Portal de Periódicos da CAPES

Advanced squirrel algorithm‐trained neural network for efficient spectrum sensing in cognitive radio‐based air traffic control application

2021; Institution of Engineering and Technology; Volume: 15; Issue: 10 Linguagem: Inglês

10.1049/cmu2.12111

1751-8636

Geoffrey Eappen, T. Shankar, R. Nilavalan,

Advanced Adaptive Filtering Techniques

IET CommunicationsVolume 15, Issue 10 p. 1326-1351 ORIGINAL RESEARCH PAPEROpen Access Advanced squirrel algorithm-trained neural network for efficient spectrum sensing in cognitive radio-based air traffic control application Geoffrey Eappen, Geoffrey Eappen School of Electronics Engineering, Department of Communication Engineering, Vellore Institute of Technology, Vellore, Tamil Nadu, IndiaSearch for more papers by this authorT. Shankar, Corresponding Author T. Shankar tshankar@vit.ac.in School of Electronics Engineering, Department of Communication Engineering, Vellore Institute of Technology, Vellore, Tamil Nadu, India Correspondence T. Shankar, School of Electronics Engineering, Department of Communication Engineering, Vellore Institute of Technology, Vellore, Tamil Nadu, India. Email: tshankar@vit.ac.inSearch for more papers by this authorRajagopal Nilavalan, Rajagopal Nilavalan orcid.org/0000-0001-8168-2039 Department of Electronics and Computer Engineering, Brunel University, Uxbridge, LondonSearch for more papers by this author Geoffrey Eappen, Geoffrey Eappen School of Electronics Engineering, Department of Communication Engineering, Vellore Institute of Technology, Vellore, Tamil Nadu, IndiaSearch for more papers by this authorT. Shankar, Corresponding Author T. Shankar tshankar@vit.ac.in School of Electronics Engineering, Department of Communication Engineering, Vellore Institute of Technology, Vellore, Tamil Nadu, India Correspondence T. Shankar, School of Electronics Engineering, Department of Communication Engineering, Vellore Institute of Technology, Vellore, Tamil Nadu, India. Email: tshankar@vit.ac.inSearch for more papers by this authorRajagopal Nilavalan, Rajagopal Nilavalan orcid.org/0000-0001-8168-2039 Department of Electronics and Computer Engineering, Brunel University, Uxbridge, LondonSearch for more papers by this author First published: 05 February 2021 https://doi.org/10.1049/cmu2.12111Citations: 3AboutSectionsPDF ToolsRequest permissionExport citationAdd to favoritesTrack citation ShareShare Give accessShare full text accessShare full-text accessPlease review our Terms and Conditions of Use and check box below to share full-text version of article.I have read and accept the Wiley Online Library Terms and Conditions of UseShareable LinkUse the link below to share a full-text version of this article with your friends and colleagues. Learn more.Copy URL Share a linkShare onFacebookTwitterLinkedInRedditWechat Abstract In the current scenario, there is a drastic increase in air traffic. The air to ground communication plays a crucial role in the air traffic control system. There is a limited spectrum available for aircraft to establish a connection with the Air Traffic Controller (ATC). With air traffic growth, the available spectrum is getting more congested. This paper proposed an Advanced Squirrel Algorithm (ASA)-trained neural network (NN) for efficient spectrum sensing for cognitive radio-based air traffic control applications. ASA is a novel metaheuristic-based training algorithm for an NN. With the proposed algorithm, it is possible to dynamically allocate the unused spectrum for air to ground communication between aircraft and ATC. The quantitative analysis of the proposed ASA-NN-based spectrum sensing is done by comparing it with the existing metaheuristic-based NN training algorithms, namely, particle swarm optimization Gravitational Search Algorithm (PSOGSA), particle swarm optimization (PSO), gravitational search algorithm (GSA), and artificial bee colony (ABC). Simulation-based evaluation shows that the proposed ASA-NN is capable of efficiently detecting the spectrum holes with high convergence rate as compared to PSOGSA-, PSO-, GSA-, and ABC-based algorithms. 1 INTRODUCTION The development in the aviation sector has resulted in the tremendous growth of the wireless communication technologies governing the air traffic control. Wide range of wireless technologies are employed to assist the on-ground surveillance and navigation of airplanes while taking off, landing, and en-route. The employed wireless devices operate at different radio channels. The radio channels in Very High Frequency (VHF) and High Frequency (HF) bands are mainly used for enabling the link between air traffic control stations and aircraft. The VHF spectrum for wireless communication between the Air Traffic Controller (ATC) and aircraft has a bandwidth of 19 MHz ranging from 118 to 127 MHz [1]. The spectral spacing of each band is 25 kHz, resulting in a total of 760 radio channels. As the flight traffic is tremendously increasing year on year, so the number of aircraft tuning to a particular station is also increasing immensely. The problem arises when the pilots of different aircraft tune the controller frequency at the same time, thus leading to frequency congestion and also the pilot may accidentally override others. This situation can lead to incorrect information delivered to the aircraft. In addition to that, different applications correspond to aircraft communication, which further leads to the congestion of radio channels, specifically in the regions of highly crowded airports. Therefore, it is important to utilize the radio spectrum available for aircraft communication. Concerning the above discussion, the recent studies suggest that the industrial, scientific and medical bands are highly congested [2, 3] . In contrast, a significant portion of the licensed radio spectrum is vacant and is used inefficiently [2]. In the radio spectrum allocated for aircraft communication, only around 12.5 % is effectively utilized [4]. In the current scenario, there exist the problem of spectrum scarcity, and at the same time, there is also the situation of inefficient spectrum utilization [5]. With the continuous increase in the air traffic over the last decade, the air traffic management system has predicted that the air traffic will reach its peak by 2020 [6]. The high traffic would result in more congested bandwidth for data transmission between aircraft and ATC. Moreover, the audio data transmission from aircraft to ATC is highly delay-sensitive, with the unavailability of proper bandwidth that can cause continuous intrusion to the transmission, which will further enhance the delay issue. Such a problem calls for the research work with emphasis made on the need for wireless communication technology capable of performing dynamic spectrum access and meeting the future requirements of aviation technology with high efficiency and precision [7]. The possible solution to the problem of spectrum scarcity and providing dynamic flexibility to the wireless communication technology employed for air traffic control is the cognitive radio network (CRN) [6]. A CRN can sense its surrounding radio environment and adjust accordingly [5]. The Federal Communications Commission's report in [8] has stated the use of the underutilized licensed spectrum to increase the effective utilization of the frequency spectrum. The CRN, with its ability to sense and adapt, can opportunistically access these underutilized licensed spectra without causing interference to the Primary Users (PU)/Licensed Users (the users having the license to utilize the licensed spectrum). The CRN makes this underutilized licensed spectrum also known as spectrum hole to the secondary users (SUs) for opportunistic access. The first and one of the most important working phases of a CRN is the spectrum sensing [9]. Through different spectrum sensing techniques, the USs find the spectrum holes and proceed further with the process of a CRN as the process of spectrum sensing is extremely vital. So, a novel Advanced Squirrel Algorithm (ASA)-trained neural network (NN) is employed for efficient spectrum sensing to improve the effectiveness of the CRN. An effective CRN would result in improved spectral and bandwidth efficiency. 2 RELATED WORKS The detection of the spectrum holes by cognitive radio (CR) devices and utilizing it opportunistically enhances the spectral efficiency and the channel bandwidth [6]. The spectrum sensing plays the pivotal role in the detection of vacant and thus it is the essential component of CR network. Conventional spectrum sensing includes intensive techniques like the Matched Filter (MF) [10], cyclostationary detector [5], and eigenvalue-based detector [11], as well as the simple method like Energy Detector [12]. The simplest spectrum sensing approach has weak performance under low signal-to-noise ratio (SNR) and are not efficiently able to detect the spectrum holes [5, 9, 13]. The cyclostationary detector and MF are highly efficient in detecting the spectrum holes, but the cyclostationary detector requires long sensing time to have high detection probability [6]. For a fixed frame period, a longer sensing time decreases the transmission time and thus reduces the overall opportunistic throughput. The MF technique requires priori knowledge of the signal for efficient detection. In the absence of accurate information of PU, the performance of the MF degrades [5]. Another important drawback associated with the MF is that it requires dedicated receiver for each PU signal type [14]. The drawbacks associated with conventional spectrum sensing technique calls for the necessity of efficient spectrum prediction by CR network. With intelligent prediction-based spectrum sensing it is possible for CR network to reduce the sensing time and improve energy efficiency by efficient prediction of channel state, thus skipping spectrum sensing for some time [15-19]. The NN forms the base for the intelligent prediction scheme. To maintain a trade-off between spectrum sensing efficiency and its complexity, an NN-based spectrum sensing is successfully employed in [16, 20, 21]. The Conventional-NN which is based on gradient descent-based back propagation (BP) method are prone to converge to local optima [22, 23] and has slow convergence rate [24]. Studies have confirmed that metaheuristic-based optimization technique can improve the efficiency of NN [25-28]. Because of the no free lunch theorem [29], different metaheuristic optimization is suited for the different objective functions. Selecting the proper optimization technique for improving the performance of the artificial neural network (ANN) is very crucial as the entire CRN working is dependent on it . The popular swarm-based optimization scheme like particle swarm optimization (PSO), artificial bee colony (ABC) Algorithm, genetic algorithm (GA), grey wolf optimization (GWO) and ant colony algorithm (ACA) lack proper trade-off between their exploration (Global Search) and exploitation (Local Search) abilities [29, 68]. The PSO lacks proper convergence ability, whereas ACA and ABC lack in exploitation [30, 67]. The GA tends to get stuck to the local best solution instead of finding the global best [31]. Such problems call for an efficient optimization scheme that has a proper trade-off between its exploration and exploitation abilities, which has a good convergence rate and can overcome local optima and converge towards global best. Therefore, advanced flying squirrel search-based algorithm is implemented and employed. The prevailing spectrum sensing studies were more focused on binary hypothesis, i.e. temporal spectrum sensing [32-34]. The works in [9, 35] have showed that temporal cooperative spectrum sensing has better performance than the temporal non-cooperative spectrum sensing. The 3D-spatial spectrum sensing is an emerging field [34, 36-39] that gives better insight about real-time implementation of CR network. The work in [34] considered 3D spatial and temporal spectrum sensing using conventional energy detector and it has its limitation in low SNR values. The 3D-spatial and temporal spectrum sensing is carried out for non-cooperative and cooperative scenario using ASA-trained ANN-based efficient spectrum prediction. The proposed technique is compared with the existing metaheuristic-based optimization technique for ANN in spectrum sensing. The major contributions of this paper are stated as under: Novel ASA-based technique for the weight optimization in NN to enhance its prediction and efficiency. ASA-NN for efficient spectrum sensing by performing effectual spectrum status prediction. 3D non-cooperative spectrum sensing (3D NCSS) scheme and temporal cooperative spectrum sensing scheme for air traffic control. National Instrument (NI) Universal Software Radio Peripheral (USRP)-based real-time implementation of the proposed technique for temporal cooperative spectrum sensing scheme. 3 SYSTEM MODELLING With the affordable airlines coming into the market, the air traffic is increasing with each passing year. Such increase in air traffic calls not only for infrastructural development but also requires high technical advancement in the field of wireless communication governing the air traffic system. As the flight traffic is enormously increasing, so, the number of aircraft tuning into a station is also increasing immensely. The problem arises when the pilots of different aircraft tune the controller frequency at the same time, thus leading to frequency congestion and pilot may also accidentally override others. This situation can lead to incorrect information delivered to the aircraft. To overcome such spectrum congestion problem, an efficient spectrum sensing-based CRN is proposed for air traffic control. The spectrum sensing efficiency is improved by incorporating novel metaheuristic algorithm ASA-trained NN. The Air to Ground (A/G) communication frameworks are basic for the aircraft's secure routing. In this way, the progress to the CR-based systems ought to be finished with most extreme consideration. While designing the CRN for the A/G communication it should make sure that its effect should be minimal on the existing A/G communication infrastructure. It is to be noticed that the existing A/G communication frameworks still depend on analogue transmission frameworks. In this work, an efficient spectrum sensing technique has been proposed for the effective CR-based Air Traffic control. This paper proposes Phase I and II for the air traffic control. For Phase I, aircrafts perform the spectrum sensing with the help of ASA optimization algorithm. As ASA behaviour which employs gliding technique to find the optimal solution depicts the landing the process of the aircraft. Moreover, the seasonal constraints in the ASA is utilized and mapped for efficient spectrum sensing by the aircraft during different weather conditions. So, ASA is efficient in assisting aircraft to detect the vacant spectrum holes during the landing process. In Phase II, the cognitive radio sensors/spectrum sensing sensors (CRs/SSs) are deployed in the ground level performs spectrum sensing aided with ASA-trained NN. The Conventional-NN technique uses gradient descent-based BP algorithm for training and has loopholes like getting stuck to local minima and large convergence time. Whereas, ASA is a squirrel-based optimization technique in which the population of squirrels moves around the search space constituted by the optimization problem in search of an optimal solution. Squirrel position varies during the process of search based on the best solution position so obtained. The ASA has excellent balance between its exploration and exploitation abilities, moreover it has fast convergence rate. So, ASA is used as the possible alternative for optimizing the weights of NN. The ASA is employed to optimize the weights of NN and thus to minimize the error in the prediction of spectrum holes. The information obtained by CRs are transferred to the ground CR base station, which then makes the final decision on the vacant spectrum availability. The ASA-NN-based ATC has training and working phase, during the training phase network is trained using ASA and in the working phase-trained network is employed for the spectrum hole detection. The illustrative figures depicting Phase I and II is shown in Figures 1 and 2, respectively. The blocks in Figure 1 comprise of CR base station, which is responsible for making the final decision on the presence and the absence of the PUs for ground to air communication. Blocks ATC and Aircrafts resemble incoming aircraft communicating with the ATC. The working in Figure 1 is explained as: in the case of allocated frequency band congestion for data transmission between Aircraft and ATC, the CR technique is employed. For the proposed model, aircraft perform spectrum sensing and establish link to ATC via the spectrum holes, simultaneously CRs deployed in grounds perform spectrum sensing so as to find the spectrum holes for ground to air data transmission. If the allocated frequency bands are not congested then the routine data transmission is carried out in the allotted frequency spectrum. FIGURE 1Open in figure viewerPowerPoint Flow diagram of the proposed CR-based ATC (Phase I) FIGURE 2Open in figure viewerPowerPoint Proposed CR-based ATC (Phase II) 4 MATHEMATICAL MODELLING The aircraft and CRs perform periodic spectrum sensing as aircraft approaches ATC. The schematic representation of the frame structure for spectrum sensing and data transmission is as shown in Figure 3. For the data transmission, orthogonal frequency division multiplexing (OFDM) scheme is employed with fast fourier transform (FFT) of length 64 . Each subcarrier has the bandwidth of 10 kHz and total bandwidth of 0.5 MHz with 50 subcarriers. Each 25 subcarriers are used for air to ground and ground to air transmission. FIGURE 3Open in figure viewerPowerPoint Proposed CR-based ATC (Phase II) The conventional spectrum sensing considers the binary hypothesis for the detection as in Equation (1) [6]: H 0 : y i ( n ) = N i ( n ) { Hypothesis 0 ( PU Absent ) } H 1 : y i ( n ) = x i ( n ) + N i ( n ) { Hypothesis 1 ( PU Present ) } , (1)where x i is the PU signal which can be modelled as the zero mean complex Gaussian with the power σ x 2 , N i is the zero mean complex AWGN (additive white Gaussian noise) with the power as σ N 2 [34], n = 1 , 2 , … m , m is the total sample number. i = 1 , 2 , … K ( K is the total number of aircraft arriving at the airport at the same time). j = 1 , 2 , … L ( L is the total number of SSs/SUs or cognitive radio users (CRu) operating in a cooperative manner). According to the binary hypothesis, a spectrum sensing can accept or reject samples and infer the presence and the absence of the PU based on the detection threshold as [40] E i = 1 K ∑ i = 1 K | y i ( n ) | 2 > H 1 < H 0 λ , here λ is the detection threshold. Energy sample value greater than the threshold would be considered as the PU present and vice versa for the energy sample value lower than the threshold. The x i ( n ) is the nth sample sensed by the ith aircraft. Similar hypothesis can be employed for the jth SS. The two phases of spectrum sensing is consideredand and is operated simultaneously for this research. The binary hypothesis holds good for the ground-based CRu performing cooperative spectrum sensing. For the aircraft to perform spectrum sensing, it is necessary that it performs spectrum sensing when it is near to ground (looking for ground clearance to land), so that it can be within the coverage area of some PU transmitter. It is because, if aircraft performs spectrum sensing outside the range of PU then there will be high false alarm. Thus, the aircraft can cause interference to the PU's transmission while performing down link data transmission to the ATC. From Figure 4, it can be seen that ( D 0 − D 1 ) is the region in which aircraft is outside the coverage area of PU and D 1 represents the radius of PU transmission range. The hypotheses H 0 and H 1 for the aircraft approaching an ATC can be validated only when it is inside the radius D 1 . The formulation of the spatial hypothesis can be written as in Equation (2): B 0 D 1 ∩ H 0 B 1 D 1 ∩ H 1 B 2 D 0 ∩ H 0 B 3 D 0 ∩ H 1 (2)where B 0 = D 1 ∩ H 0 represents that the aircraft is within the range of PU Transmitter–Receiver (Tx–Rx) coverage and the Hypothesis H 0 holds true (i.e. PU is inactive). Thus, aircraft have the spectrum access opportunity. The B 1 = D 1 ∩ H 1 indicates that PU is active and aircraft is within the PU Tx–Rx coverage. FIGURE 4Open in figure viewerPowerPoint Schematic representation of PU coverage for aircraft spectrum sensing From Equations (1) and (2), the modified hypothesis can be postulated as: B 0 : y i ( n ) = N i ( n ) 0 ≤ d i ≤ D 1 B 1 : y i ( n ) = x i ( n ) + N i ( n ) 0 ≤ d i ≤ D 1 B 2 : y i ( n ) = N i ( n ) D 1 ≤ d i ≤ D 0 B 3 : y i ( n ) = N i ( n ) D 1 ≤ d i ≤ D 0 (3)The condition B 2 and B 3 cannot be employed as the spectrum opportunity for the aircraft. Only B 0 is the available spectrum opportunity (i.e. the Aircraft is within the PU coverage and PU is inactive). 4.1 3D NCSS by aircraft The two phases of spectrum sensing is considered simultaneously. The Phase I spectrum sensing performed by the aircraft is termed as the 3D NCSS because each aircraft performs spectrum sensing without any cooperation with the other aircraft and also the spectrum sensing is performed during descent, so three-dimensional space has been considered. Figure 5 shows the schematic representation of spectrum sensing performed during the 3D NCSS scenario/Phase I. FIGURE 5Open in figure viewerPowerPoint Schematic representation of spectrum sensing for the case 3D NCSS In Phase II, the SSs are deployed in the ground near ATC performs spectrum sensing in cooperative manner (i.e. spectrum sensing data of each SUs are beamed towards the fusion centre (FC), which makes the final decision on spectrum occupancy). Therefore, Phase II is termed as the cooperative spectrum sensing scheme. For Phase I, to evaluate the performance of spectrum sensing the probability of detection and the probability of false alarm is employed. Based on Equation (3), the probability of detection and the false alarm can be calculated as in Equation (4): P f , i N c ( k ) = P ( B 1 | B 0 ) P d , i N c ( k ) = P ( B 1 | B 1 ) , (4)where, P f , i N c ( k ) is the probability of the false alarm of the ith aircraft under non-cooperative spectrum sensing scheme for the kth sensing period. The P d , i N c ( k ) is the probability of the detection of the ith aircraft under non-cooperative spectrum sensing scheme for the kth sensing period. For each aircraft as well as for the SSs, energy detector is employed for obtaining the energy samples while performing the spectrum sensing. In case of cooperative spectrum sensing scheme performed by the SUs, the samples are used for training the ASA-based NN at CR base station/ (FC). The test statistics for energy detection-based spectrum sensing is written as in Equation (5): e i ( k ) = 1 m ∑ n = 1 m x i ( n ) 2 . (5)For large number of samples, the term e i ( k ) as per the Central Limit Theorem (CLT) can be approximated as a Gaussian random variable for the Hypotheses H 0 and H 1 of Equation (1)[12, 41]. e i ( k ) ∼ N σ N 2 , σ N 4 m : H 0 N ( 1 + S N R T V r , i ) σ N 2 , ( 1 + S N R T V r , i ) 2 σ N 4 m : H 1 , (6)where S N R T V r , i is the received SNR of the PU (TV received power) at the ith aircraft. Based on the approximation in Equation (6), the P f , i N c ( k ) and P d , i N c ( k ) can be estimated as in Equations (7) and (8), respectively: P f , i N c ( k ) = Q λ − σ N 2 σ N 4 m , (7) P d , i N c ( k ) = Q λ − ( 1 + S N R T V r , i ) σ N 2 ( 1 + S N R T V r , i ) 2 ∗ σ N 4 m , (8)where λ is the detection threshold to mark the difference between the PU's presence and absence. 4.2 Cooperative spectrum sensing by SSs/SUs/CRu The energy detection samples from SUs are transferred to the FC. The FC with the help of ASA-NN makes the correct prediction about the vacant spectrum. The detailed working of ASA-NN-assisted spectrum hole prediction by FC is explained in Section 6. The cooperative spectrum sensing scheme employed by the SSs is explained as below: At FC, the linearly weighted energy values from all SUs are obtained as in Equation (9): e c s s = ∑ j = 1 L W j e j , (9)where e j is the energy detected by the jth SU, W j is the weight coefficient of the jth SU and it is calculated as shown in Equation (10): W j = S N R T V r , j ∑ l = 1 L S N R T V r , l , (10)where S N R T V r , j is the received SNR at the jth SU. In Equation (10), l ≠ j and l corresponds to other SUs. The weighted energy value e c s s can be approximated as Gaussian random variable as in Equation (11) for the Hypotheses H 0 and H 1 in Equation (1): e c s s ∼ N ( μ 0 , σ 0 ) N ( μ 1 , σ 1 ) , (11)where μ 0 = ∑ j = 1 L W j σ n 2 σ 0 = ∑ j = 1 L W j σ n 4 m μ 1 = ∑ j = 1 L W j ( 1 + S N R T V r , j ) σ n 2 σ 1 = ∑ j = 1 L W j ( 1 + S N R T V r , j ) 2 σ n 4 m . (12)Using Equations (11) and (12), the probability of false alarm and the probability of detection can be calculated as Equations (13) and (14): P f , j c s ( k ) = Q λ − μ 0 σ 0 , (13) P d , j c s ( k ) = Q λ − μ 1 σ 1 . (14)In general, the opportunistic throughput of the SU performing data transmission in the absence of PU can be calculated as in Equation (15) T o p t = P H 0 F t − S t F t ( 1 − P f ) log 2 ( 1 + S N R ) , (15)where T o p t denotes the opportunistic throughput, F t and S t are the frame time and the sensing time, respectively. The P H 0 denotes the probability that PU is inactive, generic false alarm probability, and the SNR of SU (while performing opportunistic data transmission) is represented as P f and S N R , respectively. In Figure 6, the hemispherical dome represents the PU coverage and within this coverage CRs are deployed as represented by blue circles and the approaching aircraft towards ATC is represented by black dots with coordinates. FIGURE 6Open in figure viewerPowerPoint Simulation-based representation of the proposed model 5 CONVENTIONAL FLYING SQUIRREL SEARCH ALGORITHM Flying Squirrel Search Algorithm (FSSA) is introduced by Jain et al. [42] and it is based on the effectual foraging behaviour of flying squirrels. The search for the food sources depends on weather, type of trees, and the presence and the absence of the predators. The flying squirrels are more active in warm weather than cold and are capable of obtaining better and abundant food source during that period. The flying squirrels consumes two types of food sources, i.e. Hickory nuts (Hickory tree) and Acorn nuts (Acorn tree). The Acorn nuts which are abundantly available during warm weather is immediately consumed by the squirrels whereas the Hickory nuts are stored for winter. During winter flying squirrels are inactive, so it is difficult for them to obtain new foods. Therefore, storing Hickory nuts is one of the prime motives so as to withstand the extreme weather condition. 5.1 FSSA initialization The algorithm starts with the initialization of parameters: Maximum Iteration = i t max , Size of squirrel population = M, Decision variable count/Number of dimensions = D, Probability of predator's presence = P r p , Scaling factor = s f (value ranges between 16 and 37), Gliding distance constant = G c , and Decision variable bounds. 5.1.1 Random initialization of flying squirrels ( F s ) For the population count ′ M ′ and with the upper and lower bounds as F s u and F s l , respectively, then each flying squirrels can be randomly initialized using Equation (16): F s I , z = F s z l + r a n d ( ) × ( F s z u − F s z l ) , (16)where F s I , z represents Ith flying squirrel in the zth dimension, r a n d ( ) generates random number between 0 and 1, I = 1 , 2 , 3 … M , and z = 1 , 2 , 3 … D . The fitness of a squirrel in a particular dimension represents its location and the quality of the solution. In the FSSA, optimal solution location is mapped as optimal food source (Hickory nuts) which is termed as the location of squirrel at Hickory tree. The next best solutions are termed as the location of squirrel at Acorn tree. The normal solutions (Acorn nuts) are termed as location of squirrel at normal food source, i.e. Normal tree. After random initialization of the flying squirrels, the squirrels with maximum fitness value are noted as to be on the Hickory nut tree. The next few best solutions are termed as squirrels' locations on the Acorn nut trees. The rest of the squirrels are considered to be on Normal tree. While foraging it is important to consider the probability of predator's presence ( P r p ). 5.1.2 Movement of flying squirrels towards new solutions Squirrels on Acorn tree tends to move towards Hickory nut tree, i.e. the best solution found so far using Equation (17):