A Hybrid Delay Aware Clustered Routing Approach Using Aquila Optimizer and Firefly Algorithm in Internet of Things

Abstract

Protocols for clustering and routing in the Internet of Things ecosystem should consider minimizing power consumption. Existing approaches to cluster-based routing issues in the Internet of Things environment often face the challenge of uneven power consumption. This study created a clustering method utilising swarm intelligence to obtain a more even distribution of cluster heads. In this work, a firefly optimization method and an aquila optimizer algorithm are devised to select the intermediate and cluster head nodes required for routing in accordance with the NP-Hard nature of clustered routing. The effectiveness of this hybrid clustering and routing approach has been evaluated concerning the following metrics: remaining energy, mean distances, number of hops, and node balance. For assessing Internet of things platforms, metrics like network throughput and the number of the living node are crucial, as these systems rely on battery-operated equipment to regularly capture environment data and transmit specimens to a base station. Proving effective, the suggested technique has been found to improve system energy usage by at least 18% and increase the packet delivery ratio by at least 25%.

1. Introduction

It’s possible to think of the Internet of Things (IoT) as a decentralized, autonomous ecosystem made up of a vast collection of disparate, unconnected devices [1]. The monitoring system, medicine, the military, and e-learning are just some fields that have benefited from recent technological developments in the realm of actual wireless infrastructure applications [2,3]. Low energy levels in the platform’s nodes are a significant problem that needs fixing [4]. Network performance can be enhanced by more evenly distributing electricity among nodes [5].

Some IoT service-providing nodes are fixed in the site and can’t be updated or recharged, yet they still need to operate on a restricted energy budget [6]. Receiving and transmitting data causes the most tremendous drain on a network’s power sources and is proportional to the distance among nodes [7]. Therefore, developing a dependable and low-power routing method is a primary goal in order to extend the lifespan of the network, which is one of the most critical routing concerns in the IoT [8]. The primary idea behind the clustered routing algorithm is to break the system into more manageable components or clusters. There has been a lot of study into the best ways to optimize clustered routing to lessen the energy needs of IoT devices and lengthen the lifespan of networks. Cluster Member (CM) nodes may report data to the Cluster Head (CH) node [9]. Single-hop or multi-hop connection is used to provide the aggregated and cleansed information to the Base Station (BS) [10].

One of the Non-deterministic Polynomial-time hard (NP-hard) challenges is the clustering of IoT devices to maximize system longevity [11]. Since it cannot be handled using deterministic methods in polynomial time, this problem class is impractical. Swarm Intelligence (SI) methods, however, such as Artificial Bee Colony (ABC) [12], Whale Optimization (WO) [13], Genetic Algorithm (GA) [14], Gray Wolf Optimizer (GWO) [15], and Harmony Search (HS) [16], could get close to the result in an acceptable amount of time. Aquila Optimizer (AO) is the most recent strategy for SI, announced in 2021. This method mimics distinct aquila hunt strategies for various prey types. The hunting techniques for fast-moving prey indicate the algorithm’s capacity towards global exploration, while the hunting tactics for slow-moving prey highlight its capacity at local exploitation [17]. The AO algorithm has powerful global exploration ability, excellent searching efficiency, and rapid convergence speed [18]. Considering to the limited amount of time after the algorithm’s proposal, little study has been conducted on AO.

Through simulating a population of fireflies, simulation their actions, and assigning values based on the fitness of every firefly’s position to represent the amount of firefly pigments, the Firefly Algorithm (FA) can find the best possible answer to a problem by repeatedly running the method and upgrading the worms’ positions [19]. In reality, in each repetition of the strategy, the pigment-refreshing stage and the motion stage are the two most important stages. The fireflies in the area go to the fireflies that have more pigments [20]. As a result, the group improves over time as a whole [21]. This paper proposes a Hybrid Aquila Optimizer and Firefly Algorithm (HAOFA) routing model, wherein the AO approach is used to select the best CH node to achieve energy-efficient inter-cluster interactions, and the FA method is liable for finding the best intermediate node for efficient intra-cluster interactions. In a nutshell, these are the various contributions:

• Recreating and emphasizing the AO algorithm’s distinctive features;
• Using the AO approach, we provide a solution to the IoT’s cluster-based routing problem;
• Highlighting the salient features of the FA algorithm and discussing them in detail;
• Multi-hop routing using the FA algorithm for improved power usage and throughput;
• Investigating how efficient clustered routing is.

Following is the remainder of the article’s information. In Section 2, a look at some of the prior art in the field of clustering protocols is given. In Section 3, the procedure of HAOFA is deeply discussed. The effectiveness of the HAOFA methodology is evaluated in Section 4. Section 5 presents the last thoughts and plans for further development.

2. Literature Review

Over the past few years, a plethora of routing algorithms have emerged, each with its own unique power requirements [22]. These strategies concentrate on shortening the distance between communicating nodes to cut down on power consumption and avoid energy hole issues [23]. An important benefit that has been identified is that the nodes in this approach do not all carry out the same task [24]. They employ a hierarchical layout, segmenting the system into many groups.

Hriez, Almajali [25] have developed a clustering algorithm employing a trust model that recognizes untrusted nodes operating power and data trust. In addition, the Fitness Function (FF) is utilized to pick the CH nodes from a list of reliable nodes. Inside this work, the FF is derived using the residual energy of the nodes, the length across each node and the BS, the densities, and power. During the designing of clusters, the population of CH nodes was considered to guarantee that the burden on all CH nodes was evenly distributed. In order to prolong the platform’s lifespan, this clustering method incorporated the beneficial properties of stochastic fractal search optimization. This method also became unsuccessful for heterogeneous networks.

Also, Amutha, Kannan [26] have developed a cluster manager-based CH node mechanism to solve power usage and workload balancing. This method utilized two ideas, cluster manager and CH node, with the cluster manager observing the activities of other nodes. The message was transferred across IoT nodes by the CH node. Whenever the amount of power in the CH node became depleted, the cluster management chose the fresh CH node with the highest level of power and saved the new and old CH node operations. In addition to conserving energy and bandwidth, such an approach ensured reliable throughput. Unfortunately, some of this system’s flaws were insufficient scalability and a disregard for data protection.

Saxena and Mehta [27] have proposed a method for building a hierarchy architecture and energy savings for the IoT wherein clusters were constructed using a fuzzy multi-criteria decision strategy; subsequently, the CH nodes were ideally chosen based on the hierarchy analysis procedure. During intra-cluster and inter-cluster connections, the penguin search method is applied. This method enhances production while concurrently reducing power use. The strategy’s primary shortcoming would be its disdain towards Quality of Service (QoS).

Moreover, Ahmad, Ikram [28] have developed a clustering-aware optimization method using the features of the GA and the honey bee approach. Under this method, bees display a fundamental clustering structure whose resilience and load distribution dictate their fitness. This facilitates the population’s management of topological adjustments and high-quality harvest options. This approach performs the options in terms of clustering time, and the number of clusters extends the network’s life span and produces a lower overhead than the competitors. Even though CH nodes are dispersed irregularly, this leads to significant energy usage and overload differences.

Xiuwu, Ying [29] have created an IoT clustered routing algorithm implementing mixed genetic taboo search. Inside the initial stage, the CH node of the clusters was selected by considering the residual power of every node and its distance from the BS in order to optimize the threshold function in the cluster. To preserve the total energy balancing of the system, additional nodes are joined to the cluster at the lowest possible cost based on the Cost Function (CF) in the subsequent phase. The findings show that this method improves the lifespan of the network and spends fewer power in a better-balanced way. This approach has numerous disadvantages, such as a lengthy network latency and inadequate data security.

In addition, Poluru and Kumar R [30] have published an optimization technique for finding CH nodes based on the remaining power, the length of the BS, and the node-to-node communication channel. The suggested model conserves energy, prolongs the platform’s lifetime, and selects the proper channel. This technique has weak convergence, which is a negative.

Jayalakshmi, Sridevi [31] have introduced a hybrid HS methodology with an ABC in order to achieve effective CH node allocation. This action was taken to maintain steady energy consumption while extending the platform’s lifetime. The above technique combines the global optimization capability of the HS algorithm with the local exploitation potential of the conventional ABC algorithm in order to gather a considerable number of CH nodes for the sake of conserving energy and extending the lifespan of the network. Thus, it incorporated the benefits of the harmony-adjusting function for triggering the mechanism for generating dynamic findability during the process of CH node selection to prevent the worst contenders from being selected as CH nodes. This was done to guarantee that the most qualified people were chosen as CH nodes. The proposed approach prolongs the average network’s lifespan and keeps power consumption balanced. Consequently, the cost is a critical component that must be paid specific care inside the clustering model.

Finally, Agrawal and Pandey [32] have proposed a hybrid HS and fuzzy logic technique to extend the IoT network’s lifetime. The stated method results in unequal clusters. Utilizing an HS-based approach, the efficiency of the proposed method was demonstrated. The presented method achieves higher efficiency in a variety of contexts. It boosts the system’s durability. This approach essentially performs poorly in the presence of network dynamics.

Table 1. Comparison of the different methods highlighted.

Mechanism	Approach	Advantage	Weakness
Hriez, Almajali [25]	Trust-based technique for clustering that can identify compromised nodes.	✓ Extending the lifespan of the network	Failing to account for heterogeneous network topologies
Amutha, Kannan [26]	Electing the CH node depending on the cluster manager	✓ Low energy consumption ✓ Low bandwidth usage ✓ High overall throughput	Scalability constraints Failing to account for heterogeneous network topologies
Saxena and Mehta [27]	Multi-criteria decision-making with fuzzy logic	✓ High conserving energy ✓ Improving throughput	Poor QoS
Ahmad, Ikram [28]	Hybrid optimization method using GA and honey bee approach	✓ Minimal overhead ✓ Boosting network lifespan ✓ Minimal clustering time	Disparity in energy usage Extra overload
Xiuwu, Ying [29]	Clustering mechanism using a hybrid form of genetic tabu search	✓ Reducing power consumption ✓ Extending network lifespan	Long delay Poor data security
Poluru and Kumar R [30]	Fruit fly optimization method for picking CH nodes	✓ Extending system longevity ✓ Conserving energy	Weak convergence
Jayalakshmi, Sridevi [31]	Combining the HS technique with the ABC strategy for clustering	✓ Increasing network longevity ✓ Balancing energy consumption	Steep cost
Agrawal and Pandey [32]	Algorithm for clustering based on fuzzy logic and HS	✓ Boosting system lifespan	Unfit for topological alterations

shows a side-by-side comparison of the significant advantages and disadvantages of the analyzed strategies. Power consumption and low throughput are the main problems with the aforementioned methods, although they are essential to the IoT architecture. The suggested solution makes an effort to solve these problems. This method not only addresses crucial and sensitive routing difficulties but also analyzes the energy consumption, Packet Delivery Ratio (PDR), End-to-End (E2E) delay and throughput.

3. The Proposed Mechanism

Following is a more in-depth explanation of the strategy that will be employed. First, the network model is built, then the problem statement is explained, and then the AO and FA are described in detail.

3.1. System Model

It is assumed that BS and N IoT devices exist in the ecosystem. Each IoT node has its own CH node through which it communicates with the BS. Below are certain commonly held assumptions about logic:

• The IoT devices are scattered around in a random pattern.
• Starting power is the same for all IoT nodes.
• BS is installed constantly in the core of the system.
• Endless battery life and storage space are two of the BS’s many perks.
• Each and every cluster has exactly one node in the CH.
• All connections have plenty of available bandwidth.

3.2. Problem Declaration

In order for data to be sent from the CH nodes to the BS, a multi-hop communication must be established. It is the job of CM nodes to relay information to CH nodes in a cluster. The meta-heuristic approaches are more efficient computationally because they explore broadly for the best solution. Because of these traits and AO skills, it is capable of picking the best alternatives. Additionally, FA is adaptable to problems involving both continuous and discrete parameters. System reliability is enhanced, and assets are used to their full potential when the AO method is used to pinpoint the CH nodes. By utilizing FA, the best intermediate node can be selected to increase the stability and durability of the platform.

3.3. AO Algorithm

The most newly invented SI method is known as the AO. Aquila employs four distinct hunting techniques to capture a variety of animals. In addition, Aquila may rapidly transition among multiple prey-specific hunt techniques before launching an assault, including its velocity, claws, and strong feet.

Stage 1: Expanded exploration (high soar with a vertical descent).

In this approach, whenever the Aquila establishes the position of its prey, it dives vertically after soaring high above the surface and thoroughly surveying the searching space. These Equations (1) and (2) illustration demonstrates the behaviour [33].

$$Y\left(c_i+1\right)=Y_{Best}\left(c_i\right)\times \left(1-\frac{c_i}{c_I}\right)+\left(Y_N\left(c_i\right)-Y_{Best}\left(c_i\right)\times P\right)$$

(1)

$$Y_N\left(c_i\right)=\frac{1}{M}{\displaystyle \sum }_{j=1}^M{Y}_j\left(c_i\right)$$

(2)

In the current repetition, the mean location of the Aquila is indicated by $Y_N\left(c_i\right)$, while the ideal position is denoted by $Y_{Best}\left(c_i\right)$. $c_i$ represents the current repetition, whereas $c_I$ represents the max number of repetitions. P and M represent the random value between [0, 1] and the size of the population, respectively.

Stage 2: Narrowed exploration (flying along a contour with a rapid glide assault).

Narrowed exploring is the predominant common hunting method employed by Aquila. Upon circling the target and falling within the designated place, the predator assaults the victim with quick glides. The location updating appears as Equation (3).

$$Y\left(c_i+1\right)=Y_{Best}\left(c_i\right)\times LFF\left(Dim\right)+Y_{RP}\left(c_i\right)-\left(x-y\right)\times P$$

(3)

The arbitrary location of Hawk is indicated by $Y_{RP}\left(c_i\right)$. Equation (4) Dim and LFF indicate the dimensional size and Levy flight function, respectively.

$$LFF \left(Dim\right)=s\times \frac{v\times \omega }{{\left|u\right|}^{\frac{1}{\alpha }}}$$

(4)

$$\omega =\left[\frac{\mathsf{\Gamma }\left(1+\alpha \right)\times \sin \left(\frac{\pi \alpha }{2}\right)}{\mathsf{\Gamma }\left(\frac{1+\alpha }{2}\right)\times \alpha \times 2^{\left(\frac{\alpha -1}{2}\right)}}\right]$$

(5)

α and s have constant values of 1.5 and 0.01. v and u are between 0 and 1 random numbers. The spiral pattern in the search using x and y is displayed by Equation (6).

$$\{\begin{array}{c}y=r\times sin\left(\theta \right)\\ \begin{array}{c}x=r\times cos\left(\theta \right)\\ r=r_1+0.00565\times Di{m}_1\end{array}\\ \theta =-\sigma \times Di{m}_1+\frac{3\times \pi }{2} \end{array}$$

(6)

$r_1$ represents the searching loop number between 1 and 20. $Di{m}_1$ consists of whole numbers from 1 to the size of the dimension (Dim), and 0.005 is equivalent to σ.

Stage 3: Expansion of exploitation (short flight and a slow assault fall).

Upon accurately identifying the position of the prey, the Aquila flies vertically to perform a preliminary assault in the third strategy. Here, AO exploits the selected position in order to reach and assault the target. The following activities are highlighted in Equation (7):

$$Y\left(c_i+1\right)=(Y_{Best}\left(c_i\right)-Y_N\left(c_i\right))\times \beta -P+\left(\left(Uppe{r}_{bound}-Lowe{r}_{bound}\right)\times p+Lowe{r}_{bound}\right)\times \mu$$

(7)

The exploitation modification settings for β and μ are tuned to 0.1. The lower bound and upper bound of the issue are defined by $Lowe{r}_{bound}$ and $Uppe{r}_{bound}$, respectively.

Stage 4: Narrowed exploitation (walking and capturing prey).

Using a method called “narrowed exploitation”, the Aquila chases its target while maintaining an eye all over its escape path before launching an assault from the ground [34]. The given Equation (8) represents these actions:

$$\{\begin{array}{c}Y\left(c_i+1\right)=QF\times Y_{Best}\left(c_i\right)-\left(M{P}_1\times Y\left(c_i\right)\right)\times P)-M{P}_2\times LFF\left(Dim\right)+P\times M{P}_1\\ \begin{array}{c}QF\left(c_i\right)=c_i{}^{\frac{2\times P-1}{{\left(1-c_I\right)}^2}}\\ M{P}_1=2\times P-1\end{array}\\ M{P}_2=2\times \left(1-\frac{c_i}{c_I}\right) \end{array}$$

(8)

To balance the search method, the current position is indicated by $Y\left(c_i\right)$ and the quality function result is indicated by $QF\left(c_i\right)$. $M{P}_1$ denotes the monitoring prey motion variable for the Aquila, that is a random number in the range [−1, 1]. $M{P}_2$ represents the flight angle while hunting prey.

3.4. Clustering STEP

The following criteria proved critical for the computation of the CF:

CH node balance element: This is essential for maintaining a stable cluster. Clusters of any size are possible as a result of the nodes’ unpredictable placement. Hence, this feature is taken into account to strike a reasonable balance in energy consumption. Equation (9) describes the balance component at each node.

$$F_1= {\displaystyle \sum }_{j=1}^k\frac{n}{k}-l_j$$

(9)

Mean BS distance: The average BS distance is calculated by dividing the distance from the BS to the CH by the absolute number of nodes in the concerned CH node. Because of the correlation between distance and energy use, this statistic is taken into account. Thus, in order to conserve energy, it is essential to shorten this length. The separation between the two points is given in Equation (10), and the average lengths to BS were displayed in Equation (11).

$$ED\left( p ,q\right)=\sqrt{ {\displaystyle \sum }_{r=1}^d{\left(p_r-q_r\right)}^2}$$

(10)

$$F_2={\displaystyle \sum }_{j=1}^k(\frac{1}{l_j}ED\left(C{H}_j, BS\right)$$

(11)

Remaining energy: Since the platform’s lifespan is directly tied to power consumption, finding ways to reduce that statistic is crucial. Consequently, this factor should be considered with special attention. It’s calculated by adding up the energy of the active CH nodes that were specifically chosen. In order to strike a proper balance between all objective functions, the inversion of the maximization of total power is considered as well. In Equation (12), a form for the remaining energy function can be seen.

$$F_3=\frac{1}{{{\displaystyle \sum }}_{j=1}^k\left(E_{C{H}_j}\right)}$$

(12)

Mean intra-cluster distance: Total length between each node and its associated node in the CH graph is the intra-cluster length. Lessening the average distance between nodes within a cluster can help save energy. The connection between a node and its corresponding CH node uses a lot of energy; thus, this is described as Equation (13).

$$F_4={\displaystyle \sum }_{j=1}^k\left(\frac{1}{l_j}{\displaystyle \sum }_{s=1}^{l_j}ED\left(C{M}_s, C{H}_j\right)\right)$$

(13)

It is preferred to reduce the CF mixture represented in Equation (14), as opposed to separately minimizing each function.

$$CF=w_1{F}_1+w_2{F}_2+w_3{F}_3+w_4{F}_4$$

(14)

$${\displaystyle \sum }_{k=1}^4{w}_k=1$$

(15)

The importance of each CF component is represented by its strength. The coefficients represent the perceived importance of the components $w_1$, $w_2$, $w_3$ and $w_4$; as shown in Equation (15), the sum is 1.

3.5. FA Algorithm

The FA was initially released at the end of 2007 [35]. FA is built on several rules informed by the lighting patterns and actions of fireflies:

✓

Fireflies are unisexual. This indicates that fireflies are interested in one another regardless of their sex [36].

✓

Their luminosity determines the appeal of fireflies. It indicates that fewer brilliant fireflies are drawn to more brilliant fireflies. As the space among fireflies grows, so does their desirability. In the absence of a more brilliant firefly, the motion of fireflies is arbitrary.

✓

A function for measuring the luminosity of fireflies is created.

FA discusses 2 criteria: light intensity fluctuation and attraction. Brightness contributes to an individual’s allure. Furthermore, brightness is represented through an objective function [37].

$I\left(x\right)\propto f\left(x\right)$ denotes the brightness of a firefly (I) at a particular location (x). But, depending on the length $r_{ij}$ among $firefl{y}_i$ and $firefl{y}_j$, their attractiveness varies. In addition, when a firefly is distant from the source of light, the light intensity decreases. Additionally, the medium absorbs light. Depending on the level of absorption, several attractivenesses exist. Typically, the intensity of light I(r) varies in accordance with the inverse-square rule, as shown in Equation (16):

$$I\left(r\right)=\frac{I_s}{r^2}$$

(16)

Wherein $I_s$ represents the source’s power.

Assuming a given medium with a fixed coefficient of light absorption γ, Equation (17) calculates I based on the distance r:

$$I\left(r\right)=I_0{e}^{-\gamma r}$$

(17)

Here $I_0$ represents the initial intensity of light at r = 0 to avoid singularity at r = 0 in the expression $\frac{I_s}{r^2}$, Equations (1) and (2) are calculated as Equation (18) [38].

$$I\left(r\right)=I_0{e}^{-\gamma r^2}$$

(18)

Equation (19) easily calculates β for a firefly:

$$\beta ={\beta }_0{e}^{-\gamma r^2}$$

(19)

As ${\beta }_0$ represents the attraction when r = 0 Equation (19) is approximated as Equation (20) since $\frac{1}{1+r^2}$ may be calculated more quickly than an exponential function.

$$\beta =\frac{{\beta }_0}{1+\gamma r^2}$$

(20)

The spacing among $firefl{y}_i$ and $firefl{y}_j$ is denoted by $z_i$ and $z_j$, accordingly, in Equation (21) [39].

$$r_{ij}=\Vert z_i-z_j\Vert =\sqrt{{\displaystyle \sum }_{k=1}^d{\left(z_{i,k}-z_{j,k}\right)}^2}$$

(21)

here $z_{i,k}$ is the kth element of the ith firefly’s location coordinate $z_i$.

Per Equation (22), $firefl{y}_i$ is fascinated to $firefl{y}_j$ that is more appealing:

$$z_i^{t+1}=z_i^t+{\beta }_0{e}^{-\gamma r_{ij}^2}\left(z_j^t-z_i^t\right)+\alpha {\epsilon }_i^t$$

(22)

Whereas the second statement expresses the concept of attraction. Using the third phrase, randomization is indeed produced. It is accomplished via the randomization variable. Moreover, I represent an array of random numbers. Remember that ${\epsilon }_i$ could be changed with $rand \frac{-1}{2}$ to create a random number in the range [0, 1]. In general, it is considered that β₀ = 1 and α falls between [0, 1].

FA is used to identify intermediate CH nodes in the network since this technique explores optimal global reply sufficiently. In addition, FA has a good convergence rate and adequate precision. In addition, this could achieve a balance among the global and local searches. Additionally, FA is simple to apply.

3.6. Communication Step

As the transmission of data in the proposed IoT algorithm is accomplished via multi-hop data forwarding mechanisms, the primary objective of the route discovery step is to identify a collection of intermediate CH nodes that must be chosen to build multiple paths from the source nodes to the BS. To pick the CH node, the intended multipath routing strategies utilize numerous criteria. The intermediary CH node subsequently transmits the message collected to the BS. Therefore, every intermediate CH node is assigned a group of next neighbour CH nodes depending on the number of hops and distance.

The FF ranks the solutions based on how well they optimize the plan. The most recent alpha, beta, and delta data should be used for each iteration. The novel FF given here is set up to generate a route that all CH nodes could travel to access the BS quickly and reliably. The total distance travelled via CH nodes could be determined by employing Equation (23).

$$F_1={\displaystyle \sum }_{j=1}^{m}ED\left(C{H}_j, NextCH\left(C{H}_j\right)\right)$$

(23)

Following is an expression for the overall quantity of hops between CH nodes in the system, as shown by Equation (24):

$$F_2={\displaystyle \sum }_{j=1}^{m}NextCHcount\left(C{H}_j\right)$$

(24)

The entire length to go and the overall quantity of hops must be considered while arranging a route. Reducing both the overall distance travelled and the number of hops enhances a plan’s fitness value. Therefore, the route fitness drops when both the length and the number of hops increase. A population’s preference tends to favour whatever choice has the maximum fitness level. As demonstrated in Equation (25).

$$FF=\frac{k_1}{w_1{F}_1+w_2{F}_2}$$

(25)

$${\displaystyle \sum }_{k=1}^2{w}_k=1$$

(26)

So, $k_1$ is the proportionality constant, and each of $w_1$ and $w_2$ are real numbers. Path FF is the optimal option to find a fine balance between the number of network hops and the total network length.

4. Performance Review

A number of alternative approaches to IoT clustering, such as GA [13], WOA [40], GWO [41] and Hybrid Artificial Bee Colony and Harmony Search (HABCHS) [31], are compared and contrasted with the HAOFA method below. The criteria used to select a suitable simulation program are discussed in Section 4.1. The parameters of the model are described in Section 4.2. The simulator findings are thoroughly discussed in Section 4.3.

4.1. Simulator

The efficiency of the IoT platform is assessed with NS-3 version 3.25. It could be seen in [42] NS-3 is used in numerous distinct categories of IoT scenarios due to its various advantages versus alternative simulators and software products.

▪

Its designs are more adaptable than those of rival software.

▪

Making use of an extensive range of resources in a unified way.

▪

As a result of its compatibility with other programming languages, it is simple to debug.

▪

The ability to scale as well as flexibility for multiple layers is intrinsic [43].

4.2. Model Parameters

A well-described modelling parameter is essential for obtaining reliable simulation results. The most fundamental and important considerations are listed in

Table 2. Test parameters.

Parameter Title	Value
Radius	500 m
Count of all Nodes	400
Placement of BS	(250, 250)
Size of Data packet	1024 byte
Primary Nodes Power	0.5 J
$E_{elec}$	50 nJ/bit
${\epsilon }_{fs}$	10 pJ/bit/m2
${\epsilon }_{mp}$	0.0013 pJ/bit/m4
$\mathrm{Threshold}\mathrm{of}\mathrm{Distance}(d_0)$	$\sqrt{\frac{{\epsilon }_{fs}}{{\epsilon }_{mp}}} \mathrm{m}$

.

To further augment the efforts, the evaluation in this study is conducted on a standardized IoT platform. A total of one 2000 repetitions of the test are allowed. Specifically, the system is a 500-meter-diameter circle with the BS in its epicentre. Initially, 400 IoT nodes within the network, each receiving an initial power supply of E0 = 0.5 J. The increasing data transfer speed of IoT nodes necessitates a larger data packet size; therefore, 1024 bytes it is.

4.3. Evaluation Metrics

Towards what comes after, it would look at how to measure metrics such as energy consumption, PDR, E2E delay and throughput to assess whether effectively various ways are functioning.

4.3.1. Energy Consumption

As can be seen in Figure 1, the HAOFA algorithm extends the lifetime of the network by decreasing power consumption. The figure makes it easy to comprehend how the nodes’ initial energy gradually dissipates. This is because of the use of multi-hop routing in FA-based connectivity between clusters. Power consumption in the network is decreased when using this method instead of GA, WOA, GWO and HABCHS. This happens as a result of the reality that the HAOFA employs an Aquila search-based optimization approach to make the most intelligent choice regarding CH nodes and their distribution across the surveillance region. When using this planned approach, the CH nodes are distributed evenly across the surveillance area, which in turn extends the lifespan of the network. In other words, the lifespan of the IoT network is extended due to the more efficient use of energy at each node. That’s because the planned routing method improves network dependability.

4.3.2. PDR

Node energy reduction extends the network’s life span through keeping more nodes operational for longer. Having this improvement spread out over the lifespan of the network boosts data gathering and transmission speeds. PDR reveals the ratio of data packets that were successfully transmitted and received. Whenever this proportion grows, productivity rises. The suggested scheme integrates meta-heuristic AO and FA optimization to minimize complexity and boost PDR values while lowering the number of failed nodes and the packet loss rate. Two components of the CF take into account the node’s energy consumption as a function of distance, and the amount of energy the network’s nodes still have is also a major consideration in keeping the nodes from going dying and inactive. Finally, a higher PDR index is strongly influenced by the degree of balance at the CH node. The PDR diagram of the four approaches is depicted in Figure 2. The increased strength and durability of routes are a direct result of this. The poor link quality or great distance between the two nodes might also shorten the lifetime of the communication. Thus, the provided method reduces risky connections in the system and boosts the likelihood of finding the best routes between CH nodes. The least amount of PDR is raised by 25% as opposed to the GA, WOA, GWO and HABCHS strategies.

4.3.3. E2E Delay

The system transmission latency would be another crucial metric for gauging the efficiency of an infrastructure. The latency from selecting a CH node to actually receiving the data is known as the E2E delay. The impact of this delay on the total duration of the test is graphically represented in Figure 3. This algorithm incorporates the benefits of the AO protocol for improving energy efficiency with the FA for decreasing communication delay. The HAOFA mechanism picks the best CH nodes to collect information from CM nodes. When compared to the GA, WOA, GWO, and HABCHS approaches, this one decreases delay. The following factors, among others, contribute to this shortened delay:

(1) Clustering conserves time due to the proximity among CM and CH nodes.

(2) The uniform dispersion of clusters in HAOFA equalizes the time usage.

4.3.4. Throughput

In this method, by using meta-heuristic methods, CH nodes are spread evenly, and the total number of nodes in each cluster is closer, but otherwise, energy consumption would not be evenly distributed and nodes would die faster, ultimately reducing throughput. Data transmission energy loss is decreased since the leftover energy of the node is used as a criterion for selecting CH nodes. Thereby, more packets make it to the BS than with other algorithms. As can be shown in Figure 4, the proposed approach greatly improves throughput in comparison to GA, WOA, GWO and HABCHS. The suggested approach generates paths that are ideal for node-to-node connectivity on this architecture, and it also reduces route loss. Since low-energy CH nodes won’t be helpful during route construction and will have a meagre chance of being selected as intermediate nodes; as a result, routes can already handle more traffic and have a longer shelf life. It’s also possible that the low link quality or the considerable physical separation between the nodes contributed to their short connection lifetime. Thus, the proposed method reduces risky connections in the system and boosts the likelihood of finding the best routes among CH nodes.

5. Conclusions and Future Works

This research examines the most significant issue facing the IoT infrastructure today: a deficiency of available energy in nodes. A clustering approach was proposed for maximum efficiency in terms of both power consumption and network longevity. Here, the HAOFA approach for IoT clustering and routing operations is introduced. In the described method, CH nodes are picked and clustered by employing the AO algorithm. Next, FA determines which route choices are best. Two new functions support routing and clustering. These calculations accounted for aspects like available energy, distance, node balance, and the number of hops. This population-optimal value is then used for routing and clustering purposes. The proposed HAOFA is compared with GA, WOA, GWO and HABCHS methodologies to show that the presented strategy is practical. The results show that implementing the plan helps the system function more quickly and lasts longer.

In the future, researchers may choose to systematically relocate a system’s BS, CH node, and CM node, all in an effort to better understand how such changes affect the network’s operation. It is also possible that using a more comprehensive range of parameters will help boost FF and CF’s efficiency.