Intelligent Greedy Perimeter Stateless Routing Scheme for Unmanned Aerial Vehicles

BAIG Mohsin Ayub(貝 歌), ZHANG Yihong(張義紅), 2*, YE Xiaoxian(葉曉嫻)， SULTANA Umbrin(彩 虹), AKBAR Sunila(海 榮)， LI Shuai (李 帥), AHMAD Sohail (阿門德)

1 College of Information Science and Technology, Donghua University, Shanghai 201620, China2 Engineering Research Center of Digitized Textile & Fashion Technology, Ministry of Education, Donghua University, Shanghai 201620, China3 School of Art and Design, Shanghai University of Engineering Science, Shanghai 201620, China4 Department of Electrical Engineering, NED University of Engineering and Technology, Karachi 75270, Pakistan

Abstract: The dynamic behavior, rapid mobility, abrupt changes in network topology, and numerous other flying constraints in unmanned aerial vehicle (UAV) networks make the design of a routing protocol a challenging task. The data routing for communication between UAVs faces numerous challenges, such as low link quality, data loss, and routing path failure. This work proposes greedy perimeter stateless routing (GPSR) based design and implementation of a new adaptive communication routing protocol technique for UAVs, allowing multiple UAVs to communicate more effectively with each other in a group. Close imitation of the real environment is accomplished by considering UAVs’ three-dimensional (3D) mobility in the simulations. The performance of the proposed intelligent greedy perimeter stateless routing (IGPSR) scheme has been evaluated based on end-to-end (E2E) delay, network throughput, and data loss ratio. The adapted scheme displayed on average 40% better results. The scenario has been implemented holistically on the network simulator software NS-3.

Key words: flying ad-hoc network (FANET); vehicular ad-hoc network (VANET); unmanned aerial vehicle (UAV); greedy perimeter stateless routing (GPSR); intelligent greedy perimeter stateless routing (IGPSR)

Introduction

Recent developments in small-sized unmanned aerial vehicles (UAVs) or drones have enabled them to be used in versatile applications ranging from the civilian domain to the defense sector. The UAV technology has provided noticeable benefits in rescue operations, real-time monitoring, exploring undiscovered places, traffic monitoring, security, and events coverage. Moreover, they are beneficial at remote sites where humans cannot directly access[1]. Furthermore, the UAVs are quite essential in natural disasters (earthquakes, floods, heavy rainfall, and windstorm), as they can be used as an ad-hoc network to restore the broken communication network in disastrous regions and help connect the affected communication network with the nearest cellular infrastructure[2-3]. It has been found that the potential of multiple UAVs’ network is much more than that of a single UAV[4]. Besides, remarkable modern applications of UAVs[5-8]have increased their importance more than ever, making them an essential need of contemporary societies. A group of UAVs can perform the given task rather efficaciously and competently than a single UAV[9-12]. To achieve a given operation successfully in a group, a robust communication routing protocol is utmost important. It allows multiple UAVs to connect wirelessly and enables them to communicate and share information back and forth during the flight. More than one UAV (where each UAV represents a node) is connected without following any fixed infrastructure in space, which is known as a flying ad-hoc network (FANET). In FANET, each node is self-configured and self-managed. The nodes can communicate with each other directly or by multiple hops. To send information from the source node to the destination node, nodes along the way act as routers and assist data in arriving at the destination node by following multiple hops[13]. Therefore, nodes’ availability is crucial for efficiently transferring data packets.

Unlike the vehicular ad-hoc network (VANET), the movement of nodes is two-dimensional (x,y), and FANET movement of nodes is three-dimensional (3D) (x,y,z).Moreover, the movement of UAV nodes in the air is highly dynamic and it is faster than the mobile ad-hoc network (MANET)[14-16]. Due to these constraints, designing a robust communication routing protocol for UAVs becomes a challenging job for scientists and researchers[17]. Many attempts have been made in the past in which protocols designed for VANETs are directly suggested for FANET without considering the 3D movement of nodes in FANET[18-19]. In this work, we proposed an adaptive intelligent greedy perimeter stateless routing (IGPSR) model based on a greedy perimeter stateless routing(GPSR) protocol. The suggested adaptive approach aims to improve the communication of mobile UAVs and create a robust UAV network. The new approach brings changes in the existing GPSR strategy. For instance, the classical GPSR method only uses a right-hand data forwarding scheme in the recovery mode, whereas the prescribed IGPSR technique considers the dual-mode data-carrying scheme. The other change includes the introduction of a new data routing table named an intelligent table (IT), which works along with the existing nearby inventory table (NIT).

By working in collaboration, ITs and NITs maintain the data record of the source and destination pairs and do not allow data packets to get trapped between multiple nodes. Moreover, they help the nodes to find optimized paths for the data packets to be transmitted to the destination nodes following relatively fewer hops. These alterations give intelligence to each node operating in the network that subsequently reduce end-to-end (E2E) delay and data loss ratio and increase network throughput.

The biggest advantage of GPSR is that nodes store the least amount of information and can quickly respond to changes in topology. This work proposes a routing protocol that stores detailed information. However, results show that, even though the nodes in the IGPSR technique store more information than those in the GPSR, the adaptive IGPSR technique has produced much better results compared to the classical GPSR protocol while both approaches are compared on metrics: E2E delay, network throughput, and data loss ratio.

1 Related Research Work

The UAVs have been continuously investigated in the past several years, countless pieces of research and surveys have been conducted, and different routing protocols have been suggested so far. Sangetal.[20]discussed routing protocols that emerged in recent years for UAVs and compared them with each other in detail. Bekmezcietal.[21]tried to create an infrastructure less communication among multiple UAVs with the help of satellites. Another prominent research has been carried out to create an effective routing protocol for UAVs. In this work, after getting inspired by bees’ movements, Leonov[22]suggested a nature-inspired routing algorithm for UAVs in which nodes followed the ditto movements like bees. Another similar kind of study had been done[23], where de Rango et al.[23]suggested a light way bio-inspired coordination routing protocol for UAVs. However, this protocol was especially recommended for agriculturally based applications where sudden climate change could be an important issue for the farmers.

Sangetal.[24]presented an energy-efficient opportunistic routing protocol based on trajectory prediction (EORB-TP) for UAVs. In this work, the possibility of future time moving speed of a node has been analyzed, and limited energy and buffer of nodes have been taken into account in the simulations. Another related effort has been put recently into designing a UAV routing strategy with the name, future network topology-aware routing (FNTAR) protocol[25]. In this attempt with the aid of a ground positioning system (GPS), the future position of the UAV nodes has been utilized to select the best path for the data packet to arrive at the destination node faster. Moreover, data loss problems caused by unstable link quality and dynamic network topology have also been resolved. Furthermore, Bahlouletal.[26]proposed a flocking-based demand routing protocol called Boids of Reynolds ad-hoc on-demand distance vector (BR-AODV) for UAVs. To transmit data to the next node and to maintain routing path and connectivity, Boids and Reynolds method[26]has been used. In addition, for proactive drones, automatic discovery of ground base station has been presented. Likewise, another scheme based on an ad-hoc on-demand vector has been proposed. Tanetal.[27]presented a simpler and more secure routing protocol algorithm for UAVs called improved security ad-hoc on-demand distance vector (ISAODV). By means of simulations, it has been proved that the newly introduced method can perform better compared to the former method security ad-hoc on-demand distance vector (SAODV).

On the flip side, quite a few studies and comparisons have been carried out in which routing protocols ad-hoc on-demand distance vector (AODV), destination-sequenced distance-vector (DSDV), dynamic source routing (DSR), dynamic source routing (OLSR), ad-hoc on-demand multipath distance vector (AOMDV), and hybrid wireless mesh protocol (HWMP) designed for the mobile ad-hoc network have been directly suggested for UAVs without taking into account the 3D movement of nodes[19, 28-29]. The UAV network topology changes rapidly, which is directly proportional to the quality of E2E delay, network throughput, and data loss ratio[30]. Therefore, an optimal mobility model is inevitable in order to analyze the network behavior realistically. The presented IGPSR model is compiled with the 3D Gauss Markov mobility model, and designed to imitate the 3D movement of nodes in a 3D rectangular space.

GPSR is a position-based routing protocol, and sends a packet from the source node to the destination node. It selects the nearest node in terms of distance and transfers the packet to the destinations by the following multiple hops. In a network, every node also acts as a router and assists its neighbors to transmit data to the destination[31]. In the GPSR network, every single node is equipped with a GPS[32].

The GPS is designed to find the location of objects on earth. Hence is not appropriate for finding the position or location of objects in space. However, the continuous development in the field of tracking and positioning of space objects has made it possible to locate the exact position or location of moving objects in space. In this case, each UAV has a built-in relative positioning system[33], which helps every node to locate and determine its position in space. The location information is shared by beacons from the medium access control (MAC) layer. Each node with the help of a beacon algorithm broadcasts hello packets (which contains node IDs and their coordinates’ information) at a regular interval of time, to help their neighbors keep their routing tables up to date[34].

Principally, the GPSR protocol routing technique uses dual-mode data-carrying strategies: greedy forwarding strategy and perimeter forwarding strategy or recovery mode. Coordinate information related to the position of each node is assumed to be acquired through the medium of localization or positioning systems (GPS/RPS). After assessing the beacon messages, the actual node checks the status of the mode in which the last data packet is received. If it is greedy forward, the actual node will begin seeking the nearest node to pass on the data packet which should be closer to the destination node than the actual node itself. If the existing node does not get feedback from the nearby node, it will reckon as if the communication link is broken and will remove all records from its routing table. These unique settings are called local minimum state when the actual node does not find any better suitable node in its neighborhood. At this stage, GPSR has to change its strategy from greedy forward to perimeter forward. In this approach, every radio node follows the right-hand strategy and conveys the data packet to the next closest right-hand node.

As shown in Fig.1, each node is aware of the positions of surrounding nodes with the help of beacon messages. Presupposing the network is operating in the greedy mode, nodeS(source node) is an actual node in this situation and it transmits the data packet to the destination nodeD. Carrying out this task, nodeSwill seek for a nearer node which should be closer to nodeD, then nodeSitself (communication range of nodeSis shown with the help of green circle in Fig.1). In this case, it will select nodeAas it is closer to nodeDthan nodeS. By following the greedy forwarding strategy, nodeSwill transmit the data packet to nodeA. Since nodeDis out of the communication range of nodeA(communication range of nodeAis shown with the help of red circle in Fig.1). At this moment, the network is in the local minimum state. Consequently, according to the algorithm, the network will change its strategy from greedy forwarding mode to perimeter forwarding mode. It will start following the right-hand strategy and thereby nodeAwill transmit the data packet to the right-hand nodeB. With the same forwarding mode, nodeBwill transfer the data packet to nodeC. At this point, the network will return to greedy mode because the distance between nodeCand nodeDis less than that between nodeAand nodeD(distance from nodeDto nodesAandCis shown with the help of a black circle in Fig.1). NodeCusing greedy forwarding will forward the data packet to nodeF, and then nodeFwill transmit the data packet to destination nodeD.

Fig.1 Basic working of GPSR communication protocol

2 Issues with GPSR

The packet transporting in GPSR routing technique depends on nodes positions. Because of the dynamic nature of nodes, they keep changing their positions in different circumstances, and the data packet can trap in a loop formed by the group of nodes. If the data packets trap between a group of nodes, the data packets will take much more hops to arrive at the destination node. Additionally, when the classical GPSR data routing technique encounters the local minimum state, the algorithm changes its data forwarding strategy from greedy forwarding strategy to perimeter forwarding strategy. At that moment, the classical algorithm only follows the right-hand strategy and does not consider the left-hand strategy.

When the algorithm runs into the local minimum state, the situation is considered to understand the effect of overlooking the left-hand strategy on the network’s robustness.

If the classical GPSR algorithm follows both the left-hand and the right-hand strategies, then the local minimum node not only finds the right-hand node or the left-hand node to transfer the data packet, but also selects the shortest possible path leading to the destination node, which can be achieved by selecting the right-hand or the left-hand node, depending on the given situation. As a result, the data packet could arrive at the destination node by following a comparatively lesser number of hops. Subsequently, it would mainly improve the whole network operation and increase its robustness. However, the traditional GPSR technique does not take into account the left-hand strategy which stagnates the network’s end results.

In fact, these data trapping or one-handed data forwarding contingencies can end up losing data on the way. Nonetheless, the classical GPSR routing protocol does not take into account the obstacles between the nodes which not only atrophy communication among the mobile nodes, but also extremely affect the strength and sturdiness of the utter system.

These all types of fortuitous events increase the E2E delay, the data loss ratio, and decrease the average network throughput. Consequently, not only do these factors affect the vigor of the network, but also deteriorate the overall performance of the network.

As shown in Fig.2, actual nodeS(source node) sends the data packets to nodeD. NodeDis inaccessible from nodeSbecause it is out of the communication range of nodeS(range of nodeSis shown with the help of the green circle in Fig.2). As a result, nodeScannot send data packets directly to nodeD(the destination node in this case). Assume that the network is operating in the greedy forwarding mode at the moment. NodeBis a closer node to the destination nodeDthan nodeSitself. Therefore, nodeSforwards data packets to nodeB.In the given scenario, nodeBis the closest node to the destination nodeDthan any other node in the surrounding (distance of the nodeDfrom the other nodes is shown with the help of the black circle in Fig.2). However, the nodeDis out of the communication range of nodeB(the range of nodeBis shown with the help of the red circle in Fig.2). Therefore, nodeBin this case is in local minimum state. Consequently, the algorithm changes the data forwarding strategy from greedy forwarding strategy to perimeter forwarding strategy. Subsequently, nodeBacquires the right-hand method and forwards the data packet to the closest right-hand nodeC.NodeCcan deliver the data packet to destination nodeDby following the path {C-F-A-D}. The nodes are mobile in nature, so they can change their positions. It is supposed that the data packet is transferred from one node to another, and nodeFchanges its position and gets a new positionF′ (shown in Fig.2). Now nodeCby following the right-hand rule would transfer the data packet to nodeF′ as it is the best possible option at the moment. Next, nodeF′ would forward the data packet to nodeBagain. As a result, the data packet will get trapped in a loop formed by the nodes {B-C-F′}. Eventually, the data packets would make many more extra hops arrive from the source nodeSto the destination nodeD.

Fig.2 Issues with GPSR communication protocol

3 Proposed Methodology

To resolve the serious issues connected with the former GPSR routing protocol highlighted in the previous section, an adaptive scheme IGPSR has been proposed in the following section.

The objective of this new approach is to boost the working performance of the existing GPSR protocol. The network enhancements are achieved by introducing an additional data table named IT which is an extension of the default data table called NIT. The new table IT works in correlation with NIT and helps the algorithm record information connected with the node’s positions and data packets relatively well, and each node can make the decisions intelligently in different given circumstances. Furthermore, unlike the traditional GPSR routing technique, when the network encounters a local minimum state, the IGPSR uses both left-hand and right-hand data forwarding schemes. These newly developed changes help the network overcome the problems and issues faced with the former classical GPSR routing protocol, and enhance the entire network’s performance in terms of metrics: E2E delay, network throughput, and data loss ratio significantly. The following section discusses NIT and IT.

3.1 NIT

Mobile nodes periodically send out and receive a hello packet from the single-hop neighbors. The purpose of NIT is to store and maintain neighboring nodes’ information in its inventory table which includes location information (x,y,z) of the nearby nodes, internet protocol (IP) address (node identification), and the timestamp of the last received hello packet. NIT is updateable and can be overwritten by the new entry over the last entry.

3.2 IT

IT works in correlation with NIT. This table comprises various information including IP address of neighboring nodes, vectors of IP addresses of destinations, and data forwarding strategy (Fs) which can be greedy forward (Gf), right-hand forward (Rhf), or left-hand forward (Lhf), and destination IP address (D). Since IT is an extension of NIT, both tables work in cooperation. IT refreshes its data with every new hello packet received from each neighbor. This feature allows the algorithm to work in synchronization with the current model.

In the proposed technique, when a network operates in greedy forwarding mode, the IT helps nodes select the best neighboring node to transfer data packets by verifying two conditions. Firstly, the data forwarding node verifies, and if it has not received a data packet from the data receiving node operating in the recovery mode for the same destination node, its IP address does not exist in the IT of the forwarding node. Secondly, the same data packet has not been sent to the receiving node for the same destination.

If the first condition is false, the forwarding node would not consider receiving nodes unless the IT of the forwarding node does not get refreshed. And if the second condition is false, then the forwarding node will search for another suitable neighbor. If the forwarding node does not find any reasonable neighboring node in its vicinity, then the local minimum state occurs. The purposed IGPSR protocol also enters into the perimeter mode just like the GPSR protocol. However, it follows a two-way data forwarding strategy unlike classical GPSR which only follows the one-way strategy. Nonetheless, in IGPSR, while the network operates in perimeter or recovery mode, the receiver node discards the data packet if it has already received the same data packet for the same destination unless the IT of the receiver gets refreshed. This ability reduces the packet duplication possibility and the network overhead load and hence, increases the network’s efficiency.

The IT is designed to help actual nodes find an efficient path to transmit the data packet to the destination, and enable data packets to arrive at their destinations using a lesser number of hops. Furthermore, it also helps data packets avoid the local minimum state, and improves network efficacy remarkably.

The proposed algorithm IGPSR for UAVs works as follows.Rrepresents a recipient node, andCis the set of closest single-hop nodes {C1,C2, …,Cn}. Data forwarding strategy is represented byFs, andIrepresents a unique packet identity for every single source-destination pair. Variablercontains the distance from nodeCnto the destination nodeD, anddis the data packet forDas shown in Fig.3.

Fig.3 Representation of mobile nodes and parameters

For data packets identification, identification field of IP version 4 (IPv4) is used, because it can generate 232unique number of combinations. These combinations help identify each pair of source and destination respectively. The working principle of the proposed algorithm is explained with the help of flow diagrams. Figure 4 represents the procedure while a node receives a packet. Figure 5 represents the procedure while a node forwards a packet.

Fig.4 Packet receiving flow diagram of proposed IGPSR algorithm

Fig.5 Packet forwarding flow diagram of proposed IGPSR algorithm

3.3 Greedy and perimeter forwarding with IGPSR

As shown in Fig.6, there are two source-destination pairs: nodesAandF; nodesGandF. Consider the network follows GPSR protocol and nodeA, from the first pair, deliver packetd1to nodeF. Using greedy forwarding strategy, nodeAfollows the path {A,B,C,D,F}. When the data packetd1arrives at nodeF, the packet faces the local minimum. At this moment, the network would change the data forwarding strategy to perimeter forwarding strategy. Now, by following the right-hand rule, the nodeFwill forward the data packetd1to nodeE, and the data packet will follow through this strategy from nodeF{F,E,D,C,B,A,G,H,I,J,K,M} until nodeM. At nodeM, the data forwarding strategy returns to greedy forwarding mode as nodeMis nearer to the destination nodePand nodeF(distance from nodePto nodesFandMis shown with the black circle in Fig.6). Finally, by following the path {M,N,P}, the data packetd1arrives at the destination nodeP. It means that data packetd1following GPSR routing strategy needs to make 17 hops to arrive from nodeAto nodeP. Data transferring path is represented by curved solid and dotted arrows in Fig.6.

In the GPSR technique, due to its simplicity, the data packet always follows the same path to deliver a packet between the same source and destination pair. On the other hand, the suggested algorithm IGPSR with the help of extension table IT not only enables the data packet to follow different paths while traveling between the same source and destination pair but also saves data packets from the local minimum state. For instance, in a given situation, when the data packetd1arrives at nodeF, in the proposed technique, the network also changes strategy to perimeter forwarding strategy and nodeFtransfers packetd1to nodesDandEby following left-hand and right-hand data forwarding modes. NodeD, in this case, will discard the packetd1received from nodeE, since it has already received the data packetd1from nodeFfor the same destination nodeP. Next, by following a similar step, like GPSR technique, the data packetd1arrives at the destination nodeP. The data packetd1following IGPSR routing strategy will also have to make 17 hops to arrive from nodeAto nodeP. However, when source nodeAsends the second packetd2to the same destination nodeP. In order to transfer datad2, nodeAwill not select neighborB, nodeAhas received data packetd1from nodeBwhile nodeBis being operated in a recovery mode and nodeAstill has this entry in its IT. This means the first condition of IGPSR does not verify. Thereby, to transfer data packetd2from source nodeAto nodeP, using IGPSR algorithm, the data packetd2will travel through from nodes {A,G,H,I,J,K,M,N,P} and make only 8 hops to arrive at the destination, which is, a significant reduction in hops.

The flexibility of the proposed algorithm lies in the fact that the actual node can skip a node only if the same source and destination node pair are involved in communication. This means that nodeAtransfers data packets to nodeP, and the second source and destination pair (nodesGandF) can easily transfer the data packet from the source nodeGto the destination nodeFwithout skipping nodeB{G,B,C,D,F}, since this time destination node for the packet is different. The data transferring path is represented by solid blue arrows in Fig.6. Within the set time duration, nodeAwill refresh its NIT, which will help IT erase old entries from the table, and NIT will start receiving hello packets from the neighbors.

Nonetheless, the standard GPSR approach in recovery mode only employs a right-hand data forwarding strategy, and many scenarios may be imagined in which the data packet either makes a significantly higher number of hops or fails to be safely transmitted to the destinations. Whereas IGPSR, by using left-hand and right-hand date forwarding strategies in recovery mode, can help data packets to be successfully transported to the destinations in the same situations much more effectively. The performance of the proposed IGPSR technique is evaluated in the following section.

4 Simulation Results and Discussion

The proposed algorithm is simulated in an open-source network simulator NS-3, specially designed for the executions of mobile ad-hoc networks. To keep it simple, the node velocity is considered to be constant as 25 km/h. The simulation was performed in an area of 100 m3and the duration between hello packets was kept at 1 s. The packet size for the data packet was set to 1 024 bytes and the simulation duration was 180 s. Mac layer protocol 802.11p dedicated to wireless access in vehicular environment (WAVE) has been used. To depict the realistic 3D nodes’ movement, 3D Gauss Markov mobility model is implied in this experiment. Because of its fast data transmission, user datagram protocol (UDP) is used to transfer data to the destination. The time delay between two data packets is set to 1/10 s. NIT refresh rate is 1 time every 3 s. An omnidirectional antenna with a transmission range of 275 m is set during the simulation. By varying the number of source and destination pairs from 8 to 32 and UAVs from 30 to 110, the network performance of the former GPSR has been compared with that of the prescribed IGPSR technique. To get the precise result, 30 simulations have been executed on each given situation. E2E delay, network throughput, and data loss ratio have been considered as performance matrices. To store neighboring nodes’ information, GPSR’s node uses 16 bytes of space. However, the IGPSR’s node uses 24 bytes of space for each neighbor node because of its additional table. Table 1 shows the parameters used in simulation in summarized form.

Table 1 Parameter list

4.1 E2E delay analysis

Figure 7 shows the performance comparison of GPSR and IGPSR, and there are 8 source and destination pairs considered in an experiment. On the one hand, E2E delay of GPSR is higher than that of IGPSR under any circumstances. On the other hand, the performance of both networks improves as the number of UAVs increases from 30 to 110, and it is completely apparent that the more router nodes accessible, the faster the data packets arrive at the target node. Whereas in Fig.8, 32 source and destination pairs are considered in the experiment, the IGPSR has produced out-class results. There are two reasons. Firstly, the IGPSR transfers consecutive packets to the same source and destination pair by following a fewer number of hops. Secondly, it discards the repeated packets, and thus the congestion on the network reduces. As a result, the delay decreases.

Fig.7 Average E2E delay with 8 source and destination pairs and varying numbers of UAVs on x-axis

Fig.8 Average E2E delay with 32 source and destination pairs and varying numbers of UAVs on x-axis

4.2 Network throughput analysis

Figure 9 depicts a behavior of comparison between IGPSR and GPSR data routing protocols. There were 8 source and destination pairs in a network, and there were 30 UAVs available. The proposed technique IGPSR produced good results and as the availability of nodes increased, it showed even better throughput compared to GPSR. On the flip side, in Fig.10, when the source and destination pairs increased from 8 to 32, the proposed technique displayed much better throughput. For instance, in Fig.10, with 90 available UAVs in an experiment, the GPSR protocol showed better output, and this could be justified by the fact that, GPSR had less congestion network than IGPSR especially in recovery mode, and GPSR just had to forward the data packet to the right-hand node. Whereas the proposed technique utilizesRhfandLhf. However, in IGPSR with the help of IT and NIT, a node can discard duplicated packets thereby reducing network congestion.

Fig.9 Average network throughput with 8 source and destination pairs and varying numbers of UAVs on x-axis

Fig.10 Average network throughput with 32 source and destination pairs and varying numbers of UAVs on x-axis

4.3 Data loss ratio analysis

Figure 11 displays the contrast of IGPSR and GPSR routing protocols. When there were 8 source and destination pairs used in the experiment, at the time of 30 available UAV nodes, both techniques exhibited the same results. Nonetheless as the number of UAVs increased, there was a continuous reduction in data loss in both methods and it made sense, since the number of nodes increased. Consequently, more nodes were available to act as a router to help data packets arrive at destinations conveniently. In Fig.12, when there were 32 source and destination pairs involved in observation, the prescribed method produced remarkable results. The data loss ratio reduced drastically and predictably as the availability of nodes increased from 30 onwards in both strategies. However, the given solution produced significantly superior results. The IGPSR approach with the help of IT and NIT, prevented packets from reaching a dead end and allowed data packets to arrive at the target node successfully in most situations. Whereas GPSR might remain unsuccessful in these circumstances.

Fig.11 Average packet ratio with 8 source and destination pairs and varying numbers of UAVs on x-axis

Fig.12 Average packet ratio with 32 source and destination pairs and varying numbers of UAVs on x-axis

5 Conclusions

Dynamic, unpredictable 3D movement of nodes, highly changeable network topology, and several other flying constraints make the design of a UAV routing protocol a challenging task. In this paper, a new adaptive algorithm IGPSR routing protocol has been designed and implemented for UAVs. Compared to the former GPSR technique, the presented IGPSR technique, when assessed on the metrics of the E2E delay, the average network throughput, and the data loss ratio, has produced on average 40% better results. Additionally, in the past many attempts have been made to design an effective routing protocol. However, in many investigations, scholars overlooked the importance of 3D mobility of UAVs in their works. In this paper, in order to achieve more truthful results, the 3D Gauss Markov mobility model has been considered in simulations.

In the future work, we would analyze the effectiveness of the network in presence of different source and destination node pairs. Furthermore, we would compare the suggested IGPSR algorithm with some other existing routing algorithms. Lastly, we would investigate the behavior of the network in the presence of obstacles.