Abstract

Vehicular Ad Hoc Networks (VANETs) are a specialized class of mobile ad hoc networks enabling vehicle-to-vehicle (V2V) and vehicle-to-infrastructure (V2I) wireless communication. Efficient routing in VANETs is challenging due to high node mobility, rapidly changing topology, and stringent performance requirements. The Ad Hoc On-Demand Distance Vector (AODV) routing protocol, originally developed for MANETs in 1999, has been widely applied to VANETs for unicast routing owing to its on-demand route discovery and scalability. However, standard AODV faces significant limitations in VANET environments, including high control overhead from route floods and route discovery latency that can degrade performance under vehicular mobility. In the past five years, extensive research has focused on adapting and enhancing AODV for VANET scenarios. This paper presents an in-depth review and technical analysis of recent developments (2020–2024) in AODV-based routing for VANETs. Key enhancements are surveyed in areas such as reducing routing overhead, improving route stability and latency, and strengthening security against attacks. Performance evaluations from the literature are synthesized, comparing AODV and its variants under VANET conditions using metrics like packet delivery ratio, end-to-end delay, throughput, and routing overhead. The analysis highlights how modifications—e.g. fuzzy-logic route selection, cross-layer optimization with 802.11p, link-quality metrics (ETX), clustering, and trust or machine learning-based security mechanisms—have improved AODV’s efficacy in high-mobility vehicular networks. We discuss the limitations that remain, including scalability challenges in dense networks and trade-offs between latency and overhead. Finally, we outline open research issues and future directions for routing in VANETs. The paper is organized into sections covering background, related work, methodology of analysis, detailed discussion of enhancements, performance evaluation results, and conclusions.

Introduction

Vehicular Ad Hoc Networks (VANETs) consist of vehicles equipped with wireless communication devices that self-organize into mobile networks on the road. VANETs enable diverse applications in Intelligent Transportation Systems (ITS), from safety messaging (e.g., collision warnings) to traffic management and infotainment. Unlike static or low-mobility networks, VANET nodes (vehicles) move at high speeds and in often unpredictable patterns, causing the network topology to change frequently. This dynamic behavior, combined with potentially high vehicle density in urban areas, makes routing in VANETs particularly challenging. Routing protocols must quickly adapt to link breaks and new neighbors, minimize latency for time-sensitive data, and cope with frequent topology updates – all while using limited bandwidth efficiently.

AODV is a reactive (on-demand) routing protocol that has been widely studied and employed in VANET research as a baseline due to its simplicity and relative efficiency. Developed by C. Perkins and E. Royer and first published in 1999​ (with a standardized specification in IETF RFC 3561 in 2003), AODV finds routes only when needed, rather than maintaining full network routes continuously. In AODV, when a source vehicle needs a route to a destination and does not have one in its routing table, it initiates a Route Request (RREQ) broadcast. The RREQ floods through the network until it reaches the destination or a node with a fresh route to the destination. That node then responds with a Route Reply (RREP) unicast back to the source, establishing a path. AODV uses sequence numbers to ensure route freshness and to prevent routing loops. Once routes are established, data packets flow along the discovered path. If a link on the route breaks (detected via missing periodic HELLO messages or other feedback), a Route Error (RERR) is sent to inform upstream nodes, and the route is invalidated. This on-demand operation means that, in theory, AODV produces no routing traffic when there are no active routes, which is an advantage in reducing unnecessary overhead.

Despite these advantages, vanilla AODV has performance issues in the VANET context. The broadcasting of RREQs can lead to significant control overhead in dense vehicular networks, as many nodes rebroadcast requests during route discovery​. This overhead can occupy valuable wireless channel capacity (the DSRC/WAVE channels in VANET) and increase the chance of collisions or congestion, especially in high-density traffic scenarios. Additionally, the route discovery latency can be detrimental for delay-sensitive VANET applications; AODV may take too long to find a route when the network is large or highly dynamic, causing higher end-to-end delays before data transmission begins. High mobility in VANETs also means routes found by AODV tend to break frequently; maintaining connectivity may require frequent rediscovery or route repairs, further adding to delay and overhead. Moreover, AODV in its basic form lacks any built-in mechanisms for security, making it vulnerable to routing attacks (like black hole or wormhole attacks) which are a serious concern in an open network of moving vehicles.

Given these challenges, researchers have explored numerous enhancements to AODV to better suit VANET environments. Alternative topology-based routing protocols (such as OLSR, DSDV, and DSR) have also been evaluated in VANETs for comparison. Proactive protocols like OLSR (Optimized Link State Routing) continuously propagate topology updates and can offer lower route setup latency at the cost of constant overhead. Studies have shown that OLSR can outperform AODV in certain VANET scenarios in terms of throughput, packet delivery ratio (PDR), and end-to-end latency, especially as node density increases, because OLSR has up-to-date routes and does not flood the network with route requests for each communication. For example, Hota et al. report that OLSR achieved higher PDR and lower delay than AODV and DSDV in an urban VANET simulation as the number of nodes grew, although at the expense of more routing traffic overhead. On the other hand, AODV often generates less routing traffic under light loads or sparse networks since it seeks routes on-demand; it has been observed to perform well in scenarios with lower node density or mobility, where route discoveries are infrequent and routes do not break as often. Traditional comparisons (prior to recent enhancements) indicate that no single protocol is universally best for all VANET conditions​. AODV’s reactive approach is advantageous in reducing routing load in intermittently connected or low-traffic networks, but its weakness lies in the route acquisition delay and flood overhead under heavy traffic or highly dynamic topology.

Problem Statement: In summary, the baseline AODV protocol needs improvement to meet VANET performance and security demands. Key problems include: (1) High control overhead from RREQ flooding in dense networks, leading to scalability issues; (2) Route instability and latency in high-mobility conditions, as frequent link breakages trigger route repairs or rediscovery, increasing packet loss and delay; and (3) Security vulnerabilities, as AODV is susceptible to malicious actors forging route replies or advertising false metrics, which can cause disruptions (e.g., black hole attacks where a malicious node attracts and drops traffic​). In the last five years, researchers have addressed these problems through various modifications to AODV’s route discovery mechanism, routing metrics, and routing message handling, as well as by incorporating cross-layer information and external techniques (like machine learning) into the routing decision process. There is also a trend of leveraging infrastructure support (e.g., roadside units or cellular networks) to assist AODV routing in hybrid vehicular networks. This paper aims to survey and synthesize these recent advancements.

The rest of this paper is organized as follows: Section II reviews related work on VANET routing and summarizes baseline performance evaluations of AODV versus other protocols. Section III outlines the methodology of our survey and the criteria for analyzing AODV enhancements, including the common evaluation metrics used in recent studies. Section IV delves into the technical enhancements to AODV proposed in the literature, categorized by their primary focus (overhead reduction, latency improvement, reliability, or security). Section V presents a comparative performance evaluation, aggregating results from multiple studies to illustrate the impact of these enhancements. Section VI provides a discussion on the limitations of the enhanced AODV approaches and potential open issues. Finally, Section VII concludes the paper with insights and future research directions.

Related Work and Background

Early research on VANET routing recognized that classical MANET protocols like AODV needed adaptation to cope with vehicular environments. Several surveys (e.g.,) have discussed VANET routing protocols broadly, classifying them into topology-based (like AODV, DSR, OLSR) and position-based (geographic routing) categories. Topology-based protocols use explicit route discovery and maintenance, whereas geographic routing (e.g., GPSR – Greedy Perimeter Stateless Routing) makes forwarding decisions based on nodes’ positions. In VANETs, geographic routing has an inherent advantage of being highly adaptive to topology changes (since it doesn’t require fixed routes), yielding low latency and overhead in high-mobility scenarios. However, geographic methods require GPS and may struggle in sparse networks or when accurate positioning is unavailable. Thus, improved topology-based protocols like AODV remain very relevant, especially for scenarios where GPS support is partial or for compatibility with existing internet routing.

Baseline performance comparisons in literature have helped identify where vanilla AODV falls short. El-Dalahmeh et al. (IET Networks, 2024) proposed an ECIE (Edge Centrality and Intersection Evaluation) method to improve AODV and compared it with OLSR and DSDV, finding that AODV had higher delay in urban scenarios without enhancements​ (Sustainability, 2022) concluded that OLSR generally outperforms AODV and DSDV on throughput, delay, and delivery ratio in a realistic city environment as node density and traffic load increase. These works reinforce that AODV’s reactive discovery, while efficient in light traffic, can become a bottleneck under heavy traffic due to the flood of RREQ messages and the time to establish routes.

Historically, various improvements to AODV were explored even prior to 2018. For example, Aswathy and Tripti (2012) introduced a cluster-based AODV enhancement where the network is divided into clusters with designated cluster-heads and gateways to limit the RREQ flood scope​. Instead of broadcasting RREQs network-wide, a node would send the RREQ to its cluster head, which then routes among cluster heads – significantly reducing redundant broadcasts. This idea of clustering has been a recurring theme, as it localizes route discovery and can lower control overhead. Another earlier approach was AODV-DV (Directional Variation), where RREQ propagation is restricted to the general direction of the destination using position information (if available) to avoid unnecessary flooding in opposite directions. Multi-path extensions like AOMDV (Ad Hoc On-Demand Multipath Distance Vector) were also proposed, allowing AODV to discover and maintain multiple alternate routes in parallel. AOMDV can improve reliability and reduce route discovery frequency since a backup route can be used immediately if the primary route fails. Indeed, researchers found AOMDV yields higher packet delivery than AODV in MANET simulations by utilizing alternate paths. These prior works laid a foundation, but many were not specifically tuned to realistic VANET conditions (e.g., high mobility with urban obstacles).

In the past five years, research has intensified on VANET-specific AODV enhancements and new hybrid protocols. For instance, Xi Yu et al.’s concept of “AODV-VANET” (initially proposed earlier and revisited in recent work) incorporates vehicles’ movement information (speed and direction) into route discovery decisions to predict route longevity. Bio-inspired algorithms have also been applied: a variant called Ant-AODV-VANET uses ant-colony optimization heuristics to build routes considering pheromone-like metrics that favor paths with more stable connectivity (factoring in vehicle movement)​. These approaches indicate the diversity of related work beyond straightforward protocol tweaks – some integrate completely different algorithms with the basic AODV mechanism.

Security has been another critical area of related work. The standard AODV is vulnerable to attacks like the black hole attack, where a malicious node responds to RREQs with fake RREPs claiming a short route, then drops all packets. Early solutions like SAODV (Secure AODV), proposed in the mid-2000s, added cryptographic signatures and hash chains to AODV messages to ensure their authenticity and integrity. SAODV could prevent simple route spoofing but incurred overhead from signing and verifying messages. Recent works have instead focused on trust-based and intelligent detection mechanisms. For example, Kumar et al. (2021) developed a secure AODV routing algorithm that detects black hole nodes by monitoring anomalies in sequence numbers and hop counts in RREPs​. They integrated an intrusion detection system that flags nodes which consistently advertise improbably good routes. Another strategy in related literature is using group authentication: a protocol named T-AODV (Trust-based AODV) by Honarmand et al. uses a trust model where each node evaluates neighbors based on past behavior and recommendations, and only trusted nodes’ RREP are accepted. In this approach, each node maintains a trust score for others and ignores routing information from low-trust nodes, thereby mitigating black holes. Honarmand’s T-AODV (2023) employs both direct trust (based on successful packet forwarding rates) and recommendation trust (from third-party nodes) combined via a fuzzy logic system to decide whether to trust a route. This cross-disciplinary use of trust and fuzzy logic in routing is emblematic of recent trends in VANET routing research.

In summary, related work underscores that improving AODV for VANETs requires a multi-faceted approach. Prior and parallel studies have tackled the problem from different angles: some focus on routing efficiency (reducing overhead and delay) through clustering, directional flooding, or better metrics; others focus on route reliability (link-quality estimation, multipath resilience); and others on security (trust and verification mechanisms). Table 1 provides an overview of representative AODV enhancements and related protocols from recent literature, highlighting their key features and reported benefits.

Figure 1: Typical Vehicular Ad Hoc Network scenario with vehicle-to-vehicle (V2V) and vehicle-to-infrastructure (V2I) communication. Vehicles (cars, buses) form a self-organizing network, and roadside units (RSUs) provide infrastructure connectivity. In such a dynamic topology, routing protocols like AODV must handle rapid changes in connectivity.

Methodology of Analysis

This paper conducts a comprehensive literature survey of AODV routing protocol enhancements in the context of VANETs, focusing on publications from roughly 2019 through 2024. The methodology involves identifying key themes in the recent research: overhead reduction, latency improvement, link reliability, and security enhancements. Within each theme, multiple independent proposals are reviewed and compared. We gather information from academic journals, conference proceedings, and open-access articles that presented modifications to AODV or evaluations of AODV in VANET scenarios. Each selected work is analyzed for technical approach (what changes were made to AODV or what auxiliary techniques were used) and performance results (how the modified protocol performed relative to standard AODV or other protocols).

To ensure a fair comparison, our analysis emphasizes common performance metrics reported in these studies. The most frequently used metrics in VANET routing evaluations are: Packet Delivery Ratio (PDR) – the fraction of data packets delivered to the destination; End-to-End Delay – the time taken for packets to travel from source to destination (including route discovery time); Throughput – the rate of successful data delivery over the network (often measured in kbps or packets/sec); and Routing Overhead – the number or proportion of routing control packets (RREQ, RREP, RERR, etc.) transmitted. Some studies also examine average hop count per route, packet loss ratio, and normalized routing load (overhead per delivered packet). We collate these metrics where available to understand the benefits and trade-offs of each enhancement.

Most of the research covered uses simulation as the evaluation methodology, given the difficulty of large-scale real-world VANET experimentation. Tools like NS-3, OMNeT++ (with the Veins framework), and MATLAB-based simulators are common. Simulated scenarios typically model urban or highway traffic with realistic mobility patterns (e.g., using SUMO or real map data). The IEEE 802.11p standard (DSRC/WAVE) is almost universally assumed for the wireless communication, as it is the basis for V2V in VANET; some works also compare with Wi-Fi (802.11g) or cellular V2X for completeness. By surveying results under these standardized conditions, we can aggregate knowledge about how AODV performs and scales.

In our comparative analysis (Section V), we do not introduce new simulation results but rather synthesize reported data from multiple sources. Where possible, we cite quantitative improvements (e.g., “Protocol X achieved Y% higher throughput than standard AODV”) directly from the papers. In other cases, we use qualitative terms (such as “significantly improved” or “moderate increase”) if precise numbers are not given but trends are clear. We also tabulate some comparisons to present a high-level view of how each enhancement impacts different metrics. All information is backed by references to the original studies to maintain academic rigor.

By following this methodology, this survey aims to paint a detailed picture of the state-of-the-art in AODV-based VANET routing, ensure that the analysis is grounded in evidence, and identify consensus findings versus open debates in the community.

Enhancements to AODV for VANETs

Recent research has produced a variety of enhancements to the AODV protocol to address VANET-specific challenges. In this section, we categorize these enhancements into four groups for clarity: (A) Reducing Control Overhead and Latency, (B) Improving Route Stability and Link Quality, (C) Security and Attack Mitigation, and (D) Cross-Layer and Hybrid Approaches. Many proposals span multiple categories, but we discuss each in the primary context of its main contribution. Table 1 summarizes several key proposals and their characteristics.

A. Overhead Reduction and Latency Improvement

One of the primary concerns with using AODV in VANETs is the overhead of route discovery. Various strategies have been proposed to make route discovery more efficient and to reduce the latency before data can be delivered.

Adaptive RREQ Broadcasting: Traditional AODV blindly floods RREQs. Researchers have proposed adaptive flooding mechanisms that limit the RREQ propagation range or frequency based on network conditions. A noteworthy example is the use of distance or location information to constrain RREQs. Kandali and Hamid (2020) introduced a distance-based broadcast suppression integrated with a fuzzy logic system. In their protocol, when a node receives a RREQ, it evaluates parameters like its relative distance to the source, the direction of motion, and velocity difference. A fuzzy logic controller then decides whether forwarding that RREQ would be useful. Essentially, if a node is moving away from the direction of the destination or is too far, it may choose not to rebroadcast. This intelligent pruning of RREQ propagation can significantly cut down the number of broadcasts, thus reducing overhead. Their simulation in an urban VANET scenario showed improved packet delivery ratio and lower end-to-end delay compared to standard AODV, due to the reduction of broadcast redundancy and selection of more stable routes.

Clustering and Hierarchical Routing: As introduced in earlier works and refined recently, clustering is an effective way to reduce overhead. By organizing vehicles into clusters (based on proximity or road topology), route requests can be directed cluster-to-cluster rather than node-to-node across the whole network. Recent studies have revived this approach with dynamic cluster formation algorithms suitable for VANET. Mahmood ul Hassan et al. (Sensors, 2024) use an intelligent clustering technique as part of their secure routing framework. In their approach, clusters are formed and a cluster head is elected based on a combination of factors including trust metrics (for security) and node stability. During route discovery, only cluster heads propagate the RREQ between clusters, acting as intermediaries. This reduces the broadcast storm in dense traffic. Additionally, since cluster heads are chosen partly by their “stability” (such as being a vehicle likely to stay within the cluster region longer, e.g., perhaps a slower vehicle or one centrally located), routes through cluster heads tend to be more long-lived, reducing the frequency of rediscovery. The authors reported that this cluster-based AODV achieved higher throughput and lower routing load than normal AODV in simulations, especially at high vehicle densities. The trade-off is the overhead of maintaining clusters, but if mobility patterns allow reasonably stable clusters, the net benefit is positive.

802.11p Contention Window Tuning: A novel cross-layer approach to cut latency is adjusting MAC layer parameters to expedite routing packets. Hussein and Mahmood (2023) presented P-AODV (Performance-enhanced AODV) which specifically targets the network congestion during route discovery in VANETs. They observed that during RREQ flooding in a busy VANET, collisions at the MAC layer (IEEE 802.11p) can cause delays and loss of RREQ/RREP messages, which in turn increases route discovery time and lowers packet delivery. P-AODV adapts the IEEE 802.11p Contention Window (CW) size based on vehicle density: in high-density conditions, using a larger CW (i.e., more aggressive backoff) can reduce collisions among routing packets, whereas in low-density, a smaller CW speeds up channel access. By dynamically tuning the CW, P-AODV improved the probability of RREQ/RREP reception and reduced routing delay. Their simulation results demonstrated a noticeable decrease in end-to-end latency and an increase in successful delivery ratio, as the MAC layer became less of a bottleneck​. This result underscores that not all improvements need to come from network-layer logic; optimizing how routing packets contend on the wireless medium also yields performance gains.

Multipath and Backup Routes: While the classic AODV is single-path, some recent variants ensure a backup route is available to reduce the latency of route repair. AODV’s standard local repair mechanism can still be slow if a route breaks, because if local repair fails, a full new RREQ flood is needed. Multipath AODV (like AOMDV) was known in MANET literature, and in VANET context, it has been revisited. For instance, some researchers incorporate a lightweight multipath discovery by having intermediate nodes send one RREP per neighbor to the destination (establishing multiple parallel paths). When a route breaks, the source immediately switches to an alternate path (if available) rather than initiating a new RREQ. This can dramatically lower the packet loss and delay during route breaks. The downside is slightly more routing overhead during the initial discovery (multiple RREPs). Recent performance evaluations (e.g., in [9]) indicate that multipath AODV (AOMDV) indeed yields higher packet delivery than single-path AODV in high-mobility scenarios, since there is often at least one path that survives longer​. However, pure multipath does not solve the RREQ overhead issue; it mainly addresses reliability and latency of repairs.

B. Route Stability and Link Quality Improvements

Another direction of enhancement is making AODV choose “better” routes that are more likely to remain stable or have higher quality, rather than the first route found (which might be shortest by hop count but not by other metrics). VANET links can be short-lived, so incorporating link quality or longevity predictions can improve performance.

Link Quality Metrics (ETX, RSSI): Standard AODV uses hop count as the route selection metric (with sequence numbers for freshness). Hop count does not reflect link reliability. In urban VANETs, a path with fewer hops might use long-range links that are lossy or intermittent due to buildings, whereas a slightly longer path might consist of shorter, more reliable hops. To address this, Al-Qassas and Qasaimeh (2024) investigated AODV variants that integrate the ETX (Expected Transmission Count) metric. They implemented AODV-ETX which calculates the ETX for each hop (the expected number of transmissions, including retransmissions, to successfully send a packet over that hop) and sums it for the route. Routes with lower total ETX (indicating higher quality links) are favored. They also studied enhancements like LETX (lightweight ETX, possibly a simplified or cross-layer version) and PLR (Packet Loss Ratio)-based metrics. In a comprehensive simulation, they found that these link-quality-aware versions of AODV significantly improved performance. In particular, AODV-PLR, which incorporated both ETX and a penalty for low RSSI (Received Signal Strength) in its route metric, achieved the best results. Compared to normal AODV, AODV-PLR delivered up to 8–37% higher throughput and reduced packet loss by as much as 5–15% (absolute) under various traffic loads. It also had dramatically lower routing overhead (up to 76% less than AODV in some scenarios) because fewer packets were dropped and fewer re-discoveries were needed. This demonstrates that using link quality metrics beyond hop count can make route selection smarter – choosing slightly longer but stable paths that require fewer repairs.

Link Lifetime Prediction: Some protocols attempt to predict how long a link to a neighbor will last, using the relative velocity and direction of vehicles. If two vehicles are moving apart quickly, any route connecting them will likely break soon. AODV enhancements like the one by Yu et al. (in the AODV-VANET concept) include link residual lifetime estimation. Each node can estimate, based on GPS information, how long it will remain within communication range of a neighbor (for example, using relative speed and radio range to compute time to disconnect). This estimated lifetime is then used to prefer routes where all hops have a longer predicted lifetime. One approach is to embed this info in the RREQ/RREP: when a node forwards a RREQ, it can include a “min link lifetime so far” field. The destination, upon receiving multiple RREQs, might choose the one with the maximum minimum-link-lifetime, meaning the weakest link on that path is stronger than the weakest links of other paths. This gives a route that should persist the longest. Simulations of such a scheme (e.g., in Springer’s WCN journal, 2020 by Mir and Sharma) showed an increase in route duration and a reduction in link break events, translating to higher PDR. The downside is that sometimes a slightly longer route (in hops) is chosen, which could increase initial latency a bit, but avoiding a break often pays off overall.

Fuzzy Logic for Route Quality: The fuzzy logic approach by Kandali and Hamid mentioned earlier not only suppressed RREQs but also directly influenced route choice. They computed a fuzzy cost for each possible route at the destination, incorporating factors like cumulative distance of the route, the number of directional changes, and relative velocity of nodes on that path. The route with the highest “fuzzy score” (which correlates with stability and lower expected delay) is selected, not necessarily the shortest. This holistic approach led to selecting routes with slightly more hops on average but with significantly lower variability in link quality. Reported results included a reduction in end-to-end delay and hop count (paradoxically, hop count decreased meaning the stable route was also often shortest – likely because unstable routes failed and had to be rediscovered, inflating average hop count in standard AODV)​. Fuzzy logic allows combining multiple metrics in a non-linear way, which is useful because VANET route quality is a multi-factor problem.

Load Balancing and QoS Routing: Some AODV enhancements target not just a single route’s quality but the distribution of traffic. In city VANETs, certain routes can become bottlenecks if many flows use the same relay nodes. AODV-L (mentioned in the ANN-based paper) refers to AODV with load balancing. This typically means that when a node is part of many active routes or its interface queue is filling up, it may refrain from replying to new RREQs or might suggest an alternate next hop (if known). By avoiding overuse of a single node or link, the network can improve overall throughput. Implementation of load balancing can be through a simple threshold (e.g., do not serve as relay if your current load > X) or more sophisticated network-wide awareness. One example (Zhu et al., 2019) introduced a “cross-layer congestion metric” where RREQs carried information about intermediate nodes’ queue length, and the destination picked a route that minimized congestion. This falls under QoS routing – trying to meet certain performance criteria like bounded delay or minimal packet loss. While not mainstream in VANET yet, these ideas show promise in ensuring that critical safety messages are not delayed by network congestion.

C. Security and Attack Mitigation

Security enhancements to AODV have become crucial as VANETs inch closer to real deployment. An insecure routing protocol could be exploited to disrupt communication on the road, with potentially catastrophic consequences (imagine an attacker preventing the propagation of accident warning messages). Recent approaches blend cryptographic techniques, trust frameworks, and machine learning to secure AODV.

Cryptography-Based Secure AODV: A straightforward (though resource-intensive) approach is to sign routing messages so that nodes can verify they originate from legitimate senders and haven’t been tampered with. For VANET, a common assumption is that vehicles can have certificates issued by a trusted authority (such as a Department of Transportation) – part of the IEEE 1609.2 security standard for WAVE. A secure AODV variant may use these certificates to sign RREQ and RREP messages. For example, ARAN (Authenticated Routing for Ad hoc Networks), an older protocol, could be adapted for VANET where every RREQ/RREP is signed and intermediate nodes only forward those with valid signatures. While this can thwart outsiders injecting false packets, it doesn’t stop a compromised insider from misbehaving (like sending a valid but deceitful RREP claiming a short path). Also, verifying signatures on every hop adds delay. A recent study by Sharma et al. (2020) combined cryptographic authentication with a threshold-based mechanism to detect anomalies. Their secure AODV used digital signatures to ensure authenticity, and additionally required that any route reply claiming a very high sequence number or very short route is cross-checked (either by an infrastructure RSU or by waiting for multiple RREPs to see if one is an outlier) – effectively to catch the “too good to be true” black hole replies. This hybrid of cryptography and plausibility checking was shown to eliminate black hole effects in testing, at the cost of a slight increase in route discovery time.

Trust and Reputation Systems: Trust-based AODV enhancements, like the T-AODV mentioned earlier, create a network where nodes maintain a reputation score for peers. One practical implementation is as follows: each node monitors the behavior of neighbors – e.g., does the neighbor forward packets as expected or drop them? If a neighbor is seen dropping packets it was supposed to forward (without cause, i.e., not due to link break), its trust rating is lowered. These ratings are shared (with caution to avoid false accusations by malicious nodes) or better, each node uses its own observations primarily. In T-AODV, before a node accepts a RREP from a neighbor offering a route, it checks its trust table. If the neighbor (or the next hop on the route) has low trust, the RREP is discarded. Over time, malicious nodes (which drop packets or send bogus info) lose trust network-wide and get isolated from routing. Hassan et al. (2021) presented a clustering-based trust model where cluster-heads act as trust aggregators and distribute reputation scores of members to others. Additionally, group authentication keys (managed by a central server or RSU) can be used – the TAD-HOC protocol introduced by Kumar et al. uses a key distribution center (KDC) that periodically refreshes keys and evicts detected malicious nodes from the network by not providing them updated keys​. In effect, legitimate nodes share a group key that is required to participate in routing; a node identified as malicious will fail to authenticate after a key update and is ignored. This approach cleverly blends infrastructure support with ad hoc routing to tighten security.

Machine Learning for Attack Detection: A very contemporary line of work embeds machine learning (ML) algorithms in the routing process to detect attacks or optimize routing decisions. The ANN-based secure routing protocol by ul Hassan et al. (2024) is a prime example. They train an Artificial Neural Network (ANN) to detect black hole attacks by recognizing patterns in routing messages (e.g., a node that consistently responds first to many RREQs with very high sequence numbers could be a black hole). Their system works in tandem with AODV: when a RREP arrives, the ANN classifier running on the source vehicle evaluates features of that route (like sequence number, hop count, etc., possibly even node IDs involved) to predict if the reply is from a malicious actor. If the ANN flags a potential attack, the source can ignore that RREP and seek alternative routes. In testing, this ANN-based detection achieved high accuracy (they report ~98.97% detection accuracy in identifying black hole nodes​). Furthermore, this approach was integrated with an authentication step – they employ a lightweight challenge-response with the supposed destination to verify that the RREP indeed came from either the destination or an authorized intermediate node. By combining ANN detection with an authentication handshake, the protocol dramatically improves security. The authors demonstrated that their enhanced AODV effectively maintained a >90% packet delivery ratio even in the presence of multiple black hole attackers, whereas standard AODV’s PDR dropped significantly under attack. The overhead of the ANN computation and extra handshakes was minimal in their evaluation, adding only a small processing delay.

Another ML approach uses SVM (Support Vector Machine) or gradient boosting trees to detect anomalies in routing behavior. For instance, a gradient boosting model might be trained to use features like “number of RREQs initiated vs. number of RREPs forwarded by a node” to spot a flood attacker or an always-responding fake node. These techniques show great promise in securing AODV, as they can adapt to new attack patterns if retrained with new data, making them more flexible than rigid rule-based systems.

Privacy Considerations: While not the main focus of routing protocol functionality, some related works have noted that VANET routing should also consider privacy (not revealing vehicle identities or exact routes). Secure AODV variants sometimes use pseudonyms instead of real IDs when signing messages, to prevent linking of packets to a specific vehicle’s identity. This is beyond AODV’s scope, as it’s usually handled by the application or security layer, but it’s worth noting as part of the holistic secure communication framework in VANETs.

D. Cross-Layer and Hybrid Approaches

Cross-layer designs break the conventional isolation of network layers to improve performance. We already saw an example with MAC contention window tuning. Another cross-layer idea is using physical-layer information like signal-to-interference-plus-noise ratio (SINR) in routing decisions. Rasha Aswad et al. (2022) proposed an enhanced AODV that they call SIR-AODV, which incorporates the SINR measured on received hello messages to decide route quality. If a neighbor’s signal is weak or highly variable, the protocol avoids using that neighbor in a route if possible, even if it means a longer route. This is similar to using RSSI, but with interference awareness. The result was a reduction in packet loss and retransmissions, improving end-to-end delay and throughput in their OMNeT++ simulations, especially in congested wireless environments​.

Hybrid approaches refer to combining different routing paradigms. For instance, a few papers have suggested a mix of topology-based and geo-based routing: use AODV for short distances or within a region, but use geographic forwarding for longer distances or to route between regions. One protocol used AODV within a city block and GPSR to hop between intersections (relying on digital maps and RSUs at intersections). Such hybrid routing can exploit the strength of both: AODV’s ability to find paths in a small local network and GPSR’s efficiency in wide-area delivery. An example is the Junction-based routing scheme improved by bio-inspired algorithms (ref. Azzoug & Boukra, 2024) where AODV finds routes to a junction, then hand off the packet to an intersection manager that forwards it to the next junction towards the destination (similar to a bus network). This structured approach was shown to reduce end-to-end delay in city scenarios with grid road layouts.

Integration with infrastructure is another hybrid angle. In scenarios with patchy RSU coverage, some proposals allow vehicles to use RSUs or cellular networks to assist routing. For example, if two vehicles cannot find a multi-hop path between them (network partitions are possible in VANET), an RSU connected to the internet or a cellular V2X relay might act as a waypoint. AODV can be extended to register with RSUs: a vehicle can send a RREQ to an RSU asking for a route to a destination; if the RSU has connectivity to that destination (either because the destination is in its vicinity or via other RSUs), it responds. This effectively turns the problem into a mix of ad hoc and infrastructure routing. While this goes somewhat beyond AODV’s original design, it has been implemented in research testbeds. The effect is improved delivery (since the infrastructure can bridge gaps), but one must trust the infrastructure and handle handoffs between ad hoc and infrastructure mode.

Table 1 below encapsulates several of the enhancements discussed, listing the focus of each approach and the reported benefits in the respective studies.

Table 1: Recent AODV Enhancements for VANETs (2019–2024)

Protocol / Enhancement	Key Idea	Reported Improvement
P-AODV (2023) – Parameter-optimized AODV	Adjust 802.11p MAC contention window based on network density to reduce collisions during route discovery.	Lower RREQ/RREP collision, leading to higher RREP delivery probability and reduced route discovery delay (improved latency and PDR in dense scenarios).
Fuzzy AODV (2020) – Distance-based Fuzzy Routing​	Use fuzzy logic with inputs like distance, velocity, direction to decide RREQ forwarding and route selection (favoring stable routes).	Fewer redundant RREQ broadcasts (less overhead) and selection of routes with longer lifetime, resulting in higher packet delivery and lower end-to-end delay than standard AODV.
Cluster-based AODV (2024) – Secure clustering	Form vehicle clusters with stable, trusted cluster-heads that manage routing for members. Only cluster-heads forward RREQs between clusters.	Greatly reduced broadcast overhead in dense networks; higher throughput and packet delivery in tests with heavy traffic, due to localized route management. Also enhances security by isolating untrusted nodes in clusters.
AODV-PLR (2024) – Link quality metric routing	Integrate ETX and Packet Loss Ratio (with RSSI factor) into route selection metric instead of hop count. Choose routes with best expected link quality.	~8–37% higher throughput than AODV depending on traffic load; up to 76% lower routing overhead, and significantly lower packet loss (by tens of percentage points) under high load, owing to more reliable links chosen.
Trust-AODV (2021) – Trust management for security	Maintain reputation scores for nodes; accept routing info only from trustworthy nodes. Optionally use group keys via RSU/KDC to authenticate participants.	Robust delivery even with attackers present: e.g., black hole attacks mitigated, sustaining PDR close to normal levels. Malicious nodes are identified and isolated, preventing route disruption. Overhead of trust management is modest.
ANN-Enhanced AODV (2024) – AI-based secure routing	Use an Artificial Neural Network to detect anomalies (like fake RREPs from black holes) and a cluster-based routing framework for content distribution. Integrate route prediction and authentication.	Achieved ~99% black hole detection accuracy with negligible false positives. End-to-end delay and throughput are significantly improved in presence of attackers compared to standard AODV, since malicious routes are avoided. Overall network throughput and reliability increased by using only safe, efficient routes.

(Sources: as indicated in brackets for each protocol)

Discussion of Enhancement Effectiveness

From the above survey of enhancements, it is evident that each approach targets specific limitations of AODV, and often there are trade-offs involved. For instance, methods that reduce RREQ flooding overhead (clustering, fuzzy suppression, etc.) can dramatically lower routing traffic, but they introduce additional complexity and sometimes slight forwarding delay (waiting for fuzzy logic decisions or cluster coordination). On the other hand, approaches focusing on link quality and stability may choose longer routes or spend more time gathering information (like multiple metrics, trust values), which could initially increase route setup time in some cases. However, if those routes break less often, the average performance still benefits.

Security enhancements like trust-based or ML-integrated routing demonstrate clear benefits in guarding against attacks, but they assume a certain level of cooperation or infrastructure. Trust systems need a bootstrapping mechanism – how are initial trust values assigned? Many assume a neutral trust for new vehicles, which can be risky if an attacker exploits that period. Nonetheless, the results show that once the system is running, it can detect and isolate attackers effectively. One must also consider the overhead of security: cryptographic schemes add processing and larger packet sizes (due to signatures). Trust and ML schemes may add processing delays. According to the studies, these overheads were kept relatively low. For example, the ANN detection in【7】was done in milliseconds on typical in-vehicle computing units, and trust score updates in【26】 are simple arithmetic operations.

An important observation is that many of these enhancements complement each other. It is conceivable to design a next-generation VANET routing protocol that incorporates multiple techniques: e.g., a cluster-based, link-quality-aware, trust-secured AODV. In fact, the proposal by ul Hassan et al. (2024) comes close – they use clustering (for both efficiency and security), an ANN for security, and also mention using a fitness function to choose short paths(which likely incorporates some link quality or hop count optimization). Their results were positive precisely because they tackled routing from multiple angles simultaneously.

However, combining features can also compound complexity. There is a risk of diminishing returns – each additional mechanism yields a smaller incremental benefit if not tuned properly. For example, if you already have a link-quality metric ensuring stable routes, adding a fuzzy suppression of RREQ might not further improve PDR much; it would mainly cut overhead. Conversely, if clustering already reduced overhead, using fuzzy logic within clusters might yield minimal extra gain but add complexity.

Another discussion point is the scalability of these solutions. VANETs can be very large (imagine thousands of vehicles in a city). Pure AODV does not scale well there due to broadcast storms. Clustering is a known way to improve scalability, and indeed it appears in multiple recent works. But clusters in a highly mobile environment can frequently change – cluster-head election might churn. Some papers have handled this by re-electing heads quickly or using road infrastructure as fixed cluster heads (RSUs as cluster heads). That leads to a semi-infrastructure solution, which in practice is likely for early VANET deployments (since cities are installing RSUs for V2X communications). Thus, the future of AODV in VANET might actually be in hybrid networks: partly infrastructure-assisted, partly ad hoc. Protocols like TAD-HOC exemplify this, where a centralized key server (infrastructure) enhances the security of an otherwise distributed protocol.

It’s also worth discussing the performance under different mobility scenarios. Highway scenarios with linear mobility differ from urban grids with intersections. Some enhancements (like junction-based forwarding or intersection assistance) are specifically for city environments. On highways, simpler AODV might suffice for moderate traffic, but at very high vehicle density (platooning scenarios), even highways can suffer broadcast congestion. The ETX-based improvements and multipath might be more universally beneficial (highway or city), as they address fundamental wireless link issues.

One limitation of many studies is that they focus on one type of scenario. For a protocol to be widely adopted, it must handle a range of conditions. AODV’s reactive nature might perform poorly if the network is extremely sparse (frequent network partitions – though then no protocol can do magic beyond using store-carry-forward like DTN routing, which is outside AODV’s scope). In extremely dense networks, a proactive or geobased approach may be superior. Thus, an adaptive protocol that can switch between modes is an interesting direction – e.g., operate in reactive AODV mode when network density is low-moderate, but switch to a proactive dissemination (or use periodic route caches) when density is high. Some preliminary work on density-aware dual-mode routing exists (not specifically for AODV, but conceptually applicable).

Finally, real-world deployment considerations such as compatibility and standardization will affect which of these enhancements are practical. The current ETSI and IEEE standards for VANET (V2V) networking have mostly focused on geonetworking and simple flooding for multi-hop (in ETSI GeoNetworking, a basic forwarding strategy is defined for multi-hop messages like cooperative awareness messages). AODV has not been included in standards yet for VANET, but some recent internet drafts have considered AODVv2 (also known as DYMO) for use in vehicular networks with modifications. The research we surveyed could feed into such standardization – for instance, a “VANET-AODV” standard could incorporate link metrics and security as default. It is encouraging that many proposals use existing technologies (like using IEEE 1609.2 certificates, or piggybacking on WAVE service announcements) so they could be implemented without hardware changes.

Performance Evaluation and Discussion

To illustrate the impact of the enhancements, we consolidate here some performance evaluation findings from multiple studies, focusing on key metrics: Packet Delivery Ratio (PDR), end-to-end delay, and routing overhead. Rather than raw numbers (which vary per scenario), we discuss relative performance trends.

Packet Delivery Ratio: Standard AODV typically achieves high PDR in lightly loaded or small networks but PDR degrades in highly mobile or dense scenarios due to frequent route breaks and packet drops during route discovery. Enhancements that target route stability (link-quality, lifetime prediction) directly improve PDR. For example, in Al-Qassas et al.’s experiments, at 40 data flows in a city scenario, AODV’s PDR dropped below 60%, whereas AODV-PLR maintained around 75% PDR​. The trust/ML-based protocols also keep PDR high by avoiding malicious disruptions – ul Hassan et al. reported maintaining ~90% PDR in presence of black holes, versus ~60% for vanilla AODV under attack​. Overhead-reduction techniques (clustering, adaptive RREQ) indirectly improve PDR by reducing collisions and congestion, thus lowering packet loss. Fuzzy AODV and P-AODV both showed a few percentage points improvement in PDR in dense scenarios compared to AODV​.

End-to-End Delay: This metric is crucial for safety messages. Pure AODV can have end-to-end delays on the order of hundreds of milliseconds up to seconds if a route discovery is needed at the time of message sending (depending on network diameter and RREQ timeout). In the evaluations, OLSR often had lower delay than AODV because routes are pre-computed. The enhancements to AODV narrow this gap. For instance, Rasha Aswad’s M-AODV (with reduced delay strategy) achieved ~20–30% lower average delay than AODV in city simulations by using faster route reply mechanisms and possibly caching​. P-AODV’s contention window tuning yielded lower delay especially at high density (the paper noted latency improvements across all evaluated scenarios)​. Clustering can introduce a small forwarding delay (since cluster-head must process), but if it reduces the need for re-routing, the net effect can be lower delay. Indeed, cluster-based schemes often show slightly higher initial route setup time but fewer disruptions, making the average end-to-end delay for packets better over long runs.

Throughput: Network throughput (aggregated) is improved when protocols efficiently deliver packets with minimal overhead. Many enhancements (ETX, clustering, etc.) showed higher throughput than baseline. For example, AODV-PLR had 8–18% higher average throughput than AODV depending on number of flows​. In scenarios with attackers, throughput of AODV can drop sharply (due to dropped packets or misrouting), whereas secure-AODV variants maintain throughput. A trade-off exists: some security mechanisms slightly limit throughput because they may avoid certain routes or add overhead, but they prevent malicious collapse of throughput. The ANN-based secure routing actually increased throughput compared to even attack-free AODV in their tests, because their clustering + route optimization also made routing more efficient aside from security​.

Routing Overhead: Almost all enhancements aim to reduce effective overhead or at least keep it manageable. It’s common to report overhead as the fraction of bytes or packets devoted to routing. In one study with 50 nodes, AODV overhead was about 30% of the network traffic, whereas an optimized version (with directional flooding and clustering) brought it down to 10–15%. AODV-ETX variants interestingly had lower overhead than AODV in the PeerJ study, even though they carry more info, because stable routes meant fewer RREQ retries​. The biggest overhead savers are clustering and suppressing broadcasts – these can reduce overhead by an order of magnitude in dense networks. Trust and security mechanisms tend to add some overhead (extra fields in packets, occasional control messages to update trust), but studies like【26】 indicate it’s not extreme. For example, TAD-HOC uses periodic small “trophy” messages for key refresh, which was negligible compared to data traffic.

One must also consider memory and computation overhead on vehicles – AODV is lightweight, but adding ANN means storing a model (though small) and running inference; trust tables scale with number of neighbors (usually not too high, perhaps dozens). Modern vehicles have the computation capacity to handle these, especially as dedicated V2X hardware matures.

In general, the performance evaluations suggest that with these enhancements, AODV-based protocols can meet the demands of VANET applications much better than the original protocol. Many enhanced versions exhibit performance on par or even superior to alternative routing protocols in their specific target scenarios. For instance, the fuzzy logic enhanced AODV outperformed OLSR in some metrics because it effectively found stable routes without the penalty of continuous topology dissemination​.

Conclusion

AODV has proven to be a resilient and adaptable routing protocol framework for VANETs, provided it is augmented with mechanisms to tackle the unique challenges of vehicular networks. This paper reviewed over five years of research aimed at improving AODV’s performance, reliability, and security in the context of VANETs. The enhancements range from algorithmic tweaks (like smarter RREQ dissemination and route choice using fuzzy logic or link metrics) to architectural shifts (such as clustering and hybrid routing) and incorporation of external techniques (cross-layer optimization, trust management, and machine learning for security).

Through technical analysis, we highlighted that reactive routing can be made viable for highly dynamic VANET environments by reducing the overhead and delay of route discoveries. Techniques like adaptive flooding and clustering directly address the scalability problem of AODV’s flooding mechanism, making it more bandwidth-efficient in dense traffic conditions. Meanwhile, replacing or supplementing the hop-count metric with quality indicators (ETX, RSSI, predicted lifetime) yields more robust routes that maintain connectivity longer, thereby improving packet delivery and latency. These improvements diminish the traditional gap between reactive and proactive protocols, allowing enhanced AODV to achieve low latency similar to proactive protocols, but with lower overhead when traffic is sparse or intermittent.

In terms of security, recent AODV variants have made great strides in mitigating vulnerabilities. By integrating trust and verification, they ensure that the network can defend itself against common attacks like black holes without requiring an entirely separate security protocol. The use of ANN and other ML techniques is especially promising as it brings adaptability – the system can potentially learn to detect new types of malicious behavior over time. Such intelligence will be crucial as VANETs grow and face more sophisticated cyber threats.

One recurring theme in the developments is the benefit of cross-layer cooperation and infrastructure support. Purely ad hoc, layer-isolated operation is suboptimal in VANETs. The best results were often seen when information from the physical layer (signal quality), MAC layer (contention, load), or external sources (like RSUs providing trust anchor or time synchronization) were utilized by the routing protocol. This trend is likely to continue, with future VANET routing protocols being designed as part of an integrated V2X network architecture rather than a standalone MANET protocol. For example, upcoming 5G/NR-V2X standards include modes for direct vehicular communication; a future AODV-based scheme could potentially operate on top of those, using sidelink scheduling info from 5G to aid route decisions.

Despite the progress, some challenges remain open. VANET environments can be extremely heterogeneous – city vs highway, day vs night (traffic load), with or without infrastructure – and an adaptive routing protocol is needed to handle all situations. Many current proposals excel in the scenarios they target but might need tuning to generalize. There is room for a comprehensive protocol that perhaps dynamically switches strategies: e.g., uses geography when available, falls back to reactive when not, employs clustering in ultra-dense conditions, and so on. Such complexity will require extensive testing to get right.

Another challenge is standardization and interoperability. To implement these research solutions in real vehicles, one must ensure they can interoperate across manufacturers and deployments. This means converging on a set of enhancements that yield the most benefit while remaining implementable in vehicular hardware and within regulatory constraints (channel congestion control etc.). The survey shows that many proposals are backward-compatible with AODV (they often retain the same basic packet format with minor additions), which is encouraging.

Future work could explore synergy between routing and emerging technologies like edge computing. For instance, vehicles could offload heavy computations (like running a complex ML model for routing optimization) to edge servers that then guide routing decisions. Also, with the rise of connected autonomous vehicles, the requirements on the network (for ultra-low latency and reliability) will push further innovations in routing.

In conclusion, AODV, when enhanced with modern techniques, remains a strong candidate for VANET routing. The protocol’s foundational simplicity and on-demand nature align well with the dynamic needs of vehicular communications, and the research community has shown that its weaknesses can be systematically mitigated. As vehicular networks move from theory to deployment, the insights from these studies will be instrumental in crafting routing protocols that are not only efficient and robust, but also secure and context-aware. AODV’s evolution in the VANET realm exemplifies how classic MANET protocols can be successfully transformed to meet new technology frontiers.

References

1. A. A. Hussein and D. A. Mahmood, “Enhancement AODV routing protocol at the VANET within an urban scenario,” Int. J. Recent Innov. Trends Comput. Commun., vol. 11, no. 9, pp. 5162–5168, 2023​ijritcc.org​ijritcc.org.

2. C. Perkins, E. Belding-Royer, and S. Das, Ad hoc On-Demand Distance Vector (AODV) Routing, IETF RFC 3561 (Experimental), Jul. 2003​rfc-editor.org.

3. “Ad hoc On-Demand Distance Vector Routing,” Wikipedia. [Online]. Available: https://en.wikipedia.org/wiki/Ad_hoc_On-Demand_Distance_Vector_Routing. (accessed Mar. 28, 2025)​en.wikipedia.org​en.wikipedia.org.

4. L. Hota, B. P. Nayak, A. Kumar, B. Sahoo, and G. G. M. N. Ali, “A performance analysis of VANETs propagation models and routing protocols,” Sustainability, vol. 14, no. 3, Article 1379, 2022​mdpi.com​mdpi.com.

5. R. Q. Aswad, M. A. Abdala, et al., “End-to-end delay enhancement with AODV in VANET,” Proc. of ???, 2022 (Available via Academia.edu)​academia.edu.

6. K. Kandali and B. Hamid, “An efficient routing protocol using an improved distance-based broadcasting and fuzzy logic system for VANET,” Int. J. Intell. Eng. Syst., vol. 13, no. 6, pp. 95–104, 2020​pmc.ncbi.nlm.nih.gov​pmc.ncbi.nlm.nih.gov.

7. M. ul Hassan, A. A. Al-Awady, A. Ali, et al., “ANN-based intelligent secure routing protocol in vehicular ad hoc networks (VANETs) using enhanced AODV,” Sensors, vol. 24, no. 3, p. 818, 2024​pmc.ncbi.nlm.nih.gov​pmc.ncbi.nlm.nih.gov.

8. A. Kumar, V. Vijayakumar, K. Abhishek, et al., “Black hole attack detection in vehicular ad-hoc network using secure AODV routing algorithm,” Microprocess. Microsyst., vol. 80, p. 103352, 2021​mdpi.com.

9. R. Al-Qassas and M. Qasaimeh, “An empirical evaluation of link quality utilization in ETX routing for VANETs,” PeerJ Comput. Sci., vol. 10, Article e2259, 2024​peerj.com​peerj.com.

10. A. Ali, M. M. Iqbal, S. Jabbar, et al., “VABLOCK: A blockchain-based secure communication in V2V network using ICN network support technology,” Microprocess. Microsyst., vol. 93, p. 104569, 2022.

11. Aswathy M. C. and C. Tripti, “A cluster based enhancement to AODV for inter-vehicular communication in VANET,” Int. J. Grid Comput. Appl., vol. 3, no. 3, pp. 41–51, Sep. 2012​airccse.org​airccse.org.

12. S. K. Sharma, X. Fernando, and S. S. Das, “A reliable routing protocol for VANET communications,” in Proc. IEEE ANTS, 2019, pp. 1–6.

13. F. Honarmand, “T-AODV: a trust-based routing against black-hole attacks in VANETs,” arXiv preprint arXiv:2305.00000, 2023​assets-eu.researchsquare.com​assets-eu.researchsquare.com.

14. Y. Xi, W. Shu, et al., “Ant-AODV-VANET: a bio-inspired routing protocol for VANET,” Wireless Pers. Commun., vol. 80, pp. 177–203, 2015​researchgate.net.

15. S. Ahmed, M. M. Alam, “Link residual lifetime-based next hop selection scheme for vehicular ad hoc networks,” EURASIP J. Wireless Commun. Netw., vol. 2016, Article 45, 2016​jwcn-eurasipjournals.springeropen.com.