Survey of medium access control schemes for inter-vehicle communications

ARTICLE IN PRESS

JID: CAEE

[m3Gsc;March 3, 2017;13:54]

Computers and Electrical Engineering 0 0 0 (2017) 1–23

Contents lists available at ScienceDirect

Computers and Electrical Engineering journal homepage: www.elsevier.com/locate/compeleceng

Survey of medium access control schemes for inter-vehicle communicationsR Nasser Torabi, Behrouz Shahgholi Ghahfarokhi∗ Department of Information Technology Engineering, Faculty of Computer Engineering, University of Isfahan, Isfahan, Iran

a r t i c l e

i n f o

Article history: Received 7 June 2016 Revised 21 February 2017 Accepted 22 February 2017 Available online xxx Keywords: Vehicular Ad hoc Networks (VANETs) Medium Access Control (MAC) schemes Quality of Service (QoS) Vehicle-to-Vehicle (V2V) communications Inter-Vehicle Communications (IVC)

a b s t r a c t Currently, Vehicular Ad-hoc Networks (VANETs) are attracting a lot of attention due to their favorable applications. VANETs are the key to providing safety and efficiency on the roads. The vehicles can communicate with other vehicles to inform the ongoing status of the traffic flow or critical situations like accidents. However, this would entail a reliable and efficient Medium Access Control (MAC) protocol. Due to the high speed of the nodes, the frequent changes in network topology, and particularly the lack of an infrastructure, the design of the MAC for vehicular communications turns into a more challenging task. A lot of research works has been conducted to overwhelm the vehicular MAC problems regarding Quality of Service (QoS) requirements of both safety and non-safety applications covering both Vehicle-to-Infrastructure (V2I) and Vehicle-to-Vehicle (V2V) communications. Recently, a significant number of MAC schemes has been proposed for V2V communications. In this paper, for future studies to be more effective, the outstanding proposed V2V MAC schemes are intended to come under review. Moreover, V2V MAC design approaches are discussed and a qualitative comparison is provided. A novel classification of V2V MAC schemes is then presented, and the characteristics of these schemes along with their strengths and weaknesses are studied. Finally, a comparative summary is given and some open challenges regarding the design of V2V MAC schemes are discussed. © 2017 Elsevier Ltd. All rights reserved.

1. Introduction Vehicular Ad hoc Networks (VANETs) are primarily designed to help improve safety on roads and management of traffic flow [1]. They can also be used to provide infotainment services such as Internet access, video streaming, social networking, etc. VANETs are a special type of Mobile Ad hoc Networks (MANETs) with their own specific characteristics such as frequently changing network topology, unstable links, variable density of vehicles, and the large scale network. Mobile nodes in VANETs or vehicles are characterized by their distinctive and unique features such as variety in type, high mobility, high speed, and road-constrained movements. VANETs, generally, consist of vehicles equipped with On-Board Units (OBUs) and stationary access points referred to as Road-Side Units (RSUs). The presence or absence of RSUs in vehicular communications builds up two distinguished types of communications namely Vehicle-to-Infrastructure (V2I) and Vehicle-to-Vehicle (V2V) communications, respectively. The latter is also known as Inter-Vehicle Communications (IVC). Considering the natural characteristics of VANETs and concerning the differences between the wide range of safety and non-safety applications in term of QoS requirements, it is a vital characteristic for a MAC scheme to provide appropriate R ∗

Reviews processed and recommended for publication to the Editor-in-Chief by Associate Editor Dr. M. H. Rehmani. Corresponding author. E-mail addresses: [email protected] (N. Torabi), [email protected] (B.S. Ghahfarokhi).

http://dx.doi.org/10.1016/j.compeleceng.2017.02.022 0045-7906/© 2017 Elsevier Ltd. All rights reserved.

Please cite this article as: N. Torabi, B.S. Ghahfarokhi, Survey of medium access control schemes for inter-vehicle communications, Computers and Electrical Engineering (2017), http://dx.doi.org/10.1016/j.compeleceng.2017.02.022

ARTICLE IN PRESS

JID: CAEE 2

[m3Gsc;March 3, 2017;13:54]

N. Torabi, B.S. Ghahfarokhi / Computers and Electrical Engineering 000 (2017) 1–23

Fig. 1. The IEEE WAVE protocol stack and its comparison to OSI reference model.

levels of QoS for each type of applications. Therefore, the design of a fast, reliable, scalable, efficient and fair MAC protocol is a challenging issue. The MAC protocol should be fast enough to offer predictable access delay and should be sufficiently reliable against the instability of links. It needs to be flexible enough to share bandwidth between growing numbers of vehicles in a way that leads to fair channel access. Also, the MAC protocol should cope with channel access issues like the hidden node problem which strongly affects its efficiency. On the other hand, given the advantages of employing V2V-based VANETs against V2I-based, discussed in [2], there is a great tendency among researchers for studying V2V MAC schemes. In the literature, there exist a plethora of MAC protocols that have been proposed to cover V2V communications [3–19]. Correspondingly, several papers have surveyed V2V MAC solutions. In [1], authors go through multi-channel MAC protocols. They distinguish between channel coordination and channel allocation methods and discuss some issues and challenges. Authors in [20] survey MAC solutions that are not compatible with the IEEE 802.11p or the IEEE 1609.4 standards. The authors of [21] classify the MAC protocols for VANETs into three different categories: channel partitioning, random access, and taking turns. Although their categorization is straightforward, it does not cover all VANET-specific MAC schemes. A similar survey of MAC schemes for VANETs has been addressed in [22] where the authors categorize the protocols into three categories: time-based, DSRC-based, and directional antennabased. However, their classification does not cover all types of MAC protocols. In [23,24], multi-channel MAC protocols have been surveyed. Authors in [25] go through a wide range of recently proposed Time Division Multiple Access (TDMA)-based MAC protocols for VANETs and divide them into three categories including protocols operating in a fully distributed manner, protocols operating on cluster-based topology, and protocols operating on centralized topology. With respect to the aforementioned surveys, the main contribution of this out-of-the-box study is to present a novel inclusive classification of V2V MAC design strategies in an attempt to cover all types of design approaches. Moreover, the paper focuses on detailed characteristics of the recently proposed V2V MAC schemes and discusses their strengths and weaknesses. The paper is organized as follows. Section 2 provides a summarized background of MAC layer and its standardization in VANETs. Section 3 gives a classification of the V2V MAC design approaches. Section 4 reviews the details, advantages, and limitations of the recently proposed V2V MAC schemes. Section 5 presents a comparative summary of reviewed schemes. Section 6 discusses some open challenges focusing on drawbacks of the existing V2V MAC schemes. Finally, Section 7 concludes the paper. 2. Background Following, we take a look at the standardization efforts in VANETs and then we review several QoS and security aspects in VANETs. In the last section, we briefly discuss the MAC schemes in ad-hoc networks. 2.1. Standardization efforts for VANETs The standardization of VANETs is still under research and discussion. The IEEE in the US and the ETSI in EU are two main leaders supporting the VANETs standardization process, each with separate but fairly similar protocol architectures. The IEEE protocol stack which is known as WAVE is depicted in Fig. 1. The WAVE protocol stack includes a set of standards, including IEEE 1609.1/.2/.3/.4, and IEEE 802.11p, defining several functionalities similar to Open Systems Interconnection (OSI) model, but with some additional specifications and extensions. A management plane, termed as WAVE Management Entity (WME) is also superimposed to the data plane of WAVE suit to facilitate management services throughout the stack layers. Please cite this article as: N. Torabi, B.S. Ghahfarokhi, Survey of medium access control schemes for inter-vehicle communications, Computers and Electrical Engineering (2017), http://dx.doi.org/10.1016/j.compeleceng.2017.02.022

JID: CAEE

ARTICLE IN PRESS N. Torabi, B.S. Ghahfarokhi / Computers and Electrical Engineering 000 (2017) 1–23

[m3Gsc;March 3, 2017;13:54] 3

Fig. 2. Channel access operation modes in WAVE.

The IEEE 1609.4 describes the Multi-channel Operation of the WAVE stack and IEEE 802.11p specifies the WAVE PHY and WAVE MAC layers; the physical and MAC layers of WAVE stack respectively. At the WAVE PHY, Orthogonal Frequency Division Multiplexing (OFDM) technique is used to symbolize data bits and different modulation schemes are used to transmit coded data which leads to data rates of 3, 4.5, 6, 9, 12, 18, 24, and 27 Mbps. At the WAVE MAC layer, Carrier Sense Multiple Access with Collision Avoidance (CSMA/CA) mechanism alongside a prioritized channel access scheme similar to Enhanced Distributed Channel Access (EDCA) of the IEEE 802.11e is applied to control the access to the shared medium. The concept of Basic Service Set (BSS) and Independent BSS (IBSS) of baseline 802.11 standard have been replaced with a novel communication mode referred to as Outside the Context of a BSS (OCB) or WAVE BSS (WBSS) which does not need time-consuming setup process as exists in BSS, so results in a simple and short connection establishment. According to WAVE stack, coordination between channels in a single-radio OCB-enabled vehicular device is accomplished at the WAVE MAC layer through multi-channel operation mechanism defined by the IEEE 1609.4. As illustrated in Fig. 2, in an alternating access mode, i.e. the default mode of operation, the radio alternates between Control Channel (CCH) and Service Channels (SCHs) almost every 50 ms (considering a 4 ms Guard Interval (GI) for radio re-tuning), making a nominal Synchronization Interval (SI) with a default length of 100 ms in order to meet the time limitations of majority of the safety applications. Usually, synchronization between SCH Interval (SCHI) and CCH Interval (CCHI) is achieved through Coordinated Universal Time (UTC) using Global Positioning System (GPS) signals. However, in cases of inaccessibility of GPS signal or unavailability of GPS devices on board, synchronization can be accomplished via distributed techniques, based on timing signals received from neighboring vehicles. Moreover, devices in a continuous access mode do not need any channel coordination as they tune only in one channel, CCH or one of the SCHs, all the time. However, an immediate access mode ignores the restrictions for preserving SI timing, i.e. regular switching between SCHI and CCHI. In this mode, instantaneous switching to the SCH without waiting for the beginning of SCHI is possible. In extended access mode of operation, a WAVE device is allowed to not to switch back to the CCH or SCH and it extends its current status of being in a specific channel. In this way, the exchange of large infotainment data during SCHI would be facilitated. Alternatively, a device can stay on the CCH during SCHI when there is no need to be on an SCH. For the sake of better performance in vehicular communications, Federal Communication Commission (FCC) of the United States granted a 75 MHz of bandwidth around 5.9 GHz frequency band known as DSRC which is partitioned into seven 10 MHz-wide channels including one control channel (CCH) and six service channels (SCHs) which are used to exchange control (governing), safety (critical) and non-safety (non-critical or infotainment) information. While IEEE is working on WAVE, the ETSI is standardizing a protocol architecture similar to WAVE stack which is known as the ETSI ITS Station for vehicular communications. Fig. 3 depicts the ETSI protocol stack. Among other protocols, the ETSI stack also supports IEEE 802.11p protocol at the access layer. Facilities and also networking and transport layers do the same operations as the IEEE 1609.4 and IEEE 1609.3. Similar to the IEEE 1609.2 and IEEE 1609.1 in WAVE, the ETSI also provides security and management services. Apart from US and EU, there also exist other standardization bodies all over the world working on ITS services. In japan, Association of Radio Industries and Businesses (ARIB) is the standardization organization contributing in the standardization of ITS communications. STD-T75 the descendant version of the STD-T55 is the main contribution of ARIB suggesting the frequency range 5.725–5.875 GHz with 40 MHz wide uplink and downlink channels and 80 MHz separation bandwidth for VANETs. It focuses on the structure of the OSI reference model and the standardized layers are Layer 1 (Physical Layer), Layer 2 (Data Link Layer) and Layer 7 (Application Layer). This standard emphasizes on ITS applications such as Electronic Toll Collection (ETC). With some difference from STD-T55 and STD-T75, the more recent ARIB standard, the STD-T109 intends Please cite this article as: N. Torabi, B.S. Ghahfarokhi, Survey of medium access control schemes for inter-vehicle communications, Computers and Electrical Engineering (2017), http://dx.doi.org/10.1016/j.compeleceng.2017.02.022

ARTICLE IN PRESS

JID: CAEE 4

[m3Gsc;March 3, 2017;13:54]

N. Torabi, B.S. Ghahfarokhi / Computers and Electrical Engineering 000 (2017) 1–23

Fig. 3. The ETSI protocol stack.

to realize ITS applications based on the frequency band 760 MHz (ranging from 755.5 MHz to 764.5 MHz). While aforementioned ARIB standards are substantially focusing on lower layers, the STD-T88 is another standard of ARIB which is applied to a DSRC system based on the ARIB STD-T75. The STD-T88 defines an additional application layer, known as DSRC Application Sub-Layer (ASL) over layers standardized by STD-T75. A brief comparison of discussed regional standards is summarized in Table 1.

2.2. QoS and security aspects in MAC layer of VANETs Probably it will be the largest open access network around the world if VANETs become pervasive. This makes the whole network vulnerable, because of multiple access points from inside and outside the network. The situation gets worse when due to tight connectivity between vehicles, hazards from a vulnerable and faulty vehicle affect more vehicles at once. Hence, despite unique features of VANETs which offer rich options to drivers, there exist many challenges that must be addressed before VANETs can be successfully deployed. Among these challenges is the tradeoff between provisioning security and designing efficient MAC schemes so that messages can be securely and reliably disseminated. On the other hand, the QoS is needed to be supported at several layers of the protocol stack and should involve the whole network architecture [1]. Hence, a QoS-aware MAC scheme should be of concern which can provide QoS requirements of safety and non-safety applications at this level. Predictability of access delay, transmission delay reduction, reliability, enough flexibility (scalability), and fair sharing of radio resources are among the most significant requirements that are needed to be supported at the MAC layer [1,20]. In this paper, we will compare famous MAC schemes from such QoS points of view. Another fact is that securing the vehicular communications should not impose negative effects on performance. Hence, a QoS-aware secure MAC scheme will be of paramount importance which can eliminate security threats while preserving QoS requirements of safety and non-safety applications. To this end, both hardware solutions and architectural infrastructures are needed. The OBUs should be equipped with hardware modules like Event Data Recorder (EDR) and Tamper-Proof Device (TPD) to record the vehicle’s critical data, such as location, speed, time, etc., during emergency events (similar to an airplane’s black box), and to take care of storing cryptographic materials and performing cryptographic operations, including signing and verifying safety messages, respectively. Thus, the data in EDR will help in accident reconstruction and the attri-

Table 1 Comparison of regional VANETs’ standards. Feature

Europe (ETSI ITS station)

Japan (ARIB STD-T75/-T109)

USA (DSRC/WAVE)

Radio frequency band Radio frequency bandwidth

5.8 GHz 30 + 20 (= 50) MHz (ITS-G5A + ITS-G5B) 5 10 MHz OFDM DCC access Half-duplex ∼1500 m Up to 200 km/h

5.8 GHz / 700 MHz 40 + 40 (= 80) MHz (Uplink + Downlink)/< ∼10 MHz

5.9 GHz 75 MHz

Uplink: 7, Downlink: 7 / 1 5 MHz/∼10 MHz ASK, QPSK / OFDM TDMA-FDD / CSMA/CA OBU: Half-duplex, RSU: Half- or full-duplex / Half-duplex <30 m Up to 180 km/h/up to 140 km/h

7 10 MHz OFDM CSAM/CA-FDD Half-duplex ∼10 0 0 m Up to 200 km/h

Number of channels Channel bandwidth Modulation type Radio access type Communication type Communication range Vehicle speed

Please cite this article as: N. Torabi, B.S. Ghahfarokhi, Survey of medium access control schemes for inter-vehicle communications, Computers and Electrical Engineering (2017), http://dx.doi.org/10.1016/j.compeleceng.2017.02.022

ARTICLE IN PRESS

JID: CAEE

N. Torabi, B.S. Ghahfarokhi / Computers and Electrical Engineering 000 (2017) 1–23

[m3Gsc;March 3, 2017;13:54] 5

Fig. 4. Classification of ad hoc MAC schemes.

bution of liability, while the TDP guarantees the accountability property by binding a set of cryptographic keys to a given vehicle [26]. 2.3. A short review of ad hoc MAC schemes In this section, we make a short and concise classification of ad hoc MAC schemes. Referring to Fig. 4, ad hoc MAC schemes are divided into two top-level categories termed as contention-free and contention-based. In a contention-free MAC scheme, (e.g., Polling, Token-based, TDMA, FDMA, CDMA), certain assignments are used to avoid contentions. Pure contention-free MAC schemes are more applicable to static networks and/or networks with centralized control. On the other side, there exist contention-based MAC schemes which are mainly based on competition between attending nodes competing for accessing a shared wireless channel. Competition-based schemes are categorized into random access and dynamic reservation/collision resolution protocols. In random access based schemes, such as ALOHA, a node may access the channel as soon as it is ready. A variation of ALOHA, namely Slotted ALOHA, introduces synchronized transmission time-slots similar to TDMA where nodes can transmit only at the beginning of a time-slot leading to doubled throughput as compared to the pure ALOHA, with the cost of time synchronization. The CSMA-based schemes further reduce the possibility of packet collisions and improve the throughput through introducing carrier sensing mechanisms. In order to solve the hidden and exposed terminal problems in CSMA, researchers have come up with many protocols adopting some forms of dynamic reservation/collision resolution such as Multiple Access Collision Avoidance (MACA) and MACA for Wireless LANs (MACAW) which use Request-To-Send/Clear-To-Send (RTS/CTS) control packets to prevent collisions and some others combine both carrier sensing methods and control packets such as Floor Acquisition Multiple Access (FAMA), Sensor-MAC (SMAC), and IEEE 802.11 CSMA/CA. A comprehensive classification of ad hoc MAC protocols can be found in. 3. V2V MAC design approaches V2V MAC schemes constitute the larger part of VANETs MAC solutions in the literature. The reason is two-fold; on one side, employing V2V-based communications is more advantageous against V2I-based communications [2]. On the other side, challenging issues of V2V MACs (such as synchronization between vehicular nodes, adaption to the mobility of vehicles, variety in the number of vehicles, and QoS-related issues) are more interesting for researchers. Following these, in this paper we focus on inspecting the recently proposed V2V MAC solutions. However, prior to investigating the V2V MAC schemes, in this section, we give an insight into the V2V MAC design approaches. 3.1. Single- or multi-radio The number of radios that a vehicle can use, is an important factor for designing MAC schemes and has a great impact on their performance. Assuming DSRC to be default physical layer specification, vehicles which use only one transceiver may choose between continuous, alternating, immediate, or extended access modes, each with its own advantages and disadvantages. For example, although using a single-radio device which is continuously tuned in certain channel results in poor spectrum utilization, it facilitates the in-time delivery of safety data. On the other hand, switching between CCH and SCHs lets transmitting non-safety data besides transmission of safety information, while brings more complexity and the probable expiration of real-time safety data. Generally, due to simplicity and low cost of employing single-radio devices, they could be considered as a primary choice for deployment of VANETs [1,27]. However, negative impacts of channel switching on spectrum efficiency and performance of safety applications do not make the single-radio approach a good candidate for multi-channel MAC architectures. Therefore, a multi-radio transceiver Please cite this article as: N. Torabi, B.S. Ghahfarokhi, Survey of medium access control schemes for inter-vehicle communications, Computers and Electrical Engineering (2017), http://dx.doi.org/10.1016/j.compeleceng.2017.02.022

ARTICLE IN PRESS

JID: CAEE 6

[m3Gsc;March 3, 2017;13:54]

N. Torabi, B.S. Ghahfarokhi / Computers and Electrical Engineering 000 (2017) 1–23

can be considered as a better substitution when dealing with multi-channel MACs. For example, without loss of generality in the number of radios, a dual-radio device having one radio dedicated to safety data and the other one to non-safety data promises fast and reliable delivery of real-time safety data, in addition to providing more spectrum utilization. Although higher technical complexity and more costs are included as the main weak points of such an approach, multi-radio devices will be used prominently in the implementation of VANETs in near future [1,27]. 3.2. Single- or multi-channel In a single-channel MAC solution, all the competing nodes share only one channel and all the packets from any type are transmitted through the same channel. So, the maximum throughput is constrained by the transmission capacity of the single channel. Furthermore, the competition among the nodes that share the same wireless channel leads to the increase of packet drops and decrease of the throughput. Contrary to a single-channel, a multi-channel MAC solution is based on the division of spectrum into several sub-channels and distribution of contending nodes over them. Thus, using multiple sub-channels competitions are decreased, packet collisions are reduced, and delay of packet delivery declines. In [1], wireless multi-channel MAC schemes are classified into four categories: (1) in a Common Hopping approach, nodes use one radio device. As soon as two nodes agree on a transmission, they stop hopping and after ending the transmission, they rejoin the common hopping again; (2) in a Split Phase approach, nodes use only one radio device too. Herein, time is divided into an alternating cycle of control and data intervals that lets the nodes transmit safety and non-safety information; (3) in a Dedicated Control Channel approach, each device is equipped with two radios. One radio is dedicated to one channel, namely the control channel and the other one is tuned in any of the other channels; (4) in a Parallel Rendezvous approach, independent of the number of radios used, more than one channel is dedicated to control information to overcome any possible bottlenecks of using only one single control channel. All these approaches are also potential candidates for designing V2V MAC schemes. Vehicular devices with multiple transceivers can take advantages of Dedicated Control Channel and Parallel Rendezvous approaches and make more bandwidth efficiency and transmission reliability. However, due to the high costs and more technical hurdles, these two approaches are not suitable choices for V2V MAC schemes, at least for now. While the Common Hopping scheme is the common approach in several MAC schemes in the literature, it suffers from hopping time penalty and tight synchronization which make it greatly unusable for V2V MAC. The Split Phase approach is free of such defects and its simplicity of implementation makes it the most preferable approach among the VANETs researchers. However, in case of a poor design, it may introduce non-negligible problems. In [27], several remarkable issues around the multi-channel operation of the IEEE 1609.4 protocol have been detailed. The inefficiencies of the switching scheme, the adjacent channel interference, the issues around multi-hop communications, and channel selection are among these problems. 3.3. Competitive or non-competitive In a competitive or contention-based MAC approach mobile nodes are allowed to contend for accessing the shared channel in a completely distributed manner without depending on predefined infrastructures. A preferred distributive competitive MAC scheme is the one that provides an efficient mechanism to share limited bandwidth resources fairly, reduces the delay, improves the throughput and importantly simplifies the nodes management operation. Contention-based MAC protocols mainly include protocols that are based on carrier sensing and collision detection/avoidance mechanisms. As a foible, a competitive MAC scheme behavior is much like a stochastic one; since no one can predetermine the exact time when a node will be able to transmit through the channel. However, its potential to be used in decentralized scenarios makes it greatly useful for V2V communications. On the other side of the coin, there exists a non-competitive approach which tries to have a more deterministic performance towards the channel access. Normally, non-competitive schemes divide time or frequency into several pieces of shares such that each piece is allocated to one mobile node to transmit its data. For instance, a TDMA-based MAC scheme provides sequential pieces of the time and lets nodes have one or more of them to transmit data or an FDMA-based MAC scheme offers pieces of the frequency known as sub-channels and lets nodes use one of them to transmit data. Therefore, the competition between nodes is eliminated, nodes survive from starvation, and transmission delay can be deterministic. However, since a fully non-competitive approach needs a centralized control infrastructure, e.g. to handle the allocation of resources, it is not practical for developing V2V MAC schemes. The better choice is to take the advantages of the combination of the both competitive and non-competitive approaches in order to achieve desirable performance results. Hence, the hybrid MAC solutions could be good candidates for V2V MAC schemes. For example, the IEEE 802.11p protocol in combination with the IEEE 1609.4 protocol build some kind of hybrid MAC scheme where in each sub-channel (channels defined in DSRC), time is divided into fixed-size intervals and during each interval, nodes are allowed to access the medium through the competitive CSMA/CA mechanism. 3.4. Cognitive MAC In an ordinary network, due to the strict spectrum assignment policies large sections of the available spectrum resources are heavily underutilized. Dynamic Spectrum Access (DSA) policy with the help of Cognitive Radio (CR) technology are the Please cite this article as: N. Torabi, B.S. Ghahfarokhi, Survey of medium access control schemes for inter-vehicle communications, Computers and Electrical Engineering (2017), http://dx.doi.org/10.1016/j.compeleceng.2017.02.022

ARTICLE IN PRESS

JID: CAEE

N. Torabi, B.S. Ghahfarokhi / Computers and Electrical Engineering 000 (2017) 1–23

[m3Gsc;March 3, 2017;13:54] 7

promising solution to overwhelming this problem. The main idea in a Cognitive Radio Network (CRN) is to determine the available portion of the spectrum, known as spectrum holes (or white spaces), and then using them without interfering with the other portions of the spectrum granted to licensed users. In CRNs, unlicensed users which are called Secondary Users (SUs) may not pay for using that wasted portions of the spectrum that is bought by the licensed/authorized users known as Primary Users (PUs). This is considered as a great advantage of CR approach to be accepted in VANETs. In VANETs, the IEEE 1609.4 protocol does not provide adequate spectrum utilization. The half of the time intervals of the service channels remain idle when the nodes are contending within the control channel, i.e. during the CCHI. Some cognitive MAC schemes have been developed to overcome the inefficiency of the WAVE system itself. In these schemes, the vehicular nodes with safety-related messages are defined as PUs and nodes with non-safety information are determined as SUs [7]. Moreover, there are several cognitive MAC solutions that are developed to address the insufficient capacity of DSRC channels to support the broad range of ITS services envisioned in VANETs. Hence, depending on the availability of wireless networks and location of vehicles, WiFi, WiMAX, cellular, TV and satellite bands are used alongside the unlicensed ISM bands to acquire more spectrum capacity for VANETs [19]. Despite the success of cognitive radio, there are still many technical challenges to overcome. Hidden primary users and spread spectrum primary users are serious problems which both of them lead a cognitive radio to incorrectly choose an empty spectrum block, and consequently result in interfering with the licensed users. Similar to an ordinary hidden node problem, the hidden primary user problem occurs when a secondary user undesirably destruct transmission signals between two primary users while it is being out of range of one of the primary users, i.e. the hidden primary user. On the other side, spread spectrum primary user problem stands at same insidious status. A spread spectrum user may only require a very low power signal spread across a wide bandwidth. In the worst case scenario, a cognitive radio system might test the spectrum in bands much smaller than the wide bandwidth used by the primary user. Thus, the primary user’s low power transmission may be interpreted as background noise, leading to the false identification of empty spectrum. 3.5. Clustering-based MAC Clustering is a process wherein the mobile nodes that are geographically in the same vicinity and have certain common characteristics join in creating small groups. Clustering schemes are mainly used to handle the mobility and the quick changes in the network topology. At the MAC layer of mobile ad hoc networks, clustering is used to handle the hidden node problem, provide better scalability, reduce the number of interfering nodes, limit the area of dissemination, and provide fairness. In VANETs, due to the high mobility and the varying speed of vehicles, most of the V2V MAC schemes are built based on clustering of vehicles. The main steps in clustering process are virtually grouping of the vehicular nodes that are close together (regarding speed, location, direction, etc.) and selecting a vehicle as the Cluster Head (CH) to act as the central manager of the cluster. The algorithm which is used to create the clusters has significant effects on the stability of the clusters. In a general categorization, clustering in VANETs is accomplished according to two distinctive classes of algorithms. In the first class, geographical information such as velocity, acceleration, location, and movement direction of vehicles are used to make stable clusters; and in the second class, different measurable parameters like signal strength, relative mobility, vehicle density, connectivity, etc. are considered as the factors of constructing clusters. Other classifications may classify clustering methods into active and passive methods. An active clustering method requires that all the participating nodes disseminate cluster-related control information frequently. These techniques often need to be executed prior to any MAC layer activity. However, passive clustering methods do not require dedicated control messages and instead some predefined cluster information are piggybacked by data packets. Studies on clustering-based MAC solutions have mostly focused on the minimizing of the numbers of transceivers, avoiding the inter-cluster collisions, and use of multi-hop communications. 3.6. Directional antenna-based MAC A directional antenna allows a node to transmit in a particular direction which results in reduced transmission collisions, improved channel reusability, increased transmission radius, and probably decreased signal interferences. In VANETs, many vehicular applications require only to disseminate messages in a specific direction. Under such circumstances, directional antenna-based MAC schemes can be applied to help decreasing interference on neighboring vehicles. Nonetheless, directional antenna-based MAC schemes need to overcome some technical issues. Commonly, a directional antenna-based transmitter may fail to communicate to its intended directional antenna-based receiver, because it is beam-formed towards a direction away from the transmitter. This is a common problem in directional antenna-based MAC schemes which is known as deafness. Hence, one explicit requirement in such schemes is the knowledge of the transmitter’s and the receiver’s active antennas during the information transmission. 3.7. SDMA-based MAC Partitioning the geographical area into multiple divisions is the main concept introduced by a Space Division Multiple Access (SDMA) approach. Each division is mapped to one or several numbers of channels. In VANETs, the precise geographical Please cite this article as: N. Torabi, B.S. Ghahfarokhi, Survey of medium access control schemes for inter-vehicle communications, Computers and Electrical Engineering (2017), http://dx.doi.org/10.1016/j.compeleceng.2017.02.022

JID: CAEE 8

ARTICLE IN PRESS

[m3Gsc;March 3, 2017;13:54]

N. Torabi, B.S. Ghahfarokhi / Computers and Electrical Engineering 000 (2017) 1–23

location of the vehicles is assumed to be known, either through GPS or any other geo-localization system. Therefore, each vehicle is able to find out which division it resides on at any instance of time and consequently is able to find out which channel(s) it can use while it resides or travels inside different divisions. Many advantages can be reached using SDMAbased MAC solutions including reduced access collisions, increased channel reusability, flexible underlying multiple access schemes (such as TDMA, FDMA, or CDMA), low communication/time overhead, and robust network organization. However, there are some challenging issues for the SDMA-based MAC schemes. Among them, partitioning process of the road which needs precise knowledge of road map is of great importance [28]. 3.8. Cooperative MAC Due to the plurality of signal destructive obstacles, channel impairment in VANETs is very common. Hence, a signal from the source to the destination can be destructed easily. Cooperative communication is an approach which makes use of nearby nodes to improve the transmission performance between a pair of source and destination nodes. In a cooperative MAC scheme, the nodes which help to relay the packet to the destination are referred to as helper nodes [4]. Various metrics can be used to choose the helper node. For example, it can be selected through periodic measurement of signal strength of neighbor nodes using heartbeat messages. However, the legacy cooperative schemes which are based on historical transmissions are not appropriate for dynamic environments like VANETs. Instead, distributed cooperative MAC schemes, wherein the decision of cooperation and selection of helper node(s) are made during ongoing transmissions are more applicable, especially for V2V communications. However, most of the existing distributed cooperative MAC schemes are based on the IEEE 802.11 standard and force neighbor nodes to stop their own transmissions during the cooperative transmission. To cope with this problem, newer cooperative mechanisms utilize TDMA approaches [4,18] that let neighboring nodes to continue their own ongoing transmissions while they are engaged in another cooperation. 3.9. Taking turns MAC Having a certain continual opportunity to transmit data through the channel can be ideal for a mobile node when it is participating in a transmission. Taking turns is a MAC design approach which helps to realize this through using either polling- or token-based techniques. The main idea behind a polling-based MAC solution is to serve multiple data frames after a successful contention resolution, thus alleviating signaling overhead and making the protocol more efficient. And, the idea behind token-based MAC schemes is originated from the traditional token-ring networks. Both techniques provide fairness by giving each node a minimum opportunity to transmit. Moreover, a real-time bandwidth allocation will be possible where time will not be wasted at a node if it does not intend to transmit [21]. It seems that taking turns approaches could be good candidates for developing QoS-aware MAC schemes. However, developing such schemes needs further considerations to establish more reliable connections regarding the instability of V2V connections in VANETs. In Table 2, features, advantages, and disadvantages of discussed MAC design approaches are summarized. 4. Review of recently proposed V2V MAC schemes In this section, we provide a comprehensive discussion on several outstanding recently proposed V2V MAC schemes and go through their details. However, since MAC solutions usually exploit multiple approaches, placing them in explicit categories without any overlap is far from simple. Hence, in order to avoid over-lapping between categories, we are going to categorize them based on the following metrics: •

•

•

Single-radio/Single-channel: all the governing, critical, non-critical and infotainment messages are transmitted through one channel using one radio device. Single-radio/Multiple-channel: typically one channel is used for governing and critical messages, and one or more channels for non-critical and infotainment messages. Multiple-radio/Multiple-channel: typically one radio tuned in one of the channels is dedicated for governing and critical messages, while the other radio(s) is/are used for non-critical infotainment messages.

4.1. Single-radio/single-channel MAC protocols A self-organizing TDMA MAC protocol, namely STDMA for V2V communications has been proposed by Rezazade et al. [3]. The proposed MAC is a hybrid MAC scheme which combines both CSMA/CA and TDMA techniques to provide synchronization among vehicles and to help them obtain their own time slots using a mechanism including three consecutive phases. A newly arrived vehicle, starting from phase 1 (refer to Fig. 5), waits for a certain amount of time, called Channel Empty Time (CET) whenever it senses the channel as empty then it starts to send a control packet, namely Competitive Message (CM) in order to announce to the neighboring vehicles the time slot which it wants to possess. Successful transmission of CM lets the sender node proceed to phase 3 where it should wait for another time period, called Waiting Time (WT) until other nodes have the opportunity to reserve their requested time slots. Nodes that have not possessed their time slot continue to Please cite this article as: N. Torabi, B.S. Ghahfarokhi, Survey of medium access control schemes for inter-vehicle communications, Computers and Electrical Engineering (2017), http://dx.doi.org/10.1016/j.compeleceng.2017.02.022

ARTICLE IN PRESS

JID: CAEE

[m3Gsc;March 3, 2017;13:54]

N. Torabi, B.S. Ghahfarokhi / Computers and Electrical Engineering 000 (2017) 1–23

9

Table 2 Summary of important features, benefits and drawbacks (or issues) of V2V MAC design approaches. Approach

Important features

Benefits

Drawbacks (or issues)

Single-radio

- One single radio for exchanging both safety and non-safety data - Primary choice for deployment of VANETs

- Simplicity and low cost of employing

- Poor spectrum utilization

Multi-radio

- E.g. dual-radio device, with one radio dedicated to exchanging safety and the other for non-safety data - Suitable for long-term deployments

- Fast and reliable delivery of real-time safety data - Better spectrum utilization

- Higher complexity and more cost

Single-channel

- Sharing one single channel - Transmitting all types of packets through the same channel

- Simplicity and low overhead

- Constrained maximum throughput - Increase of packet loss

Multi-channel

- Division of spectrum into several channels - Distribution of contending nodes over the sub-channels

- Decreased competitions - Improved throughput of non-safety apps - Declined delay of safety packet delivery

- Channel selection - Adjacent channel interference

Competitive

- Contention for common channel - Stochastic behavior

- Useful for decentralized (V2V) scenarios - Low control overhead

- Unbounded transmission delay - Access starvation - Hidden station problem

Non-competitive

- Division of time (e.g. TDMA) or frequency (e.g. FDMA) - Deterministic behavior

- Prevention from access starvation - Bounded transmission delay

- Centralized control (basically not suitable for V2V) - High control overhead

Cognitive

- To allow secondary users to use spectrum allocated to primary users

- Better spectrum utilization (to overcome inadequate spectrum utilization of the IEEE 1609.4) - To cope with spectrum scarcity of DSRC)

- Hidden primary users - Spread spectrum primary users (false identification of empty spectrum)

Clustering-based

- Grouping of nodes according to geographical information (velocity, acceleration, location, movement direction, etc.), signal strength, relative mobility, vehicle density, connectivity, etc. - Active and passive clustering methods

-

To handle hidden node problem Better scalability Reduced number of interfering nodes Fairness

- High control overhead - Inter-cluster collisions - Multi-hop communications

Directional Antenna-based

- To allow to transmit in a particular direction - Division of transmission space into x transmission angles of 360/x degrees

-

Reduced transmission collisions Improved channel reusability Increased transmission radius Decreased interference

- High implementation cost - Deafness problem

SDMA-based

- Partitioning geographical area into multiple divisions and assigning one or more channel(s) to each division - Current division is found out through GPS or any other navigation system and then usable channel(s) are determined

- Reduced access collisions - Increased channel reusability - Flexible underlying access schemes (such as TDMA, FDMA, or CDMA) - Low overhead - Robustness

- Dependency on precise knowledge of (electronic) road map - Dependency on navigation systems

Cooperative

- Use of nearby nodes (known as helper nodes) to improve transmission performance between source and destination nodes - Historical and distributed methods

- To cope with signal destructions and channel impairments in V2V communications

- High control overhead - Potential interruptions in transmission of helper nodes

Taking Turns

- Polling-based MAC solution: - To serve multiple data frames after a successful contention resolution Token-ring MAC solution: - Originated from traditional wired networks

Polling-based MAC solution: - To alleviate high control overhead leading to more efficiency Token-ring MAC solution: - Bounded transmission delay Both techniques: - Fairness - Real-time bandwidth allocation - Prevention from time wasting

- To establish reliable connections (given instability of V2V communications)

phase 2 where they try again to reserve a time slot via sending new CMs. Upon termination of phase 2, all the vehicles enter to the time-slotted portion of the channel at the same time. Fig. 5 illustrates a simplified picture of the synchronization and the slot allocation mechanisms of the STDMA scheme. Although the STDMA scheme benefits from its simplicity, its synchronization mechanism imposes additional delay to real-time transmissions, especially when the network is crowded. Consider a scenario including N vehicles, all of them with information to send through the shared channel. In this scheme, the first vehicle which succeeds to obtain a time Please cite this article as: N. Torabi, B.S. Ghahfarokhi, Survey of medium access control schemes for inter-vehicle communications, Computers and Electrical Engineering (2017), http://dx.doi.org/10.1016/j.compeleceng.2017.02.022

JID: CAEE 10

ARTICLE IN PRESS

[m3Gsc;March 3, 2017;13:54]

N. Torabi, B.S. Ghahfarokhi / Computers and Electrical Engineering 000 (2017) 1–23

Fig. 5. Simplified picture of STDMA protocol for three nodes A, B, and C. At the end of the phase 3, Nodes A, B, and C possess their own time slots (ts) 1, 3, and 2, respectively [3].

slot during synchronization process has to wait for N-1 vehicles to get their own time slots until it can send through the channel. Moreover, the scheme does not declare when the reservation process is renewed or even how a vehicle is informed once its CM message is received successfully by its neighboring vehicles. This may have serious effects on the reliability of the scheme. Bharati et al. [4] have suggested a distributed two-hop clustering-based cooperative MAC scheme for V2V communications, referred to as CAH-MAC which is based on ADHOC MAC scheme. The protocol is based on a fully TDMA scheme and aims at utilizing unused time slots for retransmission of failed packets. Broadcast, multicast, and unicast communication are all supported. In this scheme, each vehicle maintains a list of its one-hop and two-hop neighbors and slot reservation is done between nodes which are at the same two-hop-set. A node needs to listen to the channel for a while before it can reserve a time slot among the unreserved ones, if it is available. Different from the ADHOC MAC, the CAH-MAC introduces two new signaling fields in the packets, namely the Frame Information (FI) and Cooperation Header (COH). Time slots in each time frame are identified by an Identification Field (IDF) in the FI field. By successfully receiving FIs from all of its one-hop neighbors, a node can obtain a table which includes: (a) all of its one-hop neighbors, (b) all of its two-hop neighbors, and (c) the owner of each time slot in a frame. The cooperation process is triggered if all the following conditions are satisfied: the direct transmission fails, the helper successfully receives a packet for retransmission, the destination is reachable, and there exists a free time slot. Then a helper node offers the cooperation to the source and destination by filling the COH field. Following that, a Cooperation Acknowledgement (C-ACK) message from the destination node is transmitted during an unreserved time slot offered by the helper node, and finally the cooperative transmission is performed in the selected time slot. Fig. 6 exemplifies these steps. The CAH-MAC scheme considers a perfectly synchronized network where nodes have already reserved their own time slots and cooperation is performed by only these nodes. While this could be very facilitative, it ignores a large portion of the collisions that is the access collisions which may have great impacts on the performance. Simulations by the authors show that the CAH-MAC scheme improves throughput compared to the ADHOC MAC, while the effects of network topology changes (due to high mobility of vehicles) have been neglected. An adaptive TDMA slot assignment strategy, namely ASAS has been offered by Hadded et al. [5] for V2V communications based on the clustering of vehicles. The main goal of this scheme is to provide a MAC protocol that reduces inter-cluster interference under different load conditions. The key idea is to take the direction and position of the vehicles into consideration in order to decide which slot should be occupied by which vehicle. It is assumed that the synchronization among vehicles is performed by GPS. To provide more stable clusters, only vehicles moving in the same direction can be members of the same cluster. Vehicles within each cluster are divided into two sets: vehicles behind the CH (set B) and vehicles in front of the CH (set F). Time is divided into frames and each frame is divided into Adaptive Broadcast Service (ABS) and Contention-based Reservation Period (CRP) time divisions. In ABS, TDMA method is used where each vehicle uses its own time slot to only transmit safety or control messages while it receives safety messages through other time slots. In CRP, the CSMA/CA method is used to reserve a periodic time slot in ABS. Please cite this article as: N. Torabi, B.S. Ghahfarokhi, Survey of medium access control schemes for inter-vehicle communications, Computers and Electrical Engineering (2017), http://dx.doi.org/10.1016/j.compeleceng.2017.02.022

JID: CAEE

ARTICLE IN PRESS N. Torabi, B.S. Ghahfarokhi / Computers and Electrical Engineering 000 (2017) 1–23

[m3Gsc;March 3, 2017;13:54] 11

Fig. 6. Cooperative communication in CAH-MAC: (a) Source node S transmits a packet to the destination node D. Node D fails to receive the packet. (b) Node D announces transmission failure through its FI. Neighboring nodes detect transmission failure after examining the FI. (c) Helper node H offers cooperation via transmitting COH. (d) Node H retransmits the packet that failed to reach Node D after receiving a C-ACK message from Node D [4].

Fig. 7. The architecture of time frames in ASAS [5].

In ASAS, to avoid merging collision problems, each ABS frame is divided into two sets of time slots: the first set is used by vehicles moving in the left direction (Left) and the other is used by vehicles moving in the right direction (Right). As depicted in Fig. 7, each set of time slots is further partitioned into three subsets of time slots: time slots for vehicles of set F (L), time slots for vehicles of set B (R), and time slots in which all vehicles in the cluster remain inactive (N). To avoid inter-cluster interference problem, the order of time slot subsets (R, L, and N) differs from that of neighboring clusters. A vehicle with information to send via the channel, sends a reservation request to the CH for a time slot. A MAP concept which is obtained from periodically received Frame Information (FI) messages from neighboring CHs, is used by the CH to determine the distribution of time slot subsets. So, it can reply to the request of the vehicles by assigning an available time slot that is not a cause of access collision. Although the ASAS scheme tries to improve the convergence performance of the MAC protocol through a clustering-based approach, its strict dependency on the CH makes it vulnerable to dynamic changes in network topology. Moreover, periodic transmission of FI messages between neighboring CHs imposes undesired signaling overhead to the protocol. Furthermore, one can argue that the division of time slots into constant-length Left and Right sets may compromise fairness or bandwidth utilization, since number of vehicles with different bandwidth demand may change through the time.

4.2. Single-radio/multiple-channel MAC protocols A DSRC-based adaptive multi-channel MAC protocol for dense VANETs has been proposed by Xu et al. [6], known as VMMAC. The key purpose of VMMAC is to increase bandwidth utilization through spatial reuse. In this scheme, each vehicle is equipped with a single half-duplex transceiver which continuously alternates between the CCH and one of the SCHs. It is assumed that the synchronization is reached via periodically piggybacked beacons. A combination of unidirectional and omnidirectional modes is used by the sender and the receiver on the CCH to reduce the effects of the hidden nodes. To this end, each vehicle maintains a Beam Table (BT) to indicate the current state of beams in all DSRC channels. For channel selection, vehicles use a four-way Channel Selection (CS) handshaking mechanism (CS-RTS/CS-CTS/CS-Res/CS-Ack) to exchange the BTs, as illustrated in Fig. 8. In this way, a common unblocked SCH would be reserved for data transmission between two nodes. Moreover, the unblocked beams can be used as long as the vehicular nodes get into the unblocked state. Please cite this article as: N. Torabi, B.S. Ghahfarokhi, Survey of medium access control schemes for inter-vehicle communications, Computers and Electrical Engineering (2017), http://dx.doi.org/10.1016/j.compeleceng.2017.02.022

JID: CAEE 12

ARTICLE IN PRESS

[m3Gsc;March 3, 2017;13:54]

N. Torabi, B.S. Ghahfarokhi / Computers and Electrical Engineering 000 (2017) 1–23

Fig. 8. The VMMAC handshake mechanism between sender and receiver to select the best channel [6].

In VMMAC scheme, although one-hop neighbors of the sender and receiver become aware of the ongoing communication between these two nodes, newly arrived vehicles are still potential interferers. Furthermore, while the implementation of beam-forming antennas imposes high costs and complexities, the scheme focuses on just lane-directional transmissions which are appropriate for single-lane roads which means multi-lane highways or even city scenarios are not supported by the scheme. Chu et al. [7] proposed a cognitive radio-enabled multi-channel MAC protocol based on IEEE 1609.4, termed as CREM to address two major concerns in the WAVE standard as follows: (a) to improve the unutilized CCHI in the SCHs; and (b) to increase the transmission reliability of safety-related messages. Based on the concept of cognitive radio, the vehicular nodes with safety-related messages are defined as primary providers/users and nodes with non-safety data are determined as secondary providers/users. In order to give higher priority to safety messages, smaller Contention Window (CW) size is adjusted for primary nodes comparing to that of the secondary nodes. Moreover, the primary nodes are allowed to occupy an SCH to transmit their data in multiple SIs until the completion of data transmission. To promote the channel utilization of the WAVE system much more, an enhanced CREM (CREM-E) protocol is proposed that allows the secondary nodes to transmit their data for an additional SI if there is no primary node in the current CCHI. The CREM establishes the connection between vehicles based on the concept of WBSS in the IEEE 1609.4 standard. The higher privilege in channel negotiation is granted to the primary provider which is allowed to transmit its data for more than one SI with no need to return to the CCH during CCHI. The secondary provider, on the other side, will only be able to transmit data if there are spectrum holes within the network. In the case that one SI would not be necessary to complete the data transmission, it has to conduct channel sensing, contention, and negotiation processes within each CCHI per each SI. As depicted in Fig. 9, every node updates its perception of all the six SCHs by using its radio transceiver at the beginning of every SI and creates its own Channel Status Table (CST) which determines if the channel is in idle or busy state. Based on its CST information, an unoccupied SCH will be selected by the provider to exchange data, which will be recorded in the WAVE Service Advertisement (WSA) message. A handshake will be completed when one of the WBSS members transmits the corresponding WSA Response (WSAR) message for acknowledgement. After the completion of contention process during the CCHI, the nodes that are granted with an SCH will be switched to that channel for data transmission. The CREM scheme and its extension the CREM-E improve the channel utilization effectively since unlike the IEEE 1609.4 protocol, they make use of the SCHs during CCHI. At the same time, they preserve the time requirements of safety-related messages, as much as possible. The simulation results show that these scheme surpass the IEEE 1609.4 standard in both channel utilization and response time. However, according to the scheme the provider and user nodes in a WBSS are obligated to sense the channel at the beginning of each SI, while it may exist some nodes which they do not have any data for transmission. Moreover, since the available SCHs are not dedicated to the primary nodes, they also have to contend for channel access on the CCH. This may have undesired effects on the QoS of the primary nodes’ applications, which needs to be investigated by the authors. Lu et al. [8] have presented a dedicated multi-channel MAC protocol with an adaptive broadcasting mechanism, named as DMMAC. The DMMAC is based on a hybrid channel access mechanism exploiting advantages of both TDMA and CSMA/CA methods. The main goal of the DMMAC is to providing collision-free delay-bounded delivery of safety messages. In this scheme, similarly to the IEEE 1609.4 protocol, time is divided into 100 ms Sync Intervals (SI) which are further partitioned into equal-length 50 ms CCHI and SCHI periods. Moreover, the CCHI is divided into two consecutive periods: a variablelength Adaptive Broadcast Frame (ABF) and a Contention-based Reservation Period (CRP). The ABF is further divided into equal-sized contention-free slots where each one is used by vehicles as its Basic Channel (BCH) for collision-free transmission of the safety messages or other control messages. The CRP is used to coordinate resource allocation on the SCHs using CSMA/CA method. The SCHI which is called Non-Safety Application Frame (NSAF) is used for non-safety data transmissions. Fig. 10 demonstrates the timing architecture of the DMMAC scheme. Please cite this article as: N. Torabi, B.S. Ghahfarokhi, Survey of medium access control schemes for inter-vehicle communications, Computers and Electrical Engineering (2017), http://dx.doi.org/10.1016/j.compeleceng.2017.02.022

JID: CAEE

ARTICLE IN PRESS N. Torabi, B.S. Ghahfarokhi / Computers and Electrical Engineering 000 (2017) 1–23

[m3Gsc;March 3, 2017;13:54] 13

Fig. 9. The operation of CREM/CREM-E schemes. During CCHI, primary providers do not have to return to the CCH for a new contention once they are involving in a data transmission, while the secondary providers are obligated to return to the CCH to participate in a new contention, even they are involving in a data transmission [7].

Fig. 10. Timing architecture of DMMAC [8].

In DMMAC, several rules have been defined to managing the access behaviors in the ABF including: how a BCH slot is reserved in a distributed manner, how the length of ABF is adjusted, and when virtual slots are added into the end of the ABF to avoid collisions. In this scheme, a newly arrived node determines the unused slots in the ABF based on the report messages received from its neighbors which contain slot allocation maps of one-hop neighbors and tries to pick up a BCH considering some defined rules. The vehicles do not necessarily experience the same length of the ABF. So, due to the fixed length of CCHI, the length of the CRP depends on the ABF length of the vehicle. Therefore, vehicles may enter the CRP asynchronously. To avoid potential collisions, some vehicles may have several additional slots named Virtual Slots after entering the CRP. Moreover, each vehicle can adjust its ABF length according to slot allocation of its two-hop neighbors using Active Re-Reservation (ARR) process. In this way, the BCH slots may be freed up, so other vehicles have the chance to pick them up. During CRP, the way in which vehicles negotiate to select NSAF resources is described by some rules, where a three-way handshake mechanism is used to complete one negotiation. Even though the freeing up mechanism of DMMAC adds more length to the CRP, it cannot improve the bandwidth utilization. Because the length of the NSAF is still the same. Additionally, the fixed length of the NSAF may lead to poor bandwidth utilization. Moreover, it has not been clarified that what happens if the length of the ABF grows unintentionally such that Please cite this article as: N. Torabi, B.S. Ghahfarokhi, Survey of medium access control schemes for inter-vehicle communications, Computers and Electrical Engineering (2017), http://dx.doi.org/10.1016/j.compeleceng.2017.02.022

JID: CAEE 14

ARTICLE IN PRESS

[m3Gsc;March 3, 2017;13:54]

N. Torabi, B.S. Ghahfarokhi / Computers and Electrical Engineering 000 (2017) 1–23

Fig. 11. Division of SCHs into frames and slots and CCH into frames, slots, and mini-slots in the TC-MAC [9].

Fig. 12. Basic operation of AMCMAC-D protocol, assuming nodes (n0, n1), (n2, n3), and (n4, n5) are negotiating on CCH and communicating on SCHs [10].

no more time remains for the CRP. And, since the NSAF resource management has not been described, the performance of the scheme in case of channel utilization is not clear. Almalag et al. [9] have suggested a TDMA cluster-based MAC called TC-MAC for V2V communications. In this scheme, a traffic flow based algorithm is used to creating stable clusters. A pure TDMA based slot reservation method is used to share resources among vehicles within the clusters. The scheme aims at achieving high bandwidth utilization, preserving fairness, and maintaining reliability of safety messages. The protocol takes the advantage of multiple channels of DSRC. In this scheme, the collision-free one-hop intra-cluster communications are managed by the Cluster-Head (CH). Vehicles are assumed to be synchronized using GPS. Access time on all the seven channels is divided into consecutive non-overlapping logical time frames. Each time frame is further divided into equal-sized time slots. The length of a time frame is based on the number of vehicles in the cluster and will remain unchanged as long as the number of vehicles does not change. The CH is responsible for updating the number of active vehicles in the cluster. Time slots on the CCH are also partitioned into a number of mini-slots (equal to the number of SCHs). Fig. 11 depicts the time structure of the CCH and SCHs in TC-MAC. Vehicles are allocated a mini-slot on the CCH and a transmission time slot on one of the SCHs just after the time slot which the mini-slot resides. Each vehicle uses simple mathematical calculations to compute its transmission time slot and mini-slot in the frame. The main advantage of TC-MAC is its usage of SCHs during CCHI. This will result in improved bandwidth utilization. However, due to the slot reservation mechanism used in the protocol, unused transmission slots are unavoidable. So, the scheme can be the source of the poor bandwidth utilization, in itself. Also, although all the nodes are allocated equally, their transmission requirements are neglected. Furthermore, since the CH is involved in all the communications within the cluster, transmissions may experience delay. Last but not least, inter-cluster communications are not considered in the scheme. An asynchronous multi-channel MAC with a distributed TDMA mechanism, known as AMCMAC-D has been proposed by Han et al. [10]. As the name suggests, while the scheme relies on a single-radio transceiver per each vehicle, in contrast to synchronous approaches no strict synchronous channel switching from CCH to SCHs is required. The protocol aims at tackling the Missing Receiver Problem (MRP) which is a multichannel MAC-specific problem caused by an asynchronous receiver that is busy on a second channel, either receiving or transmitting or just listening on, while a sender is trying to send packets to it on a different channel. The AMCMAC-D solves the MRP in a manner which reduces the waiting time and decreases the collision rate on the CCH. A basic operation of the AMCMAC-D protocol is shown in Fig. 12. The AMCMAC-D scheme also intends to improve the bandwidth utilization of SCHs via addressing the multi-channel hidden node problem which occurs when a node is listening on a particular channel, while it cannot hear the communication taking place on a different channel. As well, it solves the problem of missing emergency messages through rebroadcasting Please cite this article as: N. Torabi, B.S. Ghahfarokhi, Survey of medium access control schemes for inter-vehicle communications, Computers and Electrical Engineering (2017), http://dx.doi.org/10.1016/j.compeleceng.2017.02.022

JID: CAEE

ARTICLE IN PRESS

[m3Gsc;March 3, 2017;13:54]

N. Torabi, B.S. Ghahfarokhi / Computers and Electrical Engineering 000 (2017) 1–23

15

Fig. 13. The operation of VEMMAC protocol. Nodes A, B, C, and D are communicating on CCH, SCH-1, and SCH-2 [12].

of emergency messages by nodes that have received the message. The scheme also balances the distribution of SCHs among vehicles, using a load-balancing channel selection mechanism. Moreover, a distributed TDMA algorithm is used by the protocol to reduce the high contention level on the CCH in large-scale networks which also results in improved bandwidth utilization of SCHs. Service differentiation is also enhanced via definition of Access Categories (ACs). Although the AMCMAC-D tries to not to miss any safety-related messages, there is still the probability that the emergency messages are not received by those destination nodes that are involved in another ongoing communication on an SCH. Even, rebroadcasting of the emergency messages does not completely resolve the problem. Moreover, the multi-channel hidden node problem in multi-hop communications is still remained unsolved. Furthermore, since the load-balancing mechanism is based on the random selection of the SCHs, it is not always possible to have the best SCH selection, while if a load-aware mechanism is employed the result will be more appropriate. Almalag et al. [11] have offered a modified version of TC-MAC specifically for multi-hop intra-cluster communications. Unlike TC-MAC, time frames have fixed length of 100 ms. The number of vehicles in a cluster is limited and so the number of TDMA slots is also fixed. A multi-hop intra-cluster communication is performed via the CH where safety and governance messages to nearby cluster members are disseminated. However, some vehicles are selected as the relay for those cluster members that are out of the range of the CH. Also, for the sake of reliability, the CH will repeat the same safety message in any available mini-slot on the CCH of the same TDMA frame. Unicast multi-hop intra-cluster communications are also accomplished through relay nodes. Negotiations with cluster members are done on the CCH to find a path between sender and receiver. However, this scheme encounters with several shortcomings: Firstly, limiting the maximum number of vehicles in each cluster is a challenging issue in practice. Secondly, due to its impact on the number of available transmission slots, it can lead to bandwidth deficiency, especially when the cluster is sparse. Moreover, finding a relay node can impose more delay when the density of vehicles is varying or when vehicles are moving in opposite directions or with different speeds. Dang et al. [12] proposed a multi-channel MAC, named as VEMMAC, inspiring from the alternating nature of the IEEE 1609.4 protocol. The scheme intends to increase the reliability of safety messages and to enhance bandwidth utilization. To this end, in contrast to the IEEE 1609.4 protocol, the VEMMAC allows nodes to broadcast safety messages twice, on both CCHI and SCHI in order to improve the reliability of safety messages. Besides, it allows transmission of non-safety messages on SCHs during CCHI, in order to improve the bandwidth utilization of SCHs. All vehicles are equipped with only one half-duplex transceiver. Two modes of transmissions, i.e. Normal Transmission (N-Tx) and Extended Transmission (E-Tx) are allowed. The N-Tx deals with the transmissions performed only within SCHI and E-Tx deals with transmissions performed within both SCHI and the upcoming CCHI. Two lists of information, namely Channel Usage List (CUL) and Neighbor Information List (NIL) are used to keeping the track of the channels and neighbor nodes respectively. This information helps the sender to find out the receiver’s status in the next interval and to make an agreement with the receiver on an available SCH which has the minimum density. In the VEMMAC, as depicted in Fig. 13, whenever a node wants to broadcast a safety message it tunes into the CCH and participate in a new competition, then it tries to rebroadcast the message in the next SCHI or CCHI. Whenever a node has non-safety messages to send, it checks the status of the receiver on the next SI based on its NIL. If the receiver is not on the CCH or if the sender could not find a match in its CUL with receiver’s channel in NIL, the sender has to wait and try in the next SI. Once one or more available SCH(s) are found, an SCH-REQ is sent to the receiver including the CUL and the mode of transmission (N-Tx or E-Tx). Then, the receiver selects the best SCH and sends SCH-ACK to the sender to indicate the selected SCH. In response, the sender sends SCH-RES to confirm the SCH selected by the receiver. Neighbor nodes that overhear the SCH-ACK or SCH-RES messages update their NILs and CULs. After the CCHI, the sender and receiver switch to the agreed SCH and start their data transmissions. Please cite this article as: N. Torabi, B.S. Ghahfarokhi, Survey of medium access control schemes for inter-vehicle communications, Computers and Electrical Engineering (2017), http://dx.doi.org/10.1016/j.compeleceng.2017.02.022

JID: CAEE 16

ARTICLE IN PRESS

[m3Gsc;March 3, 2017;13:54]

N. Torabi, B.S. Ghahfarokhi / Computers and Electrical Engineering 000 (2017) 1–23

Fig. 14. The structure of a Chip in CS-TDMA scheme [13].

Since every safety message is broadcasted twice, the VEMMAC scheme doubles the number of contentions for transmission of a single safety message. So the probability of successful transmission of the safety messages gets higher in comparison to the IEEE 1609.4. However, the contention-based approach for transmission of safety messages still causes extensive delay or even expiration, especially in dense network. Moreover, when a safety message is broadcasted in SCHI, vehicles which are in the E-Tx transmission mode do not hear the message, even when the message is rebroadcasted in CCHI of next SI. A scalable CSMA and self-organizing TDMA MAC protocol called CS-TDMA has been proposed by Zhang et al. [13]. The protocol combines CSMA with TDMA to improve the broadcast performance in VANETs. It also takes the advantage of SDMA region-based clustering method to build independent zones. The scheme adjusts the dwelling time ratio between CCHI and SCHI dynamically according to the traffic density to achieve reliability of safety messages and efficiency of bandwidth. In this scheme, the CCH is partitioned into time intervals, named as Chips which are composed of a Transmission (TS) period and a Reservation (RS) period, as demonstrated in Fig. 14. The time-slotted TS period contains a number of time slots that are used for transmission of safety and control messages. The CSMA-based RS period is used for reservation of time slots in the TS period. Information about the current status of the slots and some other information, known as Chip Information (CI) are carried by each transmitted packet. The CS-TDMA scheme assumes that clusters are associated with non-overlapping frequency subcarriers and subcarrier allocation maps are pre-installed in each vehicle. Also, the maximum number of vehicles in a cluster is always determined to be less than the maximum number of TS slots. A newly arrived vehicle that needs to acquire a TS time slot, starts listening to the CCH for one Chip period. If the vehicle receives a Hello-New message, it is interpreted that no vehicle in the cluster has a reserved TS slot. Hence, the Chip only consists of the RS period. Then, the vehicle transmits a Hello-New message in the RS period and updates its CI based on Hello-New messages received from other vehicles. However, if the listening time expires before the vehicle receives any message, it is interpreted that there is no other vehicle in the cluster or the number of TS slots has reached the maximum. If the reason is the latter, all the vehicles have to release the reserved slots, so the system goes back to the initial state. Otherwise, each vehicle attempts to reserve a TS slot via broadcasting a reservationrequest packet in the RS period. To determine whether the reservation is successful, the vehicle needs to wait for a Chip interval. In the case of multiple reservations for a single TS slot or re-arrangement of slots, special flags are used in the CI. In order to comply with the DSRC, time frames in CS-TDMA have constant 100 ms length. However, when the traffic density is low, the duration of CCHI is reduced to leave more time for SCHI to improve the throughput. On the other side, when the traffic density is high, the duration of the CCHI is extended to guarantee the safety message transmission by alleviating the high collision problem of CSMA. Despite such a good property, the throughput of SCHs still may go down when the density of vehicles goes up. This is due to the fact that the scheme decreases the duration of SCHI under high density conditions, so the total time that the SCHs are utilized is reduced. Furthermore, the scheme does not clarify that how traffic density is determined in a distributed manner. Moreover, the scheme does not mention the structure of SCHs and how vehicles can use them, and more importantly how inter-cluster communications are taking place. Dang et al. [14] proposed hybrid efficient and reliable MAC (HER-MAC) scheme for two-hop V2V communications which combines both TDMA and CSMA schemes. It provides reliability for transmission of safety messages through the TDMA access and a retransmission mechanism. Moreover, it allows utilizing the SCH resources during the CCHI for transmission of non-safety messages to improve the bandwidth utilization. In this scheme, vehicles use one half-duplex transceiver and they are synchronized using GPS. As illustrated in Fig. 15, time is divided into sequential 50 ms long Sync Intervals (SIs). On the CCH, each SI is further divided into Reservation Period (RP) and Contention Period (CP) divisions. The RP includes variable numbers of constant-length Emergency Slots (EmgSlots) which are used for transmission of safety messages. The CP is used to reserve an EmgSlot or to exchange other control messages. On each SCH, the SI is divided into certain numbers of constant-length Service transmission Slots (SerSlots) for transmission of non-safety message. In HER-MAC, each node broadcasts Hello messages in its EmgSlot containing the status of the one-hop neighbors’ EmgSlots. Each node has to listen to whole RP to collect the status of all EmgSlots of its two-hop neighbors before it can reserve an available EmgSlot. A Switch message is used by vehicles to switch to an available EmgSlot prior to their current EmgSlot in order to minimize the length of the RP based on the changes in the network topology. For the sake of more reliability, each node broadcasts its safety message twice in two successive EmgSlots during two continuous SIs. Also, to exchange Please cite this article as: N. Torabi, B.S. Ghahfarokhi, Survey of medium access control schemes for inter-vehicle communications, Computers and Electrical Engineering (2017), http://dx.doi.org/10.1016/j.compeleceng.2017.02.022

JID: CAEE

ARTICLE IN PRESS N. Torabi, B.S. Ghahfarokhi / Computers and Electrical Engineering 000 (2017) 1–23

[m3Gsc;March 3, 2017;13:54] 17

Fig. 15. Time structure of HER-MAC [14].

Fig. 16. Timing structure of MCTRP [16].

non-safety information, a 3-way handshake is performed on the CCH during CP in order to select a SerSlot on an agreed SCH. In order to avoid hidden node problems, a Frame Information Map (FIM) mechanism is used by vehicles to store the status of EmgSlots occupied by all its two-hop neighbor nodes, so they can choose an EmgSlot with no probable collisions. Chang et al. [15] introduced a cluster-based multi-channel V2V MAC scheme known as Earliest Deadline First based Carrier Sense Multiple Access (EDF-CSMA) in order to remedy the problem of unpredictability in channel access. It tries to avoid access collisions through dynamic adjustment of priority of real-time traffic. As well, to provide guaranteed QoS in multi-channel environments, it introduces an admission control policy according to pre-defined time constraints. The EDF-CSMA scheme has two main components namely WAVE Service Group (WSG) configuration and multi-channel access mechanism, where the former constructs the network for data transmission, while the latter focuses on QoS issues and provides better bandwidth utilization. WSG configuration component constitutes a clustering mechanism based on relative position of vehicles to construct clusters containing one Group Header (GH) and multiple Group Members (GMs). Multichannel access mechanism ensures QoS for both real-time safety information and real-time non-safety traffic by focusing on transmission delay. It also organizes the radio resources of each channel to achieve higher bandwidth utilization. The EDFCSMA scheme has been evaluated through both theoretical analysis and simulations. The analysis shows that the scheme has greater channel utilization than EDCA and outperforms EDCA in terms of throughput, mean packet delay, and loss rate. However, the scheme suffers from fair sharing of radio resources among reserved and non-reserved GMs. It imposes a rather large signaling overhead, as well. Furthermore, it does not support intercluster and multi-hop intra-cluster communications. 4.3. Multiple-radio/multiple-channel MAC protocols A DSRC-based multi-channel token ring MAC protocol called MCTRP was presented by Bi et al. [16] for V2V communications in order to provide contention-free and delay-bounded transmissions for safety messages and to improve network throughput for non-safety applications. The MCTRP is a two-radio hybrid MAC scheme where a token-passing CSMA/CAbased technique is used for transmission of intra-ring data, emergency, and coordination messages using one radio, while the CSMA/CA method is used for transmission of inter-ring data, emergency, and ring administration messages using the other radio. It organizes nodes with similar velocities into token-passing rings, each with a founder-leader node. Vehicles are synchronized using GPS and time is partitioned into fixed time periods composed of a control period and a data period, which is further divided into safety period, ring coordination period, and data exchange period, as depicted in Fig. 16. The protocol includes three main sub-protocols: (i) ring coordination protocol which is designed for ring management, (ii) emergency message exchange protocol which is used for collecting emergency messages in a ring and delivering them to other rings, and (iii) data exchange protocol which controls the token delivery in a ring. The MCTRP delivers emergency messages as fast and reliable as possible. It also provides fairness among vehicles through channel sharing and adaptation of token holding time. However, relying on continuous connectivity between all nodes within the ring in order to pass the tokens is somehow unrealistic, considering the highly dynamic environment in VANETs. Moreover, using CSMA/CA method as the access method for transmission of safety messages outside the ring causes extended delay. A TDMA-based multi-channel MAC protocol for reliable broadcast, namely VeMAC was presented by Omar et al. [17]. The scheme aims at reducing collisions owing to the high mobility of nodes. It is based on a fully TDMA scheme wherein time in all the channels is partitioned into fixed-length time frames consisting of a constant number of fixed-length time Please cite this article as: N. Torabi, B.S. Ghahfarokhi, Survey of medium access control schemes for inter-vehicle communications, Computers and Electrical Engineering (2017), http://dx.doi.org/10.1016/j.compeleceng.2017.02.022

JID: CAEE 18

ARTICLE IN PRESS

[m3Gsc;March 3, 2017;13:54]

N. Torabi, B.S. Ghahfarokhi / Computers and Electrical Engineering 000 (2017) 1–23

Fig. 17. Division of time frames on the CCH into R, L, and F sets in VeMAC [17].

Fig. 18. Structure of a time frame in MAC scheme of Yang et al. [18].

slots. As depicted in Fig. 17, on the control channel, time slots in each frame are divided into three different sets, i.e. L, R, and F, and are allocated to vehicles (in left and right directions) and RSUs, respectively. Other channels also have an integer number of frames, including a fixed number of slots per frame. Each vehicle uses two transceivers with fixed power levels on all channels and all the vehicles are supposed to be synchronized using GPS. The first transceiver is tuned in the CCH while the second one is tuned in one of the SCHs. In VeMAC, based on the directional information achieved from GPS, each time slot in the frame will be assigned to just one vehicle. Each vehicle is allowed to access the CCH only once per frame. Also, in order to avoid hidden node problem, a vehicle inserts required information in the header of its packets transmitted on the CCH. Given that VeMAC uses two radios per each vehicular node, it can resolve many problems which a single-radio MAC scheme encounters with. Moreover, by allowing vehicles which are running out of their time slots to use slots from other sets, it improves the bandwidth utilization problem which many TDMA-based schemes suffer from. However, large signaling overhead of VeMAC impacts on its performance negatively. On the other side, considering the way that a vehicle takes to access a time slot on the CCH, delayed transmission of critical messages is probable especially when the network is crowded. A multi-channel clustering-based cooperative MAC protocol for V2V scenarios has been presented by Yang et al. [18] which is similar to CAH-MAC [4]. The scheme is based on two-hop clusters wherein a vehicle is selected as the Cluster Head (CH) in each cluster. Vehicles are synced using GPS. In each cluster, one of the six SCHs is selected by the CH for intra-cluster communications, while the CCH is used for exchanging intra-cluster control messages and inter-cluster safety and non-safety messages. On the CCH and the selected SCH, time is divided into equal-size consecutive time frames. Each frame on the CCH is further partitioned into two CSMA periods; the first period is for exchanging safety and control messages, while the second one is for exchanging non-safety information. Meanwhile, each time frame on the selected SCH is divided into TDMA and CSMA periods. A time slot is assigned to each vehicle in TDMA period to transmit intra-cluster safety messages. The CSMA period is used for transmission of non-safety information. In This scheme, each vehicle is equipped with two radios. As shown in Fig. 18, the radio I is always tuned in the CCH, while the Radio II is tuned in one of the SCHs. The scheme is composed of three procedures: cluster coordination, safety message delivery, and non-safety message delivery. The cluster coordination procedure defines how a CH is elected, how a vehicle joins the cluster, what happens if a vehicle temporarily loses its connection to the cluster, how a vehicle leaves the cluster, and how two clusters merge. Safety message delivery procedure describes that how emergency packets are delivered to the cluster members or to clusters. Moreover, it defines that how helper nodes are selected to help for delivering safety messages to the CH when the channel quality is poor. The non-safety delivery procedure indicates that how a vehicle transmits non-safety messages. While the scheme has been designed elaborately, it suffers from several defects. First of all, using only one SCH per each cluster results in poor bandwidth utilization. Moreover, the scheme does not consider enough parameters to choose the helper node; it only uses the SNR of the received signal. In addition, the protocol does not explain how time slots are assigned to the vehicles on the TDMA period of the SCH. A cognitive radio-enabled vehicular communication scheme, named as CRAVE was proposed by Rawat et al. [19] to be used in heterogeneous wireless systems in order to achieve better channel utilization, higher throughput, and fairness between vehicular users. The scheme assumes that each vehicle is equipped with two radios; one for spectrum sensing and another for the exchange of VANET messages. Vehicular users are considered as secondary users and are capable of sensing, analyzing and accessing the spectrum opportunities within Wi-Fi, WiMAX or cellular networks dynamically. Spectrum ocPlease cite this article as: N. Torabi, B.S. Ghahfarokhi, Survey of medium access control schemes for inter-vehicle communications, Computers and Electrical Engineering (2017), http://dx.doi.org/10.1016/j.compeleceng.2017.02.022

ARTICLE IN PRESS

JID: CAEE

[m3Gsc;March 3, 2017;13:54]

N. Torabi, B.S. Ghahfarokhi / Computers and Electrical Engineering 000 (2017) 1–23

19

Table 3 Investigations of surveyed MAC protocols (“Yes”: the protocol meets the criterion. “No”: the protocol does not meet the criterion, and “N/A”: it is not obviously declared in the research if the protocol meets the criterion or does not). Investigations are mostly based on the statements declared in each reference, while, in some cases, inclusive deductions are also superimposed. Protocol

STDMA [3] CAH-MAC [4] ASAS [5] VMMAC [6] CREM [7] DMMAC [8] TC-MAC [9] AMCMAC-D [10] Almalag et al. [11] VEMMAC [12] CS-TDMA [13] HER-MAC [14] EDF-CSMA [15] MCTRP [16] VeMAC [17] Yang et al. [18] CRAVE [19]

Criterion DSRC channel map; one CCH and six SCHs

Single radio per device

CSMA/CA support

EDCA support

Multi-channel support

Time constraint associated with safety messages (<100 ms)

No No No Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes

Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes No No No No

Yes No Yes Yes Yes Yes No Yes No Yes Yes Yes Yes Yes No Yes Yes

N/A No N/A N/A Yes N/A No Yes No N/A N/A N/A Yes N/A No N/A N/A

No No No Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes N/A

N/A N/A N/A Yes Yes Yes Yes N/A Yes Yes Yes Yes Yes Yes Yes Yes N/A

cupancy is categorized as idle, occupied by licensed users, or occupied by vehicular users. When a channel is idle (State I), vehicular users can use a given channel of a given network. Similarly, when a channel is occupied by secondary/vehicular users (State II), other vehicular users are allowed to stay and compete for the channel. However, if the channel is actually occupied by the primary system (State III), vehicular users are prohibited from the transmission. In the case of State II, multi-access schemes such as CSMA/CA, TDMA, FDMA, etc. can be adopted. While the simulation results show the acceptable performance of the CRAVE scheme in terms of throughput and fairness, it suffers from some shortcomings. Firstly, the model uses two radios per vehicle which results in high implementation costs. Secondly, it does not cover the hidden and spread spectrum primary users, especially when there are different users from several types of networks. Moreover, the authors do not declare exactly how the proposed scheme can adopt multi-channel operation. 5. Comparative summary In this section, we give a qualitative discussion of investigated protocols from standard compatibility and QoS points of view. MAC schemes mostly make slight modifications to the IEEE 1609.4/IEEE 802.11p standards, whereas they sometimes take a completely different approach from the standard. We use six distinct criteria to investigate the compatibility of MAC schemes with the IEEE 802.11p and the IEEE 1609.4 standards, including matching with the DSRC specification, using only one radio per each vehicular node, adopting CSMA/CA mechanism as the channel access scheme, and supporting EDCA as the mechanism to prioritize different types of traffic. Furthermore, multi-channel support and preserving time constraint associated with the safety applications are suggested as two important factors that is adopted mostly by MAC schemes. It is noteworthy that the last criterion may be considered as a QoS-related one, but it is also a feature of the MAC itself and so is considered as a design parameter too. Table 3 summaries the investigations of the studied researches regarding these criteria. From QoS and performance points of view, some of the MAC schemes focus on bandwidth utilization of service channels, while some others have turned their attention into the reliability of emergency messages. Many others take timely delivery of critical information into account and provide more convenient and fast transmission of these messages. In addition, some of them address the issue of network topology and its dynamicity. Meanwhile, there exist protocols which care about the level of management and coordination required for accessing the shared channels. It is noteworthy that a few protocols have considered the issue of fairness among their main objectives. Table 4 summarizes the most remarkable QoS/performancerelated features of the surveyed protocols. 6. Lessons learned and open challenges Meeting the stringent requirements of safety and non-safety applications requires a MAC scheme to provide fair channel access, to allow for efficient and fast delivery of packets, to avoid packet losses, and to offer secure communications as much as possible. In this paper, we presented an overview of MAC design approaches for V2V communication and investigated a bundle of recently proposed MAC schemes. Although the literature is rich of V2V MAC schemes, there still remains a number of MAC research challenges that must be addressed. Please cite this article as: N. Torabi, B.S. Ghahfarokhi, Survey of medium access control schemes for inter-vehicle communications, Computers and Electrical Engineering (2017), http://dx.doi.org/10.1016/j.compeleceng.2017.02.022

ARTICLE IN PRESS

JID: CAEE 20

[m3Gsc;March 3, 2017;13:54]

N. Torabi, B.S. Ghahfarokhi / Computers and Electrical Engineering 000 (2017) 1–23

Table 4 QoS/Performance-related comparison of the surveyed MAC protocols discussed in Section 4. Protocol

QoS/Performance-related feature

STDMA [3] CAH-MAC [4] ASAS [5] VMMAC [6] CREM [7] DMMAC [8] TC-MAC [9] AMCMAC-D [10] Almalag et al. [11] VEMMAC [12] CS-TDMA [13] HER-MAC [14] EDF-CSMA [15] MCTRP [16] VeMAC [17] Yang et al. [18] CRAVE [19]

Utilization of service channels

Delivery latency of critical messages

Reliability of critical messages

Scalability

Overhead (required management messages)

Fairness

Not considered Considered Considered Considered Considered Not considered Considered Not considered Considered Considered Considered Considered Considered Considered Considered Not considered Considered

Not considered Not considered Not considered Considered Considered Considered Considered Not considered Considered Considered Not considered Considered Considered Considered Considered Not considered Not considered

Considered Considered Considered Considered Considered Considered Not considered Not considered Not considered Considered Considered Considered Considered Considered Considered Considered Not considered

Not considered Considered Considered Not considered Considered Not considered Considered Considered Considered Considered Considered Not considered considered Considered Not considered Considered Not considered

Fairly Fairly Fair High Fairly Fairly Low Fair Low Fairly Fair Fairly Fairly Fairly Fairly Fairly Low

Not considered Not considered Absolute fairness Not considered Not considered Absolute fairness Absolute fairness Not considered Absolute fairness Not considered Absolute fairness Not considered proportional fairness Absolute fairness Absolute fairness Not considered Absolute fairness

high high

low low

high high high high high low

Fig. 19. The pie chart showing the percent of approaches used by recently proposed MAC schemes that reviewed in this paper.

The pie chart in Fig. 19 demonstrates that a considerable percent of the recently proposed V2V MAC schemes which have been reviewed in this paper employ a single-radio/multiple-channel approach, which is the preferable approach by the DSRC/WAVE standard. It denotes the interest among researchers to develop standard compatible MAC schemes. However, compatibility with the standard, may not imply any predominance for the MAC scheme by itself, rather it indicates that how a MAC protocol can get the chance of being welcomed by the research community or even the market. MAC schemes based on single-radio devices have the advantage of simplicity and availability at low cost, hence they can be considered as the primary option for near future deployment of VANET. Nonetheless, issues regarding synchronization, channel switching, and bandwidth utilization need additional considerations. Also, the dynamic of VANETs is a major restriction that has not been attended adequately, especially in clustering-based methods. More intelligent single-radio MAC approaches that address the mobility of vehicular nodes is an open issue for future studies. The schemes that are based on multi-radio devices can resolve most of the issues concerning the single-radio devices. They can provide easier and instant switching between multiple channels and utilize the bandwidth more appropriately. However, missing receiver and multi-channel hidden node problems (due to asynchronous time structure) still remain cumbersome, since it is not always technically or economically practical to provide the same number of radios as the number of channels. Despite favorable merits beyond the multi-channel schemes such as reduction in access collisions and delay, there is always a risk of undesired utilization of bandwidth, especially when using single-radio transceivers. To overwhelm the bandwidth under-utilization in multi-channel schemes, the split phase approach is supposed to be the most desirable solution, as it is used by the IEEE 1609.4 standard. However, it is of high significance that the multi-channel operation itself should not waste the bandwidth or compromise transmission of critical delay-sensitive information. Leaving service channels unused during control interval and unbounded length of time frames are examples of such inefficiencies, respectively. It is argued that a single-radio multi-channel approach with an efficient multi-channel operation procedure can be a suitable option for being used in a V2V channel access scheme. Therefore, as an open issue, more researches are required in this regard. Regarding the multi-channel MAC scheme utilizing a split phase single-radio approach, it is usual to select between a purely competitive, a pure non-competitive, or a hybrid channel access approach. The approach taken by the IEEE 1609.4 standard is a hybrid approach, where time is divided into two sequential equal-sized CSMA/CA-based control and data intervals. Generally, hybrid approaches give better chances to the MAC designer to offer more useful and efficient MAC schemes. However, such schemes may have side effects. Referring to the previously discussed MAC protocols, it can be deduced that most of them require strict synchronization between vehicles. This requirement is often dispelled by utilizing Please cite this article as: N. Torabi, B.S. Ghahfarokhi, Survey of medium access control schemes for inter-vehicle communications, Computers and Electrical Engineering (2017), http://dx.doi.org/10.1016/j.compeleceng.2017.02.022

JID: CAEE

ARTICLE IN PRESS N. Torabi, B.S. Ghahfarokhi / Computers and Electrical Engineering 000 (2017) 1–23

[m3Gsc;March 3, 2017;13:54] 21

additional technologies like GPS, and if not, it exposes exchanging a lot of control messages. Unfortunately, this may lead to poor performance especially when dealing with a large dynamic vehicular network. Speaking of large dynamic networks, some of the MAC schemes surveyed in this paper try to manage quick changes in the network topology through clustering. Clustering methods can help to mitigate the effect of hidden/exposed vehicular nodes and improve QoS-related requirements through addressing the intra- and inter-cluster communications, separately. The SDMA-based schemes perform in the same way, except they usually need an electronic map of the region to determine the zones and making them more sophisticated to implement. As well, there are schemes based on directional antennas which make use of geographical specifications to deliver data. Although the use of such methods has been common in V2V MAC schemes, there are specific problems with them like prerequisite setups, dependency on further technologies and hence additional overhead. They also bring about subsidiary difficulties such as probable merging collisions or signal conflictions. The issue of multi-hop communications in clustering-based MAC schemes which is addressed by some cooperative protocols like modified version of TC-MAC [11], CAH-MAC [4], and MAC scheme of Yang et al. [18], still needs further considerations. Selection of relay (helper) node(s) is among the most considerable problems in this context. The criteria for selection of relay node(s) may have great impacts on the performance of the MAC protocol in case of reliability and delay of safety messages, and also the interference with other ongoing transmissions. The prediction methods for foreseeing the future geographical location and non-geographical situation (e.g. signal strength) of the relay node(s) could be beneficial for faster and more reliable transmission of safety messages. While on the subject, it is also worthy to mention that the way which is used for clustering of vehicles into multi-hop clusters may have serious side-effects on delivery delay of safety messages and throughput of non-safety messages; in a crowded cluster composing of vehicles mostly wistful on transmitting data, a split phase MAC scheme should compromise each of the safety or non-safety applications in favor of another. Moreover, simultaneous attempt of two (or more) vehicular nodes belonging to different neighboring clusters (to acquire the same channel at the same time), results in the access collisions which are a factor of additional delay and reduced throughput or even data loss. Moreover, resource allocation always demonstrates a tradeoff between bandwidth utilization and fairness. For example, to bring fairness to the vehicles via TDMA approach (wherein time slots are assigned according to vehicle’s requirements), fully satisfaction of vehicles may result in compromising the bandwidth utilization, since vehicles have a various amount of bandwidth requirements. With the same level of importance, the tradeoff between delay and bandwidth utilization should also be considered as the matter of subject by MACs. The restriction of the dedicated spectrum and the growing demand for supporting multimedia and entertainment traffic necessitate more attentions to opportunistic access in VANETs. As a result, the cognitive radio based MAC schemes need more attention in future since current approaches are not efficient enough. Using predictive schemes for spectrum sensing is a suggestion that could improve the performance of current approaches. Last but not the least, security challenges have still remained as an unsolved problem since there is no perfect proposal that satisfies both security and QoS requirements of a channel access scheme. 7. Conclusion In this article, we overviewed the V2V MAC design approaches and discussed their pros and cons. We also focused on recently proposed V2V MAC schemes and gave a detailed review of each alongside their strengths and drawbacks. Furthermore, we provided a qualitative performance comparison between discussed schemes and disclosed their compatibility with current WAVE standard features. We also went through MAC-related challenging issues regarding previously proposed schemes, as well as issues that have not been addressed yet by the researchers. We conclude that despite numerous existing V2V MAC schemes, there is still a lack of thorough solution for inter-vehicle communications which can provide efficient, fast, fair, and reliable transmission of critical and infotainment information. As well, we argue that the current WAVE MAC standards need further considerations, owing to their poor performance in case of bandwidth utilization, delay, fairness, reliability, and also security. References [1] Leng S, Fu H, Wang Q, Zhang Y. Medium access control in vehicular ad hoc networks. Wireless Commun Mobile Comput 2011;11:796–812. [2] Hang S, Xi Z. Clustering-based multichannel MAC protocols for QoS provisionings over vehicular ad hoc networks. IEEE Trans Veh Technol 2007;56:3309–23. [3] Rezazade L, Aghdasi HS, Ghorashi SA, Abbaspour M. A novel STDMA MAC protocol for vehicular ad-hoc networks. In: International symposium on computer networks and distributed systems; 2011. p. 148–51. [4] Bharati S, Weihua Z. CAH-MAC: cooperative ADHOC MAC for vehicular networks. IEEE J. Sel. Areas Commun. 2013;31:470–9. [5] Hadded M, Zagrouba R, Laouiti A, Muhlethaler P, Saïdane L. An adaptive TDMA slot assignment strategy in vehicular ad hoc networks. J. Mach. Mach. Commun. 2014;1:175–94. [6] Xu X, Benxiong H, Shaoshi Y, Tiejun L. Adaptive multi-channel MAC protocol for dense VANET with directional antennas. In: 6th IEEE consumer communications and networking conference; 2009. p. 1–5. [7] Jui-Hung C, Kai-Ten F, Chuah C-N, Chin-Fu L. Cognitive radio enabled multi-channel access for vehicular communications. In: Vehicular technology conference fall; 2010. p. 1–5. [8] Ning L, Yusheng J, Fuqiang L, Xinhong W. A Dedicated multi-channel MAC protocol design for VANET with adaptive broadcasting. In: IEEE wireless communications and networking conference; 2010. p. 1–6. [9] Almalag MS, Olariu S, Weigle MC. TDMA cluster-based MAC for VANETs (TC-MAC). In: IEEE international symposium on a World of wireless, mobile and multimedia networks; 2012. p. 1–6.

Please cite this article as: N. Torabi, B.S. Ghahfarokhi, Survey of medium access control schemes for inter-vehicle communications, Computers and Electrical Engineering (2017), http://dx.doi.org/10.1016/j.compeleceng.2017.02.022

JID: CAEE 22

ARTICLE IN PRESS

[m3Gsc;March 3, 2017;13:54]

N. Torabi, B.S. Ghahfarokhi / Computers and Electrical Engineering 000 (2017) 1–23

[10] Chong H, Dianati M, Tafazolli R, Xing L, Xuemin S. A novel distributed asynchronous multichannel MAC scheme for large-scale vehicular ad hoc networks. IEEE Trans Veh Technol 2012;61:3125–38. [11] Almalag MS, El-Tawab S, Olariu S, Weigle MC. A modified TC-MAC protocol for multi-hop cluster communications in VANETs. In: International conference on connected vehicles and expo; 2013. p. 832–7. [12] Dang DNM, Dang HN, Do CT, Hong CS. An enhanced multi-channel MAC for vehicular ad hoc networks. In: IEEE wireless communications and networking conference; 2013. p. 351–5. [13] Zhang L, Liu Z, Zou R, Guo J, Liu Y. A scalable CSMA and self-organizing TDMA MAC for IEEE 802.11 p/1609.x in VANETs. Wireless Pers Commun 2014;74:1197–212. [14] Dang DNM, Dang HN, Nguyen V, Htike Z, Hong CS. HER-MAC: a hybrid efficient and reliable MAC for vehicular ad hoc networks. In: IEEE international conference on advanced information networking and applications; 2014. p. 186–93. [15] Chang CY, Yen HC, Deng DJ. V2V QoS guaranteed channel access in IEEE 802.11p VANETs. IEEE Trans Dependable Secure Comput 2016;13:5–17. [16] Bi Y, Liu K-H, Cai LX, Shen X, Zhao H. A multi-channel token ring protocol for QoS provisioning in inter-vehicle communications. IEEE Trans Wireless Commun 2009;8:5621–31. [17] Omar HA, Weihua Z, Li L. VeMAC: a TDMA-based MAC protocol for reliable broadcast in VANETs. IEEE Trans Mobile Comput 2013;12:1724–36. [18] Yang F, Tang Y, Huang L. A multi-channel cooperative clustering-based MAC protocol for VANETs. In: Wireless telecommunications symposium; 2014. p. 1–5. [19] Rawat DB, Yanxiao Z, Gongjun Y, Min S. CRAVE: cognitive radio enabled vehicular communications in heterogeneous networks. In: IEEE radio and wireless symposium; 2013. p. 190–2. [20] Booysen MJ, van Rooyen GJ, Zeadally S. Survey of media access control protocols for vehicular ad hoc networks. IET Commun 2011;5:1619–31. [21] Almalag MS, Weigle MC, Olariu S. Mac protocols for vanet. In: Basagni S, Conti M, Giordano S, Stojmenovic I, editors. Mobile ad hoc networking: cutting edge directions. The Institute of Electrical and Electronics Engineers/ Inc.: JohnWiley&Sons, Inc; 2013. p. 599–618. [22] Park C-K, Ryu M-W, Cho K-H. Survey of MAC protocols for vehicular ad hoc networks. SmartCR 2012;2:286–95. [23] Kakarla J, Siva Sathya S. A survey and qualitative analysis of multi-channel mac protocols for VANET. Int J Comput Appl 2012;38:38–42. [24] Menouar H, Filali F, Lenardi M. A survey and qualitative analysis of MAC protocols for vehicular ad hoc networks. IEEE Wireless Commun 2006;13:30–5. [25] Hadded M, Muhlethaler P, Laouiti A, Zagrouba R, Saidane LA. TDMA-based MAC protocols for vehicular ad hoc networks: a survey, qualitative analysis, and open research issues. IEEE Commun Surv Tut 2015;17:2461–92. [26] Raya M, Papadimitratos P, Hubaux JP. Securing vehicular communications. IEEE Wireless Commun 2006;13:8–15. [27] Campolo C, Molinaro A. Multichannel communications in vehicular Ad Hoc networks: a survey. IEEE Commun Mag 2013;51:158–69. [28] Blum JJ, Eskandarian A. A reliable link-layer protocol for robust and scalable intervehicle communications. IEEE Trans Intell Transp Syst 2007;8:4–13.

Please cite this article as: N. Torabi, B.S. Ghahfarokhi, Survey of medium access control schemes for inter-vehicle communications, Computers and Electrical Engineering (2017), http://dx.doi.org/10.1016/j.compeleceng.2017.02.022

JID: CAEE

ARTICLE IN PRESS N. Torabi, B.S. Ghahfarokhi / Computers and Electrical Engineering 000 (2017) 1–23

[m3Gsc;March 3, 2017;13:54] 23

Nasser Torabi received the B.Sc. degree in Information Technology from University of Tabriz and the M.Sc. degree in Computer Networking from the University of Isfahan, Iran, in 2012 and 2014, respectively. His research interests are mobile ad hoc networks and vehicular ad hoc networks. Behrouz Shahgholi Ghahfarokhi received his B.Sc. in Computer Engineering (2004), his M.S. in Artificial Intelligence (2006), and his Ph.D. in Computer Architecture (2011) from University of Isfahan. He joined the University of Isfahan in 2011 and he is now an assistant professor at Department of Information Technology. His research interests are computer networks, mobile communications, and intelligent systems.

Please cite this article as: N. Torabi, B.S. Ghahfarokhi, Survey of medium access control schemes for inter-vehicle communications, Computers and Electrical Engineering (2017), http://dx.doi.org/10.1016/j.compeleceng.2017.02.022