Journal of Intelligent & Fuzzy Systems 34 (2018) 1573–1582 1573

DOI:10.3233/JIFS-169452
IOS Press

Owl search algorithm: A novel

nature-inspired heuristic paradigm
for global optimization
Mohit Jaina,∗ , Shubham Mauryab , Asha Rania and Vijander Singha
a Instrumentationand Control Engineering Division, Netaji Subhas Institute of Technology,
University of Delhi, Delhi, India
b Department of Electronics and Communication, GLA University, Mathura, Uttar Pradesh, India

Abstract. This paper presents, a novel nature-inspired optimization paradigm, named as owl search algorithm (OSA) for
solving global optimization problems. The OSA is a population based technique based on the hunting mechanism of the owls
in dark. The proposed method is validated on commonly used benchmark problems in the field of optimization. The results
obtained by OSA are compared with the results of six state-of-the-art optimization algorithms. Simulation results reveal that
OSA provides promising results as compared to the existing optimization algorithms. Moreover, to show the efficacy of the
proposed OSA, it is used to design two degree of freedom PI (OSA-2PI) controller for temperature control of a real-time heat
flow experiment (HFE). Experimental results demonstrate that OSA-2PI controller is more precise for temperature control
of HFE in comparison to the conventional PI controller.

Keywords: Nature-inspired algorithm, unconstrained optimization, two degree of freedom PI controller, Heat flow experiment

1. Introduction helps the algorithm to search in the neighbourhood of

current best solutions. There are distinct objectives
In recent years, metaheuristic optimization behind the development of modern metaheuristics
techniques have gained significant attention of such as fast and effortless handling of complex as
researchers due to successful application of these well as large problems and designing more effective
techniques in a variety of complex optimization and robust techniques [2].
problems. These techniques are found more effective There is no limitation on the source of motivation
than conventional methods which use derivative to design a metaheuristic technique. As an illus-
information of function. Two eminent features of tration, the gravitational search algorithm (GSA) is
any metaheuristic technique are exploration and inspired from law of gravitation and mass interac-
exploitation [1]. Exploration phase of algorithm, tion [3], interior search algorithm (ISA) is based on
also known as diversification, redirects the search the concepts of interior designing and decoration [4]
towards unvisited regions of the search space, in etc. Nevertheless, nature is always a primary source
order to find new but potentially better solutions. On of motivation for proposing new metaheuristic tech-
the other hand, exploitation or intensification phase niques. A brief literature review of nature-inspired
∗ Corresponding
optimization algorithms is presented in Table 1.
author. Mohit Jain, ICE Division, NSIT, Sec-
3, Dwarka, New Delhi, University of Delhi, India. Tel.: +91
Various nature-inspired optimization algorithms are
9650298558; Fax: +011 25099022; E-mails: mohit.jain@nsit.ac.in available in literature, however “no free lunch (NFL)”
and nsit.mohit@gmail.com. theorem [19] supports the present study as proposed

1064-1246/18/$35.00 © 2018 – IOS Press and the authors. All rights reserved
1574 M. Jain et al. / Owl search algorithm

Table 1
Brief literature review of nature-inspired optimization algorithms
Algorithm Inspiration Year
Genetic algorithm (GA) [5] Evolution 1975
Particle swarm optimization (PSO) [6] Bird flock 1995
Ant colony optimization (ACO) [7] Ant colony 2006
Artificial bee colony (ABC) [8] Honey bee 2006
Monkey search (MS) [9] Monkey climbing process on trees while looking for food 2007
Firefly algorithm (FFA) [10] Social behavior of fireflies 2009
Bat algorithm (BA) [11] Echolocation behaviour of bats 2010
Krill herd (KH) [12] Herding behavior of krill individuals in nature 2012
Dolphin echolocation (DE) [13] Echolocation ability of dolphins 2013
Lightning search algorithm (LSA) [14] Natural phenomenon of lightning 2015
Dragonfly algorithm (DA) [15] Static and dynamic swarming behaviours of dragonflies 2015
Multi-verse optimizer (MVO) [16] Basic concepts in cosmology 2016
Shark smell optimization (SSO) [17] Ability of shark in finding its prey by smell sense 2016
Whale optimization algorithm (WOA) [1] Social behavior of humpback whales 2016
Crow search algorithm (CSA) [18] Intelligent food hiding behaviour of crows 2016

algorithm may outperform other existing optimizers

on some problems. This paper introduces a new sim-
ple, easy to implement and powerful nature-inspired
optimization algorithm called as owl search algo-
rithm (OSA). This algorithm simulates the hunting
mechanism of barn owls which rely on their hearing
capability to find prey (vole) in the dark night rather
than sight. The effectiveness of OSA is validated on a
set of unconstrained numerical benchmark functions.
The results obtained are compared with standard opti-
mizers like GA, PSO, BA, FFA, MVO, and KH.
Further suitability of proposed technique, to resolve
real world black box optimization problems, is inves- Fig. 1. Real barn owl and its anatomy of ears.
tigated by conducting a real-time experimental study
on Heat Flow Experiment (HFE) setup.
The rest of the paper is structured as follows: Sec-
tion 2 briefly outlines the motivation of proposed
work. Section 3 presents the implementation of OSA.
Sections 4 and 5 present the comparative analysis
of results on standard benchmark functions and real-
time Heat Flow Experiment respectively. Finally, the
work is concluded in Section 6.

2. Inspiration

Owls are typically nocturnal but highly efficient

predators with an extraordinary auditory system
which helps to locate the prey (vole). Some of the
species like barn owls (Fig. 1a) have evolved with a
distinct anatomical feature of auditory system with
vertical asymmetry of ears (Fig. 1b) [20]. Due to
this unique feature, the sound reaches one ear before
the other, and location of prey is obtained. Hence
prey can be located in dark by hearing ability instead Fig. 2. Auditory map of prey sound generated by owl’s brain.
 M. Jain et al. / Owl search algorithm 1575

of sight [21]. The sound signal generated by a vole receives max intensity (for maximization problems)
(Fig. 2a) is processed in the owl’s brain in two parts as it is more close to vole. The normalized intensity
i.e. the interaural time difference (ITD), and inter- information of ith owl is utilized to update the position
aural level (loudness) difference (ILD) (Fig. 2b) to and may be calculated as:
prepare an auditory map (Fig. 2c) of prey location
fi − w
[20]. The distance of prey is estimated on the basis of Ii = (4)
time and intensity differences of sound wave arrival b−w
[22]. where
b = max fk (5)
k∈1,...,n
3. Owl search algorithm (OSA)
w = min fk (6)
Similar to other nature-inspired population based k∈1,...,n
algorithms, OSA starts the optimization process with The distance information of each owl and prey is
an initial set of random solutions which represent the calculated by the following equation:
initial position of owls in a forest (d dimensional
search space). If there exist n number of owls in a Ri = Oi , V 2 (7)
forest, then their random position is stored in a n × d
matrix as follows: where V is the location of prey which is achieved
⎡ ⎤ by the fittest owl. It is also assumed that there exists
O1,1 O1,2 · · · · · · O1,d only one vole (global optimum) in the forest. Owls
⎢O ⎥
⎢ 2,1 O2,2 · · · · · · O2,d ⎥ take silent flights while moving towards the prey.
⎢ ⎥
⎢ .. .. .. .. .. ⎥ Hence, they receive changed intensity obeying the
O=⎢ ⎢ . . . . . ⎥
⎥ (1) inverse square law of sound intensity (Fig. 3) [23].
⎢ . ⎥ The change in intensity for ith owl can be obtained as
⎢ . .. .. .. .. ⎥
⎣ . . . . . ⎦ follows:
On,1 On,2 · · · · · · On,d Ii
Ici = + Random noise (8)
where matrix element Oi,j represents the j th variable R2i
(dimension) of ith owl. A uniform distribution (Equa-
In Equation (8), R2i is used instead of 4πR2i and
tion (2)) is used to allocate the initial location of each
random noise of environment is also considered to
owl in the forest.
make the mathematical model more realistic. In the
Oi = OL + U(0, 1) × (OU − OL ) (2) real world, voles are active and hence their move-
ment forces the owls to change their current position
where OL and OU are lower and upper bounds silently. In the present work, the movement of prey
respectively of ith owl Oi in j th dimension and is designed using probability and hence new posi-
U(0, 1) is a uniformly distributed random number in tions of owls can be obtained by following position
the range [0,1]. The fitness of each owl’s location in updating mechanism:
a forest is evaluated using an objective function and 
stored in the following matrix: Oti + β × Ici × |αV − Oti |, ifpvm < 0.5
⎡ 
 ⎤ Oi =
t+1
f1 O1,1 , O1,2 , . . . , O1,d Oti − β × Ici × |αV − Oti |, ifpvm ≥ 0.5
⎢ ⎥ (9)
⎢ 
 ⎥
⎢ ⎥ where pvm is the probability of vole movement, α is
⎢ f2 O2,1 , O2,2 , . . . , O2,d ⎥
⎢ ⎥ a uniformly distributed random number in the range
f =⎢ ⎢ .. ⎥
⎥ (3) [0, 0.5] and β is a linearly decreasing constant from
⎢ . ⎥
⎢ ⎥ 1.9 to 0. β introduces large changes initially and
⎢ .. ⎥
⎣ . ⎦ promotes the exploration of search space. As the

 algorithm progresses these variations are reduced to
fn On,1 , On,2 , . . . , On,d
encourage exploitation. The proposed method has
In the present work, it is assumed that fitness value only one user defined parameter (i.e β) whereas
of each owl’s position directly relates the intensity GA, PSO, BA etc. have large number of parametric
information received through ears. Thus best owl settings.
1576 M. Jain et al. / Owl search algorithm

Fig. 3. Inverse square law of sound intensity.

4. Experimental validation functions. For impartial evaluation, common param-

eters of all the algorithms like problem dimensions,
In this section, two experimental studies are car- population size and maximum number of function
ried out to examine the efficiency, effectiveness and evaluations are considered to be the same. The stan-
stability of OSA. The first one is a simulation study dard parametric settings of algorithms are used as
conducted on ten benchmark problems and the results given in Table 2. A valid statistical analysis is per-
are compared statistically with six state-of-the-art formed by independently executing each algorithm
optimization algorithms. The second study is hard- for 30 trial runs for each benchmark function with
ware validation on a real-time engineering design maximum 30000 number of function evaluations in
problem. each run. The recorded results are presented in Table 3
in terms of their best, worst, mean value and standard
4.1. Experimental study on benchmark problems deviation (SD). It is revealed from the results that
OSA outperforms all the other six algorithms in all
In this experimental study, a benchmark suit of cases except for function TF3. In case of TF3, none of
ten standard functions [24] are considered to validate the algorithms could find global optimum solution but
the performance of OSA. This suit includes uni- MVO shows better performance in terms of accuracy.
modal, multimodal and high-dimensional benchmark For function TF7 most of the algorithms could find

Table 2
Parametric settings of algorithms
Name of Parameter GA PSO BA FFA MVO KH OSA
Crossover fraction 0.8 - - - - - -
Selection Tournament - - - - - -
Crossover Arithmetic - - - - - -
Mutation Adaptive feasible - - - - - -
c1 and c2 - 2 - - - - -
Inertia weight - 0.9 - - - - -
Loudness - - 0.5 - - - -
Pulse rate - - 0.5 - - - -
fmin , fmax - - 0, 2 - - - -
α - - - 0.25 - - -
β - - - 0.20 - - -
γ - - - 1 - - -
WEPmax , WEPmin - - - - 1, 0.2 - -
Vf - - - - - 0.02 -
Dmax - - - - - 0.005 -
Nmax - - - - - 0.01 -
β (linearly varying) - - - - - - 1.9-0
 M. Jain et al. / Owl search algorithm 1577

Table 3
Statistical results acquired from GA, PSO, BA, FFA, MVO, KH and OSA after 30 independent runs on benchmark functions
Function GA PSO BA FFA MVO KH OSA
TF1 Best 3.9630E+06 4.8215E+08 2.3094E+10 2.0214E+03 1.2844E+05 4.2321E+03 8.1166E-60
Worst 5.7634E+06 5.7003E+09 6.9382E+10 1.2078E+04 5.0108E+05 2.4338E+05 3.4006E-43
Mean 4.9323E+06 1.9034E+09 3.8444E+10 4.7586E+03 2.7324E+05 7.5525E+04 1.1361E-44
SD 4.4253E+05 1.2098E+09 1.3038E+10 2.6260E+03 8.6995E+04 4.6589E+04 6.2081E-44
TF2 Best 4.7694E+00 1.4694E+04 4.0951E+04 1.4656E+04 9.2006E+01 9.8984E+03 1.0584E-64
Worst 5.2668E+03 7.9058E+04 5.7355E+07 3.8679E+04 3.1523E+02 7.1154E+04 4.1172E-52
Mean 5.7329E+02 3.4275E+04 2.5257E+06 2.6443E+04 1.9133E+02 3.1796E+04 2.2816E-53
SD 1.1980E+03 1.7105E+04 1.0560E+07 6.6120E+03 6.4489E+01 1.7799E+04 7.5801E-53
TF3 Best 1.1848E+00 6.3753E+02 1.8503E+04 3.0471E-01 2.4877E-01 2.7577E-01 4.8354E-01
Worst 1.7022E+00 4.5505E+03 7.0283E+04 1.0157E+00 1.3518E+00 1.0163E+00 5.0000E-01
Mean 1.4350E+00 2.0341E+03 3.8745E+04 4.0939E-01 5.2591E-01 5.6240E-01 4.9895E-01
SD 1.2544E-01 8.5452E+02 1.1319E+04 1.5596E-01 2.9414E-01 2.5248E-01 3.4469E-03
TF4 Best 2.6786E+00 8.6492E+00 1.7202E+01 7.7087E-03 1.4773E-01 1.5744E-02 8.8818E-16
Worst 3.6454E+00 1.6029E+01 1.9967E+01 2.0154E-02 2.7439E+00 3.2869E+00 8.8818E-16
Mean 3.3673E+00 1.1121E+01 1.9791E+01 1.3591E-02 1.1877E+00 1.8294E+00 8.8818E-16
SD 1.8638E-01 1.5605E+00 6.5680E-01 2.6408E-03 7.2758E-01 6.2898E-01 0.0000E+00
TF5 Best 1.6724E-01 9.9499E+00 1.9501E+02 2.4279E-03 3.1657E-01 1.1253E-02 0.0000E+00
Worst 2.9838E-01 4.2255E+01 6.8387E+02 5.9629E-03 7.3709E-01 2.6511E-01 0.0000E+00
Mean 2.4063E-01 2.1289E+01 3.9058E+02 3.9206E-03 5.3260E-01 4.1584E-02 0.0000E+00
SD 3.3169E-02 7.4842E+00 1.1472E+02 9.7419E-04 1.1226E-01 5.0919E-02 0.0000E+00
TF6 Best 3.9371E+00 9.0796E+02 1.8201E+04 1.5287E-03 1.5467E-01 1.2141E-02 5.9501E-71
Worst 6.2102E+00 4.2463E+03 6.9138E+04 6.1995E-03 5.7240E-01 1.0634E+00 2.9808E-50
Mean 5.2432E+00 2.1438E+03 3.8109E+04 3.0050E-03 3.2357E-01 1.6247E-01 1.0209E-51
SD 5.7999E-01 8.6931E+02 1.1135E+04 1.0807E-03 9.2028E-02 2.0769E-01 5.4376E-51
TF7 Best 0.0000E+00 1.0000E+00 5.5040E+03 0.0000E+00 0.0000E+00 0.0000E+00 0.0000E+00
Worst 2.0000E+00 1.9100E+02 1.9767E+04 1.0000E+00 2.0000E+00 0.0000E+00 0.0000E+00
Mean 7.0000E-01 3.7367E+01 1.0845E+04 6.6667E-02 4.0000E-01 0.0000E+00 0.0000E+00
SD 7.0221E-01 5.1696E+01 3.6443E+03 2.5371E-01 6.2146E-01 0.0000E+00 0.0000E+00
TF8 Best 1.0753E-02 6.3034E-01 6.4342E+00 1.8480E-03 5.7964E-03 1.3555E-05 3.0545E-47
Worst 2.1415E+00 4.3929E+00 6.1320E+01 4.3026E-03 3.9335E-02 3.4228E-04 5.5052E-29
Mean 8.1023E-01 2.4909E+00 2.9075E+01 2.6360E-03 2.0507E-02 1.3219E-04 2.3856E-30
SD 5.6240E-01 1.0203E+00 1.2909E+01 5.6867E-04 7.7546E-03 9.7199E-05 1.0376E-29
TF9 Best 3.1326E-06 1.1657E+00 1.1656E+01 5.2233E-07 5.6727E-05 3.1980E-03 4.0810E-96
Worst 1.0842E-01 3.6964E+01 2.9962E+02 2.5962E-06 1.0868E-03 3.2933E+00 5.2759E-77
Mean 9.1356E-03 7.2906E+00 9.3712E+01 1.3848E-06 2.0769E-04 5.2154E-01 1.7823E-78
SD 2.4277E-02 7.3451E+00 8.1922E+01 5.6504E-07 1.8630E-04 7.1552E-01 9.6285E-78
TF10 Best 1.3771E-06 4.1380E-03 3.4819E+00 3.2766E-06 2.0994E-05 5.1153E-07 5.9484E-104
Worst 6.2299E-02 4.4955E+00 2.9347E+02 1.1025E-05 1.0418E-03 1.9172E-04 2.1009E-85
Mean 4.3179E-03 4.8335E-01 1.1247E+02 5.8596E-06 3.0326E-04 2.0596E-05 1.4572E-86
SD 1.3160E-02 8.6825E-01 6.6985E+01 1.9157E-06 2.1765E-04 3.6555E-05 4.8716E-86

global optimum solutions but the performance of both In spite of the statistical analysis, two algorithms
KH and OSA are found better. Thus it is revealed that may perform equally well with no significant differ-
the performance of the proposed technique is quite ence in their results. Several non-parametric statisti-
accurate. However, apart from accuracy, an optimiza- cal tests discussed in literature [25] may be performed
tion algorithm must have fast convergence rate with to differentiate their performance. In this study, most
sufficient amount of stability to get the global opti- frequently used Wilcoxon’s test is employed to find
mum results. Therefore convergence rate analysis and the difference in significance level of two algorithms.
ANOVA test are performed and few results are pre- To conduct this test, best results on the benchmark
sented in Fig. 4. It is clear from the results that the problems for 30 independent runs of an algorithm
proposed method offers very fast convergence rate. are considered and the level of significance is con-
Moreover, the results of ANOVA test for OSA are also sidered to be 95% (α = 0.05). Table 4 shows the
found satisfactory as 25th and 75th percentiles of the recorded results of Wilcoxon’s test where ‘+’ sign
samples decline toward the global optimum solution indicates that the proposed algorithm performed sig-
with a narrow interquartile range. Quantitative anal- nificantly better than the compared algorithm. On
ysis (Table 3) also confirms the stable performance the other hand, ‘−’ sign indicates that the proposed
of OSA as the value of SD is very low. algorithm is inferior to the compared technique. It is
1578 M. Jain et al. / Owl search algorithm

Fig. 4. Convergence rate comparison and results of ANOVA for benchmark functions TF1, TF4, TF5 and TF9.

revealed from Table 4 that OSA is found significantly 5. Real-time experimental validation
better in comparison to other existing optimizers as
higher number of ‘+’ counts are recorded. Next sec- Recently nature-inspired algorithms are widely
tion presents the hardware validation of the proposed used in solving real-world problems. Therefore, to
technique while solving a real-time controller design show the applicability of proposed technique, it is
problem in process industry. used to design two degree of freedom proportional
 M. Jain et al. / Owl search algorithm 1579

Table 4
Results of Wilcoxon’s test for OSA against other six algorithms for each benchmark function with 30 independent runs (α = 0.05)
Function GA vs OSA PSO vs OSA BA vs OSA FFA vs OSA MVO vs OSA KH vs OSA
p-value win p-value win p-value win p-value win p-value win p-value win
TF1 1.6911E-17 + 1.6911E-17 + 1.6911E-17 + 1.6911E-17 + 1.6911E-17 + 1.6911E-17 +
TF2 1.6911E-17 + 1.6911E-17 + 1.6911E-17 + 1.6911E-17 + 1.6911E-17 + 1.6911E-17 +
TF3 1.6911E-17 + 1.6911E-17 + 1.6911E-17 + 3.1767E-06 + 3.5062E-05 + 7.7200E-02 −
TF4 1.6911E-17 + 1.6911E-17 + 1.6911E-17 + 1.6911E-17 + 1.6911E-17 + 1.6911E-17 +
TF5 1.6911E-17 + 1.6911E-17 + 1.6911E-17 + 1.6911E-17 + 1.6911E-17 + 1.6911E-17 +
TF6 1.6911E-17 + 1.6911E-17 + 1.6911E-17 + 1.6911E-17 + 1.6911E-17 + 1.6911E-17 +
TF7 6.1856E-07 + 1.6911E-17 + 1.6911E-17 + 4.9150E-01 − 7.9701E-04 + 1.0000E-00 −
TF8 1.6911E-17 + 1.6911E-17 + 1.6911E-17 + 1.6911E-17 + 1.6911E-17 + 1.6911E-17 +
TF9 1.6911E-17 + 1.6911E-17 + 1.6911E-17 + 1.6911E-17 + 1.6911E-17 + 1.6911E-17 +
TF10 1.6911E-17 + 1.6911E-17 + 1.6911E-17 + 1.6911E-17 + 1.6911E-17 + 1.6911E-17 +
+/- 10/0 10/0 10/0 10/1 10/0 10/2

Fig. 5. Heat Flow Experiment (HFE).

integral (2-DOF PI) controller for precise temper-

Fig. 6. Laboratory set up of HFE.
ature control of a real-time Heat Flow Experiment
(HFE). HFE is made up of a fiberglass chamber fitted to control temperature profile of the duct, while the
with blower and coil based heater at one end fol- blower voltage Vb is maintained at 5V to have a uni-
lowed by three equally spaced temperature sensors form temperature profile. Figure 6 shows the snapshot
with the other end open (Fig. 5). A tachometer is of actual experimental setup of HFE, used in this
placed on the blower to measure the speed of fan. work. The HFE apparatus is interfaced to personal
Temperature inside the chamber is measured at three computer with WinCon 5.2 software, which oper-
distinct locations by platinum temperature transduc- ates the plant in real-time in MATLAB Simulink
ers. Two analog input voltage signals called as heater environment. MATLAB R2007a is configured with
voltage (Vh ) and blower voltage (Vb ), are generated, Real-Time workshop (RTW) and Microsoft Visual
and applied to HFE through data acquisition (DAQ) C++ is used as compiler.
device to control the heater temperature and blower As discussed previously, HFE apparatus is in direct
speed respectively. The temperature inside the cham- interaction with the surroundings i.e. ambient temper-
ber changes with the change in two input voltage ature is a potential source of disturbance. Therefore,
signals and this change is measured by three tem- to get a precise temperature control with such a plant,
perature sensors. The output signal of sensors can 1-DOF control scheme i.e. conventional PI controller
be accessed from three analog input channels of the is not worthwhile. This motivates for the use of 2-
DAQ device. The range of input voltage signals (Vh DOF control scheme [26, 27] for HFE. The basic
& Vb ) is from 0V to 5V only. As the plant is in direct block diagram of the implemented control scheme is
interaction with the surroundings, it becomes crucial shown in Fig. 7. The control signal uc generated by
to control its temperature in precise manner. Hence, the 2-DOF PI controller can be defined in s-domain
the temperature of chamber at any sensor i can be as follows [26]:
described as follows: 
1
d
Ti (t) = f (Vh , Vb , Ta , xi ) i = 1, 2, 3 (10) uc (s) = Kp βr(s) − Yout (s) + {r(s) − Yout (s)}
dt Ti s
where Ta is the ambient temperature and xi is the Ki
= Kp {βr(s) − Yout (s)} + {r(s) − Yout (s)}
distance of ith sensor from the heater. In this experi- s
mental study, voltage of the heater Vh is manipulated (11)
1580 M. Jain et al. / Owl search algorithm

Fig. 7. Block diagram of 2-DOF PI based temperature control of HFE.

The control signal after passing through limiter, is

given as voltage input to the heater and hence con-
trols the temperature of HFE. The voltage input to the
blower and ambient temperature are constant during
the experimentation. It is clear from Equation (11)
that in order to generate a precise control signal,
appropriate values of controller parameters such as
Kp , Ki and β are required. In the present study, a
novel method for controller tuning is proposed in
which OSA heuristically optimizes the tuning param-
eters of 2-DOF PI controller, which leads to OSA-2PI
controller. OSA is used offline on the following iden-
tified plant model of HFE:
−0.2405s + 1.721 Fig. 8. Convergence plot of OSA for performance criterion J.
P(s) = (12)
s2 + 1.17s + 0.2
by satisfying the following performance criterion:
J = w1 IAE + w2 Os (13)
where IAE is Integral Absolute Error and Os is the
percentage overshoot of the response. w1 and w2 are
the weighing factors obtained after rigorous experi-
mentation having values 0.10 and 0.90 respectively.
Fig. 8 shows the convergence plot of OSA for the
above defined performance criterion and the opti-
mal combination of controller parameters are Kp =
0.6982, Ki = 0.0725 and β = 0.9993. It is revealed
from Fig. 8 that the convergence rate of OSA is
fast enough for solving black box optimization prob-
lem with unknown search space. The experimental Fig. 9. Temperature variations of HFE controlled by the designed
results of OSA-2PI controller are compared for set controllers for set point tracking.
point tracking against Tyreus-Luyben tuned PI (TL-
PI) controller. It is revealed from the results (Fig. 9)
that OSA-2PI controller controls temperature of HFE of IAE (Fig. 11) supports the same. Thus OSA-2PI
more precisely in comparison to conventionally tuned is far better than TL-PI controller for sharp and pre-
PI controller. This is due to the reason that proposed cise temperature control of HFE. It is revealed from
controller makes precise variations in the control sig- the experimental studies that OSA provides optimal
nal (Fig. 10) due to appropriate tuning parameters. solutions with higher accuracy and fast convergence
The quantitative analysis of the results on the basis rate.
 M. Jain et al. / Owl search algorithm 1581

References

[1] S. Mirjalili and A. Lewis, The whale optimization algo-

rithm, Advances in Engineering Software 95 (2016),
51–67.
[2] E.-G. Talbi, Metaheuristics for multiobjective optimization,
Metaheuristics: From Design to Implementation (2009),
308–384.
[3] E. Rashedi, H. Nezamabadi-Pour and S. Saryazdi, GSA:
A gravitational search algorithm, Information Sciences
179(13) (2009), 2232–2248.
[4] A.H. Gandomi, Interior search algorithm (ISA): A novel
approach for global optimization, ISA Transactions 53(4)
(2014), 1168–1183.
[5] D.E. Goldberg and J.H. Holland, Genetic algorithms
and machine learning, Machine Learning 3(2) (1988),
Fig. 10. Control effort applied by designed controllers for set point 95–99.
tracking. [6] R. Eberhart and J. Kennedy, A new optimizer using parti-
cle swarm theory. In Proceedings of the Sixth International
Symposium on Micro Machine and Human Science (MHS
’95) (1995), 39–43.
[7] M. Dorigo, M. Birattari and T. Stutzle, Ant colony opti-
mization, IEEE Computational Intelligence Magazine 1(4)
(2006), 28–39.
[8] D. Karaboga and B. Basturk, A powerful and efficient algo-
rithm for numerical function optimization: Artificial bee
colony (ABC) algorithm, Journal of Global Optimization
39(3) (2007), 459–471.
[9] A. Mucherino and O. Seref, Monkey search: A novel meta-
heuristic search for global optimization. In AIP Conference
Proceedings 953 (2007), 162–173. AIP.
[10] X.-S. Yang, Firefly algorithms for multimodal optimiza-
tion. In International Symposium on Stochastic Algorithms
(2009), 169–178. Springer.
Fig. 11. IAE comparison of designed controllers for set point [11] X.-S. Yang, A new metaheuristic bat-inspired algorithm,
tracking. Nature Inspired Cooperative Strategies for Optimization
(NICSO 2010) (2010), 65–74.
[12] A.H. Gandomi and A.H. Alavi, Krill herd: A new
bio-inspired optimization algorithm, Communications in
6. Conclusion Nonlinear Science and Numerical Simulation 17(12)
(2012), 4831–4845.
The present work proposes a novel nature-inspired [13] A. Kaveh and N. Farhoudi, A new optimization method:
Dolphin echolocation, Advances in Engineering Software
algorithm, called as OSA which is a population based 59 (2013), 53–70.
technique imitates the hunting mechanism of the owls [14] H. Shareef, A.A. Ibrahim and A.H. Mutlag, Lightning
in complete darkness. The proposed method is first search algorithm, Applied Soft Computing 36 (2015),
validated on benchmark functions and the results 315–333.
[15] S. Mirjalili, Dragonfly algorithm: A new meta-heuristic
obtained are compared with the results of various optimization technique for solving single-objective, dis-
existing optimization algorithms. It is revealed from crete, and multi-objective problems, Neural Computing and
the simulation results that the proposed OSA provides Applications 27(4) (2016) 1053–1073.
[16] S. Mirjalili, S.M. Mirjalili and A. Hatamlou, Multi-verse
more accurate and stable optimization solutions as
optimizer: A nature-inspired algorithm for global optimiza-
compared to other algorithms. Moreover, to show the tion, Neural Computing and Applications 27(2) (2016),
effectiveness of the proposed technique, it is applied 495–513.
in designing of two degree of freedom PI (OSA- [17] O. Abedinia, N. Amjady and A. Ghasemi, A new meta-
heuristic algorithm based on shark smell optimization,
2PI) controller for temperature control of a real-time Complexity 21(5) (2016), 97–116.
HFE. It is revealed from experimental results that [18] A. Askarzadeh, A novel metaheuristic method for solv-
the OSA-2PI controller is more precise for temper- ing constrained engineering optimization problems: Crow
ature control of HFE in comparison to conventional search algorithm, Computers & Structures 169 (2016),
1–12.
PI controller. Therefore, the proposed algorithm can [19] D.H. Wolpert and W.G. Macready, No free lunch theorems
be further tested to solve other complex black box for optimization, IEEE Transactions on Evolutionary Com-
optimization problems. putation 1(1) (1997), 67–82.
1582 M. Jain et al. / Owl search algorithm

[20] AS MANY EXCEPTIONS AS RULES. http://biological [25] J. Derrac, S. Garcı́a, D. Molina and F. Herrera, A practi-
exceptions.blogspot.in/2011/09/do-you-have-be-ugly-to- cal tutorial on the use of nonparametric statistical tests as a
hear-well-owls.html methodology for comparing evolutionary and swarm intel-
[21] How owls hunt in the dark. https://web.stanford.edu/ ligence algorithms, Swarm and Evolutionary Computation
group/stanfordbirds/text/essays/How Owls Hunt.html 1(1) (2011), 3–18.
[22] M. Konishi, How the owl tracks its prey: Experiments with [26] V.M. Alfaro and R. Vilanova, Model-reference robust tuning
trained barn owls reveal how their acute sense of hearing of 2DoF PI controllers for first-and second-order plus dead-
enables them to catch prey in the dark, American Scientist time controlled processes, Journal of Process Control 22(2)
61(4) (1973), 414–424. (2012), 359–374.
[23] Inverse Square Law, Sound. http://hyperphysics.phy- [27] N. Pachauri, V. Singh and A. Rani, Two degree of
astr.gsu.edu/hbase/Acoustic/invsqs.html freedom pid based inferential control of continuous biore-
[24] X.-B. Meng, X.Z. Gao, Y. Liu and H. Zhang, A novel bat actor for ethanol production, ISA Transactions 68 (2017),
algorithm with habitat selection and doppler effect in echoes 235–250.
for optimization, Expert Systems with Applications 42(17)
(2015), 6350–6364.