A Method to Reduce Route Discovery Cost of UAV Ad Hoc Network

. The unmanned aerial vehicle communication networks (UAVCNs) are composed of unmanned aerial vehicles (UAVs) connected in ad hoc mode to facilitate vertical communication in 5G and beyond networks. UAVs operating in an ad hoc mode of operation mostly use reactive routing protocols to establish routes in the network to reduce the energy consumption of the nodes. In this article, a route discovery method is presented to reduce the overhead faced by reactive routing protocols during the route discovery phase. The expanding ring search (ERS) method is mostly used by reactive routing protocols in the destination discovery phase of the routing algorithm. Although the performance of the ERS method is better than simple ﬂ ooding, the presented method further reduces energy consumption and routing overhead as compared to the conventional ERS method. To achieve the task, the time to live (TTL) is modi ﬁ ed to accommodate a large number of nodes in a search attempt. We proposed variants of the proposed techniques for diverse application requirements and compared the performance with the state-of-the-art ERS technique. It has been demonstrated with the help of simulations that the presented algorithm outperforms the ERS method in terms of reduced routing overhead and reduced energy consumption.


Introduction
Unmanned aerial vehicle communication networks (UAVCNs) serve 5G and beyond applications by providing vertical connectivity to ground devices [1]. With the limited energy resources available with UAV devices, these have to leave and join the network repeatedly, which results in frequent route disconnections during communication. The problem of limited energy resources not only affects the performance in terms of time delay and energy consumption but also challenges the security of the network. In [2], the authors propose a machine learning-based framework to improve network security. The presented machine learning-based approach which is based on MLP classifiers gives 99.98% for intrusion detection. In [3], the authors provided a systematic literature review for the security of such low-power and lossy network scenarios. On the other hand, multiple solutions such as power control, resource sharing, channel allocation, and routing are proposed to reduce energy consumption and efficient bandwidth utilization. On the other hand, multiple solutions are proposed for reducing energy consumption. For instance, edge technology is used in [4] to reduce the processing burden from the nodes. Similarly, cognitive technology is applied for efficient channel sharing and bandwidth utilization [5]. However, less focus is given to improving energy by efficient routing and path management.
UAV ad hoc networks are a category of UAVCNs that are distributed in nature, and there is no central administration to control the routing and management operations; the devices themselves manage network operations [6]. In order to provide long-term connectivity among the UAV devices and to propagate an uninterrupted connection to the users connected with the other device, the energy efficiency of UAV ad hoc networks is a major requirement.
As discussed above, UAV ad hoc networks operate without any central administration, and nodes operate cooperatively to manage network operations. Therefore, the nodes act as source, destination, and/or relay nodes to facilitate communication with other UAV devices. As the nodes reduce their transmission power to save battery resources, multiple hops are required to establish routes, which increases the time delay during communication [7]. In [8], the authors propose a graph theoretical approach to establish the shortest path in nondelay torrent networks with reduced time complexity. In these dynamic networks with mobile UAVs, the nodes have to refresh their routing tables frequently. Therefore, an efficient route discovery method is extremely important to reduce routing overhead. The primary goal of this route search mechanism is to reduce routing discovery overhead to save nodes' energy by conducting a fast search for the destination.
Various types of routing protocols exhibit different overheads. In a broader sense, the routing protocols are divided into two categories: reactive, or on-demand, and tabledriven, or proactive, both having their own pros and cons. Ad hoc on-demand distance vector (AODV) routing protocol is one of the famous routing protocols used in UAV ad hoc networks that fall in the category of reactive routing protocols [9]. In UAV ad hoc networks with the AODV routing protocol, the devices request routes whenever they are required [10]. Therefore, these networks incur less overhead as compared to proactive routing methods such as DSDV, where periodic updating of the routing table is required.
In this article, we focus on the reactive routing protocols and propose a method to improve route discovery overhead and reduce energy consumption in the UAV ad hoc networks. The proposed method is an extension of the ERS method that searches a route toward the destination in multiple attempts. In ERS, a source node initiates a networkwide broadcast request if it fails in discovering the destination in a specific number of attempts. In the proposed protocol, the discs replace the rings, accommodating a larger number of nodes in a single search query before deciding to broadcast the RREQ across the entire network. The method takes into account multiple network parameters, such as disc area, network size, and the number of attempts during the destination search process. The time-to-live (TTL) field of ERS is modified to control the disc area in each attempt. A blocking mechanism is also presented to stop the transmission of the RREQ packet when the destination is searched. In addition, we also presented the optimal value for the number of attempts.
The proposed EDS algorithm has various general benefits for mobile ad hoc networks, including reduced routing overhead, faster search time, and increased coverage. In addition, the proposed method specifically benefits unmanned aerial vehicle (UAV) ad hoc networks that have a major contribution to improving the performance of many industries including the constriction industry, agriculture industry, education sector, and security and surveillance sys-tems. EDS, by fast and accurate search, improves the network coverage, reduces time latency, reduces the overhead of communication, and hence improves the performance of UAV ad hoc networks. The algorithm's low routing overhead can help reduce communication costs and increase the overall efficiency of the system. Additionally, the high coverage and search time can improve the accuracy and speed of UAV missions, making them more effective in various applications such as aerial surveillance, package delivery, and crop monitoring. These benefits can ultimately lead to cost savings for UAV companies, improved mission success rates, and increased customer satisfaction.
The rest of the paper is organized as follows: Section 2 presents the related work. The proposed routing process is explained in Section 5. Variants of ERS and EDS are also presented in this section. In Section 6, the results are presented and compared with the state-of-the-art techniques available in the literature. Finally, Section 7 concludes the work.

Related Work
A major issue in UAV ad hoc networks is routing. The algorithms used in UAS are mainly based on AODV for routing operations. In [11], a performance comparison of OLSR and AODV is presented by the authors for flying ad hoc networks (FANETs). They showed that the AODV performs better than the OLSR for low-density and highly dynamic networks. In [12], the authors compare the performance of reactive and proactive routing protocols for applications that use HTTP, such as voice applications, database queries, and video conferencing. However, modifications are required in existing routing algorithms to apply them to 5G applications such as UAVCENs, IoTs, and the UAV ad hoc network [10]. In [13], the authors proposed optimized AODV (OAODV) to optimize the AODV routing protocol with respect to some performance measures. The proposed algorithm is better than AODV in terms of latency, routing overhead, and link generation. The OAODV protocol reduces routing overhead by 35.07% and 68.93% and time latency by 47.73% and 11.55% as compared to AODV and OLSR, respectively. Furthermore, the ant colony algorithm is integrated with OAODV to enhance the throughput and reduce the number of hops in the route. The abovementioned advantages of OAODV make OAODV a suitable candidate for 5G swarm networks. In the research of [13], the authors show that during the initial path discovery stage, the time consumption of AODV is overwhelming. Another drawback of AODV is observed in [14] when applying AODV to FANETs. The authors concluded that the stability of FANETs decreases due to the high cost paid during the path discovery stage of AODV. Therefore, there is a need to reduce the cost of the path discovery phase of AODV-based routing protocols.
Route discovery is an unavoidable part of the AODV routing protocol due to its reactive nature. During the destination search phase, the RREQ packet is carried over the whole network in a practice known as flooding [15]. The broadcast cost of flooding increases linearly with the number of nodes in the network. Afterward, multiple algorithms are 2 International Journal of Distributed Sensor Networks presented to control the flooding of RREQ packets, such as the random walk protocol, gossip-based protocols, and the expanding ring search technique [16][17][18]. The source node in the random walk protocol chooses the subsequent hop at random among its neighbors before transmitting the RREQ packet. This procedure is repeated until the target is located. The random walk protocol has a significant time delay while reducing the amount of redundant RREQ packets within the network [19]. Similarly, in gossip-based techniques [20][21][22], each node when receiving an RREQ packet takes a decision to expire or forward the packet based on some probability distribution. There are many variants of gossip-based protocols, and we need additional information to estimate the probability of forwarding. The expanding ring search (ERS) approach was suggested by the author in [23] to lower the search expense of route discovery. To cut the cost of the destination search process, ERS searches the location using expanding ring-like patterns. Numerous enhancements are put forth to further boost the ERS technique's performance in terms of decreased routing overhead and time delay. The route discovery cost and search time of the AODV routing protocol are improved by the expanding disc search method we present in this research, which uses the TTL field differently from the traditional ERS approach. Table 1 summarizes the related work. Table 2 compares different variants of ERS algorithms presented in the literature.

Route Discovery in UAVs
In UAVs, route discovery refers to the process of finding the optimal path or route between the source and destination nodes. This process involves identifying the intermediate nodes that the data packets need to pass through to reach the destination node. The route discovery process is crucial in UAV networks, as UAVs are highly mobile and their movements can lead to frequent changes in network topology, making it difficult to establish and maintain reliable routes. Therefore, path planning problem for unmanned aerial vehicles (UAVs) is a critical issue in UAV applications. To achieve this objective, researchers have proposed various algorithms and techniques to optimize the UAV path planning problem. Two recent articles propose methods to solve UAV path planning and routing discovery problems. The first method, called the modified mayfly algorithm [30], searches UAV configuration space using adaptive Cauchy mutation with exponent decreasing inertia weight strategy. The second method, called intuitionistic fuzzy VIKOR [31], uses intuitionistic fuzzy membership degree to describe dynamic and uncertain attribute information of nodes and calculates the Q value as the probability density value of the node. The third method models UAV path planning as an optimization problem with fitness functions including traveling distance and risk of UAV and three constraints. An adaptive selection mutation-constrained differential evolution algorithm is proposed to solve the problem, which is shown to be competitive in experimental results.
In [32], the authors build a fitness function that includes traveling distance and risk of UAV in addition to the three constraints (i.e., height, angle, and slope) to formulate an optimiza-tion problem for UAV path planning. Then, solve the problem by proposing an adaptive selection mutation-constrained differential evolution algorithm. The proposed algorithm shows competitive results when evaluated under different scenarios.
In addition, researchers have proposed various machine learning-based approaches to solve the UAV path planning problem [24,[33][34][35][36]. For instance, a recent study proposed a deep learning-based path-planning approach for UAVs that leverages the power of deep reinforcement learning to discover optimal paths for UAVs. The proposed approach was evaluated on multiple real-world scenarios, and the results showed that it outperforms the other compared algorithms.
Overall, the path search problem for UAVs is a challenging issue that requires sophisticated optimization algorithms and techniques. The recent literature has proposed various approaches to solve this problem, including natureinspired algorithms, hybrid optimization algorithms, and machine learning-based approaches. These approaches have demonstrated promising results and hold great potential for real-world UAV applications.

Research Problem
The research problem identified is the high cost and time delay associated with the route discovery phase of AODVbased routing protocols in UAV ad hoc networks. Despite being widely used in UAV ad hoc networks, the AODV protocol suffers from drawbacks such as increased routing overhead and reduced stability of networks during the path discovery stage. Several studies have been conducted to improve the performance of AODV-based routing protocols, including the use of optimized variants such as OAODV and comparison with other protocols such as OLSR. However, modifications are still required to apply existing routing algorithms to 5G applications such as UAV-CENs, IoTs, and the UAV ad hoc network. The research problem, therefore, is to develop a new method or enhancement to the existing AODV-based routing protocols to reduce the cost and time delay associated with the route discovery phase and improve the stability and performance of UAV ad hoc networks in 5G applications.

Energy-Efficient Route Discovery
In this section, the proposed method is presented, named "expanding disc search" (EDS), that aims at improving the performance of the AODV routing protocol when implemented on UAV ad hoc networks. In Figure 1, we show a UAV ad hoc network, where multiple UAV devices provide connectivity to the users in a multihop fashion. In our system, we distributed N nodes randomly in a 3D space of b × b × b m 3 . These UAVs are connected with each other and with the vehicles, cellular subscribers, and ground stations as shown in Figure 1. It is assumed that each UAV has limited energy resources, and the transmission range is bounded based on the physical channel model and transmission power. A multihop network where the route to the destination is established via intermediate nodes is considered. We assume that all the nodes in the network 3 International Journal of Distributed Sensor Networks act cooperatively. When a source node desires a link with a destination, it triggers the route search mechanism and broadcasts an RREQ in the network. A high time-to-live (TTL) value is assigned to the broadcast packet to search the entire network in a single search. The source node can also opt for a small TTL value of 1, 2, 3, 4, etc. before initiating a broadcast packet. The TTL value determines the expiry time of the packet, e.g., the packet will expire after 2 hops if TTL = 2. ERS searches the network by dividing the entire network into multiple rings of irregular shapes. The size of the ring depends on the choice of TTL value. In the proposed solution, real-time is used as the TTL field, dividing the network into circular discs. The discs accommodate a relatively high number of nodes in the search area, which increases the probability of finding the destination in a search query and consequently reduces the route discovery cost. Figure 2 shows the formation of discs for EDS. In contrast to ERS, the RREQ packet travels multiple hops before the expiration of the TTL. The expiration of the packet is determined by  [16] 2019 Greedy mode Low Medium Optimized ERS (OERS) [17] 2022 Optimal path selection Low Medium Adaptive expanding ring search (AERS) [29] 2020 Adaptive TTL Low Low

International Journal of Distributed Sensor Networks
comparing the time stamp of the RREQ packet with the local time of the receiving node. We assume that the timing clocks of all the nodes in the network are synchronized, similar to how clock synchronization is accomplished by a global positioning system (GPS). The disc band created against each TTL value is named a "circular disc band," as shown in Figure 2. In the first search query, the nodes in CDB1 receive the RREQ packet from the source node. In the nest search query, the nodes in the CRB2 receive the packets in addition to the nodes in the CRB1. This process is repeated until the RREQ is reached at the destination or the entire network is searched.
In the following sections, we have presented several algorithms for energy-efficient route discovery based on the EDS principle. We establish the model that constructs logical discs of different sizes depending on the network requirements and compares the performance for different network parameters. In the next section, we present ERS in detail as it works as a benchmark for the proposed solution.

Expanding Ring Search.
The expanding ring search algorithm is proposed to reduce the route discovery cost of the flooding technique in which the RREQ packet is broadcast in the network. The main idea is to restrict the transmission of the RREQ packet by performing a partial network broadcast. To accomplish this, a TTL field is introduced in the RREQ packet that determines the expiry of the RREQ packet. The TTL field corresponds to the number of hops traveled by the RREQ packet before it expires. The source node can adjust the TTL value and set a high TTL value if it desires to search a larger network region. In ERS, the source node starts from small TTL values and increases the value of TTL if the destination is not discovered in the previous attempt. As the cost of searching increases exponentially with the number of attempts, the source node has to decide whether to initiate a network-wide broadcast after a certain number of search queries. The optimal value of the search threshold depends on the network area, node density, and network topology. The optimal value of this threshold is discussed in [23]. The average search cost for the ERS method can be written as [37] where qðjÞ donates the probability that the destination is located in the j-th search region, γðMÞ represents the average search cost of the ERS algorithm which is a function of the number of attempts ðMÞ, and θðkÞ is the accumulated number of nodes up to k − 1 rings, which can be written as With uniform node distribution, the probability that the destination lies in j-th ring can be written as where N is the total number of nodes in the network and the number of nodes in the j-th ring is donated by n j . The expression in (1) shows that the expected cost of ERS depends on the value of the number of attempts made before a network-wide broadcast is initiated.

Linear Expanding Disc
Search. The EDS technique uses the idea of ERS and blocking ERS and presents a more energy-efficient solution for the AODV protocol. In EDS, circular discs rather than rings are generated, which allows for a larger number of nodes in the search and raises the likelihood that a search will be successful in locating the destination. The TTL value in this technique is designed so that the region of each disc stays constant, i.e., where S di is the area of i-th disc and S d1 is the area of the first disc. The area of the n-th disc will be As the nodes are uniformly distributed in the network and the area of all discs is the same, therefore, the resulting number of nodes in the i-th disc is the same, The probability that the destination node is on i-th disc will be The expected cost for the proposed technique is calculated as By using the (6) and (7) and simplifying the above equation, we have The broadcast cost, in a network for ad hoc on-demand distance vector (AODV) algorithm, is N [38]. The normalized expected cost for the purposed technique becomes The normalized expected cost is a function of the 6 International Journal of Distributed Sensor Networks number of discs n, and it is clear that for a large value of n, the resulting normalized expected cost approaches 1/2. Therefore, as the number of discs in the purposed technique increases, the results are decreased in energy consumption.

Blocking Expanding Ring
Search. The blocking expanding ring search (B-ERS) is proposed to reduce the cost of ERS. In BERS, the search attempt starts in a similar manner as in ERS. But the next-hop neighbors wait for some time before discarding the packet. During the waiting time, if the destination is found, the source node informs the intermediate nodes about the discovery, and the packet is discarded. The procedure continues until the destination is found in the intermediate node and transmits the packet to its neighbors if the halt signal is not obtained during the waiting period. If the recipient is far from the source node, the BERS waiting time lengthens. Therefore, the BERS method reduces energy consumption at the cost of route discovery time. Mathematically, the broadcast cost of BERS can be written as where q i represents the chance that the target will be located in the i-th search query and M represents the number of efforts necessary to find the destination. The cost of BERS increases linearly with the number of trials, but the waiting time increases with distance, making EDS unsuitable for time-sensitive applications, according to the equation above.

Blocking Expanding Disc
Search. The blocking expanding disc search (B-EDS) method works in a similar fashion to that described for blocking ERS techniques but uses discs instead of rings. The routing overhead of BEDS is less as compared to EDS with the same time cost.

Nonlinear Expanding Disc
Search. The area of a succeeding disc is bigger than the size of the previous disc in this situation. Each trial takes twice as much time as the initial estimate and is delivered by where Ts i represents the time required to reach the i-th CRB and Ts is the time required to reach the extent of the network, i.e., the last attempt. The distance of each attempt will also be two times greater than the previous attempt.
where d n and d n−1 represent the radius (distance from source node) of n-th and ðn − 1Þ-th disc, the relation for the area of CRBs can be derived by using the (13), which results to S d i represents the area of i-th CRB. The number of nodes in the CRBs will also increase in the same proportion as the area because of the uniform node density in the network. Therefore, the following explains the number of node relations in CRB: n i represents the number of nodes in the i-th CRB. The probability of selecting a CRB, which contains the destination node, depends upon the number of nodes in a CRB; therefore, the probability q i will also increase in the same proportion and will be Equations (15) and (16) consist of two parts: in the first, the number of discs is greater than one, and in the second, the number of discs is equal to one. For the second case, the normalized expected cost will be unity; therefore, it is emphasized only in the case when the number of rings is greater than one. The expected cost for the case of n ≥ 2 will be This results in the following expression: The normalized expected cost is γ n ð Þ = 1 − 48 4 2n for n ≥ 2: Equation (19) shows the normalized expected broadcast cost for the second case, which approaches one with the increasing number of rings (n). International Journal of Distributed Sensor Networks the traffic of other UAVs in the network along with their own traffic. We consider that each UAV is equipped with a wireless network interface card (WNIC), which implements physical and medium access control layer protocols. The transmission range of the nodes depends on the channel conditions and the receiver sensitivity. MATLAB is used to generate the simulations of the proposed algorithm. The experimental setup is summarized in Table 3. Figure 3, the broadcast costs of EDS and ERS are examined. The graph shows that the transmission cost is proportional to the number of tries made by the source node while looking for the destination. According to the graph, the cost of ERS lowers for up to 4 or 5 iterations (number of attempts) before rising. EDS's routing search cost, on the other hand, decreases even after 5 cycles. Because EDS examines a large number of nodes at every attempt, the destination is generally confirmed before the source node decides to broadcast the RREQ throughout the whole network. The hypothetical graphs in Figure 3 equate to (1) and (9) for ERS and EDS, correspondingly. Figure 4 compares the effects of blocking ERS with inhibiting EDS. These solutions are more cost-effective in terms of anticipated broadcast costs since the blocking operation in them decreases the quantity of RREQ packets. The graph demonstrates how the BEDS further enhance the BERS' performance. Because blocking strategies typically take longer to find the target, ERS and EDS algorithms excel in this area. The nonlinear expanding disc search algorithm's normalized estimated broadcast cost is displayed in Figure 5. The pattern resembles ERS in that the cost first declines before trending higher. The trend remains constant when the whole network is examined, indicating the maximum number of tries required to locate the entire network. Table 4 shows a performance comparison of ERS with several state-of-the-art algorithms, including enhanced       International Journal of Distributed Sensor Networks ERS, blocking ERS, EDS (proposed), and blocking EDS (proposed). It can be observed that ERS has low routing overhead and packet overhead but high search time and coverage. Enhanced ERS has similar performance characteristics to ERS but with a slightly lower search time. Blocking ERS has very low routing overhead but very high search time and low packet overhead. The proposed EDS algorithm has very low routing overhead and packet overhead, high coverage, and very low search time. The proposed blocking EDS algorithm has low packet overhead, high coverage, and low routing overhead but high search time. The comparison shows that the EDS algorithm is beneficial as compared to ERS and its variants because it has a significantly lower routing overhead, making it more efficient in terms of energy consumption and network lifetime. In addition, EDS has a very low packet overhead, which means it generates fewer control packets compared to other algorithms, reducing network congestion. As the coverage rate is high, the target node can be discovered efficiently. Therefore, it is a good approach for UAV ad hoc networks, where network lifetime and network energy are critical parameters.

Conclusion
In this article, an expanding disc search method is proposed in order to enhance the efficiency of reactive routing protocols such as AODV. The proposed enhancement makes it suitable for application for future wireless networks such as 5G and beyond networks. Our algorithm was compared with existing route discovery algorithms and demonstrated to be more efficient with lower route discovery costs. The proposed EDS technique is suitable for route discovery in current networks such as UAVCNs, FANETs, and UAV ad hoc networks. Our analytical expressions and simulation results highlight the advantages of the proposed algorithm over existing state-of-the-art algorithms, making it a promising solution for improving the performance of routing protocols in modern networks. The proposed EDS algorithm has the potential to benefit industries such as UAV-based applications, disaster management, and military operations by providing reliable and efficient communication in ad hoc networks.

Data Availability
This paper does not require any dataset, whereas the simulated data is used.