The Mac Based Clustering

Published Date: 02 Nov 2017

Clustering in VANETs is nothing but organizing of vehicles themselves into groups based on some specific common characteristics. Clusters are conceptual structures in which nodes organize themselves into groups based on some specific common characteristics. The clustered network protocols and algorithms have shown its potentials of effectively reducing data congestion, and satisfying QoS requirements. Clustering can simplify such essential functions as routing, bandwidth allocation, and channel access. According to [5] clustering has many benefits like optimizing the bandwidth utilization, allocation of resources to the cluster member thus avoiding retransmission. Additionally, most protocols only use Vehicle-to-Vehicle (V2V) communication to gather and transmit information, so those data can hardly be converged and processed in centralization. As a fixed infrastructure, the road side unit (RSU) should be fully utilized to collect traffic information and use the information to perform central control.Several heuristic clustering techniques have been proposed to choose cluster heads in ad hoc network.The recent literature on cluster-based MACs and routing schemes for VANETs, also present similar low-maintenance clustering algorithms. One among the vehicles is elected as cluster head (CH) and is responsible for coordinating the members of the cluster. According to [2] clustering provides three basic benefits: 1) spatial reuse of resources, 2) emergence of a virtual backbone, 3) improved network stability and scalability from the viewpoint of a regular member.However, the highly dynamic topology of VANETs presents challenges for cluster formation and reconfiguration, thus increasing the cluster instability. Therefore, any clustering algorithm designed for VANETs must strive to maintain cluster stability for as long as possible, otherwise the frequent disconnections will result in degraded performance.However, there are some weaknesses in those protocols. Firstly, most protocols are too complicated to use in real applications, such as the cluster establishment, the cluster head (CH) selection, the cluster member’s joining and leaving and so on. There are different types of clustering in VANETs which are divided based on clustering in layers and cluster selection based on centralised or decentralised. The two types of clustering in layers are clustering in MAC and clustering in routing. In centralised clustering is cluster head selection based on RSU and decentralised clustering is based on V2V communications.

MAC Based Clustering

In [1] Hierarchical Clustering Algorithm (HCA), a fast randomized clustering and scheduling algorithm. HCA creates hierarchical clusters with a diameter of at most four hops. Additionally, the algorithm handles channel access and schedules transmissions within the cluster to ensure reliable communication. Unlike other clustering algorithms for VANETs, HCA does not rely on localization systems which contributes to its robustness. HCA is not suitable for real time communications or for safety applications . Authors of [2] uses RSU for centralised channel allocation in order to minimize channel allocation time and management overhead. In [2] RSU divides the limited bandwidth allocated to a region into prefixed overlapping spatial clusters and the channel in each cluster is divided into time slots. A time slot is allotted to a vehicle in accordance to the priority of the request and availability of the channel. Due to centralised channel allocation reliability and fairness is lowered. An RSU centric cluster based MAC protocol, CMAC [3], was presented for communication between vehicles in a VANET. The RSU allocates channels to the moving vehicles based on their clusters and enables channel reuse in nonadjacent clusters. The RSU broadcast is heard by all the vehicles in the RSU region and this solves the problem of hidden/exposed terminals and results efficient utilization of the allocated spectrum by avoiding contention. The lack of contention for channel acquisition and priority list at the RSU allows the protocol to ensure timely delivery of safety messages. However, for the protocol may not scale at high vehicular traffic and would not work in ad hoc mode in areas where there are no RSUs. Rawshdeh and Mahmud propose a hybrid media access technique for cluster-based vehicular networks in [4]. Furthermore, they propose a new approach to group vehicles in a cluster and try to minimize the total number of clusters. In [5], the authors propose a distributed mobility-based clustering algorithm focusing on cluster stability, where stability is defined by the duration of the cluster head and the cluster members. Those protocols generally use V2V communications to achieve the CH election and the cluster formation. Nevertheless, as a fixed oncoming road side infrastructure, the RSU will be widely furnished on the highway as well as other strategic locations, and should be fully utilized to carry out the traffic functions. In [6], the authors propose a distributed cluster-based multi-channel communication scheme, which integrates the clustering with contention-free and contention-based MAC protocols. However, the method in [5] needs an extra transceiver to accomplish the goal, which increases the total system expense and reduces the extensive applicability of the protocol. In paper [7] and [8], Gunter et al introduce a concept of medium access control and present a medium access scheme for VANET based on clustering of vehicles to minimize the effects of the hidden station problem. However, the protocol cannot fit for very dense traffic situations, because the cluster’s stability will decrease with the increment of the traffic density. In paper [9], the authors present a new multi-level clustering algorithm for VANET focusing on the formation of stable and long living clusters for reliable communication, called the Density Based Clustering (DBC) algorithm. In [13] ,Region-based Clustering Mechanism (RCM) to improve the performance of MAC operations for VANET. In RCM, we divide the service area into a set of region units, and limit the number of vehicles in each region unit for the contentions of radio channels. Each region unit is then associated with a non-overlapping radio channel pool. Since the number of vehicles in each region unit is limited, the contention period is reduced and the throughput is increased. However , the method in [13] provides low bandwidth utilisation in case of sparse traffic. DUCHA [10] utilizes dual-channel to separate control packets and data packets. RTS and CTS are transmitted on a separate control channel to avoid the collisions with data packets. Fast Collision Resolution (FCR) algorithm [11] redistributes the backoff counters to speed up the collision resolution. The FCR algorithm uses a smaller contention window for each station with successful packet transmission and reduces the backoff counter exponentially when a station detects a number of consecutive idle slots. MAC-SCC [12] schedules data transmissions to reduce the backoff time. The control channel is used to schedule data transmissions by using two different Network Allocation Vectors (NAVs) for the data channel and the control channel, respectively. The MAC protocols for radio channel access among vehicles are effective under light traffic load. However, when the number of vehicles in the vicinity is large, the protocols may not be able to ensure the desired service due to lack of radio resource (e.g., more contentions among vehicles for random access based protocols like CSMA/CA, and less chance to be allocated a time slot for TDMA based protocols like RRALOHA), and cause a longer contention period to obtain radio resource. In the current literature, several methods (e.g., [9], [10], [11]) have been proposed to reduce the contention period. Other important MAC protocols include (1) ADHOC MAC [14], [15] that are designed for an European project and relying on a Time Division Multiple Access (TDMA) based protocol called Reliable R-ALOHA (RR-ALOHA) for radio access control; (2) Space Division Multiple Access (SDMA) [16], [17] where the geographical area is divided into multiple space division units and one radio channel is dedicated to serve the vehicle in a space division unit. VeMAC [18] supports efficient one-hop and multi-hop broadcast services on the control channel by using implicit acknowledgments and eliminating the hidden terminal problem. The protocol reduces transmission collisions due to node mobility on the control channel by assigning disjoint sets of time slots to vehicles moving in opposite directions and to road side units. Important issues related to the MAC issue of VANET include mobility (i.e., the MAC protocol should support vehicles to leave and join inter-vehicle communications at high speed), delay bounded (i.e., the communication must be delay bounded and real-time), scalability (i.e., VANET should scale itself according to the number of vehicles present), bandwidth efficiency (i.e., the radio resource should be utilized in an efficient and fair manner), cost (i.e., for cost-efficient and reliable communications, VANET should be fully decentralized), and fairness (i.e., every vehicle should get a fair chance to get the radio channel). The challenge of successfully deploying VANET services is to ensure timely and reliable data delivery for mobile vehicles.

Routing Based Clustering

In cluster-based routing, a virtual network infrastructure must be created through the clustering of nodes in order to provide scalability. A well-known mobility-based clustering technique is MOBIC [19], which is an extension of the Lowest-ID algorithm [20]. In Lowest-ID, each node is assigned a unique ID, and the node with the lowest ID in its two-hop neighbourhood is elected to be the cluster head. MOBIC [19], calculates the variance of relative mobility of a mobile node with each of its neighbours, where a small value of variance indicates the mobile node is moving relatively less than its neighbourhood. Cluster head re-election only occurs when two cluster heads move within range of one another for a certain contention interval. When a cluster member moves out of range of its cluster head, it joins any current cluster head in its neighbourhood, or forms a new cluster. However, in the case in which few neighbour nodes move differently, the method still results in dramatic increase in the variance. In [8], authors addresses mobility by first classifying nodes into speed groups, such that nodes will only join a CH of similar velocity. The above clustering techniques are lacking in cluster stability, because they do not attempt to select a stable CH during initial cluster head election. . In [93] Cluster Head election is based on estimated travel time and speed deviation. This is not well suited in city environment because as the vehicle crosses the junction more vehicles join the cluster with more estimated travel time and thus becomes the CH and this leads to frequent re-clustering. In [94] the position based clustering approach is followed. In this algorithm each road is divided into cells and in each cell some anchor points are defined. The vehicle which is near to that anchor point is elected as CH and the cluster is dismissed when CH crosses that cell. This algorithm fails to address the cluster maintenance and stability issue which is also not suited for data dissemination. In [95] clustering is based on geographical data collection. The road is divided into segments and in each segment CH will be elected. But this protocol creates more over head for V2V communication and depends on availability of infrastructure and fails to address cluster stability and cluster maintenance. In [96] each vehicle entering into the network collects the neighbour vehicles information, assuming precedence to each vehicle and polls each vehicle individually (according to precedence) to check whether it is CH or not and then joins the cluster. Also every vehicle in the network collects 2-hop neighbor’s information along with 1-hop neighbor’s information from the CH through periodic polling. These two information collection leads to more overhead in V2V communication. A modified version of Distributed and Mobility Adaptive Clustering (DMAC) algorithm [24] is proposed for VANETs in [97]. MobHiD [99] estimates the future mobility of nodes predicting the probability that the current neighbourhood of a mobile node will remain the same through the received signal power to estimate the distance between two nodes. The drawback of the prediction method is the lack of accuracy in some cases. It does not consider the fast mobility that occurs within VANETs. In the other hand the selection of the CH could be determined by the transmission speed and the quantity of mobile nodes [100]. Nevertheless, the authors do not consider the patterns and behaviour of the vehicles. DGMA [101] is a distributed and adaptive clustering algorithm, which is implemented within a mobile environment called Reference Region Group Mobility model. It can be adapted to high speed environment; it also makes difference between micro and macro changes, taking actions only when a macro change arises. DGMA [101] is the approach that better considers the behavior of the vehicles, using the speed and direction parameters. Despite all the efficient considerations of the DGMA, it does not take into consideration the destination of the vehicles, and this parameter is very important to prolong cluster lifetime and cluster stability because vehicles with the same destination have the same route and can easily travel in groups. In paper [21], authors proposed a new cluster head election policy for direction based clustering algorithm C-DRIVE. This policy facilitates to attain better stability and thus accurate density estimation within the clusters. The cluster stability is obtained by electing fewer cluster heads in the network. This supports for a better accuracy in density estimation with fewer overheads. However, the reliability and fairness within the cluster is reduced. Adaptable Mobility-Aware Clustering Algorithm based on Destination in vehicular networks (AMACAD) [22] to enhance the clustering stability for VANETs scenarios. AMACAD takes into account the destination of the vehicles to arrange the clusters and implements an efficient message mechanism to respond in real time and avoid global re-clustering. Though MC-DRIVE[23] proposed a stable clustering algorithm, it is used for calculating the density of vehicles in a particular region around the junction. To effectuate this and to extend this algorithm for many applications this MC-DRIVE is modified so that cluster is formed all along the road. In [90], the clusters are formed based on mobility metric and the signal power detected at the receiving vehicles on the same directed pathway. The received signal strength (RSS) is used as criteria to assign weights to the vehicles and based on this weights the cluster head<as is elected. Through such method this protocol helps in forming stable clusters. However, it does not consider the losses prevalent in the wireless channel. In practical scenario effects of multipath fading are bound to affect the cluster formation method and thus the stability. These effects of multipath fading are taken into account in the density based clustering algorithm described in [91]. The cluster formation is based on the weight metric which takes into consideration the link quality and the traffic conditions. The results show an improved performance in terms of stability as compared to protocol mentioned in [90]. An improvement to the above algorithm is a new aggregate local mobility (ALM) clustering algorithm presented in [92]. In this mechanism instead of received signal strength, the position of the vehicle is used for calculating the weight associated to the vehicles. This algorithm is a beacon based aimed at prolonging the cluster lifetime in VANETSs. It displays a better performance in terms of stability as compared to the previous method. However since the vehicles are highly dynamic in nature the position of the vehicles change very fast and hence may induce an computational overhead in calculating the weight associated with the vehicles. A position based clustering technique is proposed in [93] where the cluster structure is determined by the geographic position of the vehicle and the cluster head is elected based on priorities associated with each vehicle. A hash function based on the estimated travel time is used to generate this priority for the vehicle. The stability of the system is improved by electing the vehicles having a longer trip as the cluster heads. Though this solution gives a stable cluster structure its performance is not tested in sparse and jammed traffic conditions which are very frequent in traffic scenarios. A similar approach is defined in [94] where a cluster of vehicles is formed based on their position on the road. This algorithm does not address the cluster maintenance and header election scheme, making its performance vulnerable to these factors. Another position based clustering algorithm is proposed in [95]. This is a cross layer protocol, based on hierarchical and geographical data collection and dissemination mechanism. The cluster formation in this protocol is based on the division of road segment. In this they account for the position of the vehicles at a particular segment instead of the individual positions. However this protocol incurs more overheads for V2V and V2I communication. Thus its performance is affected based on the availability of an infrastructure. A variation from the position based approach is described in [96] where a utility based cluster formation technique is used. In the utility function position and velocity, closest to a threshold, are used as the input parameters. The threshold is computed based on the previously available traffic statistics. A status message is periodically sent by all the neighbouring vehicles. After receiving this information, each vehicle chooses its clusterhead based on the results produced by the utility function. In [98] the cluster formation is based on direction of vehicle at the approaching intersection and the first vehicle moving in that particular direction will be elected as CH. In this case if vehicles are moving at variable speed then frequent the network.[23] is a modification of [98] where cluster is formed based on distance and direction of vehicle it takes after crossing the junction. UFCM [100] and VWCA [5] enforce a weight cluster mechanism with a backup manager. These algorithms operate in similar way. UFCM considers the position, direction, speed and range of the nodes to perform the algorithm. On the other hand, VWCA takes into consideration the number of neighbours based on the dynamic transmission range, the direction of vehicles, the entropy, and the distrust value parameters. VWCA works with an adaptive allocation of transmission range (AATR) technique, where hello messages and density of traffic around vehicles are used to adaptively adjust the transmission range among them. However, both proposals, as DGMA, do not consider the destination of the vehicles as a determinant parameter to arrange clusters. A suitable solution to prolong the cluster lifetime, stability, fairness , avoid congestion and overhead considering the vehicular behaviour is essential.

RSU centric Clustering

On the highway, it is widely recognized in traffic flow theory that vehicles in the traffic generally follow a "platoon pattern" [1]. Vehicles in a platoon generally move with similar velocities and are likely to sustain a stable wireless communication in clusters. The clusters are independently controlled and dynamically reconfigured as the vehicles moving. The clustered network protocols and algorithms have shown its potentials of effectively reducing data congestion, and satisfying QoS requirements. However, there are some

weaknesses in those protocols. Firstly, most protocols are too complicated to use in real applications, such as the cluster establishment, the cluster head (CH) selection, the cluster member’s joining and leaving and so on.

Additionally, most protocols only use Vehicle-to-Vehicle (V2V) communication to gather and transmit information, so those data can hardly be converged and processed in centralization. As a fixed infrastructure, the road side unit (RSU) should be fully utilized to collect traffic information and use the information to perform central control. Finally, some protocols need additional devices to accomplish the goal, which will increase the vehicle’s cost and reduce the feasibility of protocols. In order to overcome the abovementioned weaknesses and achieve the goal of stable. RSU centric clustering is based on centralised clustering. Moreover, RSU collects information from all the vehicles in the road and divides the vehicles into different cluster groups. In addition, it selects the CHs for different cluster groups. Furthermore, RSU will act like a centralised controller for entire network. In general, the amount of information to be transmitted is relatively small (the movement information of each vehicle), but the transmission reliability is fundamental. Active Safety applications for VANETs need to establish reliable communications with minimal transmission collisions. Vehicles move on predetermined strait jacket roads at a high speeds and enter/exit RSU region in short intervals of time. At a time, the numbers of vehicles in an RSU region can vary rapidly from a few vehicles to a large number of very short interval of time. A protocol must be distributed or should require partial RSU support with an efficient handoff from one RSU to another to satisfy these characteristics. The motion of the vehicles is confined to the roads and directional antenna would be suitable for communication via RSUs. The nodes broadcast radio frequencies with transmission channels, each one considered as a common medium over which two neighboring terminals cannot transmit simultaneously because a transmission collision occurs. So, in order to efficiently share the medium, MAC protocol is needed and is beset by contention delay. However, a protocol must ensure that critical messages are delivered within a prescribed time frame. The protocol must not suffer without the hidden/exposed terminal or deafness problem to ensure reliable message delivery. Although the RSU is an extra infrastructure, it will be furnished on the highways extensively and applied in VANET in the near future. Therefore, compared with great and lifelong benefit, the RSU’s expense is of trifling importance at all. The efficient MAC can provide a more stable communication than a solution using fixed infrastructure. In order to overcome the abovementioned weaknesses and achieve the goal of stable information transmission. We propose a protocol that takes advantages of fixed infrastructure and optimise the problem.

V2V based Clustering

V2V based clustering is a decentralised clustering where clusters are formed based on communication between vehicles. Moreover, the CH election will be based on V2V communication. Cluster head selection carried out through inter-cluster communications has some limitations, e.g., large communication and computation cost, quite complex algorithms, requirement of additional devices and so on. Another potential problem is that the connection between two adjacent cluster heads may be lost due to vehicle’s high mobility, which drastically degrades the communication quality.

Advatages of V2V compared to V2R

While there has been a large body in the literature studying both V2V [2]–[6] and V2R [7], [8] networks, there are several advantages of using V2V-based VANETs as compared with the V2R-based VANETs. First, the V2V-based VANET is more flexible and independent of the roadside conditions, which is particularly attractive for most developing countries or remote rural areas where roadside infrastructures are not necessarily available/furnished. Second, V2V-based VANET is less expensive than the V2R-based one since it does not need expensive roadside infrastructures. Third, V2V-based VANET can avoid

the fast fading, short connectivity time, high frequent handoffs, etc., that are caused by the high relative-speed difference between the fast-moving vehicles and the stationary base stations. Finally, the V2V-based VANET much better fits vehicle-related applications, which only needs to exchange data/information among neighboring vehicles within their nearby areas. Motivated by the aforementioned observations, in this paper, we will focus on V2V-based VANETs.

.

Advantages of clustering

In this vision, most applications targeting VANETs rely heavily on broadcast transmission both to discover nearby vehicles, and to disseminate traffic-related information to all reachable vehicles within a certain geographical area, rather than only to routing-selected hosts, like in MANETs. On the other side, broadcasting packets may lead to frequent contention and collisions in transmission among neighboring vehicles. This problem is sometimes referred to as the broadcast storm problem [2], [3]. This affects inter-vehicle communications, since redundant rebroadcasts, contention and collisions can be largely increased. When a vehicle rebroadcasts a message, it is highly likely that the neighboring vehicles have already received it, and this results in a large number of redundant messages. Although multiple solutions exist to alleviate the broadcast storm effect in the usual MANET environment, only a few solutions have been proposed addressing the VANET context [4], [5]. In this paper, we propose an effective broadcast method for safety alerts in VANETs, called as Selective Reliable Broadcast (SRB) protocol. The main aim is to reduce the broadcast storm problem, since SRB selects only one vehicle within a cluster –namely, a cluster-head– in order to efficiently rebroadcast emergency and control messages. SRB technique is able to detect the well-known car platoons, which cause traffic congestions and delays, in a fast way and with low overhead, in order to recommend alternative paths to other vehicles. One of the many challenges in VANETs is the dynamic and dense network topology. The dynamic topology causes significant re-routing difficulties and thus congestion, while the dense network leads to the hidden terminal problem. A clustered structure can make the network appear smaller and more stable in the view of each node. By clustering the vehicles into groups of similar mobility, the relative mobility between communicating neighbouring nodes will be reduced, leading to intra-cluster stability; in addition, the hidden terminal problem can be diminished by clustering [9].

The motivation to rely on this 802.11p capability is twofold in the view of supporting multihop data delivery: (i) most of the packets transmitted in the VANET could be of interest to many vehicles; (ii) most of the applications for traffic monitoring and efficiency can significantly benefit from data aggregation schemes [11], which allow vehicles to aggregate the reports received from neighbors with their own collected view of the road, progressively, before forwarding them.

Things to consider for clustering

Inside cluster

Inter cluster communication or to RSU

The main challenge in clustering is the overhead introduced to elect the cluster head and to maintain a stable cluster size. To optimize the communication range and hence the cluster size is very difficult especially in a highly dynamic environment such as VANETs. In [12], the authors showed how vehicles dynamics affect the network density and hence the reliability and throughput of VANETs’ safety applications. While in [2] and [13], the authors derived the relationship between the communication range and the network density, message sending rate, message size, data rate and channel conditions. Since each vehicle in the network has its own view of the network density and channel conditions, finding the optimal network parameters is difficult. Therefore, our main goal is to find the cluster size and hence the communication range that maintains a high network stability and reliability, increases the life time of a path, and at the same time decreases the time delay for an emergency message to reach its intended distance. Most implementations of these existing algorithms focus mainly on how CHs are elected. The communication overhead for the formation and maintenance of clusters have not been taken fully into account. There has been few contributions that assess analytically the communication overhead incurred in

hierarchical routing. In particular, Fan et al. [13] show that the overhead incurred by DISCA [13] is bound by a constant per node per time step, avoiding expensive re-clustering chain reactions; hence, this overhead increases with the number of nodes. Since a CH acts as a coordinator in a cluster, if it is absent

for any reason, the clustering architecture has to be reconfigured; this will significantly increase the message overhead. However, in our research, we believe that a more efficient way to form a stable clustering architecture, with

reduced overhead, is that a mobile node should be associated to a cluster and not to a CH. Indeed, replacing CH is considered only as an incremental update and does not require a whole reconfiguration of the cluster structure; this will definitively increase the lifetime of the clustering architecture. The resulting clusters are stable and exhibit long average CM duration, long average CH duration, and low average rate of CH changes. In this paper, we propose a new stability-based clustering algorithm protocol (SBCA) aiming to reduce the communication overhead that is caused by the cluster formation and maintenance, as well as to increase the lifetime of the cluster. SBCA makes use of (1) The cluster configuration protocol that is based on the velocities’ differences among neighboring vehicles to select a primary CH (PCH) for each cluster; and (2) The election of a secondary CH (SCH) for each cluster; similarly to the clustering scheme proposed in [5], SCH works as a backup for PCH.

SCALABILITY AND FLEXIBILITY

A number of vehicles may depend on an area. For example, in a rural area, where the number of vehicles is quite low, it becomes very difficult to maintain network connectivity without roadside units (RSUs), infrastructure units implemented to support communication. Deployment of RSUs requires large investments. Some researchers make use of less stringent power constraints by expanding communication range with higher transmission power to make each vehicle reachable even without RSU support. In contrast, a city area is normally very crowded. Therefore, the number of vehicles is normally higher than in a rural area. When the number of vehicles is high, routing protocols need to minimize overhead or control packets as much as possible, since a lot of vehicles need to communicate with others. In fact, a communication channel should be dedicated for safety communication rather than control overhead. For example, a dynamic topology makes communication routes unstable and routing maintenance difficult, resulting in high latency, low reliability, non-scalability, inflexibility, low fault tolerance, and security issues.

IV. OPEN CHALLENGES

Generally two different approaches for vehicle clustering can be observed. The first is location service dependant and uses information such as speed, location and movement direction for clustering. The other one uses different measurable parameters such as radio propagation, relative mobility, vehicle density, connectivity etc., but not many of them together. Both approaches are based on mathematically measurable parameters and ignore sociological aspects [23] such as why the driver is on the go and in what context the drive is taking place so there is plenty of room for new proposals. Research on clustering solutions with combined metrics should also be done. Location services might not provide the needed accuracy everywhere or will not be available at all [24], [25] so more work is needed on location independent clustering solutions. Providing highly accurate digital maps that are needed by some solutions presents a challenging task and could slow down the deployment so pros and cons of map based solutions should be researched. Many of the presented algorithms use metrics derived from the same input parameters where among them location and radio signal strength are the most popular. Focus should be put on evaluation of those common metrics to highlight the most useful ones, merge similar ones etc. This would allow researchers to concentrate their work on extending and optimizing the most prospective ones. Self learning or self adapting metrics should also be studied. Presented clustering algorithms are optimized for different goals e.g. cluster stability, fast cluster formation, overhead minimization etc. where the most popular among them is definitely the cluster stability. More research effort should be put in defining and ranking the goals that clusters and clustering algorithms should try to achieve. We believe that one of the most important goals is providing good quality of experience (QoE) for safety and emergency services which is closely related to excellent connectivity, communication reliability and limited maximum latency. An analysis like the [26] focused on clustering would be highly welcome. More work should also be put in multi-hop and multi-homing capable clustering solutions. For performance evaluations of clustering algorithms common metrics are used such as cluster head stability, cluster head changes, average cluster stability etc. These

terms are quite generic so their scientific definition and explanation with VANET specifics is needed to provide consistency between different scientific researches. Their correlations and effects between them should also be analyzed and presented. Fair comparison of different clustering solutions is a hard task due to non-existent standard testing procedures and scenarios so more work and standardization is needed in this area. Different network simulators should be evaluated and presented with all the relevant parameters including MAC, radio signal range, packet size, bit rate etc. Test scenarios with different vehicle movement patterns should be provided.

Losing contact with the cluster head temporally:

Two types of transmission collision on time slots can happen on channel c0 [13]: access collision and merging collision. An access collision happens when two or more nodes within two hops of each other attempt to acquire the same available time slot. On the other hand, a merging collision happens when two or more nodes acquiring the same time slot become members of the same two-hop set5 (THS) due to node activation or node mobility. The difference between the two types of collision

is that access collisions occur among nodes which are trying to acquire a time slot, while merging collisions occur among nodes which have successfully acquired a time slot. In VANETs, although merging collisions can happen among vehicles moving in the same direction due to acceleration or deceleration, it is more likely to occur among vehicles moving in opposite directions (approaching each other) or between a vehicle and a stationary RSU since they approach each other with a

much higher relative velocity as compared to vehicles moving in the same direction.

The algorithm finds clusters that minimize both relative mobility and distance from each cluster head to its cluster members.The cluster size is controlled by a predefined maximum distance between a cluster head and its members. Clusters are independently controlled and dynamically reconfigured as nodes move.Our clustering scheme will form clusters with both minimum distance and minimum relative velocity between each cluster head and its members.The objective of the optimization model is to maximize the utility of the vehicular nodes in a cluster and to minimize the cost of reserving exclusive-use channel while the QoS requirements of data transmission (for vehicle-to-vehicle and vehicle-to-roadside communications) are met, and also the constraint on probability of collision with licensed users is satisfied.an RSU centric cluster based MAC (CMAC) protocol has been proposed. In this protocol, the channel allocation and management is transferred to the RSU to obviate channel contention for reliable delivery of messages. The RSU region is divided into prefixed clusters with frequency reuse in non adjacent clusters. The frequency reuse increases the available bandwidth and reduces the waiting time of a node for channel allocation.

One, vehicles move on predetermined strait jacket roads at a high speeds and enter/exit RSU region in short intervals of time. At a time, the numbers of vehicles in an RSU region can vary rapidly from a few vehicles to a large number of vehicles. The stay period of vehicles in an RSU region is

very short. A protocol must be distributed or should require minimal RSU support with an efficient handoff from one RSU to another to satisfy these characteristics.

An RSU centric cluster based MAC protocol, CMAC, was presented for communication between vehicles in a

VANET. The RSU allocates channels to the moving vehicles based on their clusters and enables channel reuse in nonadjacent clusters. The RSU broadcast is heard by all the vehicles in the RSU region and this solves the problem of hidden/exposed terminals and results efficient utilization of the allocated spectrum by avoiding contention. The lack of contention for channel acquisition and priority list at the RSU allows the protocol to ensure timely delivery of safety messages. However, for the protocol may not scale at high vehicular traffic and would not work in ad hoc mode in areas where there are no RSUs.

This modification isdesigned to reduce the clusterhead changeovers that occur due overtaking of the vehicles and thereby reduce thenumber of clusterheads within the cluster. Further this algorithm acts as a platform to estimate the density ofvehicles approaching the intersection. With this density information many efficiency and resource discoveryapplications can be realized in VANETs.

In this, the clusters are formed based on mobility metric and the signal power detected at the receiving vehicles on the same directed pathway. The received signal strength (RSS) is used as criteria to assign weights to the vehicles and based on this weights the clusterhead is elected. Through such method this protocol helps in forming stable clusters. However, it does not consider the losses prevalent in the wireless channel.

In practical scenario effects of multipath fading are bound to affect the cluster formation method and thus the stability.

For example, the cluster formation interval is fixed, which implies a synchronous formation of clusters. This does not allow for effective cluster reorganization. In [3] the authors describe a theoretical analysis of directional based clustering algorithm. The clusters in this algorithm are based on the following mobility metrics (a) moving direction (b) leadership duration (c) projected distance variation of all the neighboring vehicles over time. The evaluation of the protocol showed the stability of the cluster with fewer overheads on the network. From the presented solutions it is apparent that, stability is one of the major factors to be considered in designing the clustering algorithm for vehicular environments. Moreover the application requirement influences the clustering

technique.

A generic weight is allocated to each vehicle based on the position and other set of vehicle parameters like connectivity, mobility, etc. The vehicle with the highest weight is elected as the Cluster Head amongst the neighbors. The main idea here is to avoid re-clustering when vehicles move in different directions. Hello message is periodically exchanged between the vehicles which are used for link creations and to predict the freshness as well as the

time for which the vehicles might be in contact with the Cluster Head. The results of the analysis show that the

stability of the clusters is increased. However it is observed the proposed modifications to DMAC algorithm increases the overhead on the network.

Some cluster-based approaches have been applied in VANETs, because the clusters reduce the overhead and delay, solving the scalability problem, providing an efficient resource consumption and load balance in large scale networks. However, in a high mobility environment the clusters usually are unstable and the clustering/de-clustering is constantly executed.

Therefore, our clustering algorithm takes into account the destination of vehicles, including the current location, speed, relative destination and final destination of vehicles as parameter to arrange the clusters. In this manner, the clustering algorithm resembles a natural model of location references, which helps to manage the mobility by improving the lifetime of the cluster and decreasing the number of cluster head changes and the number of cluster re-affiliations. The

information is disseminated by groups enhancing the communication delay, reliability, low data delivery and

congestion issues, making the vehicular networks accurate and efficient.

The VANET has to overcome the challenges of communication delay, low delivery rate, reliability, scalability

and congestion. The clustering dissemination is a prominent solution to overcome these challenges. However the current clustering algorithms do not exploit the vehicular behavior and group mobility patterns taking into account the final destination of the vehicles to provide cluster stability despite of the mobility.

In most cluster-based protocols, however, the cluster head selection carried out through inter-cluster communications has some limitations, e.g., large communication and computation cost, quite complex algorithms, requirement of additional devices and so on. Another potential problem is that the connection between two adjacent cluster heads may be lost due to vehicle’s high mobility, which drastically degrades the communication quality.

However due to high mobility a stable cluster, within a vehicular framework, is difficult to implement.

Within the constrained scenario of VANETs clustering manifests itself as a potential concept for realizing many efficiency applications. Clustering allows for efficient resource consumption and load balancing which will improve the performance of safety and efficiency application. In context of using a cluster structure, stability of clusters is a critical issue. In dynamic environment like VANETs, cluster reconfigurations and clusterhead changes are unavoidable, thus affecting the stability. Hence one of the important criterions for any clustering method in VANETs is to allow for formation of stable clusters.

Inter-vehicle communications are expected to significantly improve transportation safety and mobility on the road. Several applications of inter-vehicle communications have been identified, from safety and warning applications, up to traffic control and driver assistance applications [1]. Many of these applications require multicast routing protocol to a group of vehicles satisfying a geographical criterion.

cluster can better coordinate its transmission events with the help of a special mobile node, such as a clusterhead, residing in it. This can save much resources used for retransmission resulting from reduced transmission collision. clustering is important for a network to achieve scalability in the presence of a large number of mobile nodes and high mobility. However, a cluster-based MANET has its side effects and drawbacks because constructing and maintaining a cluster structure usually requires additional cost compared with a flat-based MANET. The cost of clustering is a key issue to validate the effectiveness and scalability enhancement of a cluster structure.

In our scheme, the elected clusterhead vehicle functions as the coordinator to collect/deliver realtime safety messages within its own cluster and forward the consolidated safety messages to the neighboring cluster heads. In addition, the cluster-head vehicle controls channel assignments for cluster-member vehicles transmitting/receiving nonreal-time

traffics, whichmakes the wireless channels more efficiently utilized for vehicle-to-vehicle (V2V) nonreal-time data transmissions. Our scheme uses the contention-free MAC within a cluster and the contention-based IEEE 802.11 MAC among cluster-head vehicles such that the real-time delivery of safety messages can be guaranteed.

While there has been a large body in the literature studying both V2V [2]–[6] and V2R [7], [8] networks, there are several advantages of using V2V-based VANETs as compared with the V2R-based VANETs. First, the V2V-based VANET is more flexible and independent of the roadside conditions, which is particularly attractive for most developing countries or remote rural areas where roadside infrastructures are not necessarily available/furnished. Second, V2V-based VANET is less expensive than the V2R-based one since it does not need expensive roadside infrastructures. Third, V2V-based VANET can avoid the fast fading, short connectivity time, high frequent handoffs, etc., that are caused by the high relative-speed difference between the fast-moving vehicles and the stationary base stations. Finally, the V2V-based VANET much better fits vehicle-related

applications, which only needs to exchange data/information among neighboring vehicles within their nearby areas. Motivated by the aforementioned observations, in this paper, we will focus on V2V-based VANETs.

On the other hand, clustering [9], [10] is an efficient technique to reduce data congestion and support

QoS over wireless networks. To provision QoS over our V2V networks and reduce data congestion, under the DSRC multichannel architecture, we propose a distributed clusterbased multichannel communications scheme that integrates the clustering with contention-free/contention-based medium access control (MAC) protocols.

Driver behavior, constraints on mobility, and high speeds contribute unique characteristics in VANET. In particular, [3] quantifies these differences as rapid topology changes, frequent fragmentation, small

effective network diameters, and limited redundancy. Due to the relative speeds of vehicles, the topology of VANET keeps rapidly changing even though the movement of vehicles is somewhat predictable, i.e., they must stay on the roadway and have the same moving direction. A message path can typically survive 1 min while maintaining a transmission range of 500 m [9], [18]. Consequently, a VANET may suffer frequent partition,

and hence, costly overheads for exchanging new topology information and reconfiguring each node [19]. Aoki and Fuji [3] also show that a typical effective network diameter is no larger than 9 hops.

This cluster-based cognitive vehicular network can be used for both vehicle-to-vehicle (V2V) communications

(e.g., communication among multiple clusters) and vehicle-to-roadside (V2R) communications (e.g., communication

with an RSU). In such a network, decisions have to be made on opportunistic access of shared-use channels and

reservation of exclusive-use channel by the vehicular nodes.

Challenges of clustering in vanets

One of the many challenges in VANETs is the dynamic and dense network topology. The dynamic topology causes significant re-routing difficulties and thus congestion, while the dense network leads to the hidden terminal problem. The main challenge in clustering is the overhead introduced to elect the cluster head and to maintain a stable cluster size. To optimize the communication range and hence the cluster size is very difficult especially in a highly dynamic environment such as VANETs. The cluster application assures the scalability of networks, where high mobility of the shifting nodes within the networks causes lots of challenges to face. The cluster based routing in VANET is beneficial; it necessitates better routing and scalability among hundreds or thousands of vehicles [14]. The characteristics of VANETs support the argument of clustering application, as the vehicles with relatively high mobility speed, can pose challenges for flat networks stability.

Many researchers such as [6]-[11], have proposed clusterbased multi-channel medium access control (MAC) protocols to improve the performance and reliability of VANETs. In these protocols, clustering is used to limit channel contention and provide fair channel access within the cluster. On the other hand, multi-channel is used to increase the network capacity by the spatial reuse of the network resources and reduce the effect of the hidden terminal problem. The main challenge in clustering is the overhead introduced to elect the cluster head and to maintain a stable cluster size. To optimize the communication range and hence the cluster size is very difficult especially in a highly dynamic environment such as VANETs. In [12], the authors showed how vehi- cles’ dynamics affect the network density and hence the reliability and throughput of VANETs’ safety applications. While in [2] and [13], the authors derived the relationship between the communication range and the network density, message sending rate, message size, data rate and channel conditions. Since each vehicle in the network has its own view of the network density and channel conditions, finding the optimal network parameters is difficult. Therefore, our main goal is to find the cluster size and hence the communication range that maintains a high network stability and reliability, increases the life time of a path, and at the same time decreases the time delay for an emergency message to reach its intended distance.

One of the many challenges in VANETs is the dynamic and dense network topology. The dynamic topology causes significant re-routing difficulties and thus congestion, while the dense network leads to the hidden terminal problem. A clustered structure can make the network appear smaller and more stable in the view of each node. By clustering the vehicles into groups of similar mobility, the relative mobility between communicating neighbouring nodes will be reduced, leading to intra-cluster stability; in addition, the hidden terminal problem can be diminished by clustering [9].

Node mobility should play an integral part in cluster creation in order to achieve stability. In [2], mobility is addressed during clusters’ collisions; when two CHs come within range, the winning CH will be the one with both lower relative mobility and closer proximity to its members. The algorithm used for cluster formation is based on CBLR [6]. Alternatively, Kayis et al. [15] address mobility by classifying nodes into speed groups, such that nodes will only join a CH of similar velocity.

The cluster application assures the scalability of networks, where high mobility of the shifting nodes within the networks causes lots of challenges to face. The cluster based routing in VANET is beneficial; it necessitates better routing and scalability among hundreds or thousands of vehicles [14]. The characteristics of VANETs support the argument of clustering application, as the vehicles with relatively high mobility speed, can pose challenges for flat networks stability. The cluster heads form a fundamental responsibility of the network and act as local manager to their respective members. [13] Clustering provides availability of vast structure for data distribution, where information is propagated through the cluster heads.

CLUSTER SIZE

The major objective of clustering is to achieve relatively stable cluster structure, because frequent cluster reconfiguration generates tremendous communication load, which significantly reduces available bandwidth for message dissemination. Effective cluster size is both related to radio transmission range and vehicle traffic density. Therefore, cluster size may limit radio efficiency and throughput. When the distance

between two cluster heads is detected to be less than or equal to a predetermined threshold, D(D ≤ L), the cluster with less members is dismissed. Each of the nodes in the dismissed cluster finds a new cluster to join.

CLUSTERING RELIABILITY

CLUSTER STABILITY

Clustering Performance Metrics

To evaluate the cluster stability and overall performance of

our algorithm, we use the following metrics:

1) Average Cluster Head Duration: Long cluster head

duration is important for MAC schemes where the cluster head

is the central controller and scheduler.

2) Average Cluster Member Duration: This metric judges

the overall stability of the initial clustering.

3) Average Rate of Cluster Head Change: This metric is

useful since it takes into account both cluster head duration

and the number of clusters formed.

4) Average Number of Clusters: To effectively decrease

network contention, fewer clusters is desirable.

CLUSTER DELAY

Cluster head selection

To increase the stability of cluster heads in MDS, vehicles having a longer trip are more qualified for being elected as cluster heads. A vehicle, which would travel longer time, is assigned higher priority; hence, at the very beginning of starting its travel, the expected travel time of a vehicle is calculated and announced using its desired driving speed and the geographic information system once its driver sets the destination.

2) To avoid elected cluster heads losing connectivity with their neighbors very soon, the eligibility of a vehicle

should decrease quickly when its velocity has big difference from the average speed. Thus, a vehicle with large

speed deviation is assigned lower priority. 3) Once a cluster head is elected and a cluster is formed,

recalculating priorities is necessary only if the cluster is dismissed, and therefore, each node should compute its

new priorities following the previous rules.

Cluster Reconfiguration

If the distance between two cluster head nodes is detected less than the dismiss threshold D, the cluster with fewer members is dismissed to reduce communication overheads while its members join other clusters. Each node of this cluster launches a new registration stage to join other clusters. The threshold determines the rate of cluster reconfiguration, and also, depends on the radio transmission range.

Cluster Stability

Dismiss Threshold

one can expect that a larger dismiss threshold leads to a higher rate of cluster head changes and higher probability of cluster reconfiguration. On the contrary, if L increases, the probability decreases. Since the dismiss threshold is related to the transmission range, then, the probability of cluster changes is also related to the transmission range. Larger transmission provides longer distance for cluster heads to detect each other, and therefore, more frequent cluster reconfigurations occur. In [4], mobility is addressed during cluster collision; when two cluster heads come within range, the winning CH will be the one with both lower relative mobility and closer proximity to its members.

Cluster overhead

Deusto protocol

When detecting a problem on the road, a single vehicle could send a broadcast alert message to a group

of potential receivers in the Risk Zone (RZ). Since the risk zone may be larger than the transmitting range of

wireless devices, the message should be relayed by the intermediate vehicles to extend the horizon of the

message.

Every node (i.e. vehicle or RSU) is equipped with a global positioning system (GPS)

receiver and can accurately determine its position and moving direction using GPS. The current position of each node is included in the header of each packet transmitted on channel c0, and synchronization among nodes is performed using the 1PPS signal provided by any GPS receiver. The rising edge of this 1PPS is aligned with the start of every GPS second with accuracy within

100ns even for inexpensive GPS receivers. Consequently, this accurate 1PPS signal can be used as a common time reference among all the nodes.

In our scheme, the elected clusterhead vehicle functions as the coordinator to collect/deliver realtime safety messages within its own cluster and forward the consolidated safety messages to the neighboring cluster heads. In addition, the cluster-head vehicle controls channel assignments for cluster-member vehicles transmitting/receiving nonreal-time

traffics, whichmakes the wireless channels more efficiently utilized for vehicle-to-vehicle (V2V) nonreal-time data transmissions. Our scheme uses the contention-free MAC within a cluster and the contention-based IEEE 802.11 MAC among cluster-head vehicles such that the real-time delivery of safety messages can be guaranteed.

Most (if not all) of the high priority safety applications proposed for VANETs are based on one-hop broadcast of information. For instance, for V2V communication based applications such as the pre-crash sensing, blind spot warning, emergency electronic brake light, and cooperative forward collision avoidance, each vehicle periodically broadcasts information about its position, speed, heading, acceleration, turn signal status, etc, to all the vehicles within its one-hop neighbourhood [2].

Similarly, for V2R communication-based applications, such as the curve speed warning and traffic signal violation warning, an RSU periodically broadcasts to all the approaching vehicles information related to the traffic signal status and timing, road surface type, weather conditions, etc [2]. As the precision of the safety applications is directly related to the safety of people on road, the need of a medium access control (MAC) protocol which provides an an efficient broadcast1 service is crucial for

VANETs.

A finite-state machine (FSM), as shown in Fig. 3, is employed to precisely describe the principle and operating process of our proposed scheme. Each vehicle operates under one and only one of the following four states at any given time: 1) cluster head; 2) quasi-cluster head; 3) cluster member; and 4) quasicluster member.

The functions of the four states are described as follows:

First, in the state of CH, the vehicle’s Transceiver I works on the ICC channel to forward consolidated safety messages to the neighboring clusters, and the Transceiver II is tuned to the CRC channel to collect/broadcast safety messages from/to cluster members. Second, the quasi-cluster-head state represents that this vehicle is neither a cluster head nor a cluster member. In the quasi-cluster-head state, while Transceiver II is turned off,

the Transceiver I of the vehicle works on the ICC channel, so that it can also receive and send the safety messages. In fact, the quasi-cluster-head vehicles function as real cluster heads, except for the ability in forming clusters. Third, when entering the cluster-member state, the vehicles tune their Transceiver II’s to the CRC channel where the cluster-member vehicles receive the consolidated safety messages and send their own safety messages as well as data channel reservation requests. Each cluster-head vehicle uses the centralized multichannel control algorithm (e.g., the Random Scheduling algorithm [12]) to assign appropriate CRD

or ICD channels to cluster members after receiving the data channel reservation requests. According to the decision on the assignment of CRD/ICD channels by the cluster-head vehicle, the cluster-member vehicles set their Transceiver I’s to either the corresponding CRD channels for communications with other cluster members within the cluster or the ICD channel for nonreal-time traffic data packet exchange among clusters. Finally, the function of the quasi-cluster-member state is to guarantee that the cluster-member vehicles can receive and

transmit the safety messages by switching Transceiver I’s to the ICC channel, even if it temporally loses contact with the cluster heads. In the quasi-cluster-member state, the vehicle’s Transceiver II still monitors the previous CRC channel and tries to resume the communications with the previous cluster-head vehicle.

Research states that the driver reaction time to traffic warning signals, such as brake lights, can be on the order of 700 ms and longer [13]. Consequently, the update interval of safety messages should be less than 500 ms. Otherwise, the safety system is useless in helping the driver deal with emergency situations. Hence, a safety message of 200 bytes is typically updated every 500 ms and should be delivered before the generation of the next new message. Therefore, we set Hsafety = 200 bytes, Tupdate = 200 ms, Tsafety = 200 ms, and T = 80 ms in this paper.

In particular, [3] quantifies these differences as rapid topology changes, frequent fragmentation, small

effective network diameters, and limited redundancy. Due to the relative speeds of vehicles, the topology of VANET keeps rapidly changing even though the movement of vehicles is somewhat predictable, i.e., they must stay on the roadway and have the same moving direction.

Cluster Size and Minimal Dominating Set

The major objective of clustering is to achieve relatively stable cluster structure, because frequent cluster reconfiguration generates tremendous communication load, which significantly reduces available bandwidth for message dissemination. Effective cluster size is both related to radio transmission range and

vehicle traffic density.