Abstract
Today, intelligent transportation systems (ITS) have received a lot of attention due to their significant impact on increasing the safety, efficiency, and convenience of transportation. One of the main applications of ITS is vehicular ad hoc networks (VANETs). In particular, a more flexible, reliable, real-time, and scalable routing scheme across the large urban areas is one of the most critical issues for VANETs. Past VANET routing methods have various technical issues with VANET evolutions. On the other hand, clustering improves the reliability and scalability of routing schemes in VANETs. In this paper, a cluster-based, Traffic-aware and Low-Latency Routing Schema (TaLAR) is proposed for VANETs. In the proposed scheme, the Harris hawks optimization (HHO) algorithm is used to select cluster head (CH) nodes by considering appropriate parameters such as intracluster distance, link reliability, and relative speed of vehicles. The path between source and destination CH nodes is identified by using the HHO algorithm; it chooses the appropriate route based on link reliability and intercluster distance. Also, in the interconnection area, a traffic-aware and reliable route is identified by using a digital map of the streets and the Dijkstra algorithm. The performance evaluation of the proposed scheme is analyzed in terms of packet delivery rate (PDR), average end-to-end delay, and throughput. The output of the proposed scheme is compared with the Clustering Routing Based on PSO (Particle Swarm Optimization) (CRBP) and Grey Wolf Optimization Based Clustering in Vehicular Ad Hoc Network (GWOCENT) methods. Simulation results show that the proposed scheme improves PDR (22 and 19%), throughput (25 and 21%), and average end-to-end delay (23 and 18%).
1. Introduction
In recent years, VANETs as a special type of Mobile Ad hoc Networks (MANETs) [1] are aimed at transmitting safety information from source to destination vehicles to improve passengers’ safety and prevent accidents. One of the critical issues in VANETs is the real-time and reliable information transmission from the source to the destination vehicles to allow drivers to make appropriate and timely decisions that increase road and passengers’ safety [2]. Recently, VANETs support different types of applications and services, such as mobile vehicular cloud services, route discovery, traffic monitoring, and context-sensitive infotainment [3]. Due to the increasing trend of critical applications of VANET and its dynamic nature (high mobility of vehicles), an efficient, real-time, and reliable routing scheme is essential and plays a key role in supporting safety and improving the overall quality of service (QoS) in most applications of VANETs [4–9]. However, VANETs have special limitations and characteristics, such as unsteady connectivity, frequent exchange of information, fast changes in network topology, and high vehicular node mobility that can significantly impact data transmission schemes [10–12]. VANET data transmission schemes use V2V (vehicle-to-vehicle), V2I (vehicle-to-infrastructure), and hybrid methods. In VANETs, the existing routing schemes can be classified into the following subcategories: (a) broadcast-based schemes are used when the vehicle node that will receive the message from other nodes is out of transmission range. This scheme uses flooding data transmission, and every vehicle node in the VANETs will receive the safety and security messages. This scheme improves the packet delivery rate. (b) Position-based schemes are selected when the positions of the source and destination nodes and the optimal route to the destination vehicle node can be known using the global positioning system (GPS). This type of routing scheme does not need to maintain any routing table. (c) Topology-based schemes are based on the topology of VANET nodes and maintain a routing table in each vehicle node. This type of routing scheme is slower than other schemes. (d) Cluster-based schemes increase scalability and decrease the number of control messages via data aggregation in CH nodes. CH nodes of two clusters send data to each other and can make a proper decision on routing data packets. Moreover, several cluster-based routing schemes have been proposed for VANETs [13–15]. Figure 1 shows a cluster-based routing scheme in VANET. In this figure, CH gathers data from its members, aggregates, and transmits that to another CH [16]. Clustering in VANETs is an NP-hard problem, and metaheuristic algorithms are appropriate to solve this problem [17, 18].
This paper proposes a cluster-based routing scheme for VANETs using the HHO algorithm. The main objective of the proposed scheme is to improve the QoS (quality of service) parameters such as packet delivery rate, average end-to-end delay, and throughput. In the proposed scheme, the HHO algorithm is used for the clustering process (CH selection) due to its less computational and timing complexities and high scalability features. In the interconnection area, the proposed scheme obtains an adjacent matrix from the map of that area. Then, identifies the shortest path between the vehicles in the interconnection area using the Dijkstra algorithm.
Briefly, the main contributions of this paper are as follows: (i)The HHO algorithm is used for selecting appropriate CH nodes considering proper parameters such as link reliability, the relative speed of the vehicle, and intracluster distance(ii)The shortest route between source and destination CH nodes considering link reliability and intercluster distance is identified using the HHO algorithm(iii)The Dijkstra algorithm is used for selecting a traffic-aware and reliable routing scheme in the interconnection area(iv)Average end-to-end delay, packet delivery rate, and throughput are improved because of the appropriate CH and path section using the HHO algorithm
The rest of this article is organized as follows: Section 2 explains the related work in the field of routing in VANETs. Section 3 presents the proposed TaLAR method. Section 4 discusses the performance and comparative analysis of the proposed scheme with existing methods. Finally, the conclusion of this article is given in Section 5.
2. Literature Review
There are various research works have been made on cluster-based routing schemes for VANETs. Some researchers from related works are discussed in this section.
In [19], the authors proposed a grey wolf optimization-based clustering algorithm for VANETs using the social behavior and hunting scheme of grey wolfs for creating optimal clusters. The special characteristic of grey wolf nature makes it converge earlier, which creates the optimal number of clusters. Simulation results show that it is an optimal and cluster-based robust routing protocol for VANETs, which is suitable for highway scenarios and can improve the PDR.
In [20], using a Tabu search algorithm, a reliable and multilevel routing scheme was introduced. Even if the topology constantly changes, multilevel clustering-based routing enables the organization and sustains the path. Furthermore, it solves the problem of creating a local optimum trap using the Tabu search. Tabu search uses an appropriate fitness function to choose the solution among a set of solutions. The effective parameters of this method are selecting the best route, including the distance between nodes, the speed of nodes, node angle, link sustainability, and link reliability. The evaluation of this method shows that PDR and average end-to-end delay are improved.
In [21], a reliable and cluster-based routing algorithm is proposed for VANET. In this method, for a proper clustering process, the authors consider different parameters such as node degree and node coverage by using an imperialist competitive algorithm. For selecting optimal CH nodes, this method uses Radiation-Based Function (RBF) and considers appropriate parameters such as free buffer size and speed of vehicles.
In [22], the authors proposed a hybrid genetic algorithm (GA) to improve cluster maintenance in the Weighted -medoid Clustering Algorithm (WKCA). In this model, the Tabu search algorithm was combined with the GA (genetic algorithm) to allow the search of the whole state space and obtain the most suitable solution and avoid trapping the local optima. This model enhances the allocation of nodes to clusters, which makes effective vehicle communication and ensures the reliability of the cluster architecture. According to the evaluation criteria, the results of this model are more stable and robust than the original WKCA and other approaches.
In [23], the authors presented a hybrid scatter Tabu search (HSTS) method to allocate cluster members to the proper CH nodes. In this study, the WKCA (Weighted -medoid Clustering Algorithm) is used to form the cluster and to obtain the optimal solution. The Tabu search is integrated into the internal scatter searching process to locate the global minimum. The simulation results show that this model improves network sustainability.
In [24], effective cluster-based routing scheme in VANET was presented to improve V2V performance using PSO algorithm. Three basic factors in selecting clusters are considered for the sustainability of clusters: alignment with nodes, more neighbors, and similar velocity and position with other nodes. First, the CH nodes are chosen. Second, to optimize the necessary routing, particle path, node’s speed coding rules, iteration rules, and proper functions are planned. Third, the routing method is used in clusters and among clusters. The results show that the number of nodes, communication radius of nodes, and maximum hops between the CH nodes and each normal node in the cluster have an important impact on the proposed scheme performance.
In [25], the author presented an enhanced cluster-based AODV for the Internet of Vehicles (AODV-CD) to get a stable and efficient clustering to simplify routing and ensure QoS. To achieve cluster stability, this method applies two unique messages, i.e., the HELLO packet and the CH packet, in the route discovery phase and the path conservation process in the clustering algorithm for the AODV protocol. The route replay (RREP) message is transmitted to nodes if a path is available; otherwise, the CH will send the RREQ message. Due to reduction in RREQ messages, this method reduces congestion and network overload.
In [26], the authors presented a destination-aware context-based routing (DACR) method using a clustering process with soft computing for VANETs. This scheme consists of two phases. In the first phase, they combined geographic and context-based clustering scheme. The object of this clustering scheme was to prevent network congestion and reduce clustering overhead. In the second phase, a routing algorithm based on destination is presented for data transmission between two cluster nodes which improves the overall PDR and end-to-end delay.
In [27], the authors proposed a fuzzy-logic-based routing scheme for VANETs to improve the QoS parameters. They have defined two important parameters to identify the forwarding vehicle node: channel quality factor (CQF) or “” and communication expiration time or “.” This routing scheme includes two main parts: (a) determining the forwarding vehicle node in the road based on the fuzzy logic and (b) road selection at the road junction to choose the right route to reach the data to the destination node. The authors have compared their proposed scheme with other well-known protocols such as MoZo, BRAVE, and OFAODV. Simulation results show that this proposed scheme has higher performance in terms of average end-to-end delay and control packet overhead.
In [28], the authors proposed a new hybrid cluster-based routing algorithm using a modified -means algorithm with continuous Hopfield network and maximum stable set problem (KMRP). In this scheme, the maximum stable set problem combines with a continuous Hopfield network for selecting optimal clusters. Then, each cluster member nodes assign to CH nodes using the -means algorithm considering link reliability and distance. Finally, the cluster head nodes are selected using an appropriate fitness function with different parameters such as the amount of free buffer space, the speed, and the node degree. KMRP increases throughput metric by reducing traffic congestion. In addition, the KMRP method reduces the average end-to-end latency and increases the PDR.
In [29], the author proposed a cluster-based routing protocol for hybrid VANET-WSN communication (PRAVN) for road safety applications. The author used improved water wave optimization (IWWO) algorithm for clustering the vehicle nodes. An important contribution of PRAVN is the introduction of a rider optimization (RO) algorithm to choose neighbor vehicle nodes to improve the network lifetime and lossless connections. This method by forwarding valid data from the sender to the receiver vehicle nodes will improve road safety.
In [30], the authors proposed a new improvement to the scheme of OLSR (Optimized Link State Routing Protocol), named CACA (Cluster-based Adept Cooperative Algorithm). In this scheme, each node identified a reliable route between source and destination using the cluster-based QoS algorithm. The CACA is aimed at maintaining long-lived paths for which the most sustainable path is adaptively selected considering different parameters such as signal strength and vehicle node mobility. It decreases control message overhead and updates the routing table.
The work in [3] proposes an Innovative Cluster-Based Dual-Phase Routing Protocol Using Fog Computing and Software-Defined Vehicular Network (ICDRP-F-SDVN). The authors combined fog computing technology and Software-Defined Networks (SDN) and proposed a reliable and robust architecture that overcomes critical challenges arising from advanced technological development and fast escalation of intelligent vehicle nodes. This scheme can reduce the overhead of control messages.
Table 1 shows the compression of the routing scheme for vehicle ad hoc network literature.
Using an intelligent routing scheme can effectively improve the quality of service parameters such as average end-to-end delay, PDR, and throughput in VANETs. However, for critical applications of VANETs, PDR and average end-to-end delay should be improved. The Harris hawks optimization (HHO) algorithm finds the optimal path with the search and hunt mechanism. The advantages of HHO are high accuracy, fast convergence, and easy realization appear. Therefore, this paper uses the HHO algorithm to select optimal CH nodes and identifies optimized paths among clusters and improves the quality of service parameters.
3. The Proposed Scheme
The proposed method has the following two main phases. In the first phase, an appropriate clustering scheme is proposed using an HHO algorithm with a new fitness function to select proper CH nodes. In the second phase, a reliable and real-time routing scheme is proposed using the HHO algorithm to find a reliable and real-time path between source and destination CH nodes. This section explains the proposed scheme in detail.
3.1. Clustering Phase
Variable description used in all equations is given in Table 2.
We assume that the vehicles are moving at the same speed and . For appropriate CH selection, a new fitness function is obtained using the following parameters:
3.1.1. Intracluster Distance
This important parameter is the average distance between each cluster head node (CH) and its cluster members and can be calculated using Equation (1). If this parameter is minimum, then data transmission latency is also minimum. Hence, the intracluster distance is , which can be minimized as follows: wherein the Euclidean distance is computed using Equation (2) as follows: where we have
3.1.2. Relative Speed of Vehicles
The relative velocity is an important parameter for selecting CH nodes and can be calculated using the following equation: whereis the angle between nodesand.
3.1.3. Link Reliability
In VANETs, due to node mobility and high relative velocity of vehicle nodes, finding a reliable path is a complicated task. In this regard, the link reliability is an important metric for discovering optimum path between vehicle nodes. Therefore, the link reliability among of the vehicles can be calculated using the following equation:
Therefore, two vehicle nodes with higher link reliability is a more preferable choice as a CH. It enhances the packet delivery rate (PDR) and reduces the packet loss. Hence, our other objective in terms of link reliability is , which can be minimized using the following equation:
All of the above-mentioned parameters of fitness function must be converted into a single fitness function. Therefore, the weighted sum approach is applied and the fitness function can be obtained using the following equation: where , , and are the weights assigned to each objective. The value of these weight parameters is between 0 and 1, and The values of these parameters are as 0.3, 0.45, and 0.25, respectively. The link reliability is considered as high priority to select CH node, and so the value of is 0.45.
As we know that all the above-mentioned parameters have different units, therefore, min–max normalization function is applied to each parameter using the following equation: where is the value of the function, is minimum value, is maximum value, and is the normalized value between 0 and 1.
After selecting the CH nodes, each CH node advertises itself as a candidate node to the neighbor nodes. The neighbor vehicle nodes are allocated to the CH nodes with less distance.
3.2. Routing Process Using HHO
The process of finding the shortest path from the source vehicle to the destination vehicle is done using HHO [31]. The purpose of using the HHO method is to find the optimal and reliable path based on link reliability and interclusters distance. Hawks have to select paths that are more reliable and shorter than other paths. The fitness function of selecting the optimal next CH node for data transmission can be obtained from the following equation: where , , and are link reliability, distance between two neighbor CH nodes, and weighting parameter, respectively. Min–max normalization function is applied to each parameter of Equation (9) using Equation (8). The value of is 0.6, and this weighting parameter is considered for taking link reliability as higher priority in selecting next CH node as forwarding node.
The steps of the proposed HHO routing algorithm are as follows.
3.2.1. Exploration and Transformation Phases
In this phase, the Harris hawks are often placed in special positions, wait, search, and detection of prey based on the following equation [31]:
Here, and show the current and the next iteration, respectively. In addition, , , , and show the position vectors of the hawks, as is the random number in the range of [0,1]. The average position of the hawks can be calculated using the following equation [31]: where and indicate the positions of each vehicle in iteration and the overall number of vehicles. One of the important stages of HHO is the transition from the exploration phase to the exploitation phase; this change is expected between the different simulated exploitative behaviors based on the evading energy factor of the prey (), as it reduces dramatically during the evading behavior. The energy of the prey can be calculated using the following equation [31]: where , , and show the initial escape energy, total number of iterations, and the escape energy, respectively.
3.2.2. Exploitation Phase
(1) Soft Besiege. The HHO soft besiege behavior can be calculated using the following equation [31]: where is the jump intensity of escaping process that can be demonstrated as .
(2) Hard Besiege. In this case, the current locations are obtained using the following equation [31]:
(3) Soft Besiege with Progressive Rapid Dives. We assumed that hawks could compute (decide) their next run based on the following equation [31]:
The above model calculates the next moving stage () of the hawks. Hawks also dive to attack prey according to the following equation: where LF represents the levy flight function and is the random vector for size ; at this point, the position is updated as follows [31]:
(4) Hard Besiege with Progressive Rapid Dives.
From here, and are estimated using the following equation [31]:
Algorithm 1 shows the routing process of the proposed method.
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||
3.3. Traffic-Aware Routing in Interconnection Area Using HHO
Each vehicle can obtain a graph adjacent using a digital street map that models the city map. Whenever a vehicle node wants to send a packet to another node, it must first add two source and destination vertices to the matrix. It then calculates the shortest route using the Dijkstra algorithm and sends the packet to the destination vehicle. The marked intersections of the city map in the circles and the street sections with length are given in Figure 2. In the graph of this map, the intersections with the vertices and the streets are shown as edges. Also, the corresponding proximity matrix is shown in Figure 3 [32].
Using the Dijkstra algorithm on the graph, which shows the city map, the list of intersections is calculated. The weight graph of each edge is determined by connecting that segment of the street. The weight of each edge is proportional not only to its length but also to the network connection, and the matrix values are redefined in Figure 3 as , where
A weak connection is indicated by the value of due to low traffic. The vehicles update values. The next street is chosen randomly by considering the number of vehicles. The vehicle node sets the nearest neighbor for everyone based on the next intersection [32]. The connection to pass through the street section between the intersections of and and subsequently and will be updated using the HHO algorithm when the intersections of and are registered in the two consecutive connection points based on the following equation [32]. where the variable is computed based on the following equation [32]: where , , and are the time spent, the minimum delay recorded by the vehicle for that street, and a fixed value, respectively. Since the minimum is less than or equal to , the variable is between zero and 0.5. Therefore, the required values for variable will be less than 0.5, and for variable less than 1. Also, the initial value for the variable will be less than 1. Each time is renewed, so the value of decreases. At constant intervals of each vehicle, reduces all streets using the following equation [32]. where . After a period of time, we reduce the pheromone with Equation (22). So, we multiply the previous pheromone by .
Therefore, the route chosen by the vehicle of the source will be appropriate to the traffic conditions of that area. Also, each vehicle node regularly broadcasts its ID and location on its radio range in that area. Each vehicle sets its timer so that fewer monitors are needed.
4. Performance Evaluation
The proposed TaLAR approach had been simulated, and its performance was evaluated in Network Simulator version 2 (NS-2) on Linux Ubuntu 18.04 LTS environment. This section indicates the performance evaluation of the proposed scheme to validate its efficiency. The simulation results of the proposed scheme were compared with both methods (CRBP [24] and GWOCENT [19]).
4.1. Performance Metrics
The proposed scheme and other two methods were compared based on three metrics PDR, throughput, and average end-to-end delay.
PDR: the total number of data packets that arrived at the receiver vehicle is divided by the total number of packets sent from the source vehicle and can be calculated through the following equation [33]:
Throughput: this important metric is the ratio of the total size of data packets delivered to the
receiver to the total simulation process time and can be obtained from the following equation [34]:
Average end-to-end delay: this metric is the average time interval between receiving and sending times for a packet from a transmitter to receiver node and can be calculated through the following equation: where , , and are receiving time, sending time of a packet, and the total number of packets, respectively.
4.2. Simulation Results
The basic simulation parameters used in the performance evaluation phase are listed in Table 3.
PDR: PDR has decreased due to path breakdown, congestion, and short lifespan link. As shown in Figure 4 and Table 4, the PDR rate of the proposed TaLAR method is higher than CRPB and GWOCNET in all three scenarios by changing the transmission range. The reason is that the proposed TaLAR method considers appropriate fitness function with several significant parameters such as link reliability, relative speed, and intracluster distance for clustering and routing process. Selecting CH nodes using the HHO algorithm with proper fitness function based on link reliability and relative vehicle speed will increase the stability of networks. Stable and reliable routes will increase the packet delivery rate. The use of these important parameters and the HHO algorithm contributes to the selection of a reliable, noncongested short route to increase PDR. In Figure 4(a), in the transfer range equal to 10, the proposed TaLAR method performs better than the CRPB and GWOCNET methods and has a higher packet delivery rate. In Figures 4(b) and 4(c), this superiority is maintained as the transmission range increases. In the CRPB method, only clusters based on PSO are selected to send data, and practically, no other strategy for transmitting data is considered. In the GWOCNET method, only clusters are selected based on the grey wolf algorithm and only with the distance criterion.
(a)
(b)
(c)
Throughput: Figure 5 and Table 5 show the throughput versus the number of vehicle nodes for three other schemes: TaLAR, CRPB, and GWOCNET. The results of the simulation of TaLAR show that the proposed method in this criterion is better than the other two methods. In the scenario with a transfer range of 10 in Figure 5(a), the proposed method has a much higher throughput than the CRPB and GWOCNET. This is because the proposed scheme determines the optimal CH nodes based on several significant parameters such as link reliability, the relative speed of vehicles, and intracluster distance. The proposed method determines the optimal route based on important QoS parameters such as link reliability using the HHO algorithm. In general, more vehicle nodes in the VANETs increase the throughput metric. This is because a vehicle gets more forwarder vehicle nodes with high reliability to forward its data. However, this significant improvement and the result are the idea of using the HHO with several significant parameters for the clustering and routing process. This combination determines reliable and robust CH nodes and routes between source and destination vehicle nodes.
(a)
(b)
(c)
Average end-to-end delay: Figure 6 and Table 6 show that the proposed TaLAR method performs better in terms of average end-to-end delay than the CRPB and GWOCNET methods and has less latency in all three scenarios. The CRPB method is considered an only node and link stability in the network to reduce latency. In the GWOCNET method for selecting CH nodes, only the distance criterion is considered. In the proposed method, the use of the HHO algorithm with considering several significant parameters such as link reliability, intracluster distance, and the relative speed of vehicles avails reliable and real-time path from source to destination vehicle, which improves the end-to-end delay. Therefore, for these reasons, in the proposed TaLAR method, with increasing the data transfer rate and the number of vehicles, the average delay compared to other methods has decreased significantly. In the proposed scheme, considering the link reliability in selecting the next CH node for data transmission decreases the packet loss and hence, the retransmission of lost packets will be decreased significantly. This reduction of retransmission decreases network congestion and improves the average end-to-end delay.
(a)
(b)
(c)
5. Conclusion
VANET requirements are having more reliable, scalable, and well-connected routes. This paper has proposed a cluster-based reliable and real-time routing scheme (TaLAR) for VANETs using the HHO algorithm. In the proposed scheme, a new clustering scheme was presented considering several significant parameters such as the relative velocity of vehicle nodes, intracluster distance, and link reliability. The proposed method has the following two main phases routing scheme for road and interconnection scenarios. The main goal of the first phase was to find a reliable and real-time path during the route discovery process between source and destination CH nodes. Also, the goal of the second phase was to identify the shortest path between vehicle nodes using a digital street map and the Dijkstra algorithm. The proposed scheme is scalable and applicable for V2V and V2I and hybrid architectures of VANETs. The results of simulation in NS-2 showed that the TaLAR method has much better performance than CRBP and GWOCENT methods in terms of PDR (22 and 19%), throughput (25 and 21%), and average end-to-end delay (23 and 18%), respectively.
In the future, we will propose an SDN-based architecture for the proposed scheme to make it more robust and reliable.
Data Availability
No data were used to support this study.
Conflicts of Interest
The authors declare that they have no conflicts of interest.