Distributed Channel Coordination in Cognitive Wireless Vehicle-to-Vehicle Communications (Invited Paper) Kazuya Tsukamoto *, Shinya Matsuoka **, Onur Altintas ***, Masato Tsuru *, Yuji Oie * * ** Department of Computer Science and Electronics, Kyushu Institute of Technology, *** TOYOTA InfoTechnology Center Co., Ltd. E-mail: *{tsukamoto, tsuru, oie}@cse.kyutech.ac.jp, **
[email protected], ***
[email protected] Abstract- Currently, a dedicated frequency band is assigned to each of a number of wireless technologies, artificially causing spectrum shortage for new services. It is expected that cognitive radio technology, which can adaptively detect spatial and temporal changes in use over various frequencies, will facilitate achieving efficient wireless resource sharing. However, in ad hoc cognitive communications, such as vehicle-to-vehicle (V2V) communication, since the existence of a common control channel cannot be assumed due to a lack of infrastructure, distributed channel coordination is invariably needed. As a preliminary step toward cognitive wireless V2V communication, in this paper, we focus on one-hop V2V communication and propose a distributed channel coordination scheme that exploits the different characteristics of various frequencies, in terms of both the data transmission rate and the range. Furthermore, in order to evaluate the effectiveness of the proposed scheme, we develop a channel utilization model in which the utilization of each channel changes temporally and spatially due to both primary (licensed) and secondary (unlicensed, opportunistic) usage. The simulation results demonstrate that the proposed scheme can reliably utilize unused frequency(-ies), even under temporal and spatial changes.
I. INTRODUCTION Currently, dedicated frequency bands are assigned for many types of wireless technologies (2.4/5 GHz: WLAN, 810 MHz/1.9 GHz (Japan): Cellular, 2.5 GHz: WiMAX, etc), exclusively. In the future ubiquitous network, a number of emerging wireless technologies, such as vehicular networks, will be deployed and new dedicated frequency bands will be required for these technologies, thereby inducing further shortage of wireless resources (frequency bands). On the other hand, the utilization of different frequency bands exhibits independency and dissimilarity. Figure 1 shows an example of actual measurement results of spectrum usage for a certain time and location [1]. Although the amplitude of signal strength of some frequency bands, such as those used for cellular networks, indicate heavy use, other parts indicate sparse or medium use, which are commonly referred to as ``white spaces'' [1]. The utilization of each band changes temporally and spatially. For example, the ratio decreases during the nighttime but increases during business hours, reflecting temporal change. On the other hand, the ratio in urban areas is
quite high, whereas that of the rural areas is low, reflecting spatial change. Therefore, a new communication paradigm, cognitive radio [1], has been proposed in order to facilitate efficient wireless resource management.
Fig.1: Spectrum utilization [3].
In cognitive radio, there are two types of users: the primary user and the secondary user. The primary user holds a license for the dedicated frequency band, but secondary user does not hold any license and uses cognitive radio to access the temporarily-unused frequency band. Therefore, secondary users should vacate the current frequency band and switch to a new band, when the primary user begins communication. In the United States, the Federal Communications Commission (FCC) has recently approved the unlicensed use of TV white space spectrum for wireless applications and devices. IEEE 802.22 [2] is the first standardization effort to define unlicensed operations in the TV spectrum, and the IEEE 1900 Standards Committee [3] has been discussing unlicensed operation in the overall spectrum bands. In cognitive wireless ad-hoc networks, since there is no infrastructure such as access points of WLANs and base stations of cellular networks, centralized control is not appropriate for cooperation among nodes. On the other hand, the predetermination of communication parameters, such as communication channel and data rate, and further coordination, such as switching to a new band, caused by the appearance of primary users is extremely difficult. Furthermore, spatial
changes in available frequency bands become more drastic in V2V communications due to the movement of nodes. The Dedicated Short Range Communication (DSRC) band [4], intended for vehicular communications in high-speed vehicular environments, has been already assigned in the US, Europe, and Japan for the Intelligent Transportation System (ITS) to provide safety information and road traffic information. Frequency range of DSRC is limited to several tens of MHz in the 5.8-5.9 GHz band, thereby causing limited coverage and transmission rate. For quick delivery of small amount of data to the vehicle, which DSRC mainly targets, this limitation can be acceptable. On the other hand, in the near future, various infotainment applications, such as large data transfer (i.e. elastic, long-lasting traffic) and voice/multimedia streaming (i.e. Constant Bit Rate (CBR) traffic), is expected to be used over V2V ad-hoc network, in addition to the conventional ITS applications. As a result, to fulfill the wide range of requirements from various V2V applications, we need to investigate a new way to efficiently exploit the unused communication channels in a decentralized manner. In this paper, as a preliminary step toward achieving comprehensive V2V communication, we focus on one-hop V2V communication and propose a new distributed channel coordination scheme in order to effectively and reliably maintain communication by exploiting cognitive radio technology. In the proposed scheme, we consider differences in communication characteristics in terms of the range and data rate among various frequencies, and the appropriate communication channel is selected based on these differences. The selected channel for communication is then dynamically altered in response to changes in the wireless link condition and primary user appearances. Furthermore, in order to evaluate the effectiveness of the proposed scheme, we develop a channel utilization model in which the utilization of each communication channel changes temporally and spatially due to both primary and secondary users. II. RELATED RESEARCH In this section, we summarize and compare existing works that are relevant to the proposed scheme, mainly from the viewpoint of spectrum sharing. There are two approaches for supporting spectrum sharing: centralized control and distributed coordination. As a centralized control approach, IEEE 802.22 was the first standardization effort to share available bands in TV spectrum. In 802.22, a base station determines the availability of a TV channel based on the input from nodes. In DIMSUMnet [5], the spectrum broker manages large portions of the spectrum and assigns its portions to individual users in a relatively large geographic region. In DSAP [6], the centralized controller handles the spectrum access by offering long-term leases to secondary users in a limited geographic region. However, these centralized control approaches cannot be applied to ad-hoc V2V communication due to the lack of infrastructure. In contrast, several MAC-layer protocols have been proposed to manage the overall spectrum. MMAC [7], and
LCM-MAC [8] use only a single interface and frequently switch the interface among multiple channels. Although these MAC protocols can potentially distribute load on them, switching an interface from one channel to another incurs a delay due to the coordination required. Thus, frequent switching may adversely affect performance. On the other hand, HMCP [9], xRDT [8], and KNOWS [10] use multiple (more than two) interfaces. They employ one of the interfaces as a control interface, which is always tuned to a fixed control channel (common to all nodes), and the other interfaces are assigned as data channels. Traditional coordination uses a common control channel that is known to all users. However, secondary users in the cognitive radio network observe spectrum heterogeneity, that is, spectrum availability fluctuates over time and with location. In such cases, common channels cannot exist. Recently, HDMAC [11] was proposed in order to solve this problem. HDMAC is a group-based coordination scheme which forms a distributed group setup. Users dynamically select the coordination (control) channel based on local conditions (rather than a common channel). This approach can significantly improve scalability. However, in HD-MAC, node mobility and ad-hoc communication have not been considered, and details on the dynamic decision of the control channel have not been reported. Furthermore, to the best of our knowledge, none of these approaches has taken into account the differences inherent in various frequencies, such as data rate and range. III. PROPOSED SCHEME In this paper, we deal with the basic one-hop V2V communication towards developing multi-hop multi-user distributed coordination. We consider two mobile nodes, such as vehicles on a roadway, willing to communicate with each other and are in one-hop reach of each other. Note that, while we focus on two nodes, we assume other secondary users also exist in one-hop reach. We will later use this proximity property, both in spatial and temporal terms, to efficiently and reliably determine the channel(s) for data communication. For this, we assume the availability of a location determination system such as the Global Positioning System (GPS) to acquire location and time information. Furthermore, we consider the disparate requirements of control information and application data. Specifically, the control information should be transmitted reliably due to its importance, while the application data should be transmitted efficiently due to its possibly large volume. Therefore, we choose to keep the control channel(s) separate from data channel(s) (i.e., out-band). When available, control channel(s) are selected from among relatively low-frequency with longer range and lower data rate; and later, data channel(s) are selected from among relatively high-frequency with higher data rate and shorter range. Hence the proposed scheme is able to efficiently determine the data channel(s) based on the information over the reliable control channel. Later, these channels are dynamically switched in response to temporal and spatial changes in wireless link conditions. In our
data to transmit (receiver) wait for the communication request packets from the sender over the available channels, as shown in Fig. 3. On the other hand, the sender transmits (broadcasts) probe messages over each available channel. When the receiver receives the probe message, it returns probe-ACK messages to the sender on the received channel, as shown in Fig. 5. Note that exchange of probe/probe-ACK messages indicates that the channel is available. If sender and receiver cannot establish a control channel from within the candidate channels until a predetermined timer expires, they change the hash function in a predetermined sequence and obtain other candidate channels again. If there is no available candidate channel, nodes wait for the next reception of GPS information and then select other candidate channels by using next GPS information through the hash functions. In this way, the proposed scheme scans the candidate channels only, thereby drastically shortening the scanning period. As a result, the sender determines a single control channel with the longer transmission range.
proposed scheme, each node employs multiple (more than two) interfaces for control and data channels, such as in KNOWS [10].
Sender
Receiver Step1. Decision of control channel Step2. Decision of data channel Step3. Start of data communication Step4. Adaptive change in control/data channel(s)
Fig. 2: Outline of the proposed scheme.
Figure 2 outlines how the proposed algorithm establishes and updates control and data channels during the communication period (from the start of communication between the sender and the receiver to the end). A. Communication establishment procedure To achieve one-hop communication between two nodes (sender and receiver) in ad-hoc V2V network, nodes that are in the ``same vicinity'' should use the same communication channel at the ``same time'' (channel coordination). However, since the spectrum scanning (sensing) of all channels requires a large amount of time (larger than 10 seconds: 400MHz to 6 GHz), the nodes need to be able to perform adaptive and prompt channel coordination with a distributed manner. For this, first, both sender and receiver periodically (e.g., 1 second) obtain time and location information via GPS. We hypothetically divide the entire area into much smaller regions, which we refer to as ``granules'', such that all of the nodes in the same granule resolve to the same time and location particulars based on the signals received from the GPS satellites. Note that the granule size is slightly greater than the square of the transmission range of the lowest frequency channel. Furthermore, we assume that the sender and the receiver share several common hash functions and use the pair of as the input keys to those hash functions whenever nodes receive GPS information. That is, nodes are synchronized with the reception of GPS information. Each hash function selects one specific channel from among lower frequency channels and then returns n (e.g., five) contiguous channels, as candidates of the control channel (called as candidate channels). Note that since a common wireless interface can receive messages within certain channels without reconfiguration, we choose to identify contiguous channels. 1.
Decision of control channel: Both nodes first scan the candidate channels obtained from the hash function and then find available channels within the candidate channels, respectively. After that, nodes that do not have
2.
Decision of data channel: After deciding the control channel, both nodes periodically exchange the following three pieces of information over the control channel: available communication channels, suitable communication channels for data transmission, and location. The sender first reports the list of suitable channels for data transmission to the receiver. The sender then determines data channel based on the information reported by the receiver. Naturally, selecting and using two or more channels simultaneously is possible depending on the radio capabilities, the existence of other secondary users, and application requirement. We assume the following two applications: real-time streaming (CBR) and non real-time large data transfer f1f2 f3 f4 f5
Candidate channels
：Candidate channels after scanning ：Candidate channels from hash function ：Suitable range for control channel ： Overall frequency range
low (FTP).
Frequency
high
Fig. 3: Decision of control channel. Sender Manage/Refer
Channel coordination Ch1
Ch2
Ch3 ・・・
Phy
Phy
Phy
Phy
Phy
Phy
Phy
Ch1
Ch2
Channel Availability Table, etc
Chn
Phy
Ch3 ・・・
Channel coordination
Information Reference Base
Information Reference Base
Chn
Channel Availability Table, etc Manage/Refer
Receiver
Fig. 4: Architecture of wireless interface.
： Probe messages
Receiver Scan is done at decision of control channel
fa
fc
fb
fd fe
Probe messages do not reach
Sender
low
Control channel range
Frequency
：multiple candidate channels ：Suitable range for control channel ： Overall frequency range
Data channel range
high
: Available channel for data and control channel ：Suitable range for data channel
Fig. 5: Detection of data channel.
z
z
3.
CBR: The channel with the highest data rate available that can satisfy the application requirements is selected. FTP: All channels that can be employed as data communications are selected. In the proposed scheme, we assume that one physical interface in both communication nodes can simultaneously use multiple channels for data communication through periodical switching process. Figure 4 depicts the conceptual architecture of simultaneous data transmissions over multiple data channels.
Start of data communication: Although the data channel is decided using the control channel with a longer range, the probability that the range of data channel is shorter than the distance between the two vehicles is not null. Therefore, vehicles should determine the start of data communication by taking the distance due to their movement into account. When the distance between nodes calculated from exchanged GPS information is shorter than the range of the data channel, the sender transmits probe messages, as shown in Fig. 5. The receiver then returns probe-ACK messages on those channels it received the probe messages correctly and data transmission begins subsequently.
B. Communication maintenance procedure 4. Adaptive change in control/data channel: Both selected control and data channel might become unavailable very soon due to spatial and temporal changes. Therefore, we always prepare back-up control and data channel(s) based on the change in the surrounding radio environments. More specifically, since the lowest (highest) frequency is selected as the control (data) channel for reliable (efficient CBR) communication, the second lowest frequency and the second highest (CBR)/other all available (FTP) frequency(-ies) are selected as the back-up control and data channel(s) in order to maintain the reliable and efficient data communication, respectively. To achieve this, all nodes always scan the surrounding radio conditions and periodically update the information of available channels by sending the probe/probe-ACK messages on all available channels for the adaptive detection of change in
available channels. That procedure works independent from the decision procedure of control and data channels, described in step. 1 to 3. If the back-up control and data channel(s) become unavailable, one notification packet is exchanged on the control channel before the data/control channel changes. As a result, the proposed scheme can reliably update the back-up control and data channels.
Overtake
Catch up Node1
V1 [km/h]
Maximum range: 100m
Node1
Node2 V2 [km/h]
Node2
Sub area
50m
50m
Fig. 6: Simulation model.
Fig. 7: Hierarchical subarea model. TABLE I Simulation parameters. Number of frequency IDs 100 Transmission range 10 (ID:100) – 100 (ID:1) [m] Transmission rate 100 [Kb/s] (ID:1) - 50 [Mb/s] (ID:100) Control/Data channel ID:[1-40]/[60-100] Granulation of GPS information Location (0 – 110 [m]) Time (0 – 1 [s])
IV. SIMULATION MODEL In this section, we first describe the simulation environment, and then propose a new channel utilization model used in evaluating the proposed scheme. Later, we describe two other schemes and relevant performance measures for comparative evaluation. A. Simulation environment Figure 6 shows the simple simulation model employed herein in order to examine the basic effectiveness of our proposed scheme. Two nodes (sender and receiver) move in the same direction at speeds of 50 Km/h and 45 Km/s, respectively. The entire area is divided into multiple granules as described in Sect. 3. We assume that these two nodes receive the same GPS information (in the same granule), thereby sharing the same candidate frequencies obtained from a common hash function.
B.
Channel utilization model The utilization of individual frequency IDs at a certain location is drastically different from each other. Furthermore, this difference varies temporally. In the present study, we assume that the utilization of each frequency ID by the primary users follows the actual measurement results in Fig. 1 ([1]) and then determine, by Eq. (1), the unavailable period (USED) of individual frequency IDs during a period of 100 seconds.
USED pri = − log(1 − U ) ×
1 t
Ideal Proposal Random
80 Coordination Time [s]
Simulation parameters are listed in Table 1. Since each frequency ID has different characteristics in terms of range and data rate, we divide the simulation area into multiple subareas, as shown in Fig. 7, considering that the range of each frequency ID and the channel utilization of these subareas are drastically different (spatial changes). Note that the frequency ID 60-100 (1-40) are treated as the data (control) channel. That is, these nodes can establish data (control) communication whenever the distance between them is less than the range of ID 60 (1).
70 60 50 40 30 20 10 0
0
10
20
30 40 50 60 Average usage ratio
70
80
90
C.
Compared schemes and performance measures We implement the following two schemes for comparison with the proposed scheme.
(1) Fig. 9: Coordination Time (CT).
where U is the uniform distribution from 0 to 1. The USEDpri period obeys an exponential distribution of average t, which is set based on the survey results in [1]. As a result, channel utilization by the primary users is very similar to that of Fig. 1, as shown in Fig. 8. As there will be other secondary users sharing the available communication channels in a cognitive wireless network increasing the overall channel utilization, we take into account the overall usage of communication channels caused by both primary and secondary users. Therefore, as shown in Fig. 8, we determine the USEDall which both the primary and secondary users induce, as follows:
z
Random scheme: Each of the nodes randomly sets the frequency ID (channel) for communication and tries to start communication on that band. Since a wireless interface can receive packets within a certain range of channels without reconfiguration, communication can be established when the selected band between the nodes is close (say, within five IDs). However, the performance would be drastically degraded due to the inefficiency of the random selection.
z
Ideal (hypothetical) scheme: We assume that the nodes have complete knowledge of the change in channel utilization and select the appropriate channel according to the application requirements based on the knowledge of ``what will happen next'' in the wireless channel. That is, the performance reaches the upper limit under any wireless link condition.
USEDsec = (100 − USED pri ) × sec ondary ratio [%] (2)
USEDall = USEDpri + USEDsec
(3)
We examine the performance by employing the following two criteria as performance measures.
Fig. 8: Channel utilization model.
z
Coordination time (CT): Communication should be started when the nodes enter the area in which the transmission rate can satisfy the application requirement. The time interval between the establishment time of data communication and the arrival time of the nodes into that data communication area is defined as the coordination time (CT).
z
Duration of data transmission (DDT): The total duration of data transmission over a data channel within the simulated time. In other words, DDT is the
Amount of transmitted data (ATD): The total amount of successful data transmission over all data channels during simulation time. ATD is employed to evaluate the performance of FTP application. V. EVALUATION AND DISCUSSION
In this section we examine the performance of the proposed scheme in a radio environment where both primary and more than one secondary users share the spectrum. We compare the effectiveness of the proposed method with the two methods described above (Random and Ideal). Note that, from the node movement model used herein, the sender and receiver can communicate for a maximum of 65.847 seconds without radio interference from other users, when ID 60 with maximum range is always selected. Figure 9 shows how the average, maximum, and minimum values of CT change when the secondary usage ratio is varied from 0% to 90%. The random scheme has remarkably large average CT value (more than 10 seconds) even when there are no other secondary users (0%). The average value subsequently increases with the increase in the secondary usage ratio. In particular, the maximum value reaches 60 seconds even under a secondary usage ratio of 20%. In other words, the random scheme cannot start communication, once a certain secondary ratio is passed (in this case 20%). In contrast, the ideal scheme immediately starts data communication until the secondary usage ratio reaches 20%, but CT gradually increases with the increase of the secondary usage ratio indicating that there are no available data channels at the start time. The proposed scheme shows a relatively stable coordination time, and, like the ideal scheme, the difference between the maximum and minimum times is smaller than that of the random scheme. In particular, the proposed scheme can start data communication in 10 seconds, even under the worst condition (90%), while the random scheme needs 30 seconds on average to start communication. Next, we examine the change in the DDT when the secondary usage ratio is varied (Fig. 10). DDT obtained from the random scheme is rather small and decreases gradually with the increase in the secondary usage ratio. In contrast, the ideal scheme maintains a high DDT value under a low secondary usage ratio, and the average value decreases with the increase of the ratio, while the difference between the maximum and minimum values becomes large. Here, the proposed scheme can provide near-optimal DDT performance under a low secondary usage ratio (less than 30%). The DDT decreases linearly in response to the increase of the secondary usage ratio due to the increase of latency for data channel selection. Figure 11 shows the change in ATD when the secondary usage ratio is varied. ATD obtained from random scheme is extremely limited even under a low secondary usage ratio. In contrast, the ideal scheme decreases drastically under a high
Duration of Data Transmission (DDT) [s]
z
secondary usage ratio, while achieving high ATD under a low secondary usage ratio (0%). Finally, the proposed scheme can provide near-optimal ATD performance as in the ideal scheme, irrespective of the secondary usage ratio. The ATD decreases in response to the increase of the secondary usage ratio due to the increase of latency for the detection of the change in the available channels. From these results, we can demonstrate that the adaptive detection of available data channels by exploiting the message exchange over control channel is effective for data communication, thereby achieving better performance for both CBR and FTP applications. Ideal Proposal Random
80 70 60 50 40 30 20 10 0
0
10
20
30 40 50 60 Average usage ratio
70
80
90
Fig. 10: Duration of data transmission (DDT): CBR. Amount of Transmitted Data (ATD) [MB]
period that the CBR application requirement is satisfied.
5000
Ideal Proposal Random
4000
3000
2000
1000
0
0
10
20
30 40 50 60 Average usage ratio
70
80
90
Fig. 11: Amount of transmitted data (ATD): FTP.
VI. CONCLUSION We presented a new distributed channel coordination scheme exploiting cognitive radio technology for vehicle to vehicle communications. To provide stable and efficient communication in a spatially and temporally varying radio environment, the proposed scheme first determines a control
channel from lower frequency with relatively longer transmission ranges obtained from a common hash function. Then, a data channel is selected from higher frequency with relatively higher data rates based on message exchange over the control channel. Afterward, these channels are switched adaptively according to the spatial and temporal changes of utilization in the wireless environment. We also proposed a new spectrum (channel) utilization model in order to emulate characteristics such as temporal and spatial changes in the spectrum utilization. The simulation results demonstrated that the network performance obtained by the proposed scheme is remarkably higher than that of the random scheme under the proposed spectrum utilization emulation. Our subsequent work will include extending the one-hop V2V channel coordination scheme presented here to a multi-user multi-hop environment. ACKNOWLEDGMENT This work was supported in part by the Japan Society for the Promotion of Science, Grant-in-Aid for Scientific Research (S) (18100001). REFERENCES [1]
I. F. Akyildiz, W. Y. Lee, M. C. Vuran, and S. Mohanty. Next generation/dynamic spectrum access/cognitive radio wireless networks: a survey.Elsevier Computer Networks, 50(5):2127-2750, Aug. 2006. [2] IEEE 802.22 WRAN WG, http://www.ieee802.org/22/. [3] IEEE 1900, Standards Coordinating Committee 41 (SCC41), http://www.scc41.org. [4] DSRC Working Group, October 2007, http://www.ieee802.org/11/. [5] M. M. Buddhikot, P. Kolodzy, S. Miller, K. Ryan, and J. Evans. Dimsumnet: New directions in wireless networking using coordinated dynamic spectrum access. In IEEE WoWMoM, pages 78-85, Jun. 2005. [6] V. Brik, E. Rozner, S. Banarjee, and P. Bahl. Dsap: A protocol for coordinated spectrum access. In IEEE DySPAN, pages 611-614, Nov. 2005. [7] J. So and N. H. Vaidya. Multi-channel mac for ad-hoc networks: Handling multi-channel hidden terminals using a single transceiver. In ACM Mobihoc, pages 222-233, May 2004. [8] R. Maheshwari, H. Gupta, and S. R. Das.Multichannel mac protocols for wireless networks. In IEEE SECON, pages 393-401, Sep. 2006. [9] P. Kyasanur, J. So, C. Chereddi, and N. H. Vaidya. Multi-channel mesh networks: Challenges and protocols. IEEE Wireless Communications, 13(2):30-36, Apr. 2006. [10] Y. Yuan, P. Bahl, R. Chandra, P. A. Chou, J. I. Ferrell, T. Moscibroda, S. Narlanka, and Y. Wu. Knows: Kognitiv networking over white spaces. In IEEE DySPAN, Apr. 2007. [11] J. Zhao, H. Zheng, and G. H. Yang. Distributed coordination in dynamic spectrum allocation networks. In IEEE DySPAN, pages 259-268, Nov. 2005.
