Reliable routing in MANET with mobility prediction via long short-term memory

Abstract

A MANET consists of a self-configured group of transportable mobile nodes that lacks a central infrastructure to manage network traffic. To facilitate communication, govern route discovery, and manage resources, all moving nodes in multi-hop wireless networks (MANETs) work together. These networks struggle with dependability, energy consumption, and collision avoidance. The goal of this research project is to establish a new, dependable MANET routing model, where the selection of predictor nodes comes first. For selecting predictor nodes based on factors like distance, security (risk), Receiver Signal Strength Indicator (RSSI), Packet Delivery Ratio (PDR), and energy, the adaptive weighted clustering algorithm (AWCA) is used in this case. Using the Interfused Slime and Battle Royale Optimization with Arithmetic Crossover (IS&BRO–AC) model, the node with the lower weight is selected as the Cluster Head (CH). Additionally, mobility prediction is carried out, in which the node mobility is forecast using Improved Long Short Term Memory (LSTM) while taking distance and Receiver Signal Strength Indicator (RSSI) into account. Based on the forecast, trustworthy data transfer is implemented, ensuring more accurate and dependable MANET routing. The examination of RSSI, PDR, and other metrics is completed at the end.

Keywords

Mobile Ad Hoc Network (MANET) routing Cluster Head Receiver Signal Strength Indicator (RSSI)Packet Delivery Ratio (PDR)Interfused Slime and Battle Royale Optimization with Arithmetic Crossover (IS&BRO)–AC Algorithm

1. Introduction

Table 1
Nomenclature

Abbreviation	Description
AWCA	Adaptive Weighted Clustering Algorithm
ANFIS	Adaptive Neuro-Fuzzy Inference System
BRO	Battle Royale Optimization
BS	Base Station
CA	Coot Optimization
CH	Cluster Head
CH	Cluster Head
DTRP	Decision Tree Based Routing Protocol
e2e	End-To-End
EO	Equilibrium Optimizer
ECC	Elliptic Curve Encryption
FSMR	Fuzzy Enhanced Secure Multicast Routing
FMF	Fuzzy Membership Function
GPS	Global Positioning Systems
HM-MPR	Hybrid Model Based Multi Path Routing
(IS&BRO)	Interfused Slime and Battle Royale Optimization with Arithmetic Crossover
LSTM	Long Short Term Memory
MANET	Mobile Ad Hoc Network
PSO	Particle Swarm Optimization
PS-ROGR	Particle Swarm Based Resource Optimized Geographic Routing
PDR	Packet Delivery Ratio
PA	Power Amplifier
RSSI	Receiver Signal Strength Indicator
ReCoMM	Resource Aware Cooperation Model
SMA2AODV	SMA Integration With AODV Protocol
SMA	Slime Mould Algorithm
SSA	Shark Smell Algorithm
SSOA	Social Spider Optimization Algorithm
TMF	Triangular Membership Function
WCA	Weighted Clustering Algorithm

41] that operate decentralized to allow for spontaneous communication and interaction amongst a larger count of mobile nodes. In the past ten years, various mobile operations in both commercial and defense settings have centered on MANETs [4,43,44]. A system of wireless sensors installed in an oil field, for example, or a network of smart gadgets carried by soldiers on a battlefield are examples of these deployments. Additionally, the capacity of MANET to organize and adapt without relying on the underlying infrastructure allowed for their quick deployment in unconventional settings like disaster recovery [6,16].

It is challenging (though not impossible) under such circumstances to reinstall the equipment or offer wireless access to the backbone network. Additionally, all linked nodes in MANETs are peer nodes (client and agency nodes 1) with comparable features and abilities that enable them to serve as a mobile router as well [3,11,13]. User nodes have unrestricted movement and are typically unsure about their future positions. A packet is transmitted in multi-hops from sender to receiver involving a lot of intermediary nodes in MANETs because nodes may maintain routes and forward packets while having a restricted communication range. Because of this distinctive method of message transmission, MANETs depend on node cooperation to function. In this method of collaboration, the MANET control system receives constant location information about user nodes from GPS. Along with the range and speed of their wireless connections, this information describes the user nodes [9,19,37].

This infrastructure needs flexibility, meanwhile, comes at a cost of challenges and concerns that are not present in typical wireless networks. In reality, issues with MANETs include ad-hoc addressing, self-configuration, and adaptive reconfiguring to address the impact of mobile nodes on the network architecture [26,33,34]. Additionally, data flow is subject to scheduling limits, necessitating proactive routing [17,24], and involves continuous ad hoc network appliances. This adaptability in infrastructure needs, meanwhile, comes with concerns and problems that are not yet identified in traditional wireless networks [8,12,45]. In reality, problems with MANETs include ad-hoc addressing, higher energy limitations, and self-configuration and adaptable reconfiguration to clarify the effect of the mobility of nodes on the network architecture [25,32]. To overcome certain drawbacks the proposed model is used.

The contributions are as follows:

This research work plans to develop a new reliable MANET routing model, where, predictor node selection is done via IS&BRO–AC algorithm.

Deploys new distance evaluation, where, the total absolute distinction among the 2 vectors is used to determine the Manhattan distance.

Moreover, mobility prediction is done using improved LSTM.

Here, Section 2 analyses conventional works on MANET routing. Section 3 provides an overview of MANET routing. Sections 4 and 5 depict predictor node selection and predicting mobility with optimized LSTM. Sections 6 and 7 depict results and conclusions.

2. Literature review

2.1. Related works

Singh et al.’s [31] innovative DTRP, a data mining approach, was suggested in 2020 for the source-to-destination route selection process. The one-hop neighbors are chosen using the suggested DTRP method based on factors including speed, Link Expiration Duration, journey time, and node lifetime. So choosing secure sone-hop neighbors along the way to the destination improves the performance of a route-finding technique. The simulation results demonstrate that adopting the suggested DTRP routing mechanism as opposed to existing routing methods increases the lifespan of the route, minimizes data loss, and end-to-end latency, enhancing network performance.

In 2020, Wang et al. [29] presented an effective PS-ROGR strategy in MANET to enhance resource optimization and network longevity. Each particle’s (or mobile node’s) locally renowned position in the search area initially governs how it moves through a network (i.e. geographic location). The entire particle in the system can communicate with one another through the PSO using the least amount of energy possible. Therefore, the optimal position for optimizing the network resources is shared by all of the particles. In this way, the PS-ROGR approach increases network longevity while consuming the least amount of energy possible.

Rahimizadeh et al. [18] establishment of the ideal loop-free path using fuzzy logic and an EO in 2020 ensures the continuity of development. Five steps are described in the proposed algorithm: route discovery, routing information, path recovery, reactive path setup, and multipath routing setup. The EO technique using fuzzy logic was first utilized to determine the best route for the data packets during the route research phase. This outcome shows that the suggested strategy works better than other current options.

The nodes were organized using a clustering algorithm by Pan et al.’s [21]. Using PSO, this protocol will forecast where a node will be in the future. Before finding the optimal path, the link lifespan, position, distance, and speed of the nodes are determined. To discover malicious network nodes and minimize packet loss, the trust values of nodes are calculated from their neighbors. The packets are encoded utilizing ECC to stop data from hostile nodes. Because of trust value, updating routing information is simple and increases the throughput of the network.

An adaptable routing for MANETs was introduced in 2020 by Singh et al. [42] that constantly sets up the routing function about the metrics: (1) The different requirements’ parameters and (2) according to the required application context, the contextual characteristics. Different efficiency, safety, and functional parameters are included in the requirement models. Contrarily, contextual aspects include node/group mobility, node trust values, node resource limits, geographical context, individual node roles, etc. Extensive simulation test scenarios are used to assess our routing system, and the protocol’s effectiveness is documented.

To improve network and data security in 2020, Penna et al. [7] employed the FSMR strategy. The FSMR is created based on abnormal and normal behaviors from the statistics data to assess the existence of misbehaving nodes. The suggested solution uses certificate-less routing and uses the idea of key generation to authenticate the data. Without being aware of certificate routing, packets are sent and verified. The study of the suggested system shows that it has superior data packet authentication while using less energy.

To create efficient routing, Penna et al. [38] developed the ReCoMM in 2020. ReCoMM allows for the assessment of the connection stability of the node. The ReCoMM model looks at the variables that affect cooperative and node state change using a Markov process. The Markov process is employed to alter connection stability and node longevity. With the calculation of collaboration value, the Markov process assists in identifying the upper and lower boundaries of cooperation. Additionally, it performs far better than earlier models regarding average routing overhead and e2e latency.

The presentation of extracting features and a categorization model based on ANFIS by Kostenko et al. [30] was developed in 2020. The retrieved characteristic was trained and then categorized using the ANFIS classifier. This research also suggests SMA2AODV identify flooding assaults for MANETs to combat floods through energy-preserving routing. Following detection, the hybrid approach ACO and FDR PSO were coupled for energy optimization. To extend node lifespan and assure energy-efficient routing, ACO-FDR PSO determines the energy-efficient routing and reduces the network’s energy consumption.

Shajin et al. [35] have deployed safe and effective communication between sender and destination nodes, this study seeks to assess the direct trust value for each node, compute the trust value of all nodes meeting the criteria, and update the trust value and value per trust update interval. Thus, a Trusted Secure Geographic Routing Protocol (TSGRP) that takes into account the trust value for a node obtained by combining location trusted information and direct trusted information has been proposed for detecting attackers (the presence of the hacker).

Theerthagiri et al. [39] have presented ARIMA model uses autocorrelation in conjunction with the random walk (RW) model to predict node mobility in the future. The Akaike information criterion (AIC) and auto-correlation function (ACF) are evaluated for the anticipated mobility model. When comparing the ARIMA- and RNN-predicted mobility datasets, the mean square error (MSE), a measure of error performance, is significantly lower for the ARIMA-predicted mobility datasets.

Jalade et al. [20] have deployed stable node prediction; stability measure establishment, route finding, and data transmission are the four main stages of the suggested methodology. The ADRKGN model is first used to predict the stable node and pick the stable neighbors’. The stability measure is then established, according to which the stable node and neighbor node for routing are similar in the Routing Table (RT). The suggested method has improved end-to-end delay, routing overhead, throughput, energy consumption, and network longevity. For 50 seconds, the throughput of 1178.79 was achieved using our suggested method.

2.2. Research gap

MANET research has been influenced by the widespread accessibility of wireless connectivity services and quickly growing deployment requirements over the past couple of decades. A MANET is a wirelessly linked network of mobile devices that continually configures itself. Due to the mobility, self-organization, quick deployment, and lower-cost communications of MANETs, they may be used for a wide range of purposes, together with environmental monitoring, disaster assistance, and military communications. Because of the multiple-hop manner in which data is transmitted, decentralized networks often tend to be more resilient than centralized networks. For instance, if a BS ceases operating in a cellular network, coverage would decrease. However, because data may go through several channels in a MANET, the likelihood of a single failure point is greatly diminished. The MANET design can address problems like isolation or network disconnection since it changes over time. However, the usage of an open wireless medium, changing network architecture, and resource limitations provide several security and performance difficulties for MANET routing. As a result, the academic community is very interested in creating an effective and secure routing system for MANET.

3. Mobility prediction in MANET: An overview

The purpose of this study is to create a new, dependable MANET routing model. Predictor node selection will be the first step in this process. Here, AWCA is used to pick the predictor nodes based on a variety of requirements, including proximity, security, RSSI, PDR, and energy. The IS&BRO–AC algorithm is used to select CH in the best possible way. Additionally, Mobility prediction is carried out, in which Improved LSTM is used to forecast the node mobility. Prediction-based trustworthy data transfer is used to support more dependable MANET routing. The illustrative depiction of IS&BRO–AC model is exposed in Fig. 1.

Fig. 1.

PictorialModel of IS&BRO–AC method.

4. Predictor node selection: Detailed description

Here, we have used AWCA to anticipate the out-of-range of the mobile node. All of these factors are taken into account by AWCA when selecting the CH. The cluster head that manages the trade-off between numerous cluster heads with a variable number of clusters, latency, power consumption, and data computational capabilities of each node is chosen in computing based on several viewpoints and the application domain. One of the indicated nodes is chosen as the CH, and it is given the name predicting node. LetMANET is represented as an undirected graph denoted by $H = (U, E)$ . Here, U refers to nodes and E refers to a set of links, D refers to the vertex of H or neighbor $M [U]$ and $D \subseteq U (H)$ .

This work aims to choose a CH for which, parameters such as distance, security, RSSI, PDR, and energy are considered.

4.1. Objective model

This project aims to accelerate data transmission while reducing distances and risks between and within clusters. Instead, the system’s residual energy, RSSI, and PDR ought to be at their greatest levels after a successful data transfer. The objective of the IS&BRO–AC model is revealed in Eq. (1). Here, $W 1$ – $W 5$ refers to the weight values, which were evaluated by means of FMF as exposed in Eq. (2), in which, a (lower bound), b (medium bound), c (upper bound), refers to vertices of TMF $T (f)$ . Here, $a = 1$ , $b = 0.5$ and $c = 0$ . $\begin{array}{c} (1) & \begin{aligned} Ob & = MIin [W 1 * (di) + W 2 * (se) + W 3 * (1 - rssi) \\ + W 4 * (1 - pdr) + W 5 * (1 - en)] \end{aligned} \\ (2) & W = \{\begin{array}{ll} 0; & if x < a \\ \frac{x - a}{b - a}; & if a ⩽ x ⩽ b \\ \frac{c - x}{c - b}; & if b ⩽ x ⩽ c \\ 0; & c ⩽ x \end{array} \end{array}$

The constraints in Eq. (1) are described as follows.

Distance: The evaluation of the neighboring node distance follows Eq. (3). The distance is calculated in our study using the Manhattan formula, which is shown in Eq. (4) [1]. The total absolute distinction among the 2 vectors is used to determine the Manhattan distance. $\begin{array}{c} (3) & di = \sum_{U^{'} \in M (U)}^{m} {di (U^{'}, U)} \\ (4) & di (U^{'}, U) = | U_{1}^{'}, U_{1} | + | U_{2}^{'}, U_{2} | + \dots | U_{m}^{'}, U_{m} | \end{array}$

Security (Risk): Eq. (5) exposed the estimation of risk, in which, $sd$ and $sr$ corresponds to security rank and security requirements. $\begin{matrix} (5) & se = \{\begin{array}{ll} 0; & if sd - sr ⩽ 0 \\ 1 - e^{\frac{sd - sr}{2}}; & if 0 < sd - sr ⩽ 1 \\ 1 - e^{\frac{3 (sd - sr)}{2}}; & if 1 < sd - sr ⩽ 2 \\ 1; & if 2 < sd - sr ⩽ 5 \end{array} \end{matrix}$

RSSI [ 2 ]: It is a measure of the strength of a radio signal that has been received. For users of receiving equipment, RSSI is frequently invisible. After antenna and potential cable loss, RSSI shows the power level that the receiving radio is receiving. Therefore, the signal gets stronger with a higher RSSI number. As a result, whenever an RSSI value is expressed negatively, the higher the received power, when the closer the number is to 0.

PDR [ 22 ]: Measuring the effectiveness of the routing algorithm in a network is crucial. The effectiveness of the technique is dependent on the simulation’s numerous parameter settings. The most important factors are packet size, node count, transmission range, and network topology. The PDR can be calculated by dividing the total amount of data that have reached their destinations by the number of packets that have been sent from sources. PDR is, in other words, the proportion of packets transmitted from the source to those received at the destination. It is shown in Eq. (6), which, $PRDN$ refers to packets received by every destination node and $PSSN$ refers to packets send by every source node. $\begin{matrix} (6) & pdr = \frac{Σ (PRDN)}{Σ (PSSN)} \end{matrix}$

Energy Model [ 23 ]: The fundamental issue with WSN is energy extraction [19]. The WSN battery cannot supply energy if the batteries run out since it does not utilize a re-energizing mechanism. Furthermore, data transfer from all SNs to the BS is expertly made with the aid of additional resources. For data transmission, energy efficiency is vital. The network truly needs more energy because of all of its transmission, reception, aggregation, and sensing processes. The quantity of energy needed for data broadcasting is implied by Eq. (7), in which ${en}_{el}$ refers to electronic energy, ${en}_{ea}$ which refers to energy for data time aggregation. ${en}_{el}$ is shown in Eq. (8). ${en}_{TX} (Z : di)$ refers to the whole energy essential to converse Z packet bytes at distance $di$ .

In Eq. (10), the energy needed for amplification is referred to as ${en}_{am}$ and Eq. (9) disclosed the energy needed at the receiver ${en}_{RX}$ to obtain Z packet bytes at $di$ . $\begin{array}{c} (7) & {en}_{TX} (Z : di) = en \{\begin{array}{ll} {en}_{el} * Z + {en}_{rs} * Z * {di}^{2}, & if di < {di}_{0} \\ {en}_{el} * Z + {en}_{pw} * Z * {di}^{2}, & if di ⩾ {di}_{0} \end{array} \\ (8) & {en}_{el} = {en}_{TX} + {en}_{ea} \\ (9) & {en}_{RX} (Z : di) = {en}_{el} Z \\ (10) & {en}_{am} = {en}_{fr} {di}^{2} \\ (11) & {di}_{0} = \sqrt{\frac{{en}_{fr}}{{en}_{pam}}} \end{array}$

Here,

${en}_{pam} \to$ Energy of PA

${di}_{0} \to$ Threshold distances

${en}_{fr} \to$ Crucial energy for free spaces

${en}_{1} \to$ Energy for complete immobile state

${en}_{C} \to$ Energy cost for complete sense phase

Equation (12) displays the total energy required to broadcast data. $\begin{matrix} (12) & F 1 = {en}_{total} = {en}_{TX} + {en}_{C} + {en}_{RX} + {en}_{1} \end{matrix}$

The node with a smaller weight is chosen as CH using IS&BRO–AC model. Here, CH is said to be the predictor node.

As seen in Fig. 2, the CHs are here selected in the best possible way using the IS&BRO–AC method.

Fig. 2.

Solution encoding.

4.2. Proposed IS&BRO–AC algorithm

The best possibilities are identified by the current SMA [27], however, it is not particularly precise. As a result, IS&BRO–AC is coupled with BRO [15] to overcome the inadequacies of conventional SMA [33]. Some search issues have purportedly been solved via hybridized optimization methods [5,10,36,40].

The activities taken in the predicted model are as follows:

The search agent’s population $Pop$ is initialized in Step 1. Furthermore, the maximum iteration ${max}^{itr}$ and current iteration $itr$ are initialized.

The search agent’s population is initialized $X_{i}$ ; $i = 1, 2, \dots, N$

Determine the search agent’s fitness using Eq. (1).

Update the best fitness $X_{best}$ .

Perform while $itr < {max}^{itr}$ do

Approach to food: The approach food phase is carried out in step 6 to update the solutions. The air odor in this case is used by the search agents to update the solutions. Equation (13) is typically used to represent the mathematical equation for the approaching food phase in the SMA model. As per IS&BRO–AC, it is modeled based on BRO update as shown in Eq. (14) and Eq. (15), in which, $ra$ , r refers to random integer among 0, 1 and $X^{dam, d}$ refers to the location of the injured warrior in dimension d. $\begin{array}{c} (13) & X^{t + 1} = \{\begin{array}{ll} X_{best} (t) + b . (W . X_{A} (t) - X_{B} (t)) & r < p \\ c . X (t) & r ⩾ p \end{array} \\ (14) & X^{dam, d} = X^{dam, d} + λ^{dam, d} \\ (15) & λ^{dam, d} = {ra}_{1} . X_{best, d} + {ra}_{2} . (λ^{dam, d} - X^{dam, d}) \end{array}$

The parameters b that were collected inside the range $[- a, a]$ and c imply a parameter that declines from 1 to 0] are shown in Eq. (13). The searching agent that was found in the area with the most odors is designated as $X_{best}$ . Additionally, 2 searching agents are presented and denoted by, $X_{A}$ and $X_{B}$ . The weight of the search agent is shown as W. Additionally, the mathematical expression for p is indicated by Eq. (16) $\begin{matrix} (16) & p = tanh | Ob (i) - Best | \end{matrix}$

Here, $Ob$ stands for the search agent’s fitness, which is calculated using Eq. (1). The notation $Best$ also indicates the best fitness that has been attained thus far.

Conventionally, the mathematical formula a is denoted as per Eq. (17). As per IS & BRO–AC a is modeled as in Eq. (18), wherein ${itr}_{max}$ implies maximal iteration. $\begin{array}{c} (17) & a = arctan h [- (\frac{t}{{itr}_{max}}) + 1] \\ (18) & a = 2 {(\frac{1 - itr}{{itr}_{max}})}^{(2 . itr / {itr}_{max})} \end{array}$

Calculate W by means of Eq. (19). $\begin{matrix} (19) & W (SI (i)) = (\begin{array}{ll} 1 + rand . log (\frac{{opt}_{fit} - O (i)}{{opt}_{fit} - {worst}_{fit}} + 1) & condition \\ 1 - rand . log (\frac{{opt}_{fit} - O (i)}{{opt}_{fit} - {worst}_{fit}} + 1) & others \end{array} \end{matrix}$

Here, $SI = Sort (O)$

The randomized value is represented by $rand$ and is produced between $[0, 1]$ . Additionally, the terms ${opt}_{fit}$ , ${worst}_{fit}$ designate the best and poorest fitness, respectively.

Wrap food: Initialize the values p, b, c for each search agent.

Adjust the search agent’s position to match Eq. (20). $\begin{matrix} (20) & X^{t + 1} = \{\begin{array}{ll} ran . (ub - lb) + lb & rand < z \\ X_{best} (t) + b . (W . X_{A} (t) - X_{B} (t)) & r < p \\ c . X (t) & r ⩾ p \end{array} \end{matrix}$

Here, $ran$ is an arbitrary integer produced in the range of $[0, 1]$ , $ub$ and $lb$ implies upper and lower bounds.

Arithmetic crossover is performed for the attainment of objectives.

End for

$itr = itr + 1$

End while

Return $X_{best}$

5. Predicting mobility with an optimized LSTM model

After CHS, the node with higher mobility is predicted using the DL algorithm. The CH will estimate the mobility information, RSSI, etc. of all nodes inside the cluster. The prediction inside LSTM is carried out based on distance and RSSI.

5.1. LSTM

It includes [46]: a forget gate, an input gate, and an output gate. Assume that variables Y and A are concealed and cell states. $(D_{t}, A_{t - 1}, Y_{t - 1})$ and $(Y_{t}, A_{t})$ be input and output layers.

LSTM made use of $G_{t}$ to set up the data as publicized in Eq. (21), here, $σ \to$ leaky relu activation function, $(P_{YG}, G_{YG})$ and $(P_{IG}, F_{IG})$ implied weight and bias for mapping input and concealed layers to forget gate, time $\to t$ , the forget gate $\to G_{t}$ , input gate $\to I_{t}$ and output gate $\to L_{t}$ . Here, the leaky relu activation function is deployed. $\begin{matrix} (21) & G_{t} = σ (P_{IG} X_{t} + P_{YG} Y_{t - 1} + F_{IG} + F_{YG}) \end{matrix}$

$I_{t}$ is exploited in LSTM as in Eq. (22)–(24), here, $(P_{YV}, L_{YV})$ and $(P_{IV}, V_{IV}) \to$ weight and bias factors for mapping input and hidden layers to cell gate. $(P_{YI}, F_{YI})$ and $(P_{II}, F_{II})$ imply weight and bias constraints for mapping input and concealed layers to $I_{t}$ . It gets output concealed layer from $L_{t}$ as in Eq. (25) and (26), in which, $(P_{YL}, F_{YL})$ and $(P_{IL}, F_{IL}) \to$ weight and bias to map $L_{t}$ . $\begin{array}{c} (22) & V_{t} = tanh (F_{IV} + P_{IV} D_{t} + F_{YV} + P_{YV} Y_{t - 1}) \\ (23) & I_{t} = σ (P_{II} D_{t} + P_{YI} Y_{t - 1} + F_{II} + F_{YI}) \\ (24) & A_{t} = G_{t} A_{t - 1} + I_{t} V_{t} \\ (25) & F_{t} = σ (L_{IF} + J_{IF} X_{t} + L_{YF} + J_{YF} Y_{t - 1}) \\ (26) & Y_{t} = L_{t} tanh (A_{t}) \end{array}$

In this work, the cross entropy-based loss is computed as shown in Eq. (27). $\begin{matrix} (27) & cross entropy = - \prod_{i = 1}^{N} [t_{i} log (x_{i}) - (1 - t_{i}) log \frac{1}{N} (1 - x_{i})] \end{matrix}$

5.2. Mobility control

Following mobility prediction, anytime the target node broadcasts outside of its transmission range, CH sets a flag. A different path will be selected and transmission can proceed depending on that flag value.

6. Results and discussions

6.1. Simulation setup

Python was used to execute this project. The number of nodes used is 100 and 200. The maximum number of rounds used is 2000. The length of the data packet is 4000. The efficacy of IS&BRO–AC was demonstrated over WCA [28], HM-MPR [14], SSA, SSOA, CA, SMA, and BRO. The examination was done on the topic of throughput, energy, PDR, delay, and distance. The sample model of MANET and simulation set up for 100 and 200 nodes is exposed in Fig. 3 and Fig. 4.

Fig. 3.

Sample model of MANET for nodes (a) 100 and (b) 200 before clustering.

Fig. 4.

Sample model of simulation set up for nodes (a) 100 and (b) 200 after clustering.

Table 2

Statistical study on delay

Methods	Minimum	Maximum	Mean	Median	Std Dev
For node count = 100
WCA [28]	$2.17 \times 10^{- 07}$	$3.81 \times 10^{- 07}$	$3.72 \times 10^{- 07}$	$3.81 \times 10^{- 07}$	$2.47 \times 10^{- 08}$
HM-MPR [14]	$3.81 \times 10^{- 07}$	$3.81 \times 10^{- 07}$	$3.81 \times 10^{- 07}$	$3.81 \times 10^{- 07}$	$1.88 \times 10^{- 10}$
SSA	$3.81 \times 10^{- 07}$	$3.81 \times 10^{- 07}$	$3.81 \times 10^{- 07}$	$3.81 \times 10^{- 07}$	$9.73 \times 10^{- 11}$
SSOA	$3.81 \times 10^{- 07}$	$3.81 \times 10^{- 07}$	$3.81 \times 10^{- 07}$	$3.81 \times 10^{- 07}$	$1.05 \times 10^{- 10}$
CA	$1.96 \times 10^{- 07}$	$3.81 \times 10^{- 07}$	$3.61 \times 10^{- 07}$	$3.81 \times 10^{- 07}$	$3.73 \times 10^{- 08}$
SMA	$3.81 \times 10^{- 07}$	$3.81 \times 10^{- 07}$	$3.81 \times 10^{- 07}$	$3.81 \times 10^{- 07}$	$9.74 \times 10^{- 11}$
BRO	$3.81 \times 10^{- 07}$	$3.81 \times 10^{- 07}$	$3.81 \times 10^{- 07}$	$3.81 \times 10^{- 07}$	$9.62 \times 10^{- 11}$
IS&BRO–AC	$1.65 \times 10^{- 07}$	$3.81 \times 10^{- 07}$	$2.64 \times 10^{- 07}$	$2.62 \times 10^{- 07}$	$5.79 \times 10^{- 08}$
For node count = 200
WCA [28]	$1.73 \times 10^{- 07}$	$3.81 \times 10^{- 07}$	$3.45 \times 10^{- 07}$	$3.81 \times 10^{- 07}$	$5.00 \times 10^{- 08}$
HM-MPR [14]	$3.81 \times 10^{- 07}$	$3.81 \times 10^{- 07}$	$3.81 \times 10^{- 07}$	$3.81 \times 10^{- 07}$	$1.86 \times 10^{- 10}$
SSA	$3.81 \times 10^{- 07}$	$3.81 \times 10^{- 07}$	$3.81 \times 10^{- 07}$	$3.81 \times 10^{- 07}$	$2.05 \times 10^{- 10}$
SSOA	$3.81 \times 10^{- 07}$	$3.81 \times 10^{- 07}$	$3.81 \times 10^{- 07}$	$3.81 \times 10^{- 07}$	$1.94 \times 10^{- 10}$
CA	$1.73 \times 10^{- 07}$	$3.81 \times 10^{- 07}$	$3.45 \times 10^{- 07}$	$3.81 \times 10^{- 07}$	$5.00 \times 10^{- 08}$
SMA	$3.81 \times 10^{- 07}$	$3.81 \times 10^{- 07}$	$3.81 \times 10^{- 07}$	$3.81 \times 10^{- 07}$	$1.75 \times 10^{- 10}$
BRO	$3.81 \times 10^{- 07}$	$3.81 \times 10^{- 07}$	$3.81 \times 10^{- 07}$	$3.81 \times 10^{- 07}$	$1.86 \times 10^{- 10}$
IS&BRO–AC	$1.65 \times 10^{- 07}$	$3.71 \times 10^{- 07}$	$2.37 \times 10^{- 07}$	$2.29 \times 10^{- 07}$	$5.06 \times 10^{- 08}$

Table 3

Statistical study on distance

Methods	Minimum	Maximum	Mean	Median	Std. Dev.
For node count = 100
WCA [28]	43.50911	80.07961	77.93173	80.00305	5.640636
HM-MPR [14]	79.93464	80.07112	80.00278	80.00252	0.039062
SSA	79.93031	80.08008	80.00246	80.00295	0.020441
SSOA	79.94139	80.08158	80.00822	80.00828	0.022042
CA	41.40409	80.0685	75.84266	79.98814	7.678664
SMA	79.93583	80.06591	80.00209	80.00207	0.020464
BRO	79.9346	80.0712	80.00063	80.00001	0.020185
IS&BRO–AC	34.71884	79.98743	55.39329	54.83917	11.88409
For node count = 200
WCA [28]	35.69869	80.07943	72.26056	79.93206	10.97808
HM-MPR [14]	79.93461	80.07113	80.00164	80.00036	0.039724
SSA	79.93032	80.07999	80.00659	80.00793	0.043179
SSOA	79.94151	80.08149	80.01067	80.00975	0.040274
CA	37.2942	80.07948	72.34836	79.81675	10.72897
SMA	79.9359	80.06581	80.00142	80.00224	0.037943
BRO	79.93466	80.07116	80.00091	80.00057	0.039345
IS&BRO–AC	34.72342	79.42794	49.88095	48.25885	10.66915

Table 4

Statistical study on RSSI

Methods	Minimum	Maximum	Mean	Median	Std. Dev.
For node count = 100
WCA [28]	−212.789	−179.643	−211.646	−212.755	3.51763
HM-MPR [14]	−212.785	−212.724	−212.754	−212.755	0.01764
SSA	−212.789	−212.722	−212.754	−212.754	0.009198
SSOA	−212.79	−212.727	−212.757	−212.757	0.009918
CA	−212.784	−184.718	−209.679	−212.745	5.483677
SMA	−212.783	−212.724	−212.754	−212.754	0.009209
BRO	−212.785	−212.724	−212.753	−212.753	0.009091
IS&BRO–AC	−212.661	−177.424	−193.99	−193.851	9.362664
For node count = 200
WCA [28]	−212.789	−178.789	−206.691	−212.722	8.361381
HM-MPR [14]	−212.785	−212.724	−212.753	−212.753	0.017958
SSA	−212.789	−212.722	−212.755	−212.755	0.019579
SSOA	−212.79	−212.727	−212.758	−212.758	0.018484
CA	−212.789	−178.264	−206.944	−212.722	8.163961
SMA	−212.782	−212.724	−212.753	−212.753	0.016662
BRO	−212.785	−212.724	−212.754	−212.755	0.017618
IS&BRO–AC	−212.249	−177.425	−189.152	−187.723	8.315499

6.2. Statistical analysis on delay, distance, and RSSI

Tables 2, 3, and 4 depict the study on delay, distance, and RSSI via IS&BRO–AC oriented model over WCA [28], HM-MPR [14], SSA, SSOA, CA, SMA, and BRO. The met heuristic schemes are stochastic, and to substantiate its fair evaluation, each model is analyzed quite a lot of times to accomplish less delay, distance, and high RSSI. For minimum case, a less delay of $1.65 \times 10^{- 07}$ is gained using IS&BRO–AC, while WCA [28], HM-MPR [14], SSA, SSOA, CA, SMA, and BRO have acquired comparatively high delay of $2.17 \times 10^{- 07}$ , $3.81 \times 10^{- 07}$ , $3.81 \times 10^{- 07}$ , $3.81 \times 10^{- 07}$ , $1.96 \times 10^{- 07}$ , $3.81 \times 10^{- 07}$ and $3.81 \times 10^{- 07}$ for minimum case for node count of 100. Also, for the median case, the IS&BRO–AC gained fewer values for distance and delay. Moreover, a high RSSI of −212.661 is obtained for a node count of 100. Thus, it is proven that the adopted IS&BRO–AC approach offered lesser distance and delay with high RSSI over WCA [28], HM-MPR [14], SSA, SSOA, CA, SMA, and BRO. The improvements in CHS and improved LSTM offered minimal delay, distance, and high RSSI for the proposed mobility prediction.

6.3. Convergence study

The convergence rate of IS&BRO–AC algorithm over WCA [28], HM-MPR [14], SSA, SSOA, CA, SMA, and BRO is exposed in Fig. 5. In Fig. 6, analysis is done for 100 nodes and 200 nodes. Here, a lesser cost of 0.45 is achieved by IS&BRO – AC from the 3rd to 8th iterations when the node count is 200. While the node count is 100, a lesser cost of 0.57 is achieved by IS&BRO–AC from the 5th to 8th iterations. Thus, a node count of 200 exposed lesser values than a node cunt of 100. During the prime iterations, even the IS&BRO–AC approach revealed high-cost values. Nevertheless, with raise in iterations, the cost is lessened, and at last, the less cost is attained by IS&BRO–AC. The improvements in CHS and improved LSTM offered better results for the proposed mobility prediction.

Fig. 5.

Convergence study of IS&BRO–AC based mobility prediction for nodes (a) 100 and (b) 200.

6.4. Analysis of energy and PDR

The assessment of proposed IS&BRO–AC over WCA [28], HM-MPR [14], SSA, SSOA, CA, SMA, and BRO regarding energy and PDR are given in Fig. 6. In the case of energy from Fig. 6, it is noted that energy is narrowed with raise in the count of rounds, nevertheless, IS&BRO–AC offered high energy over WCA [28], HM-MPR [14], SSA, SSOA, CA, SMA, and BRO. At around 0, the energy is superior by about 0.58, but with an increase in rounds, the energy is decreased to about 0.43 when round is 2000. Similar outcomes are observed for IS&BRO–AC when node counts are 200. In Fig. 7, PDR is computed. The PDR for IS&BRO–AC is high than WCA [28], HM-MPR [14], SSA, SSOA, CA, SMA, and BRO. Particularly, the PDR at the 2000th round is high (0.992) than at other rounds for a node count of 100. For a node count of 200, the PDR from the 500th round to the 700th round is higher (0.993) than at other rounds. The improvements in CHS with improved distance evaluation and improved LSTM offered better energy and PDR for the developed mobility prediction model.

Fig. 6.

Energy analysis on nodes (a) 100 and (b) 200.

Fig. 7.

PDR analysis on nodes (a) 100 and (b) 200.

6.5. Analysis on throughput

The analysis of throughput using IS & BRO–AC is specified in Fig. 8 for 100 and 200 nodes. Here, the throughput in Fig. 8(a) is fluctuating for all rounds. Particularly, high throughput of 3650 is gained by IS&BRO–AC at the 1500th round, while WCA [28], HM-MPR [14], SSA, SSOA, CA, SMA, and BRO gained relatively less throughput. A minimal throughput of 2900 is observed for WCA at the 2000th round for 100 nodes. The throughput graph in Fig. 8 (b) is also showing fluctuating outputs for all rounds. At the 500th round, high throughput of 3680 is observed using IS&BRO–AC, which is higher than WCA [28], HM-MPR [14], SSA, SSOA, CA, SMA, and BRO. In the case of 200 nodes, HM-MPR [14] and CA have exposed minimal throughput values than WCA [28], SSA, SSOA, SMA, and BRO. The improvements in CHS with improved distance evaluation and improved LSTM offered better throughput for the developed mobility prediction model.

Fig. 8.

Analysis of throughput for nodes (a) 100 and (b) 200.

7. Conclusion

This research work developed a new reliable MANET routing model, where, the initial phase was predictor node selection. Here, AWCA was adopted for predictor node selection based on parameters like distance, security (risk), RSSI, PDR, and energy. The node with a smaller weight was chosen as CH using IS&BRO–AC model. Moreover, mobility prediction was done, during which the node mobility was predicted using improved LSTM on considering RSSI and distance. Based upon the prediction, reliable data transmission was carried out that ensured better reliable MANET routing. In the end, analysis was done on RSSI, PDR, and so on. A minimal throughput of 2900 was observed for WCA at the 2000th round for 100 nodes. The throughput graph was also showing fluctuating outputs for all rounds. At the 500th round, high throughput of 3680 was observed using IS&BRO–AC, which was higher than WCA, HM-MPR, SSA, SSOA, CA, SMA, and BRO. In the case of 200 nodes, HM-MPR and CA have exposed minimal throughput values than WCA, SSA, SSOA, SMA, and BRO. The advantage of the proposed model is performance efficiency is high and the disadvantage is needed to consider the routing tables to reduce the data exchange in MANET. In the future, time analysis has to be concerned more.

References

https://logicplum.com/knowledge-base/manhattan-distance/.

https://en.wikipedia.org/wiki/Received_signal_strength_indication.

S.A.

Alghamdi, Three-tier architecture supporting QoS multimedia routing in cloud-assisted MANET with 5G communication (TCM5G), Mobile Netw Appl 25 (2020), 2206–2225. doi:10.1007/s11036-020-01657-6.

Amuthan,

Sreenath,

Boobalan and

Muthuraj, Dynamic multi-stage tandem queue modeling-based congestion adaptive routing for MANET, Alexandria Engineering Journal (3 May 2017).

M.M.

Beno,

I.R.

Valarmathi,

S.M.

Swamy and

B.R.

Rajakumar, Threshold prediction for segmenting tumour from brain MRI scans, International Journal of Imaging Systems and Technology 24(2) (2014), 129–137. doi:10.1002/ima.22087.

Bhardwaj and

El-Ocla, Multipath routing protocol using genetic algorithm in mobile ad hoc networks, IEEE Access 8 (2020), 177534–177548. doi:10.1109/ACCESS.2020.3027043.

Brindha,

Karthikeyan and

Manimegalai, Fuzzy enhanced secure multicast routing for improving authentication in MANET, Cluster Comput 22 (2019), 9615–9623. doi:10.1007/s10586-017-1282-9.

Chaudhry,

Tapaswi and

Kumar, Forwarding zone enabled PSO routing with network lifetime maximization in MANET, Appl Intell 48 (2018), 3053–3080. doi:10.1007/s10489-017-1127-5.

Chen,

Zhou,

Wu and

Cheng, An adaptive on-demand multipath routing protocol with QoS support for high-speed MANET, IEEE Access 8 (2020), 44760–44773. doi:10.1109/ACCESS.2020.2978582.

10.

Devagnanam and

N.M.

Elango, Optimal resource allocation of cluster using hybrid grey wolf and cuckoo search algorithm in cloud computing, Journal of Networking and Communication Systems 3(1) (2020), 31–40.

11.

V.S.

Devi and

N.P.

Hegde, Multipath security aware routing protocol for MANET based on trust enhanced cluster mechanism for lossless multimedia data transfer, Wireless Pers Commun 100 (2018), 923–940. doi:10.1007/s11277-018-5358-5.

12.

M.K.

Dholey,

Sinha,

Mukherjee,

A.K.

Das and

S.K.

Sahana, A novel broadcast network design for routing in mobile ad-hoc network, IEEE Access 8 (2020), 188269–188283. doi:10.1109/ACCESS.2020.3030802.

13.

Durr-e-Nayab ,

M.H.

Zafar and

Altalbe, Prediction of scenarios for routing in MANETs based on expanding ring search and random early detection parameters using machine learning techniques, IEEE Access 9 (2021), 47033–47047. doi:10.1109/ACCESS.2021.3067816.

14.

N.S.S.

Farheen and

Jain , Improved routing in MANET with optimized multi path routing fine tuned with hybrid modeling, Journal of King Saud University – Computer and Information Sciences.

15.

T.R.

Farshi, Battle royale optimization algorithm, Neural Comput & Applic 33 (2021), 1139–1157. doi:10.1007/s00521-020-05004-4.

16.

Garaaghaji and

Alfi, Robust performance rate control to enhance MANET networks routing issue, J. Electr. Eng. Technol. 15 (2020), 477–486. doi:10.1007/s42835-019-00218-6.

17.

Hao,

Zhang and

Song, A stable and energy-efficient routing algorithm based on learning automata theory for MANET, J. Commun. Inf. Netw. 3 (2018), 52–66. doi:10.1007/s41650-018-0012-7.

18.

Hemalatha,

Umamaheswari and

Jothi, LF distribution and equilibrium optimizer based fuzzy logic for multipath routing in MANET, Wireless Pers Commun 120 (2021), 1837–1861. doi:10.1007/s11277-021-08537-6.

19.

Hurley-Smith,

Wetherall and

Adekunle, SUPERMAN: Security using pre-existing routing for mobile ad hoc networks, IEEE Transactions on Mobile Computing 16(10) (2017), 2927–2940. doi:10.1109/TMC.2017.2649527.

20.

S.C.

Jalade and

N.B.

Patil, Adaptive deep Runge Kutta Garson’s network with node disjoint local repair protocol based multipath routing in MANET, Evolving Systems (2023), 1–25.

21.

S.S.

Jamaesha and

Bhavani, A secure and efficient cluster based location aware routing protocol in MANET, Cluster Comput 22 (2019), 4179–4186. doi:10.1007/s10586-018-1703-4.

22.

V.D.

Khairnar and

Kotecha, Simulation-based performance evaluation of routing protocols in vehicular ad-hoc network, International Journal of Scientific and Research Publications 3(10) (2013), ISSN 2250-3153.

23.

Khan,

Shankar,

Jaisankar,

Balamurugan and

Patan, A hybrid model for security-aware cluster head selection in wireless sensor networks, IET Wireless Sensor Systems (2018). doi:10.1049/iet-wss.2018.5008.

24.

B.H.

Khudayer,

Anbar,

S.M.

Hanshi and

Wan, Efficient route discovery and link failure detection mechanisms for source routing protocol in mobile ad-hoc networks, IEEE Access 8 (2020), 24019–24032. doi:10.1109/ACCESS.2020.2970279.

25.

Kim,

Na and

Cho, Dual-channel-based mobile ad hoc network routing technique for indoor disaster environment, IEEE Access 8 (2020), 126713–126724. doi:10.1109/ACCESS.2020.3008682.

26.

S.V.

Kumari and

Paramasivan, Defense against sybil attacks and authentication for anonymous location-based routing in MANET, Wireless Networks 23(3) (2017), 715–726. doi:10.1007/s11276-015-1178-7.

27.

Li,

Chen,

Wang,

A.A.

Heidari and

Mirjalili, Slime mould algorithm: A new method for stochastic optimization, Future Generation Computer Systems 11 (2020).

28.

C.R.

Mamatha and

Ramakrishna, Mobility prediction based energy efficient QoS routing, International Journal of Future Generation Communication and Networking 13(1) (2020), 716–731.

29.

Nallusamy and

Sabari, Particle swarm based resource optimized geographic routing for improved network lifetime in MANET, Mobile Netw Appl 24 (2019), 375–385. doi:10.1007/s11036-017-0911-0.

30.

Nandi and

Anusha, An optimized and hybrid energy aware routing model for effective detection of flooding attacks in a manet environment, Wireless Pers Commun (2021). doi:10.1007/s11277-021-09079-7.

31.

O.S.G.

Prakasi and

Varalakshmi, Decision tree based routing protocol (DTRP) for reliable path in MANET, Wireless Pers Commun 109 (2019), 257–270. doi:10.1007/s11277-019-06563-z.

32.

S.J.G.

Rajadurai,

Veerappan and

Ramasamy, Optimization of multicast ad hoc on-demand routing protocol based on genetic algorithm with backup paths in MANET, Wireless Pers Commun 94 (2017), 2095–2124. doi:10.1007/s11277-016-3363-0.

33.

Sarhan and

Sarhan, Elephant herding optimization ad hoc on-demand multipath distance vector routing protocol for MANET, IEEE Access 9 (2021), 39489–39499. doi:10.1109/ACCESS.2021.3065288.

34.

Satheeshkumar and

Sengottaiyan, Defending against jellyfish attacks using cluster based routing protocol for secured data transmission in MANET, Cluster Computing (2017), 1–12.

35.

F.H.

Shajin and

Rajesh, Trusted secure geographic routing protocol: Outsider attack detection in mobile ad hoc networks by adopting trusted secure geographic routing protocol, International Journal of Pervasive Computing and Communications 18(5) (2022), 603–621. doi:10.1108/IJPCC-09-2020-0136.

36.

S.K.M.

Shareef and

R.S.

Rao, A hybrid learning algorithm for optimal reactive power dispatch under unbalanced conditions, Journal of Computational Mechanics, Power System and Control 1(1) (2018), 26–33.

37.

Tabatabaei, A new routing protocol for energy optimization in mobile ad hoc networks using the cuckoo optimization and the TOPSIS multi-criteria algorithm, Cybernetics and Systems 52 (6) (2021), 477–497. doi:10.1080/01969722.2021.1899597.

38.

Theerthagiri, ReCoMM: Resource-aware cooperation modelling using Markov process for effective routing in mobile ad hoc networks, Sādhanā 46 (2021), 209. doi:10.1007/s12046-021-01743-9.

39.

Theerthagiri, Mobility prediction for random walk mobility model using ARIMA in mobile ad hoc networks, The Journal of Supercomputing 78(14) (2022), 16453–16484. doi:10.1007/s11227-022-04503-6.

40.

Thomas and

M.J.S.

Rangachar, Hybrid optimization based DBN for face recognition using low-resolution images, Multimedia Research 1(1) (2018), 33–43.

41.

S.A.

Thorat and

P.J.

Kulkarni, Uncertainty analysis framework for trust based routing in MANET, Peer-to-Peer Netw. Appl. 10 (2017), 1101–1111. doi:10.1007/s12083-016-0510-z.

42.

B.K.

Tripathy,

S.K.

Jena,

Bera et al., An adaptive secure and efficient routing protocol for mobile ad hoc networks, Wireless Pers Commun 114 (2020), 1339–1370. doi:10.1007/s11277-020-07423-x.

43.

Vadivel and

V.M.

Bhaskaran, Adaptive reliable and congestion control routing protocol for MANET, Wireless Netw 23 (2017), 819–829. doi:10.1007/s11276-015-1137-3.

44.

G.N.

Vivekananda,

P.C.

Reddy and

Ilknur, A congestion avoidance mechanism in multimedia transmission over MANET using SCTP multi-streaming, Multimed Tools Appl 79 (2020), 16823–16844. doi:10.1007/s11042-019-7260-x.

45.

Zhang,

Zhao and

Cheng, Multipath routing and MPTCP-based data delivery over manets, IEEE Access 8 (2020), 32652–32673. doi:10.1109/ACCESS.2020.2974191.

46.