1. Introduction
Recently, immense increase in the number of automobiles on the road made driving difficult and unsafe. Roads are routinely replete with vehicles, and therefore, safe-separation distance and sensible speeds do not seemed to be valued anymore. For instance, according to NHTSA (National Highway Traffic Safety Administration) [
1] in 2006, there were an estimated 42,642 traffic-related casualties. The percentage (2%) of casualties in 2005 was even higher than that of casualties in 2006. There are some advanced active and passive vehicle safety-related devices such as airbags, crumble zones, and anti-lock brakes invented to reduce causalities. The number of casualties in spite of all these latest devices has remained more than 40,000 per year for the last 15 years. Accidents do happen, but sight cannot be lost of the concept of road safety. Top car manufacturers and nationwide government agencies are determined to design solutions that assist drivers in anticipating danger and avoiding bad traffic zones. One such effort is the exploitation of wireless technology, to direct traffic issues, in the form of wireless access for vehicular environment (WAVE) devoted to the vehicular ad hoc networks (VANETs) [
2‐
4]. The aim of WAVE is to overcome traffic issues and make driving efficient by giving timely guidance to drivers and vehicles that are not available through driver observations and independent sensors [
1,
5‐
7].
Vehicular communication is possible through either vehicle to vehicle (V2V) or vehicle to infrastructure (V2I) or both. The goal of VANETs is to build an intelligent transportation system (ITS) [
8]. It supports a wide variety of applications including prevention of accidents, traffic flow control mechanisms, information services, real-time alternate route computations, and provision of Internet ccess to the vehicles on motion [
4,
8‐
13]. Vehicular communication is an offshoot of mobile ad hoc network (MANET) [
9,
12,
14]. VANETs have certain characteristics that change them from MANETs. In VANETs, >high-speed vehicular nodes have intermittent connectivity and they frequently change the network topologies. Physical factors restrict node movements and network topologies. The movement of vehicles is along roadways, and their mobility is restrained by traffic policies, such as traffic light signals, speed constraints, and road/traffic conditions. Power expenditure is not a vital concern in vehicular nodes, as vehicles can generate enough power to run the communication devices. In contrast, MANET's nodes are characterized by limited storage capacity, restricted battery power, low processing, and random movements with unpredictable mobility patterns [
8,
9,
12,
13].
This work realizes the potential of VANETs to enhance traffic safety and traffic organization and to facilitate driving through a smart transportation system. In this work, we explored routing features of VANETs. We analyzed the previous routing literatures in VANETs. We have presented their contribution and limitations. By using the unique characteristics of VANETs, we have developed a novel position-based routing protocol called traffic flow-oriented routing (TFOR) protocol for VANET multi-lane roads. It considers the vehicle's position, direction, and speed to decide the junction and dispatch the data packets. It consists of two mechanisms: (1) new junction selection mechanism and (2) routing based on two-hop neighbor information. The new junction selection mechanism determines the directional and non-directional traffic flows based on real-time traffic in city environment. Determination of these flows provides shortest rich-density routing paths, which increase packet-delivery ratio and decrease end-to-end delay. We have also proved that forwarding, based on two-hop neighbor information, is a better choice than one-hop neighbor information, which also reduces the end-to-end delay and make routing more efficient. TFOR uses road topology and traffic density for efficient relaying of data in the network. It is useful for both delay-sensitive (like accident alerts, on-vehicle chat) and delay-tolerant (like infotainment) applications. The major contributions of this manuscript are as follows:
1.
We provide a brief technical survey to analyze, compare, and present limitations of existing position-based routing protocols in VANETs.
2.
We propose a novel position-based routing algorithm with new concepts of directional routing and non-directional routing to rout the packet through a shortest rich-density city road to increase packet-delivery ratio and decrease end-to-end delay and routing overhead.
3.
We propose a forwarding technique based on two-hop neighbor information and its importance in the reduction of end-to-end delay to obtain optimum performance.
4.
We implement, analyze, and compare the behavior of our approach with existing approaches (greedy-perimeter stateless routing (GPSR), geographic source routing (GSR), and enhanced greedy traffic-aware routing (E-GyTAR)) using VanetMobiSim and Glomosim simulators.
The rest of the paper is organized as follows. Section 2 describes the existing routing techniques along with their limitations. It also elaborates the motivation of the research work. Section 3 illustrates the proposed routing strategy. Analysis of the data and simulation outcomes are presented in Section 4. Finally, conclusions are presented in Section 5.
The available routing protocols in vehicular communication networks are broadly categorized into topology-based and position-based routing. The topology-based routing techniques can be reactive (on-demand), proactive (table-driven), and hybrid. On-demand routing protocols (e.g., DSR [
15], AODV [
16]) maintain only those routing paths that are currently in use. While table-driven routing protocols (e.g., OLSR [
17,
18]) maintain all the available paths in the network topology. The maintenance of routing paths affects the performance of protocols in a highly dynamic network [
4,
8,
9,
13,
14]. In position-based routing protocols, each vehicle contains a GPS receiver or other positioning abilities so that the vehicular nodes can have precise knowledge of their geological positions, movement directions, and speed. The location of destination node can be found using location services (e.g., HLS [
19], GLS [
20], and RLS [
21]). There is no need of path maintenance [
4,
8,
11,
22] as each node has to memorize its one-hop neighbors through beaconing. These characteristics of position-based approach motivate us to focus on position-based routing for dealing the routing-related issues in VANETs. The existing position-based routing protocols are categorized as directional and non-directional. As the name suggests, the directional routing protocols [
10,
23,
24] focus on the direction of vehicles while routing the packets towards destination. The non-directional routing protocols [
4,
8,
11,
14,
22,
25] do not focus on the direction of the vehicles while routing. Examples of both the categories are listed in Table
1.
Table 1
VANET routing protocols
| N | N | Highway | No | N | Distributed | Dense | Position | Greedy | Y |
Forwarding |
| N | N | City | No | N | Distributed | Sparse | Position | Greedy | Y |
Forwarding |
| N | N | City | No | N | Distributed | Sparse | Position | Restricted | Y |
Greedy |
Forwarding |
| N | Y | City | No | Y | Distributed | Sparse | Position | Greedy | Y |
Forwarding |
| N | Y | City | Yes | Y | Distributed | Both | Position | Improve | Y |
Greedy |
Forwarding |
| Y | N | Highway | No | Y | Distributed | Both | Position | Greedy | Y |
Forwarding |
| Y | N | Highway | No | Y | Distributed | Both | Position | Greedy | Y |
Forwarding |
| Y | N | Highway | No | Y | Distributed | Both | Position | Greedy | Y |
Forwarding |
| Y | N | Highway | No | Y | Distributed | Sparse | Position | Directional | Y |
Forwarding |
| Y | Y | City | Yes | Y | Distributed | Both | Position | Improve | Y |
Greedy |
Forwarding |
A non-directional protocol is the greedy-perimeter stateless routing (GPSR) [
14] that has been actually designed for highway scenarios. It works well in a highly dense network and operates in two phases, viz. greedy phase, and perimeter phase. In the greedy phase, a node sends the packet to one of its one-hop neighbors that is closest, among all the one-hop neighbors and of course the forwarding node itself, to the destination. If a node is having no one-hop neighbor that is closest to the destination than itself, the greedy phase meets the local maximum. Perimeter phase overcomes the local maximum situation. The perimeter phase consists of two modes: namely, the graph planarization and the right-hand rule. The perimeter phase causes delay in relaying the packet from the source to the destination [
4] and may result in routing loops in the network. In addition, it cannot consider obstacles and, hence, shows poor performance in the city environment [
4]. Furthermore, the graph planarization fails in city scenarios and may cause partitioning of the network due to obstacles. Last but not the least, the fact is that GPSR is not a traffic-aware routing protocol [
22].
One of the first attempts to handle the routing issues in city scenarios was the geographic source routing (GSR) [
4] which employs the position-based knowledge with the topological knowledge of the network. It runs Dijkstra's algorithm to locate the shortest route connecting the source and the destination. GSR computes a sequence of junctions based on the shortest route from source to destination using a street map that packets must have to traverse. It employs greedy forwarding to dispatch the data packets from source to target node. The greedy forwarding approach, as already stated, has the tendency to stick onto a local maximum. In this, eventuality, GSR uses a carry-and-forward approach as a recovery strategy. This protocol does not consider the number of vehicles between the junctions before forwarding packet. End-to-end connection is difficult in case of low traffic density along a preselected path [
25]. It degrades the performance of GSR. In addition, it is not a traffic-aware routing protocol.
As against GSR, the greedy-perimeter coordinator routing (GPCR) [
8] is a map-independent routing technique. It includes restricted greedy forwarding and repair strategy. Restricted greedy forwarding prefers to choose the coordinator (the node on a junction) over a non-coordinator node when deciding the next hop, even if the former is not the geographically nearest node to destination. The routing decision is, thus, made at the coordinator node that decides the street the packet should follow next. The repair strategy overcomes the local optimum problem. It consists of a perimeter mode without any graph planarization phase. It is assumed that the graph planarization is natural in a city environment. So, there is no need of computing graph planarization, because it may cause a partitioning in the network. Similar to GSR, the GPCR neglects low traffic density case. Furthermore, it is not a traffic-aware routing protocol.
The anchor-based street and traffic-aware routing (A-STAR) is a traffic-aware routing protocol [
25], as opposed to GPSR and GSR. From these latter two protocols, A-STAR has two main peculiarities. Firstly, it uses a statically or dynamically rated map for traffic awareness and uses these maps to identify paths having higher number of vehicles. Secondly, A-STAR has a new local recovery technique, considered to be better than those of GSR and GPSR [
22], for the packets got stuck in local maximum. The path selected by A-STAR on the basis of anchors may not be the shortest, due to which it may have higher end-to-end delay [
4,
24].
Designed for city environment, the greedy traffic-aware routing protocol (GyTAR) [
25] selects the junctions dynamically as against the static approach of GSR and A-STAR. GyTAR exhibits three mechanisms: (i) a completely decentralized scheme, named the infrastructure free traffic information system (IFTIS) [
21], which estimates traffic density between the junctions in the urban roads; (ii) a mechanism for dynamic junction selection, i.e., when deciding the next destination junction, the source vehicle, or an intermediate vehicle in a junction finds the position of the neighboring junctions using the map, which allocates a score to each neighboring junction assuming the curve metric distance to the destination and the traffic density; and the next junction is the one that has highest traffic density and closest to destination vehicle, (iii) which applies an improved greedy forwarding mechanism to forward the packet between the two involved junctions. The wrong junction selection mechanism compels the packet to get stuck onto local maximum, especially at low traffic density or when all the vehicles moved away from the destination on the selected street, which is its failure [
26]. The directional routing protocols [
4,
8,
11,
14,
22,
25] focus on the direction of the vehicles while routing. The technical detail of these routing protocols is presented below. The main problem with the non-directional routing protocols is the routing loop formation while routing the packet, which may cause delay [
10]. Sometimes, the packet is sent to a vehicle that is moving away from the destination, resulting in packet loss.
Directional greedy routing protocol (DGR) [
10] resolves the issues outlined above by taking into consideration the direction of the vehicle and assigns higher weight to the vehicle that is moving towards the destination. It uses a carry-and-forward approach when packets get stuck onto a local maximum and works well in a highway scenario.
Predictive directional greedy routing protocol (PDGR) is an extended version of DGR that employs a predictive compared to the latter which only takes into account the current neighbors while calculating the weighted score. PDGR determines the weighted score for the packet carrier, its present neighbors, and the possible expected neighbors in the very near future. It decides the next hop based on these weighted scores. The packet carrier obtains the information of possible future neighbor based on the two-hop neighbor information. This information is achieved through periodical sending of hello messages. The use of a prediction mechanism makes the PDGR outperform the DGR in terms of end-to-end delay and delivery ratio.
The reliable directional greedy routing protocol (RDGR) [
27] minimizes link breaks, improves route reliability, and enhances packet-delivery ratio. In DGR and PDGR, the likelihood of packet loss increases if the neighbor node moving in the direction of destination has higher speed as compared to source node or intermediate forwarder node. RDGR enhances DGR by introducing the new metric of link stability. In RDGR, a source node or an intermediate forwarding node chooses the next neighbor node having a higher speed as well as a stable link.
The position-based directional vehicular routing (PDVR) protocol deals with straight and curvy highway roads [
23]. It selects stable and efficient route for routing packet to the destination based on two rules. First, the neighbor selected for forwarding packet should move in the same direction as the source or the intermediate packet-forwarding node. Secondly, its direction must be similar to that of the destination. PDVR may not work well in a city environment because of some obstacles. It is not a traffic-aware routing protocol.
The DGR, PDGR, and RDGR protocols take into account only the highway scenarios [
27]. The enhanced greedy traffic-aware routing (E-GyTAR) protocol [
26] selects a routing path by using an enhanced junction selection mechanism. In this mechanism, the vehicular node on the current intersection selects the next intersection considering the number of vehicles that are moving in the direction of the destination. If there is completely an opposite flow of the vehicles and none is moving toward destination, then it cannot select a routing path that may reduce packet-delivery ratio and enhance end-to-end delay.
In general, most of the above routing approaches are not traffic aware. As a result, they forward the packets along the city street where moving vehicles are absent. In such situation, packets meet local optimum and are discarded. These problems can be solved by having a mechanism that provides timely information about traffic on the city streets. Some of these routing protocols (like A-STAR, E-GyTAR) are traffic aware but they are unable to use real-time vehicular traffic density properly. As a result, these are inefficient in routing. In between successive junction, most of the above routing protocols (GSR, GPSR, A-STAR, etc.) use simple greedy forwarding. However, simple greedy does not consider neighbors' speed and direction. In addition, it uses only one-hop neighbor information. Hence, it misses some suitable candidate vehicles for packet forwarding.
We propose a routing protocol that presents a solution to the aforementioned problems. The protocol accomplishes robust routes within urban environment. The protocol envisioned to work well for different types of VANET applications and ensure user connectivity. These applications include road-safety services (like traffic flow control mechanism, issuing driving alerts like traffic jams, accident warnings, road's condition, etc.) and comfort services (like gas station location, Internet access, music downloading, games, etc.).
5. Conclusions
In this paper, we have provided a brief technical analysis of the existing routing studies along with contribution and comments. We have also proposed, ‘traffic flow-oriented routing (TFOR) protocol’, a routing protocol for VANETs. It includes two major phases: first, it selects the next junction optimally, based on directional as well as the non-directional density, and secondly, it uses two-hop neighbor information for routing between the junctions. The comparative study of TFOR with other existing approaches concludes that our routing protocol performs significantly better than the other routing approaches in VANETs. Our simulation outcomes confirm that the TFOR outperforms E-GyTAR, GPSR, and GSR. TFOR performed best in terms of packet-delivery ratio, with an increase of 7.2% compared to E-GyTAR, more than 16% as compared to GPSR, and 9% as compared to GSR. In case of average end-to-end delay, TFOR performed best, with delays of 15.3% lower than GPSR, 12% lower than GSR, and 7.5% lower than E-GyTAR. The proposed improve forwarding mechanism based on two-hop neighbor considerably lowers the average routing overhead as well compared to existing solutions.
In the future, it would be interesting to examine the behavior of TFOR in the presence of one-way roads. A possible research direction could be to design a routing technique that can work in both environments (city and highway).
Competing interests
The authors declare that they have no competing interests.