appendix presents the EnT _ elec for transmitter 16.7 nJ/bit
pseudocode for the SA algorithm. EnR _ elec for receiver 36.1 nJ/bit
Enamp for amplifier circuit 1.97 nJ/bit
S Node Number of sensor nodes 8 nodes
2.3 Proposed hybrid PSO and SA Number of sink nodes 1 nodes
I Current DC current for transmitter 10.5 mA
Hybridization of the PSO and SA algorithms has been carried DC current for receiver 18 mA
!
out in the current study, in which the position vector X can be V Voltage Supply voltage minimum 1.9 V
obtained using global best function and can be further updated f Frequency Total frequency 2.4 GHz
using the simulated annealing algorithm. The modified posi- p Packet size Total packet size 1024 bytes
tion vector is used to calculate the global fitness value, as r Rounds Total number of rounds 8000 rounds
displayed in Fig. 2.
 N. Bilandi et al.

Table 5 Results of hPSO-SA for

unimodal BFs Benchmark Function Parameters

Mean SD Worst Best Wilcoxon

BF1 1.98607E-18 1.08E-21 7.6218E-07 1.09655E-11 7.55693E-10

BF2 0.00017926 0.000245974 0.001402925 2.83301E-06 7.55693E-10
BF3 1.08E-11 1.34E-11 67.32451462 4.514468049 7.55693E-10
BF4 1.06E-12 1.89E-10 1.075852668 0.329840141 7.55693E-10
BF5 63.51699669 0.047119235 379.9994511 8.738392644 7.55693E-10
BF6 3.06E-22 1.19E-19 1.0984E-06 3.28532E-11 7.55693E-10
BF7 0.00025314 0.00029843 1.0932E-07 1.2011E-15 7.55693E-10

Table 6 Convergence curves and trial solutions of hPSO-SA for unimodal BFs

Unimodal BF
BF
Curve for PSO vs. PSO-SA Variation of fitness during trial runs

BF1

BF2

BF3
hPSO-SA: hybrid particle swarm optimization-simulated annealing algorithm for relay node selection in...

Table 6 (continued)

BF4

BF5

BF6

BF7

network will be. In our proposed work, the main focus is to Additionally, two scenarios are considered, namely, S1
find a route for data transmission that incurs the least en- (all sensors are static, and data are sent directly to the sink)
ergy loss, i.e., in simple words, an optimal and energy- and S2 (all sensors are static, and data are sent to the sink
efficient route. For this purpose, an objective function is through the relay sensor node). Therefore, we propose an
designed to find the optimal route for data transmission, in hPSO-SA algorithm whose main target is to minimize the
which the sensor relay node is chosen. energy consumption within our WBAN. hPSO-SA was
 N. Bilandi et al.

Table 7 Comparison of unimodal BF results

Algorithm Parameters Unimodal BF

BF1 BF2 BF3 BF4 BF5 BF6 BF7

PSO [29] Mean 0.000136 0.04214 70.1256 1.08648 96.7183 0.00010 0.12285
SD 0.00020 0.04542 22.1192 0.31703 60.1155 8.28E-05 0.04495
GSA [30] Mean 2.53E-16 0.05565 896.534 7.3548 67.5430 2.50E-16 0.08944
SD 9.67E-17 0.19407 318.955 1.74145 62.2253 1.74E-16 0.0433
FEP [31] Mean 0.00057 0.0081 0.016 0.3 5.06 0 0.1415
SD 0.00013 0.00077 0.014 0.5 5.87 0 0.3522
ALO [32] Mean 2.59E-10 1.84E-06 6.07E-10 1.36E-08 0.3467724 2.56E-10 0.0042924
SD 1.65E-10 6.58E-07 6.34E-10 1.81E-09 0.109584 1.09E-10 0.005089
SMS [33] Mean 0.056987 0.00684 0.95986 0.27659 0.08534 0.12532 0.00030
SD 0.01468 0.00157 0.8234 0.00573 0.14014 0.08499 0.00025
BA [34] Mean 0.773622 0.334583 0.115303 0.192185 0.334077 0.778849 0.137483
SD 0.528134 3.816022 0.766036 0.890266 0.300037 0.67392 0.112671
FPA [35] Mean 1.06E-07 0.00062 5.67E-08 0.003837 0.7810 1.08E-07 0.0031052
SD 1.27E-07 0.000176 3.90E-08 0.002186 0.366891 1.25E-07 0.001367
CS [47] Mean 0.00650 0.2120 0.2470 1.12E-05 0.00719 5.95E-05 0.00132
SD 0.00020 0.03980 0.0210 8.25E-06 0.00722 1.08E-06 0.00072
FA [48] Mean 0.03961 0.05034 0.04927 0.14551 2.17589 0.0587 0.00085
SD 0.0144 0.01234 0.01940 0.03117 1.44725 0.01447 0.00050
GA [49] Mean 0.11884 0.14522 0.1390 0.15795 0.71415 0.16791 0.01007
SD 0.12560 0.05322 0.12116 0.86202 0.97271 0.86863 0.00326
MFO [39] Mean 0.000117 0.000639 696.7309 70.68646 139.1487 0.000113 0.091155
SD 0.00015 0.000877 188.5279 5.275051 120.2607 9.87E-05 0.04642
MVO [40] Mean 2.08583 15.92479 453.2002 3.12301 1272.13 2.29495 0.05199
SD 0.64865 44.7459 177.0973 1.58291 1479.477 0.63081 0.02961
hPSO-SA Mean 1.99E-18 0.000179 1.08E-11 1.06E-12 63.517 3.06E-22 0.00025314
SD 1.08E-21 0.000246 1.34E-11 1.89E-10 63.47763 1.19E-19 0.00029843

selected because it combines the exploration capabilities of Scenario 1: All the sensor nodes are static, and data
PSO and the entrapment avoidance capability of the SA are directly sent to the sink, called hPSOSA-S1.
algorithm. Therefore, the combined strength of both algo-
rithms helps the hybrid approach detect the most optimal Position 1: In this case, eight sensor nodes (SN) are deployed
route. The following are the steps in our approach for de- on a person’s body at different positions, and all of them have
termining the optimal route. similar energy consumption potential. The sink sensor is station-
ary in the middle of the subject’s body. Various tasks (disease)
I Discover the best possible solution for the proposed net- are assigned to each sensor node for the scheme as follows:
work scenarios.
J Analyze the problem effectively. & SN1 for blood pressure (BP) monitoring
K Evaluate the effect of network parameters. & SN2 for body temperature monitoring
L Find solutions of our problem. & SN3 for electroencephalography (EEG)
& SN4 for lactic acid monitoring
& SN5 for electromyography (EMG)
2.4.1 Network model for the WBAN problem & SN6 for blood oxygen monitoring
& SN7 for electrocardiography (ECG)
The WBAN problem includes sensor nodes (S1 to S8) and & SN8 for glucose monitoring
one sink sensor node, as displayed in Fig. 3. After the data
are sensed by sensor nodes on the patient’s body, they can Position 2: This scheme has a WBAN topology in which
be transmitted to the sink to find the optimal solution. To all the sensor nodes are directly connected to the sink.
address this issue, collation of the simulation outcomes of However, the distance between sensor node S1 and the sink
hPSO-SA along with those of PSO obtained using simula- is increased due to the position change, which causes a higher
tion software for two different scenarios (hPSOSA-S1 and energy demand for data transmission.
hPSOSA-S2) is performed to check the efficiency of hPSO-
SA in terms of energy consumption. The results of these Scenario 2: All sensor nodes are static, and data are sent
two scenarios are compared. to the sink by the relay sensor, called hPSOSA-S2.
hPSO-SA: hybrid particle swarm optimization-simulated annealing algorithm for relay node selection in...

Table 8 Results of the hPSO-SA algorithm for multimodal BFs

Benchmark Function Parameters

Mean SD Worst Best Wilcoxon

BF8 −6380.72 1173.462 −3597.6 −8917.54 7.56E-10

BF9 1.00E-11 1.11E-11 91.53603 22.88407 7.56E-10
BF10 1.18E-23 1.26E-18 1.646224 1.41E-06 7.56E-10
BF11 3.5591E-04 0.0025 0.0178 0 7.56E-10
BF12 0.002073 0.014661 0.103669 1.64E-13 7.56E-10
BF13 4.87E-18 3.69E-09 0.010987 4.70E-12 7.56E-10

Position 1: In this scheme, there are sensor nodes that re- Position 2: This scheme is used to illustrate the selection of
quire high energy at some positions. The relay sensor node is the relay sensor node, represented by the red color sensor node,
used to bridge the gap between the farthest sensor nodes and for the scenario given in Fig. 3 (Scenario 2; Position 2). The
the sink. position of the relay sensor node in Scenario 2 (Position 1) differs

Table 9 Convergence curves and trial solutions of hPSO-SA for multimodal BFs

Multimodal BF
Test Function
Curve for PSO vs. PSO-SA Variation of fitness during trial runs

BF8

BF9

BF10
 N. Bilandi et al.

Table 9 (continued)

BF11

BF12

BF13

from that in Scenario 2 (Position 2) according to the cost function The energy used for transmitting the data can be evaluated
and position of the subject. In this way, any sensor node can be using Eq. (11).
selected as a relay sensor node depending on the situation.
Initially, all the sensors in the WBAN have the same ener- Entrans ¼ EnT elec  p þ Enamp  p  d 2 ð11Þ
gy. A description of the sensor positions on the person’s body
The energy consumed on the receiver side can be calculat-
is given in Table 2. The following assumptions were consid-
ed as per Eq. (12).
ered for the deployed sensor nodes on the body:
Enrec ¼ EnR elec p ð12Þ
1 All the sensor nodes are static.
2 All the sensor nodes have bidirectional links. The total energy consumed in transmitting the data after
3 Each sensor node has an equal amount of energy. adding path loss ‘n’ can be calculated by Eq. (13).
4 Each sensor node is aware of its distance from neighboring
Entrans ¼ EnT elec  p þ Enamp  n  p  d 2 ð13Þ
nodes and from the sink.
where,
Entrans Transmission energy consumption
2.4.2 Energy-optimized model for the WBAN problem Enrec Reception energy consumption
EnT _ elec Energy required by the required by
Every sensor in the WBAN remains continually active to sup- the transmitter circuitry
ply energy to sense, process, and transmit the data from the EnR _ elec Energy needed by the receiver circuitry
sensor nodes. The flowchart of proposed hPSO-SA for Enamp Amplifier energy
WBAN problem is shown in Fig. 4. p Packet size
hPSO-SA: hybrid particle swarm optimization-simulated annealing algorithm for relay node selection in...

Table 10 Comparison of multimodal BF results

Algorithm Parameters Multimodal BF

BF8 BF9 BF10 BF11 BF12 BF13

PSO [29] Mean −4.84E+03 4.67E+01 2.76E-01 9.22E-03 6.92E-03 6.68E-03

SD 1.15E+03 1.16E+01 5.09E-01 7.72E-03 2.63E-02 8.91E-03
GSA [30] Mean −2.82E+03 2.60E+01 6.21E-02 2.77E+01 1.80E+00 8.90E+0
SD 4.93E+02 7.47E+0 2.36E-01 5.04E+0 9.51E-01 7.13E+0
FEP [31] Mean −1.26E+04 4.60E-02 1.78E-02 1.60E-02 9.20E-06 1.59E-04
SD 5.28E+01 1.20E-02 2.10E-03 2.20E-02 3.6E-06 7.29E-05
ALO [32] Mean −1.61E+03 7.71E-06 3.73E-15 1.86E-02 9.75E-12 2.00E-11
SD 3.14E+02 8.45E-06 1.50E-15 9.55E-03 9.33E-12 1.13E-11
SMS [33] Mean −4.21E+0 1.33E+0 8.88E-06 7.06E-01 1.23E-01 1.35E-02
SD 9.36E−16 3.27E-01 8.56E-09 9.08E-01 4.08E-02 2.88E-04
BA [34] Mean −1.07E+03 1.23E+0 1.29E-01 1.45E+0 3.96E-01 3.87E-1
SD 8.58E+02 6.86E-1 4.33E-02 5.70E-01 9.93E-01 1.22E-1
FPA [35] Mean -1.84E+3 2.73E-1 7.40E-3 8.50E−2 2.66E-4 3.67E-06
SD 5.04E+1 6.86E-2 7.10E-3 4.00E-2 5.53E-4 3.51E-06
CS [47] Mean -2.09E+03 1.27E-01 8.16E-09 1.23E-01 5.60E-09 4.88E-06
SD 7.62E-03 2.66E-03 1.63E-08 4.97E-02 1.58E−10 6.09E-07
FA [48] Mean -1.25E+03 2.63E-01 1.67E-01 9.97E-02 1.26E-01 2.13E-03
SD 3.53E+02 1.83E-01 5.08E-02 2.45E-02 2.63E-01 1.24E-03
GA [49] Mean −2.09E+03 6.59E-01 9.56E-01 4.88E-01 1.11E-01 1.29E-01
SD 2.47E+00 8.16E-01 8.08E-01 2.18E-01 2.15E-03 6.89E-02
MFO [39] Mean −8.50E+03 8.46E+01 1.26E+00 1.91E-02 8.94E-01 1.16E-01
SD 7.26E+02 1.62E+01 7.30E-01 2.17E-02 8.81E-01 1.93E-01
MVO [40] Mean −1.17E+4 1.18E+2 4.09E+0 9.40E-1 2.46E+0 2.21E-1
SD 9.36E+2 3.92E+1 5.52E+0 6.00E-2 7.90E-1 9.05E-2
hPSO-SA Mean −6380.72 1.00E-11 1.18E-23 3.5591E-04 0.002073 4.87E-18
SD 1178.156 1.11E-11 1.26E-18 0.0025 0.014661 3.69E-09

n Path loss coefficient To validate the proposed hPSO-SA algorithm, eighteen

d Distance from a source node to benchmark functions (BFs) [41] were considered and are
destination node shown in Appendix (Tables 22, 23 and 24). Table 22 de-
picts the unimodal BFs, Table 23 illustrates the multimod-
al BFs and Table 24 depicts the fixed dimension BF prob-
3 Experimental setup lems. Additionally, five specific engineering design prob-
lems were used to test the performance of the proposed
To validate the stochastic behavior of the proposed hPSO-SA algorithm (pressure-vessel, welded-beam, three-bar truss,
algorithm, the results of 30 trial runs are analyzed to obtain the cantilever-beam design, and tension/compression-spring
average, worst, and best values along with the respective stan- design), and the proposed algorithm was also tested on
dard deviations, and the Wilcoxon rank test is performed for a an energy optimization problem (i.e., application of the
significance level of 0.05. In this complete research study, a total proposed algorithm to the wireless body area network de-
of 30 search agents were considered, and the algorithm was sign problem) with different scenarios to detect the most
executed for a maximum of 100 iterations. The various parame- optimal route. Table 4 lists the WBAN parameters used to
ters used for the proposed hPSO-SA are demonstrated in Table 3. solve this problem.

Table 11 Results of the hPSO-SA algorithm for the fixed-modal BFs

Benchmark Function Parameters

Mean SD Worst Best Wilcoxon

BF14 0.004432 0.000156 0.0209 3.0750E-04 7.56E-10

BF15 0.000856 0.000125 0.001883 0.000328 7.56E-10
BF16 −1.03162 2.27E-16 −1.03163 −1.03163 3.81E-12
BF17 0.397887 3.36E-16 0.397887 0.397887 1.54E-12
BF18 3 1.03E-15 3 3 1.28E-10
 N. Bilandi et al.

3.1 Evaluation on standard benchmarks shown in Table 8. Table 9 shows the convergence curves and
trial solutions of hPSO-SA. Table 10 shows the results obtain-
To prove the exploitation phase of hPSO-SA, unimodal ed by executing the proposed hPSO-SA in comparison with
benchmark functions BF1, BF2, BF3, BF4, BF5, BF6 and PSO [29], GSA [30], FEP [31], ALO [32], SMS [41], BA
BF7 were evaluated. Table 5 shows the results on the [34], FPA [35], CS [36], FA [37], GA [38], MFO [39] and
unimodal BFs obtained using the hPSO-SA algorithm. The MVO [40]. In this table, the mean and standard deviation were
convergence curves along with the trial solutions of hPSO- calculated on six multimodal benchmark functions that are
SA for the unimodal BFs are shown in Table 6. good for evaluating the exploitation capability of the algo-
Table 7 shows the results obtained by executing the proposed rithm. It can be observed that the mean values generated by
hPSO-SA in comparison with PSO [29], GSA [30], FEP [31], the proposed hPSO-SA are significantly lower than those gen-
ALO [32], SMS [41], BA [34], FPA [35], CS [36], FA [37], GA erated by the other methods. It can also be observed that the
[38], MFO [39] and MVO [40]. In this table, the mean and standard deviation values of the proposed approach are signif-
standard deviation were calculated on seven unimodal bench- icantly lower than those of the other methods.
mark functions that are good for evaluating the exploitation The fixed dimension benchmark function BF12, BF13,
capability of the algorithm. It can be observed that the mean BF14, BF15, BF16, BF17 and BF18 test results are shown
values generated by the proposed hPSO-SA are significantly in Table 11. All trial solutions along with the convergence
lower than those generated by the other methods. It can also curves of hPSO-SA for the fixed dimension BFs are shown
be observed that the standard deviation values of the proposed in Table 12. Table 13 shows the results obtained by executing
approach are significantly lower than those of the other methods. the proposed hPSO-SA in comparison with PSO [29], GSA
The results on multimodal BF7, BF8, BF9, BF10, BF11, [30], FEP [31], ALO [32], SMS [41], BA [34], FPA [35], CS
BF12 and BF13 obtained using the hPSO-SA algorithm are [36], FA [37], GA [38], MFO [39] and MVO [40]. In this

Table 12 Convergence curves and trial solutions of hPSO-SA for fixed dimension BFs

Test Fixed Dimension BF

Functio
n Curve for PSO vs. PSO-SA Variation of fitness during trial runs

BF14

BF15
hPSO-SA: hybrid particle swarm optimization-simulated annealing algorithm for relay node selection in...

Table 12 (continued)

BF16

BF17

BF18

table, the mean and standard deviation were calculated on five are provided in the subsequent sections. A comparative anal-
fixed-modal benchmark functions that are good for evaluating ysis of the results on the five engineering design problems
the exploitation capability of the algorithm. It can be observed obtained using the proposed hPSO-SA with different simula-
that the mean values generated by the proposed hPSO-SA are tion parameters is given in Table 14.
significantly lower than those generated by the other methods.
It can also be observed that the standard deviation values of 3.2.1 Three-bar truss design problem
the proposed approach are significantly lower than those of
the other methods. The first problem of three-bar truss design was conducted to
compare the results of the proposed hPSO-SA algorithm for
3.2 Real engineering design optimization problems engineering optimization problems in which the key goal
for the fitness feature is to minimize a weight [50]. The
To substantiate the performance of the hPSO-SA algorithm various constraints for the three-bar truss design problem
for multidisciplinary engineering design optimization prob- are the stress constraint, deflection constraint, and buckling
lems, five multidisciplinary project optimization problems constraint. The mathematical formulation of the aforemen-
are addressed. Detailed illustrations of these design issues tioned problem is as follows:
 N. Bilandi et al.

The overall truss structure is shown in Fig. 5. 3.2.2 Pressure vessel design problem
Table 15 shows the results for the three-bar truss problem in
comparison with the results of some standard algorithms that are The second problem of pressure vessel design was conducted
already applied in this field. The three-bar truss design problem to compare the results of the proposed hPSO-SA algorithm for
has been solved by MBA [51], PSO-DE [50], MFO [39], CS engineering optimization problems [50] in which the key goal
[52], Ray and Sain [53], Tsa [54], DEDS [55] and the proposed is to reduce the total cost (including material, forming and
hPSO-SA algorithm. The results are presented in Table 15 and welding) of a cylindrical vessel. For the pressure vessel design
demonstrate that the hPSO-SA algorithm is capable of success- problem, four different variables are taken into consideration:
fully solving very constrained problems. The results were aver- thickness of the shell (Ts), thickness of the head (Th), liner
aged over 30 different runs, where each run had 100 iterations. radius (R) and length of the cylindrical section without

Table 13 Comparison of fixed dimension BF results

Algorithm Parameters Fixed Dimension BF

BF14 BF15 BF16 BF17 BF18

PSO [29] Mean 3.627168 0.000577 −1.03163 0.397887 3

SD 2.560828 0.000222 6.25E-16 0 1.33E-15
GSA [30] Mean 5.86E+00 0.003673 −1.03163 0.397887 3
SD 3.83E+00 0.001647 4.88E-16 0 4.17E-15
FEP [31] Mean 1.220 0.0005 −1.030 0.3980 3.02E+00
SD 00.560 0.0003 4.90E-07 1.50E-07 1.10E-01
ALO [32] Mean 1.51E-04 1.46E+01 1.75E+02 3.16E+02 4.40E+00
SD 3.82E-04 3.22E+01 4.65E+01 1.30E+01 1.66E+00
SMS [33] Mean 776.484 873.752 9.61E+02 899.857 740.965
SD 5.21E-12 9.71617 6.72E+01 1.99E-05 0.785
BA [34] Mean 182.47 487.202 588.193 756.975 542.200
SD 117.024 161.410 137.786 160.09 220.201
FPA [35] Mean 3.37E-01 18.23309 2.24E+02 362.0262 10.1592
SD 2.36E-01 3.074685 5.03E+01 54.01816 1.393908
CS [47] Mean 0110 140.606 2.90E+02 4.03E+02 2.14E+02
SD 110.050 92.8032 8.61E+01 9.83E+01 2.07E+02
FA [48] Mean 150.1696 314.4654 734.5372 818.5732 133.5203
SD 97.15906 92.93417 203.9693 109.9663 215.6027
GA [49] Mean 114.6139 95.46331 325.4427 466.3074 90.36913
SD 26.96248 7.163383 51.66827 29.56841 13.72977
MFO [39] Mean 8.25E-31 66.73272 119.0146 345.4688 10.4086
SD 1.08E-30 53.22555 28.3318 43.11578 3.747669
MVO [40] Mean 10.00017 30.00705 50.00061 190.3 160.5312
SD 31.62288 48.30615 52.70461 128.6659 158.2887
hPSO-SA Mean 0.004432 0.000156 −1.03162 0.397887 3
SD 0.00811 0.000125 2.27E-16 3.36E-16 1.03E-15
hPSO-SA: hybrid particle swarm optimization-simulated annealing algorithm for relay node selection in...

considering the head (L). Four constraints are put on the afore- The mathematical formulation of the abovementioned
mentioned design issue. problem is as follows:

The overall pressure vessel design structure is shown in 3.2.3 Tension/compression spring design problem
Fig. 6.
Table 16 shows the results for the pressure vessel design The third problem of tension/compression spring design was
problem in comparison with the results of some standard al- conducted to compare the results of the proposed hPSO-SA
gorithms that are already applied in this field. The pressure algorithm for engineering optimization problems [50] in
vessel design problem has been solved by GWO [41], GSA which the key goal is to reduce the weight of a spring. For
[30], PSO [56], GA [57], ES [58], DE [59], ACO [60], the solution of the problem, three design variables are taken
Lagrange [61], branch-bound [62] and the proposed hPSO- into consideration, i.e., the number of active coils (N), mean
SA algorithm. The findings show that the costs obtained are coil diameter (D) and wire diameter (d), and the problem is
lowest for hPSO-SA compared with the other algorithms subject to surge frequency constraints, shear stress constraints
mentioned. and minimum deflection constraints. The mathematical for-
mulation of this problem is as follows:
 N. Bilandi et al.

Table 14 Solution of engineering design problems using hPSO-SA

Engineering Design Problem Best value Mean value Worst value SD Median Wilcoxon p value

Three-bar truss problem 263.89598 264.53092 282.84271 3.4585456 263.89873 7.56E-10

Pressure vessel problem 5894.7899 6.04E+03 7216.4691 330.89676 5909.8847 7.56E-10
Tension/compression spring design problem 0.0126751 0.012728 0.012871 3.90E-05 0.0127235 7.56E-10
Welded beam problem 1.7259158 1.7275081 1.7326591 0.0015716 1.7268292 7.56E-10
Cantilever beam design 1.3032759 1.3033278 1.3034016 3.16E-05 1.3033284 7.56E-10

The overall spring design structure is shown in Fig. 7. 3.2.4 Welded beam design problem
Table 17 shows the results for the tension/compression
spring design problem in comparison with the results of some The fourth problem of welded beam design was conducted to
standard algorithms that are already applied in this field. The compare the results of the proposed hPSO-SA algorithm for
tension/compression spring design problem has been solved engineering optimization problems [50] in which the key goal
by GWO [41], GSA [30], PSO [56], GA [57], ES [58], DE is to reduce the fabrication cost of a welded beam. The prob-
[59] and the proposed hPSO-SA algorithm. The results again lem consists of four variables: the length of the weld (L), the
show that the proposed hPSO-SA algorithm can solve real thickness of the bar (T), the breadth of the bar (B) and the
constraint problems effectively. thickness of the weld (H).
The mathematical formulation of the abovementioned
problem is as follows:
hPSO-SA: hybrid particle swarm optimization-simulated annealing algorithm for relay node selection in...

Random [63] and the proposed hPSO-SA algorithm.

Compared to the other techniques, the table confirms the
better efficiency of hPSO-SA.

3.2.5 Cantilever beam design problem

The fifth problem of cantilever beam design was conducted to

compare the results of the proposed hPSO-SA algorithm for
engineering optimization problems [64] in which the key goal
Fig. 5 Three-bar truss design problem
is to reduce the weight of a beam. In Fig. 9, each element is
defined by one variable in this design problem, and the overall
structure has five structural variables. The beam thickness is
The overall welded beam structure is shown in Fig. 8. taken as a constant. In the proposed design, a vertical load is
Table 18 shows the results for the welded beam design added to the beam’s free end (node 6), and the beam’s right
problem in comparison with the results of some standard side (node 1) is strictly maintained.
algorithms that are already applied in this field. The The mathematical formulation of the abovementioned
tension/compression spring design problem has been problem is as follows:
solved by GWO [41], GSA [30], GA [57], ES [58],

Table 15 Comparative analysis of the three-bar truss problem results

Algorithm hPSO-SA MBA [51] PSO-DE [50] MFO [39] CS [52] Ray and Sain [53] Tsa [54] DEDS [55]

Optimum values for variables A1 0.788 0.788 0.788 0.788 0.789 0.795 0.788 0.788
A2 0.408 0.408 0.408 0.409 0.409 0.395 0.408 0.408
Optimal weight 263.896 263.895 263.895 263.8959 263.972 264.3 263.68 263.895

Fig. 6 Pressure vessel design problem

 N. Bilandi et al.

Table 16 Comparative analysis of the pressure vessel problem results

Algorithm hPSO- GWO [41] GSA [30] PSO [56] GA [57] ES [58] DE [59] ACO [60] Lagrange [61] Branch-bound [62]
SA

Optimum Values Ts 0.79 0.81 1.13 0.81 0.81 0.81 0.81 0.81 1.13 1.13
Th 0.39 0.43 0.63 0.44 0.43 0.44 0.44 0.44 0.63 0.63
R 40.67 42.09 55.99 42.09 40.32 42.10 42.10 42.10 58.29 47.70
L 195.31 176.76 84.45 176.75 200.00 176.64 176.64 176.57 43.69 117.70
Optimum Cost 5894.79 6051.5 8538.8 6061.08 6288.75 6059.75 6059.73 6059.09 7198.04 8129.10

Fig. 7 Tension/compression
spring design problem

Table 17 Comparative analysis of the tension/compression spring design problem results

Algorithm hPSO- GWO [41] GSA [30] PSO [56] GA [57] ES [58] DE [59]
SA

Optimum Values d 0.050 0.052 0.050 0.052 0.052 0.052 0.052

D 0.317 0.357 0.324 0.358 0.364 0.352 0.355
N 14.052 11.289 13.525 11.245 10.891 11.632 11.411
Optimum Weight 0.013 0.013 0.013 0.013 0.013 0.013 0.013

Table 19 shows the results for the cantilever beam design

problem in comparison with the results of some standard al-
gorithms that are already applied in this field. The cantilever CFðiÞ ¼ DðiÞ=RE ðiÞ ð14Þ
beam design problem has been solved by ALO [32], SOS
[65], CS [52], MMA [64] and the proposed hPSO-SA algo-
rithm. The findings show that the weights obtained are lowest
for hPSO-SA compared with the other algorithms mentioned.

3.2.6 Wireless body area network design problem using

hPSO-SA

The proposed algorithm was tested on an energy optimization

problem, i.e., the proposed algorithm was applied to the wire-
less body area network design problem [10, 20, 29] (Fig. 3) in
which the key goal is to minimize the energy consumption of
the network using a relay node, i.e., an intermediate node. In
this regard, to find the optimal path for minimizing the energy,
hPSO-SA is used for the relay node selection problem. The
idea is to find the perfect relay sensor node for specific sensor
nodes by calculating the cost of transmitting data using the Fig. 8 Welded beam design problem
cost function (CF) given by Eq. (14).
hPSO-SA: hybrid particle swarm optimization-simulated annealing algorithm for relay node selection in...

Table 18 Comparative analysis of the welded beam problem results

Algorithm hPSO- GWO [41] GSA [30] GA [57] ES [58] Random [63] Simplex [63] David [63] APPROX [63]
SA

Optimum Values H 0.2048 0.2057 0.1821 0.2489 0.2442 0.4575 0.2792 0.2434 0.2444
L 3.4969 3.4784 3.857 6.173 6.2231 4.7313 5.6256 6.2552 6.2189
T 9.0367 9.0368 10 8.1789 8.2915 5.0853 7.7512 8.2915 8.2915
B 0.2057 0.2058 0.2024 0.2533 0.2443 0.66 0.2796 0.2444 0.2444
Optimal Cost 1.7259 1.7262 1.88 2.4331 2.3807 4.1185 2.5307 2.3841 2.3815

Fig. 9 Cantilever beam design problem

Table 19 Comparative analysis of the cantilever beam design problem results

Algorithm hPSO-SA ALO [32] SOS [65] CS [52] MMA [64] GCA_I [64] GCA_II [64]

Optimal values for variables x1 5.9854 6.0181 6.0188 6.0089 6.01 6.01 6.01
x2 4.87 5.3114 5.3034 5.3049 5.3 5.304 5.3
x3 4.4493 4.4884 4.4959 4.5023 4.49 4.49 4.49
x4 3.5172 3.4975 3.499 3.5077 3.49 3.498 3.49
x5 2.1187 2.1583 2.1556 2.1504 2.15 2.15 2.15
Optimum weight 1.3033 1.33995 1.33996 1.33999 1.34 1.34 1.34
NFE 500 14,000 15,000 2500 N/A N/A N/A

Fig. 10 Convergence curve of residual energy vs. round (r) Fig. 11 Convergence curve of network lifetime vs, round (r)
 N. Bilandi et al.

Table 20 Comparison of PSO, hPSOSA-S1 and hPSOSA-S2 for WBAN problems

Algorithm PSO hPSOSA- hPSOSA-

S1 S2
Parameter

First sensor died (round) 2450 4410 3895

Last sensor died (round) 8974 8965 9990
Packets received at round 10,000 22,500 20,700 31,520
Residual energy (J) At round 2000 2.7 2.8 3.4
At round 4000 1.6 1 2.25
At round 6000 0.8 0.4 1.3
At round 8000 0.3 0.2 0.3
Path loss (dB) At round 2000 423 375 255
At round 4000 263 104 255
At round 6000 263 104 255
At round 8000 210 104 70

where
CF Cost Function
D (i) Distance from the ith sensor node to the sink
RE (i) Residual energy of the ith sensor node

The proposed hPSO-SA aims to compare the essential pa-

rameters of WBAN problems, such as the stability period,
throughput, residual energy, and path loss, for PSO,
hPSOSA-S1, and hPSOSA-S2.

Residual energy A multihop communication technique is

used to expand the hPSO-SA algorithm capability to ex-
amine the remaining energy at each sensor node during
Fig. 12 Convergence curve of throughput vs. round (r) network operation. On the basis of the cost function, a
relay sensor node is evaluated for energy consumption
minimization. By saving energy, the execution time of
the WBAN is increased. In PSO and hPSOSA-S1, a
few sensor nodes are drained early because of the intense
load on the sink. hPSOSA-S2 chooses a different relay
sensor node in each round to balance the load of sensor
nodes by using multihop communication. This result in-
dicates that hPSOSA-S2 is highly energy efficient and
produces a maximum throughput rate, which is depicted
in Fig. 10.

Stability period The stability period is the time until the first
dead node that contributes to the network stability period is
encountered, and the time until the last dead node is en-
countered corresponds to the lifetime of the network.
Figure 11 and Table 20 show the stability period for the
Fig. 13 Convergence curve of path loss vs. round (r) three proposed algorithms. The period between the first
and last sensor node deaths is called the ‘unbalanced
hPSO-SA: hybrid particle swarm optimization-simulated annealing algorithm for relay node selection in...

Table 21 Comparative analysis of the proposed algorithm and existing algorithms on the WBAN problem

Algorithm Proposed Genetic Artificial Bee Hybrid PSO- PSO Genetic Ant Colony
→ hPSO-SA Algorithm [14] Colony [21] GA [22] [20] Algorithm [66] Algorithm [21]

Improved network 97% 95% 96% 90% 94.5% 96.76% 91%

lifetime

period.’ It is observed that all the algorithms work well, but From Table 21, it is evident that the proposed hPSO-SA
the stability period for algorithm hPSOSA-S2 is better than has a better network lifetime compared to other existing algo-
that for PSO. It can be clearly seen in the graph that for the rithms for the WBAN problem. Since the proposed hPSO-SA
proposed hPSO-SA, the first node dies at round 3750, and ability to explore the search space is much greater than that of
for PSO, the first node dies at round 2400. The lifetime of PSO, GA, ABC and hPSO-GA, it optimizes the fitness func-
the first node with hPSO-SA is approximately 1.5 times tion more effectively and rapidly detects the shortest path
longer than that with PSO, which is sufficient to maintain within the WBAN for data transmission.
the stability for a longer period.

Throughput The throughput is the total number of packets that

are successfully received by the sink. For the three algorithms, 4 Conclusions and future work
Figure 11 shows the maximum throughput of hPSOSA-S2
compared to those of PSO and hPSOSA-S1. For the success- In this research, the exploitation phase of the existing PSO
ful implementation of a WBAN, there is a need for an algo- algorithm has been successfully enhanced by incorporating a
rithm that has the least amount of packet drop and the maxi- local search-based mechanism known as SA. The proposed
mum rate of throughput. The number of packets obtained at hybrid PSO-SA has been tested in three different stages. In the
the sink can be calculated according to the number of live first stage, the authors validated the performance of the pro-
sensor nodes, which leads to the maximum throughput of posed hPSO-SA algorithm on eighteen different benchmark
the WBAN. The proposed algorithm hPSO-SA obtained the functions; then, in the second stage of analysis, the effective-
maximum rate of throughput, as shown in Fig. 12, compared ness of the proposed hPSO-SA for five engineering design
to PSO. problems was analyzed, and finally, the proposed algorithm
was implemented for the selection of the relay node in a
Path loss The path loss is the power loss while transferring WBAN model to minimize the total energy consumption.
data of the sensor node from the transmitter to the receiver, The results clearly indicate the superior performance of the
which helps in depicting the significant role of the link. The proposed hPSO-SA algorithm in all three stages of evaluation
remoteness of the sensor nodes from the sink is considered at a when the results are compared against those of various state-
frequency of 2.4 GHz. The path loss can be measured as the of-the-art metaheuristic algorithms. The experimental results
distance between the transmitter and the receiver for commu- indicated that the proposed algorithm aids in increasing the
nication. By using the multihop technique, the proposed meth- network lifetime and energy compared to other algorithms for
od hPSOSA-S2 reduces the transmission distance and thus, as the WBAN relay node selection problem. In the future, this
Fig. 13 illustrates, leads to the least path loss compared to PSO work can be further extended to reduce the computation time
and hPSOSA-S1. of the algorithms and to the selection of the relay node in a
The proposed algorithm hPSOSA-S2 leads to a greater scenario with a larger number of deployed sensor nodes.
energy with a maximum number of live sensor nodes, which
results in a longer period of stability and thus less accumulated Compliance with ethical standards
path loss. A comparative analysis of PSO, hPSOSA-S1, and
hPSOSA-S2 with regard to the different simulation parame- Conflict of interest The authors declare that there is no conflict of
ters is given in Table 20. interest.
 N. Bilandi et al.

Appendix

Table 22 Unimodal BFs

3-D view
Function Dim Range
f min

30 [-100 , 100] 0

30 [-10 ,10] 0

30 [-100 , 100] 0

30 [-100 , 100] 0

30 [-30 ,30] 0

30 [-100 , 100] 0

30 [-1.28 , 1.28] 0
hPSO-SA: hybrid particle swarm optimization-simulated annealing algorithm for relay node selection in...

Table 23 Multimodal BFs

Function Dim Range f min 3-D View

 N. Bilandi et al.

Table 24 Fixed dimension BFs

hPSO-SA: hybrid particle swarm optimization-simulated annealing algorithm for relay node selection in...

Fig. 14 Pseudocode for Particle Swarm Optimization algorithm

Fig. 15 Pseudocode for SA algorithm

 N. Bilandi et al.

References 21. Yan J, Peng Y, Shen D, Yan X, Deng Q (2018) An artificial bee
colony-based green routing mechanism in WBANs for sensor-
based E-healthcare systems. Sensors 18:3268
1. Cai X, Li J, Yuan J et al (2014) Energy-aware adaptive topology
22. Agnihotri A, Gupta IK (2018) A hybrid PSO-GA algorithm for
adjustment in wireless body area networks. Telecommun Syst 58:
routing in wireless sensor network. In: In 2018 4th international
139–152
conference on recent advances in information technology. IEEE,
2. Elias J (2014) Optimal design of energy-efficient and cost-effective Dhanbad, pp 1–6
wireless body area networks. Ad Hoc Netw 13:560–574 23. Bilandi N, Verma HK, Dhir R (2019) PSOBAN : a novel particle
3. Javaid N, Ahmad A, Nadeem Q, Imran M, Haider N (2015) IM- swarm optimization based protocol for wireless body area net-
SIMPLE: IMproved stable increased-throughput multi-hop link ef- works. SN Appl Sci 1:1492
ficient routing protocol for wireless body area networks. Comput 24. Samal TK, Patra SC, Kabat MR (2019) An adaptive cuckoo search
Human Behav 51:1003–1011 based algorithm for placement of relay nodes in wireless body area
4. Wang J, Cho J, Lee S et al (2010) Hop-based energy aware routing networks. J King Saud Univ – Comput Inf Sci:14
algorithm for wireless sensor networks. IEICE Trans Commun 25. Raj AS, Chinnadurai M (2020) Energy efficient routing algorithm
E93–B:305–316 in wireless body area networks for smart wearable patches. Comput
5. Tauqir A, Javaid N, Akram S et al (2013) Distance aware relaying Commun 153:85–94
energy-efficient: DARE to monitor patients in multi-hop body area 26. Singh P, Satvir M (2019) Improved artificial bee colony
sensor networks. In: Proceedings - 2013 8th international confer- metaheuristic for energy-efficient clustering in wireless sensor net-
ence on broadband, wireless computing, communication and appli- works. Artif Intell Rev 51:329–354
cations, BWCCA 2013. Compiegne, France, pp 206–213 27. Mittal N (2019) Moth flame optimization based energy efficient
6. Jing L, Ming L, Bin Y, Wenlong L (2015) A novel energy efficient stable clustered routing approach for wireless sensor networks.
MAC protocol for wireless body area network. China Commun 12: Wirel Pers Commun 104:677–694
11–20 28. Wolpert DH, Macready WG (1997) No free lunch theorems for
7. Yuan X, Li C, Yang L, Yue W, Zhang B, Ullah S (2016) A token- optimization. IEEE Trans Evol Comput 1:67–82
based dynamic scheduled MAC protocol for health monitoring. 29. Kennedy J, Eberhart RC (1995) Particle swarm optimization. In:
EURASIP J Wirel Commun Netw 2016:125 Perth A (ed) Proceedings of the IEEE international conference on
8. Kim J, Song I, Choi S (2015) Priority-based adaptive transmission neural networks, pp 1942–1948
algorithm for medical devices in wireless body area networks 30. Rashedi E, Nezamabadi-Pour H, Saryazdi S (2009) GSA: a gravi-
(WBANs). J Cent South Univ 22:1762–1768 tational search algorithm. Inf Sci 179:2232–2248
9. Vimalarani C, Subramanian R, Sivanandam SN (2016) An en- 31. Yao X, Liu Y, Lin G (1999) Evolutionary programming made
hanced PSO-based clustering energy optimization algorithm for faster. IEEE Trans Evol Comput 3:82
wireless sensor network. Sci World J 2016:1–12 32. Mirjalili S (2015) The ant lion optimizer. Adv Eng Softw 83:80–98
10. Dhadwal MK, Jung SN, Kim CJ (2014) Advanced particle swarm 33. Cuevas E, Echavarría A, Ramírez-Ortegón MA (2014) An optimi-
assisted genetic algorithm for constrained optimization problems. zation algorithm inspired by the states of matter that improves the
Comput Optim Appl 58:781–806 balance between exploration and exploitation. Appl Intel 40:256–
272
11. Xu G, Wang M (2014) An energy-efficient routing mechanism
34. Yang X-S (2010) A new metaheuristic bat-inspired algorithm. In:
based on genetic ant Colony algorithm for wireless body area net-
Nature inspired cooperative strategies for optimization (NICSO
works. J Networks 9:3366–3372
2010). Springer, Berlin, pp 65–74
12. D’Andreagiovanni F, Nardin A (2015) Towards the fast and robust
35. Yang X-S (2012) Flower pollination algorithm for global optimiza-
optimal design of wireless body area networks. Appl Soft Comput J
tion. In: Unconventional computation and natural computation.
37:971–982
Springer, Berlin, pp 240–249
13. Sangwan A, Bhattacharya PP (2016) An optimization to routing 36. Yang X (2012) Multiobjective firefly algorithm for continuous op-
approach under WBAN architectural constraints. Intell Syst timization, pp 13–15
Technol Appl 385:75–89 37. algorithm YXF (2010) Levy flights and global optimization.
14. Kaur N, Singh S (2017) Optimized cost effective and energy effi- Swarm Intell bio-inspired Comput 2013:49–72
cient routing protocol for wireless body area networks. Ad Hoc 38. Holland J (1992) Adaptation in natural and artificial systems. MIT
Netw 61:65–84 Press, Cambridge
15. Qais MH, Hasanien HM, Alghuwainem S, Nouh AS (2019) Coyote 39. Mirjalili S (2015) Moth-flame optimization algorithm: a novel
optimization algorithm for parameters extraction of three- diode nature-inspired heuristic paradigm. Knowledge-Based Syst 89:
photovoltaic models of photovoltaic modules. Energy 187:116001 228–249
16. Hayyolalam V, Asghar A, Kazem P (2020) Engineering applica- 40. Mirjalili S, Mirjalili SM, Hatamlou A (2016) Multi-verse optimizer:
tions of artificial intelligence black widow optimization algorithm : a nature-inspired algorithm for global optimization. Neural Comput
a novel meta-heuristic approach for solving engineering optimiza- Appl 27:495–513
tion problems. Eng Appl Artif Intell 87:103249 41. Mirjalili S, Mirjalili SM, Lewis A (2014) Grey wolf optimizer. Adv
17. Gao W (2020) New ant Colony optimization algorithm for the Eng Softw 69:46–61
traveling salesman problem. Int J Comput Intell Syst 13:44–55 42. Jamili A, Shafia MA (2011) A hybrid algorithm based on particle
18. Garg H (2016) A hybrid PSO-GA algorithm for constrained opti- swarm optimization and simulated annealing for a periodic job shop
mization problems. Appl Math Comput 274:292–305 scheduling problem. Int J Adv Manuf Technol 54:309–322
19. Asghar A, Mirjalili S, Faris H, Aljarah I (2019) Harris hawks opti- 43. Wang X, Li J (2004) Hybrid particle swarm optimization with
mization : algorithm and applications. Futur Gener Comput Syst 97: simulated annealing. In: Proceedings of the third international con-
849–872 ference on machine learning and cybernetics, Shanghai, In, pp 26–
20. Wu T, Lin C (2015) Low-SAR path discovery by particle swarm 29
optimization algorithm in wireless body area networks. IEEE 44. Shi Y, Eberhart R (1998) A modified particle swarm optimizer.
Sensors J 15:928–936 1998 IEEE Int Conf Evol Comput proceedings IEEE World
Congr Comput Intell (Cat No98TH8360), pp 69–73
hPSO-SA: hybrid particle swarm optimization-simulated annealing algorithm for relay node selection in...

Naveen Bilandi He completed

45. Url S, Archive TJ, Archive T (2007) Optimization by simulated
his B.Tech from Rajasthan
annealing. Science (80) 220:671–680
University, Jaipur and M.Tech in
46. Mafarja MM, Mirjalili S (2017) Hybrid whale optimization algo- Computer Science and
rithm with simulated annealing for feature selection. Engineering from National
Neurocomputing 260:302–312 Institute of Technology (NIT),
47. Mareli M, Twala B (2017) An adaptive Cuckoo search algorithm Jalandhar. At present, he is pursu-
for optimisation. Appl Comput Informatics 14:107–115 ing Ph.D. from NIT, Jalandhar.
48. Yang XS (2008) Firefly algorithm. Eng Optim pp 20:79–90 He joined DAV University,
49. Kazarlis SA, Bakirtzis AG, Petridis V (1996) A genetic algorithm Jalandhar as Assistant Professor
solution to the unit commitment problem. IEEE Trans Power Syst in the Department of Computer
11:83–92 Science and Engineering in
50. Liu H, Cai Z, Wang Y (2010) Hybridizing particle swarm optimi- 2013. He has a teaching experi-
zation with differential evolution for constrained numerical and ence of more than ten years. His
engineering optimization. Appl Soft Comput 10:629–640 research interests are Wireless
51. Sadollah A, Bahreininejad A, Eskandar H, Hamdi M (2013) Mine Communication, Cloud computing, machine learning Internet of thing.
blast algorithm: a new population based algorithm for solving He has published twenty five research papers in international/national
constrained engineering optimization problems. Appl Soft journals/conferences. He has supervised more than thirty M.Tech thesis.
Comput J 13:2592–2612
52. Hossein A, Yang GX (2013) Cuckoo search algorithm : a
metaheuristic approach to solve structural optimization problems.
Eng Comput 29:17–35
53. Taylor P, Ray T, Saini P (2007) Engineering design optimization
using a swarm with an intelligent information sharing among indi-
viduals. Eng Optim 33:735–748
54. Taylor P, Tsai J (2007) Global optimization of nonlinear fractional
programming problems in engineering design. Eng Optim 37:399– Harsh Kumar Verma Professor,
409 Department of Computer Science
55. Zhang M, Luo W, Wang X (2008) Differential evolution with dy- and Engineering, Dr B R
namic stochastic selection for constrained optimization. Inf Sci 178: Ambedkar National. (h index
3043–3074 16). His research interests are
56. He Q, Wang L (2007) An effective co-evolutionary particle swarm Software Systems, Numerical
optimization for constrained engineering design problems. Eng Computing, Computer Networks
Appl Artif Intell 20:89–99 and Information Security. He has
57. Coello CAC (2000) Use of a self-adaptive penalty approach for supervised seven PhD thesis.
engineering optimization problems. Comput Ind 41:113–127
58. Taylor P, Mezura-montes E, Coello CAC (2008) An empirical
study about the usefulness of evolution strategies to solve
constrained optimization problems. Int J Gen Syst 37:443–473
59. Li LJ (2007) A heuristic particle swarm optimizer for optimization
of pin connected structures. Comput Struct 85:340–349
60. Kaveh A (2010) An improved ant colony optimization for
constrained engineering design problems. Eng Comput 27:155–
182
61. Kannan BK, Gradient RC, Kramer SN (1994) An augmented
Lagrange multiplier based method for mixed integer discrete con- Renu Dhir Associate Professor,
tinuous optimization. J Mech Des 116:405–411 Department of Computer Science
62. Sandgren E (1990) Nonlinear integer and discrete programming in and Engineering, Dr B R
mechanical design optimization. J Mech Des 112:223–229 Ambedkar National. Research in-
63. Phillips DT (1976) Optimal design of a class of welded structures terests are Image Processing, pat-
using geometric programming. J Manuf Sci Eng 98:1021–1025 tern Recognition, Information
64. Journal I, Numerical FOR, In M et al (1996) Structural optimization S e c u r i t y, A I a n d S o f t
using a new local approximation method. Int J Numer Methods Eng Computing. She has supervised
39:829–846 eight PhD thesis.
65. Cheng MY, Prayogo D (2014) Symbiotic organisms search: a new
metaheuristic optimization algorithm. Comput Struct 139:98–112
66. Ravi KKGRSV (2019) Genetic algorithm based sensor node clas-
sifications in wireless body area networks (WBAN). Cluster
Comput 22:12849–12855

Publisher’s note Springer Nature remains neutral with regard to jurisdic-

tional claims in published maps and institutional affiliations.