1. Introduction
Unmanned aerial vehicles (UAVs) have received much attention in the current decade because of their ability to assist humans in a variety of tasks. A Flying Ad-Hoc Network (FANET) is a critical component of an ad hoc network and UAVs are commonly used to form FANETs. UAVs are small, self-contained drones that can be controlled remotely [
1,
2]. UAVs have been employed in a variety of industries, including warfare, agriculture, medicine, photography, and environmental applications, among others. UAVs were initially only employed by the military for surveillance and rescue missions. Nowadays, with the advancement in technology, UAVs are widely employed in a variety of fields, including product shipping and delivery, soil analysis, agricultural monitoring, and so on [
3,
4]. With the introduction of FANETs, a variety of applications have emerged, including a cargo of goods, domestic package delivery, crop monitoring, agricultural surveillance, and rescue operations. The latest techniques of FANETs are mobility, modeling, and theoretical proofs of communications in FANETs; traffic models and network control for FANETs; security, privacy, and trust in FANETs; performance, scalability, reliability, and efficiency of FANETs; UAV-assisted packages; energy efficiency in FANETs; the emerging Internet of Things (IoT) and cloud applications with FANETs; routing strategies for FANETs; application of future Internet architectures in FANETs; the energization of FANETs. Furthermore, this work explores into the modeling and performance evaluation of drones and UAVs in emerging applications, describing novel strategies, new models for energy, communication, and routes for UAVs, as well as interesting applications.
Load balancing is the act of dispersing system traffic flow over various servers, and localization is the process of estimating the location of unknown nodes placed in FANETs. As computing moves more and more to the network, load balancing plays an increasingly critical security function [
5]. In such networks, the accuracy of the node’s position estimation is determined by two key processes. The first is node location estimation, which involves calculating the position of unidentified nodes, and the second is node position confirmation, which involves comparing the computed location to the real site. To expand the correctness of the localization process, several expressions have been developed.
Congestion is always an undesirable circumstance in wired and wireless networks since it can degrade the communication environment. At the media access control address level or above this level, congestion can cause packet loss and retransmission. The task of planning and executing a traffic congestion control algorithm along with a firefly algorithm is difficult since numerous elements must be considered. The congestion control system is not equipped to deal with the unique characteristics of shared wireless multi-UAVs in the network. Packet loss and retransmissions can occur for a variety of reasons, including route pause due to mobility and concealed terminal problems in the wireless networks. Unpredictable packet delivery ratio and the loss of packet rates are caused by affected fluctuations in topology and a weak wireless network, posing a challenge for congestion control in FANETs [
6]. Furthermore, due to the random link-state induced by UAVs’ fast mobility, it is hard to control overall transmission delay within a specific threshold. Consequently, it is required to plan an accurate traffic congestion control of flying nodes algorithm that can adjust to the different status of the link, and guarantees the interruption prerequisites and different parameters.
To the best of our knowledge, only a few earlier studies have looked into the load-balancing problem in FANETs. In [
3], the flying ad hoc networks are described along with mobility models, features, and routing protocols. In [
7], the Geographical Position Mobility Oriented Routing protocol (GPMOR) is described where the main aim is to reduce the number of hops based on the Gauss-Markov (GM) mobility model, which further improved the performance of routing by efficient packets. In summary, we make two contributions:
For FANETs, we propose the specific functions F1 and F2 with the optimization process by taking different parameters such as end-to-end delay (EED), packet delivery ratio (PDR), fuel emission, and throughput. The technique is used to resolve the constraints of the optimization problem with the firefly algorithm, which is used to estimate the exact match of the dynamic network topology. The primary problem is therefore converted into a distributed solvable problem, allowing senders to compute the attractiveness of flying nodes to execute congestion control.
To reach the best solution, we present a distributed traffic congestion control algorithm that incorporates the delay constraints. We propose Rs, Rd, and Rp variables for all flying nodes to verify the incoming flow of the flying nodes and outgoing flow of the flying nodes probability to exploit network utilization and decrease transmission delay in a circulated manner. Finally, we examine the optimization method’s performance and demonstrate its convergence using a simulator.
In this paper, we use the GPMOR protocol and the firefly algorithm to develop a load-balancing technique for position estimation in FANETs.
Section 2 discusses related research in this field. The suggested network model, shown in
Section 3, is completed in two phases: (i) optimizing the flying nodes using the firefly method in
Section 3.1; (ii) using the traffic congestion control algorithm to predict the position of all unknown nodes in
Section 3.2.
Section 4 contains the findings and analyses. Lastly,
Section 5 concludes the research paper.
2. Related Work
There are several assumptions such as device hardware, network configuration, operating setting, signal propagation models, time management, communication costs, error requirements, timing, energy requirements, and node mobility used to address the localization problem in FANET.
Some studies formalized congestion control as an effectiveness maximization issue [
8], which was then explained using optimization approaches to diminish network congestion and to propose an innovative cross-layer architecture for optimizing congestion control, direction-finding, and control in ad hoc networks, to address the routine limits imposed by the nonexistence of collaboration between layers. The research paper [
9] proposed a framework for wireless networks based on the concept of utility maximization of the network that considers routing and congestion as well as power controls all at the same time while taking into account the continuing or temporary random attenuation of ad hoc networks. Khodaian et al. [
10] observed the delay in multiple access networks, addressed the delay-based effectiveness maximization issue, and tried to find the best balance of delay, PDR, and energy parameters. Li et al. [
11] explained congestion control in two modes, complex mode and non-complex mode of delay traffic, and proposed an innovative decentralized system for the same problem. Under predicted contact limits, [
12] investigated the combined enhancement of power control as well as congestion control in other ad hoc networks. A disseminated cross-layer enhancement framework was planned at the same time to meet the simultaneous design aim of power control and blockage control from one source to another node at the physical layer and transport layer, respectively.
In [
13], authors discussed a new paradigm, software-defined networking, for the load-balancing technique to deal with the issue of load irregularity in the organization, and they planned unified programming characterized by LTE remote access network structure. To overcome the problem of throughput expansion, [
14] suggested an algorithm for congestion control-based scheduling. In heterogeneous QoS-aware applications, [
15] solves the rate assignment problem to optimize the system value of closure in a certain time frame and enable every data flow to satisfy the criteria of lengthy average end-to-end delay constraints. Alaei et al. [
16] suggested a distributed congestion control strategy that holds the congestion issue in wireless multimedia sensor networks. When it is discovered, a local binary tree is created at the congestion link to resolve the issue. The established tree is used to transfer packets to the next possible congestion node, whereas the mobile sink node is used to facilitate communication when the established tree is unable to reduce congestion.
Silva et al. [
17] investigated congestion management in time lag and disruption-tolerant networks (DTNs) and a novel disruption-tolerant network congestion control system that modifies its operation dynamically based on the underlying network aspects. The authors of [
18] proposed a novel, impartial and dispersed congestion control method for neighborhood area networks to assure network access equality. Lubna et al. [
19] proposed improving throughput by applying a unique strategy for dynamically controlling the reduction factor of the congestion control algorithm in multipath TCP based on the interval between packet losses. Verma et al. [
20] investigated a novel congestion control policy for the Internet of Things, which aims to reduce packet drops and improve throughput by promptly adapting the data transfer rate whensoever the existing bandwidth and delay changes. The description of traffic congestion control with parameters of FANET is shown in
Table 1 below:
4. Results and Discussion
Simulations have been run to evaluate the performance of the proposed scheme. Because standard FANET simulator tools have limitations, realistic mobility models are very useful in designing and implementing routing protocol performance. To reflect the effects of a routing protocol based on environmental variables, a flying mobility model should be proactively easily adjustable. The results are validated using the NS2 simulator as a network simulator and the cbrgen utility to generate traffic files. Setdest is a program in the NS2 simulator that has an average pause of 2 s between node movements. The simulation is terminated after 200 s, and the topology perimeter is defined as 500 × 500.
In this paper, we use the NS2 simulator to pretend the network scenario. NS-2 packages are made up of the following components:
geo utility.h contains geometrical utility functions such as the projection of a 3D graph to a 2D graph and the communications network between two flying nodes;
geo pkt.h contains the new geo packet header definition;
geo node.h and geo node.cc files define and implement the geographic node;
geo.h and geo.cc files contain the definition and implementation of the geographic agent prototype;
the proposed algorithms are defined and implemented in geo next node.h and geo next node.cc.
In the beginning phase, flying nodes are dispersed arbitrarily, the altitude of UAVs is 40 m, and the directional gain is 10 dBi with a frequency range of 2.4 GHz. The value of the transmission power is 0.005 W for each session is set to the speed of the UAVs, which varies up to 60 m/s. In addition, we use the queue type as the priority queue to simulate a wireless physical medium channel and further to estimate the link quality of the nodes. The detailed definitions of simulation parameters can be found in
Table 3. The goal of this study is to regulate the congestion level of the entire network by satisfying various characteristics of flying nodes.
The network is initialized with the help of multiple flying nodes with the altitude of UAVs at 70 m. The transmission power is constrained by the connectivity between terrestrial base stations and UAVs. To prevent these, UAVs can communicate with one another using purely ad hoc architecture. In
Figure 1, the initialization of the flying nodes is defined. The syntax for defining a node is set as n0 (
$ns node). We created a node that is represented as node 0 by the variable n0. When we refer to that node in the code, we will use
$n0. After we have defined a few nodes, we can define the links that connect them, and so on. In ns2, a node’s output queue is implemented as a component of each link whose input is that node. The definition of the link then includes a method for dealing with queue overflow.
Figure 2 depicts the specific source and destination nodes as defined by RoadSide Units (RSU). Furthermore, we need to connect the traffic source to the traffic sink. This maintains the RSU’s proper infrastructure for sending route packets to flying nodes. It manages traffic signals; the RSU is intended for use at road intersections and serves as a resource for information.
Furthermore, such wireless communication may be used to enable multi-node communications and other applications if a data packet needs to be delivered to another node that is outside of the range. In the network, the node with name 26 and node 30 are described as the root nodes mentioned in
Figure 3.
In such a case, each node selects a random destination, then travels with a random velocity and pauses at the destination. When the stop time expires, the node chooses a random destination with a random velocity and a similar pause duration based on set probability. Furthermore, the root node sends a Route Reply Packet (RRP) to the next neighbor node as shown in
Figure 4. A few entries of statistical data from the source node to the destination with appropriate time are shown in
Figure 5 below.
In the simulation, a large number of flying nodes with defined direction or speed changes are used. We investigated various degrees of flying node density, velocity, and network activity from source to destination during the simulation.
Furthermore, in
Figure 6, the source node starts sending data to the furthest node, accordingly, to identify the traffic in the entire network. Then, the source node multicast route request packets (
Figure 7) and the root node again sends a Route Reply Packet to the next neighbor node (
Figure 8). The neighbor node’s data with the exact distance value are shown in
Table 4 below. Finally, the source node starts sending data to the next proceeding nodes to achieve the target for traffic balance in-network as shown in
Figure 9.
Different performance metrics (parameters) are to be taken further in this research work such as delay analysis, fuel emission analysis, packet delivery ratio analysis, and throughput analysis of flying nodes. There is a comparison between two routing protocols such as GPMOR and Greedy Perimeter Stateless Routing (GPSR).
Delay: The average time it takes for data packets to move across the network from the source flying node to the destination flying node is referred to as delay. The delay of a communications network is a significant design and performance aspect. The processing and transmission delays of a network link are all included in end-to-end delays.
Figure 10 depicts the working of a network in terms of end-to-end delay when the number of UAVs, speed, and area magnitudes are varied. The X-axis depicts the time in m/s, while the Y-axis denotes the delay in seconds. The end-to-end delay decreased with the number of UAV nodes. This is because packets are more likely to be routed rather than captured in the suspension buffer. Once the delay for each of these measures was compared, GPSR (purple bar) had the longest delay, even when the region was smaller. This was because when a route request was given, the destination answered to every RREQs that it received, making determining the least populated route take longer. When compared to the GPSR protocol, GPMOR had the shortest delay.
Fuel Emission: In fuel emission, to accomplish connection dependability and the quantity of stored fuel in terms of energy and input buffer, the system chooses a route based on the present processing status of a node.
Figure 11 shows that GPSR released more fuel than the GPMOR technique. Several requirements must be satisfied, such as the minimum fuel necessary to process packets in kilobytes. The current processing state of nodes calculated in terms of fuel and input buffer. Node priority is based on a threshold value route selection which is in terms of fuel and if a node meets the threshold criteria to participate in routing. To avoid a node becoming a bottleneck, the optimal information capacity of a metric node concerning traffic and remaining fuel is employed.
Packet Delivery Ratio (PDR): A network generates certain data packets, which are then delivered through a routing mechanism. A data packet is considered delivered when it is received in full and without loss by the destination node.
In the simulation results analysis, we discovered that the network connectivity giving a packet delivery ratio of more than 95% is dependent on the network characteristics of the GPMOR protocol as opposed to the GPSR protocol.
Figure 12 shows the simulation results of a network of varied nodes with the transmission power of a flying node configured.
As the simulation results show, the number of packets should range from 0 to 120, and the duration should range from 0 to 12 m/s, to ensure the requisite connection between flying nodes. This is represented by two lines in the outcome, the purple and green. The green line denotes the maximum amount of data packets that must be sent to the destination. In this case, increasing the number of nodes does not enhance data quality according to the GPSR protocol; however, it does improve the data quality of flying nodes according to the suggested protocol (GPMOR). As a result, the appropriate packet delivery ratio values were obtained using the simulation settings and network configuration utilized. When the number of nodes is increased to 120, simulation results demonstrate that FANET connection with a packet delivery ratio greater than 95 percent is obtained.
Throughput: Throughput is an important measure for measuring network performance. Throughput can be affected by the distance between the transmitter and the receiver. Throughput is defined as the average data probability of a successful data packet or message passing across a communication connection from the source flying node to the destination flying node in a given time unit. Because the flying nodes’ positions may be changed, the distance between two nodes can be modified, and the capacity of the related link can be tuned to increase network throughput. Here, each flying node provides its position and user location information to the ground station, which utilizes all of the flying nodes’ current positions.
Figure 13 depicts the analysis of a network in connection with throughput when the number of UAVs, area sizes, and speed is varied. The X-axis denotes the time of simulation in m/s, whereas the Y-axis denotes throughput in bits per second (bps). The network’s throughput grew as the number of UAVs increased, as did its performance.
Thus, when UAVs’ speeds were reduced to 0 m/s and 12 m/s, the GPMOR protocol beat the GPSR protocol, with the number of UAVs growing to almost 100, as shown in
Figure 13. This is because GPMOR allocates time slots for packet transfers to prevent network congestion.