Dynamic Channel Coordination Schemes for IEEE 802.11p/1609 Vehicular Networks: A Survey

IEEE 802.11p/1609-based vehicular networks utilize a multichannel architecture to support vehicle-to-vehicle and vehicle-to-infrastructure communications. In the multi-channel architecture, the available channels in the 5 GHz spectrum are divided into one control channel (CCH) and multiple service channels (SCHs). Multiple SCHs are defined for nonsafety data transfer, while the CCH is used to broadcast safety messages called beacons and control messages (i.e., service advertisement messages). According to the channel coordination scheme, a radio interface alternately switches between the CCH and a specific SCH. The intervals during which a radio interface stays tuned to the CCH and SCH are called CCH and SCH intervals, respectively. Both intervals are set to a fixed value (i.e., 50 ms) in the standard. However, since the fixed-length intervals cannot be effective for dynamically changing traffic load, some dynamic interval division protocols have been recently proposed to support the dynamic adjustment of the CCH/SCH intervals for improving channel utilization. In this paper, we therefore provide a survey of dynamic interval division protocols for VANETs, discuss the advantages and disadvantages of them, and define some open issues and possible directions of future research.


Introduction
Intelligent transportation systems (ITS) apply advanced information and communication technologies to transport infrastructure and vehicles in order to enhance the current transportation systems. ITS provides a wide range of applications such as safety, security, congestion alleviation, environmental monitoring and protection, productivity and operational efficiency, comfort, and convenience. Due to potential benefits of ITS applications, many academic institutes, automotive industries, and governments over the world have paid increasing attention to ITS over the last few decades. In addition, until now, many standard activities have provided the overall system architecture and communications framework. A variety of groups including IEEE in the US, and the C2C-CC, ETSI, and ISO in Europe and other parts of the world are participating in the standardization [1].
ITS can be implemented based on various underlying network architectures. Among them, a vehicular ad-hoc network (VANETs), a special type of mobile ad-hoc networks, is one of the candidate network architectures. In a VANET, moving cars and roadside units create a mobile network in a self-organized manner without any permanent infrastructure. Due to its decentralized nature, a VANET is highly preferred by a variety of safety applications which cannot easily obtain help from central nodes, such as cooperative collision avoidance, blind sport warning, and approaching emergency warning [2].
A VANET is mainly characterized by high mobility and the restricted movement patterns governed by roads and traffic rules [3]. These characteristics lead to many challenges in designing vehicle-to-vehicle (V2V) and vehicle-toinfrastructure (V2I) communication protocols. Particularly, to cope with the unstable nature of V2V and V2I links caused by high mobility, IEEE 802.11p has been proposed as an approved amendment to the IEEE 802.11 MAC/PHY standard to support wireless access in vehicular environments [4]. It introduces a modification to IEEE 802.11a at the PHY layer in order to cope with the fast-fading propagation environment. At the MAC layer, IEEE 802.11p relies on the prioritized channel access of IEEE 802.11e MAC. In addition, it simplifies the authentication and association operations, which are considered time-consuming for vehicular communications. In addition, the IEEE 1609 family of standards (denoted by  IEEE 1609.x) [5], a higher layer standard on which IEEE 802.11p is based, is being standardized to support ITS applications. Currently, it consists of 5 published standards and 2 unpublished standards under development [6]. This standard defines architecture, communications model, management structure, security mechanism, and how it works with the physical layer and media access layer for high speed (up to 27 Mb/s), short range (up to 1000 m), and low latency wireless communications in the vehicular environment. Collectively, IEEE 802.11p and IEEE 1609.x are called wireless access in vehicular environments (WAVE) standards.
Specifically, IEEE 1609.4 specifies extensions to the IEEE 802.11p MAC for multichannel operations. In WAVE systems, there are two types of channels: control channel (CCH) and service channel (SCH). The available channels are divided into one CCH and multiple SCHs. The safety and control messages are exchanged between devices in the CCH and messages for nonsafety application services are exchanged in SCHs. According to the coordination scheme, each device should alternate between the CCH and SCHs. The intervals during which a device stay tuned to the CCH and SCHs are called CCH and SCH intervals, respectively. Both intervals are set to a fixed value (i.e., 50 ms) in the WAVE standard.
Recently, many researchers investigated the performance of multichannel operations of the WAVE standard [7] and demonstrated that the fixed-length CCH/SCH intervals cannot be effective for dynamically changing traffic load. In a congested vehicular traffic condition, the fixed-length CCH may not be able to handle a large amount of safety packets and control packets. In addition, in sparse networks, the CCH channel can be wasted due to infrequent transmissions from vehicles. On the other hand, some applications consuming a large amount of bandwidth, such as video download and map update, cannot obtain sufficient SCH resources due to excessive contention. Therefore, many protocols have been proposed to address the dynamic adjustment of CCH/SCH intervals. We call them dynamic interval division protocols in this paper. In these protocols, CCH/SCH intervals are dynamically adjusted according to their own criteria.
In this paper, we therefore provide a survey of dynamic interval division protocols for VANETs. First of all, we introduce the IEEE 802.11p/1609.4 multichannel operations mentioned above in detail in Section 2. The operations of the existing dynamic interval division protocols are summarized in Section 3. In Section 4, we qualitatively analyze the protocols and point out some open issues and possible direction of future research. Finally, Section 5 concludes this paper. Figure 1 depicts the multichannel operations defined in IEEE 1609.4. The seven 10 Mhz-wide channels are available in the frequency band of 5.85-5.925 GHz: one control channel (CCH) and six service channels (SCHs). The channel access time is divided into synchronization intervals with a fixed length of 100 ms. A synchronization interval is further divided into CCH and SCH intervals. Each of intervals lasts 50 ms long. According to the channel switching scheme, all devices must stay tuned to CCH during the CCH interval for exchanging safety and control messages. A device can actively switch from the CCH to a specific SCH for its desired nonsafety application services. IEEE 1609.4 defines a guard interval (GI) at the beginning of both the CCH and SCH as shown in Figure 1. GI accounts for the radio switching delay and the time synchronization error. Typical values for the guard interval are between 4 and 6 ms. During the guard interval, nodes are not allowed to send or receive data packets.

IEEE 802.11p/1609.4 Multichannel Operations
Since two or more devices which want to exchange data should stay tuned on the same channel, time synchronization  is required for accurate multichannel operations. IEEE 1609.4 makes the CCH and SCH intervals to be synchronized to an external time reference, the Coordinated Universal Time (UTC) which is often provided by the Global Positioning System (GPS). However, if a device fails to get the UTC from its local GPS, it should get time information from other nodes over the air. This becomes possible by using wave time advertisement (WTA) frames, which is available in the IEEE 802.11p specification. Given the UTC, a node aligns the start of the CCH interval with the UTC or a multiple of 100 ms after the UTC.
In the IEEE 1609.4 specification, the channel switching scheme described above is called alternating access. Besides the alternating access scheme, the IEEE 1609.4 additionally defines 3 more channel switching schemes as shown in Figure 2: continuous, immediate, and extended access. The continuous access scheme allows a node to always stay tuned to the CCH in order to exchange only safety-related data. Hence, this scheme is not suitable for nodes who are interested in both safety and nonsafety applications. The immediate access scheme allows nodes to start communications over the SCH immediately without waiting for the next SCH interval. The extended access scheme allows nodes to keep communicating over the SCH without switching to the CCH. Both immediate and extended access schemes have been designed to improve the delivery performance of bandwidth-demanding nonsafety applications, requiring a huge amount of data to be transferred. However, they can be only beneficial to those vehicles which are not interested in safety applications such as a cooperative collision avoidance. Consequently, the alternating access scheme is the best solution for supporting both safety and nonsafety applications. We only focus on the alternating access scheme in this paper.
For providing a nonsafety service, a node (called WAVE provider) initializes a basic service set (BSS). A WAVE provider could be a either a roadside unit or a vehicle. Each WAVE provider advertises its presence and offered services by periodically broadcasting WAVE Service Advertisement (WSA) messages during the CCH interval. WSAs contain the information about the offered services and the network parameters necessary to join the advertised BSS (its identification, its SCH, its EDCA parameter sets, configuration parameters needed to access the Internet, etc.). For reliability purposes, the standard suggests that each WAVE provider sends WSAs several times in the CCH interval. A WAVE provider also should choose the least congested SCH for its BSS in order to reduce interference between nearby BSSs. However, a specific scheme for selecting the least congested SCH is not specified in the standard.
Nodes (called WAVE users) interested in the services offered by the WAVE provider should monitor the CCH to learn about the existence and the operational parameters of available BSSs. When a WAVE user receives a WSA frame from a WAVE provider, it simply switches into the SCH advertised in the WSA and starts to exchange data with the WAVE provider.

Dynamic Interval Division Protocols
There has been substantial interest in multichannel MAC protocols for multihop ad-hoc networks [8]. On the other hand, only a few researchers have discussed the multichannel MAC protocols for VANETs. In this section, we review existing multichannel MAC protocols for VANETs. In particular, we only address protocols that handle issues for the dynamic adjustment of CCH and SCH intervals in this paper.    [9], for the dynamic adjustment of CCH/SCH intervals. As shown in Figure 3, DID-MMAC further splits the CCH interval into three phases based on a type of different messages: Service Announce Phase (SAP), Beacon Phase (BP), and Peer-to-Peer Reservation Phase (PRP). WSA and beacon messages are transmitted in SAP and BP, respectively, and the exchanges of control messages for SCH reservations are performed in PRP. For dynamically adjusting the intervals of SAP and BP in a distributed manner, DID-MMAC assigns different channel access priorities to different messages by differentiating the contention window (CW) and the interframe space (IFS). Table 1 summarizes CWs and IFSs assigned to WSA and beacon messages. According to the medium access control strategy of IEEE 802.11p (i.e., CSMA/CA), a node wishing to initiate a transmission invokes the carrier-sense mechanism to determine whether the medium is busy or not. If the medium is busy, the node defers the transmission until the medium is determined to be idle for a period of time equal to IFS. After IFS, the node will wait for an additional random time called backoff time before starting the transmission in order to avoid collisions. The backoff time is determined by * SlotTime (SlotTime is a basic time unit in the backoff process. Its value in the idle case depends on the duration that is required by different PHY techniques (e.g., slot time is 9 s for the OFDM-based 802.11a) to detect the medium state), where is the random number selected between 0 and CW. Therefore, since IFS SAP is smaller than IFS BP , senders which have a WSA message to send always win the contention for accessing the medium against the other senders which have a beacon to send. Consequently, after all the WSA messages are transmitted, beacon transmissions can be allowed to proceed within the CCH interval. In DID-MMAC, SAP and BP end when the channel is idle during IFS SAP + 7 * SlotTime and IFS BPP + 15 * SlotTime, respectively.
PRP starts right after BP. The duration of PRP is calculated based on the network real-time traffic load. In order to make a node to be aware of the traffic load, DID-MAC adds two new fields called Service Indication (SI) and Traffic Indication (TI) into WSA and beacon messages, respectively. The SI field indicates the total data size of service and the TI field indicates the service status in vehicle side. Whenever a node receives WSA and beacon messages, it updates the traffic load based on SI and TI information in the received message. The duration of PRP is adjusted adaptively according to the estimated traffic load which is affected by the underlying peer-to-peer communication negotiation protocols. DID-MMAC adopts the modified MMAC [10] as the negotiation protocol. It is noted that since we are only interested in how to dynamically adjust the CCH/SCH intervals, we omit SCH interval (100 − T cch ) Figure 4: Operation of VCI.
the explanation on operations related to the negotiation protocol in this paper. Readers who are interested in the details of operations regarding the negotiation protocol can refer to [9]. [11][12][13] multichannel MAC protocol to enhance the saturation throughput of SCHs while ensuring the transmissions of safety messages. Similar to DID-MMAC, VCI also divides the CCH interval into safety interval and WSA interval as shown in Figure 4. Periodic beacons are transmitted in the safety interval. During the WSA interval, nodes exchange control messages for SCH reservations. Service providers broadcast WSA packets and piggyback service information and the identities of SCHs to be used. Nodes which need the service can optionally respond to the WSA packet with an acknowledgement (ACK). Furthermore, a service user can initiatively send a request for service (RFS) packet to make an agreement with a service provider. In VCI, the length of safety interval ( safety ) is determined by (1) where represents the total number of nodes sending safety packets, cch is the data rate of CCH, is a predefined factor according to current vehicular environment, and is the sending frequency of safety messages. In this equation, is the only variable and safety is affected mainly by . However, VCI does not mention how to calculate in a dynamically changing traffic condition

VCI: Variable CCH Interval. Wang et al. proposed a Variable CCH Interval (VCI)
The residual synchronization interval except for the safety interval is divided into WSA and SCH intervals. VCI calculates the optimal ratio between WSA and SCH intervals for maximizing the channel utilization of SCHs. Ideally, the optimality is satisfied when the number of successful reservations equals the number of packets transmitted on all SCHs within the same contention domain. To achieve this goal, VCI analyzed the behavior of a single node and the stationary probability that the node transmits WSA or RFS packets during the predefined short time interval by using a Markov chain model. The analyzed results are used to calculate the average time consumed on the CCH for the negotiation of service packet transmission. Finally, VCI derives the optimal ratio between WSA and SCH intervals based on the average reservation time. As stated in Section 3.1, we do not address the explanation on operations related to the negotiation protocol in this paper. For the more detailed explanations, readers can refer to [11].
Basically, the CCH interval (safety and WSA intervals) is calculated by the roadside unit (RSU). The RSU broadcasts a packet (VCI packet) containing the length of the CCH interval to the nodes under its transmission range. For the reliable delivery, each VCI packet will be broadcasted at least twice. Under heavy traffic or congested conditions, the VCI packets may not be heard by some nodes. To tackle this issue, VCI adds a field representing the latest CCH interval information into the WSA/RFS packets. Moreover, when there exists no RSU, VCI selects a leader among nodes within one hop and allows the leader to broadcast the VCI packet.
3.3. DCI: Dynamic CCH Interval. D. Zhu and D. Zhu proposed a Dynamic CCH Interval (DCI) MAC protocol for the dynamic adjustment of CCH/SCH intervals [14]. DCI works identically to VCI except for the calculation of the WSA interval. Different with VCI that is based on the average time consumed on the CCH for the negotiation of service packet transmission, DCI calculates the optimal WSA interval based on the probability distribution of the reservation time for service packet in the CCH interval.
DCI defines as the maximum number of service data packets which can be transmitted on the given SCH interval. And then, it derives the minimum WSA interval for reserving the transmission of service packets. Finally, DCI obtains the optimal WSA interval by finding the optimal to minimize the difference between the sum of WSA and SCH interval and the residual synchronization interval except for the safety interval.

DSI: Dynamic Safety Interval. Yoo and Kim proposed a
Dynamic Safety Interval (DSI) protocol for calculating the optimal safety interval under dynamic traffic conditions in a distributed manner [7]. Different with protocols mentioned before, DSI only addresses the issue of the adjustment of the safety interval. In particular, DSI calculates the safety interval considering the presence of hidden nodes. Note that DID-MMAC, VCI, and DCI do not consider the presence of hidden nodes.
DSI aims to allow the safety interval to accommodate transmissions from nodes within the same contention domain (a contention domain is a section of a network where data packets can collide with one another when being sent on a shared medium) as well as nodes hidden from them. The region within which hidden nodes reside is affected by three types of ranges related to packet transmissions in the IEEE 802.11 MAC: transmission range ( ), carrier sensing range ( ), and interference range ( ). Here, represents the range within which a packet can be successfully received by a node if there exists no interference from other nodes. is mainly affected by the transmission power and radio propagation models. is the range within which a transmitter triggers carrier detection. It is usually determined by the antenna sensitivity. is the range within which nodes in a receiving mode interfere with transmissions from other nodes. These ranges are tunable parameters that can significantly affect the MAC performance. Measurement studies such as [15] demonstrate that < < . In addition, many studies assume that and are typically more than twice [16]. In particular, and have default values of 2.2 times in the ns-2 simulator [17]. Assuming that the ratio of to is 1: / and the ratio of to is 1: / , / and / are regarded as tunable system parameters. DSI assumes that / and / are equal to 2.2. Hidden nodes refer to the nodes located within of the intended destination and out of of the sender. When a receiver is receiving a packet, if a hidden node tries to start a concurrent transmission, collisions can happen at the receiver. For example as shown in Figure 5, node B is located within interference ranges of both nodes A and D so that node D's transmission interferes with the transmission from node A to node B. On the other hand, node A's is not overlapped with node E's , since nodes A and E are separated by a distance denoted by sr . Therefore, both nodes can concurrently transmit their packets without interfering with each other. sr is called spatial reuse distance and should be larger than the sum of and . DSI allows each node to share the safety interval with nodes whose distance from is smaller than sr . In DSI, the extended contention domain (ECD) is defined, which is the region where nodes sharing the safety interval reside. Given ECD, each vehicle calculates the number of vehicles (denoted by ) within ECD and derives the safety interval based on , interframe space, the average backoff time, and the transmission time of a beacon.

Qualitative Analysis
In this section, we qualitatively analyze existing dynamic interval division schemes in terms of various aspects listed in what follows: DID-MMAC adjusts the CCH/SCH intervals based on the traffic load in a distributed manner without the help of RSU. In order to allow each node to measure the traffic load, it causes the message overhead by adding new fields called SI and TI into each WSA and beacon message. However, the message overhead per a message is negligible (i.e., 1-byte) as compared to the size of control and beacon message ranging from 100 bytes to 300 bytes. The operation of DID-MMAC is designed to work well in the saturated traffic condition. However, under the unsaturated traffic condition, if WSA, beacon, and control messages are randomly generated within the synchronization interval, each interval (i.e., SAP, BP, or PRP) will be repeated multiple times in a random pattern during the synchronization interval. In this situation, since nodes cannot detect the end of BP, it is impossible for nodes to estimate the PRP interval accurately. Regarding interference from hidden nodes, in DID-MMAC, hidden nodes outside of a given contention domain (CD) also transmit their beacons during their BP. BPs of nodes within and outside the CD may be overlapped in time. Therefore, transmissions from the safety interval are continuously exposed to the interference from hidden nodes. VCI uses the number of vehicles within the contention domain and the average time ( nego ) consumed on the CCH for the negotiation of service packet transmission as criterion for calculating the CCH/SCH intervals. VCI assumes the saturated traffic condition when it calculates nego so that it cannot avoid the performance degradation under the unsaturated traffic condition. Fundamentally, the CCH/SCH interval is calculated by RSU. Moreover, when there exists no RSU, a node is selected as a leader, and it calculates the intervals. The RSU or the leader broadcasts the VCI message periodically in order to advertise the calculated intervals to nodes within the coverage of them, which leads to the message overhead. However, the size of the VCI message is not specified in [11]. VCI also does not consider a network scenario with potential hidden nodes, and tacitly means the total number of nodes within the same contention domain. Therefore, VCI also cannot avoid interference from hidden nodes as in DID-MMAC. Regarding the SCH scheduling method, VCI introduces a new SCH negotiation (denoted by WSA/RFS/ACK exchange in Table 2) protocol, while DID-MMAC exploits an existing protocol called the modified MMAC.
As stated in Section 3.3, DCI works identically to VCI except for the calculation of the CCH/SCH intervals. DCI calculates the WSA and SCH intervals based on the probability distribution ( nego ) of the negotiation time for service packet in the CCH interval. DCI defines the predefined threshold in the algorithm to calculate the duration of the CCH interval. Hence, DCI's performance is affected by the threshold. However, DCI does not include the method to determine the most appropriate threshold value.
DSI calculates the safety interval according to the number of vehicles (denoted by ECD ) within the extended contention domain (ECD) in order to allow the safety interval to accommodate transmissions from nodes within the same contention domain as well as nodes hidden from them. The concept of ECD is newly introduced in [7]. Each vehicle calculates ECD in a distributed manner without the help from the infrastructure such as RSU. In the process of calculating ECD , each vehicle exchanges the latest information on the vehicle density within its proximity with its neighboring vehicles, which leads to the message overhead. In DSI, every beacon message includes the vehicle density information whose size is lower than 10 bytes. In addition, since DSI does not assume a specific traffic condition, it works well under any traffic conditions. Table 2 summarizes the analysis results. We also introduce some open issues and possible directions of future research related to the dynamic interval division protocols for vehicular ad-hoc network as follows: (i) Most existing dynamic interval division protocols are defining a system parameter in their algorithm to calculate the CCH/SCH intervals. Hence, if the system parameter is not adjusted according to the dynamically changing network condition, the operation of the protocols cannot be guaranteed to work appropriately under the dynamic VANET environment. Therefore, a dynamic interval division protocol that is independent of such a parameter is needed.
(ii) The authors of the DID-MMAC, VCI, and DCI protocols assumed the saturated traffic condition when they analyze the behavior of their SCH negotiation protocols. And then, based on the analysis results, they calculate the SCH interval. However, in vehicular ad-hoc networks, the traffic condition is dynamically varying with time and space [18]. Therefore, the analysis of the behavior of their SCH negotiation protocols should be done under the various traffic conditions.
(iii) The interference from hidden nodes is known to deteriorate the MAC layer performance [19]. Although DSI considers the interference from hidden nodes when it calculates the safety interval, the DSI design is based on the deterministic propagation model where transmission/interference ranges are defined as the ideal circles as shown in Figure 5. Therefore, a new dynamic interval division protocol should consider the interference from hidden nodes under the probabilistic propagation model [20] which is more realistic than the deterministic model.
(iv) All the existing dynamic interval division protocols assume that each vehicle periodically broadcasts 8 International Journal of Distributed Sensor Networks a beacon message with the fixed rate and power. Although the use of the fixed rate and power is regarded as the default option of the typical VANET systems, many researchers are studying the dynamic use of the rate and power of beacon transmissions. Therefore, a new dynamic interval division protocol should calculate the safety interval considering the variable rate and power of beacon transmissions [21] which can affect the total load of the beacon traffic.
(v) Except for the alternating access scheme, the WAVE standard additionally defines 3 more access schemes: continuous, immediate, and extended accesses. The analysis and tuning of them is an open issue.

Conclusion
In this paper, we have performed an extensive survey of various dynamic interval division protocols for IEEE 802.11p/1609-based VANETs. We summarized the operation of the existing protocols and discussed the advantages and disadvantages of them. Furthermore we also defined some open issues and possible directions of future research related to dynamic interval division protocols for vehicular ad-hoc networks.