Enhancing VANET Performance by Joint ... - Semantic Scholar

IEEE TRANSACTIONS ON PARALLEL AND DISTRIBUTED SYSTEMS, VOL. XX, NO. X, MONTH 2011

1

Enhancing VANET Performance by Joint Adaptation of Transmission Power and Contention Window Size Danda B. Rawat, Member, IEEE, Dimitrie C. Popescu, Senior Member, IEEE, Gongjun Yan, Member, IEEE, and Stephan Olariu, Senior Member, IEEE Abstract—In this paper we present a new scheme for dynamic adaptation of transmission power and contention window (CW) size to enhance performance of information dissemination in Vehicular Ad-hoc Networks (VANETs). The proposed scheme incorporates the Enhanced Distributed Channel Access (EDCA) mechanism of 802.11e and uses a joint approach to adapt transmission power at the physical (PHY) layer and quality-of-service (QoS) parameters at the medium access control (MAC) layer. In our scheme, transmission power is adapted based on the estimated local vehicle density to change the transmission range dynamically, while the CW size is adapted according to the instantaneous collision rate to enable service differentiation. In the interest of promoting timely propagation of information, VANET advisories are prioritized according to their urgency and the EDCA mechanism is employed for their dissemination. The performance of the proposed joint adaptation scheme was evaluated using the ns-2 simulator with added EDCA support. Extensive simulations have demonstrated that our scheme features significantly better throughput and lower average end-to-end delay compared with a similar scheme with static parameters. Index Terms—Vehicular networks, VANETs, broadcast, contention window adaptation, message differentiation, transmission power adaptation, QoS, medium access control protocol, 802.11e EDCA, intelligent transportation system.

F

1

I NTRODUCTION

W

are witnessing an unmistakable convergence of Vehicular Ad-hoc Networks (VANET) and Intelligent Transportation Systems (ITS) that is poised to bring about a revolutionary leap by making our roadways and streets safer and the driving experience more enjoyable [1]. Working in tandem with the fielded ITS infrastructure, VANET is expected to enhance the awareness of the traveling public by aggregating, propagating and disseminating up-tothe-minute information about existing or impending traffic-related events. In support of their mission, VANET communications, employing a combination of Vehicle-to-Vehicle (V2V) and Vehicle-to-Infrastructure (V2I) wireless communication are expected to integrate the driving experience into a ubiquitous and pervasive network that will enable novel traffic monitoring and incident detection paradigms [2], [3]. It is worth noting that the vast majority of the traffic advisories are of a general interest and, therefore, benefit from being broadcast. For instance, when a E

• D. B. Rawat and D. C. Popescu are with the Department of Electrical and Computer Engineering, Old Dominion University, Norfolk, VA 23529. E-mail: {drawa001,dpopescu}@odu.edu. • G. Yan is with the Department of NMIS, Indiana University Kokomo, IN 46904, Email: [email protected]. • S. Olariu is with the Department of Computer Science, Old Dominion University, Norfolk, VA 23529. E-mail: olariu @cs.odu.edu. This work was supported in part by the NSF under grant CNS-0721586 and was presented in part at the 70th IEEE Vehicular Technology Conference – VTC 2009 Fall. Manuscript received 11 May 2010; revised 1 Oct. 2010. Recommended for acceptance by I. Stojemenovi´c. For information on obtaining reprints of this article, please send e-mail to: [email protected], and reference IEEECS Log No. TPDS-2010-05-0290. Digital Object Identifier no.

traffic incident occurs, all the vehicles on the road benefit from timely and accurate information dissemination allowing the drivers to make informed decisions. Thus, reliability and low delay are extremely important factors in VANET safety applications. It is widely known that, due to high-speed mobility, V2V and V2I communication links tend to be shortlived. Thus, it is important to propagate traffic-related information toward a certain region of interest instead of sending to a particular vehicle; moreover, one of the best ways of propagating traffic-related advisories towards a particular region is some form of (controlled) broadcast transmission. One strategy of increasing duration of communication links in VANET is by increasing the transmission range in sparse traffic conditions where only a few vehicles may be present on the road. However, increasing the transmission range may generate high levels of disruptive interference and high network overhead in dense traffic conditions. It follows that dynamic adaptation of transmission power in response to changing traffic density is a critical requirement in VANET. In addition, in order to propagate emergency messages in a timely manner, VANET must support some form of message differentiation, similar in spirit to service differentiation for QoS in the contentionbased channel access mechanism EDCA of 802.11e [4]. To implement this strategy, different priority levels can be assigned to various traffic-related messages according to their urgency or delay requirements. For example, messages related to an incident on the roadway should be propagated to the target region on time and in an accurate manner in order to avoid congestion and potential secondary accidents.

c 2011 IEEE 1045-9219/11$26.00 ⃝


1.1

Related Work

Because of the advantages it offers, the IEEE 802.11 wireless standard is used by a host of protocols for information forwarding and dissemination in VANET. In 2004, the IEEE Task Group p (TGp) started to develop IEEE 802.11p [5] by amending the IEEE 802.11 standard to include vehicular communication operating at speeds up to 200 km/h and with a communication range as large as 1000 m. The IEEE 802.11p standard was designed to operate at 5.9 GHz and data rates up to 27 Mb/s with the stated goal of supporting V2I communication or V2V communication in the context of the FCC-mandated Dedicated Short Range Communication (DSRC) [6]. To set the stage for the proposed scheme, it is appropriate to note that the use of the EDCA mechanism of 802.11e in the context of VANET was discussed by Suthaputchakun and Ganz [7] where a prioritybased scheme for V2V communications was proposed. However, while the authors of [7] proposed incorporating of the EDCA mechanism of 802.11e in VANET, they did not address the problem of adapting QoS parameters or that of adapting transmission power according to local traffic conditions. To the best of our knowledge the problem of adapting transmission power in VANET based on vehicle density was first discussed by Artimy [8], while the problem of dynamically adapting the CW size for reliable broadcast in VANET was discussed by Balon and Guo [9]. However, the authors of [9] only considered the channel access time according to the urgency of messages and their delay requirements, without considering the adaptation of transmission power, or the prioritization of messages according to their urgency, or the adaptation of the CW size for transmission opportunity, which can enhance system throughput while reducing end-to-end message delay [10], [11]. 1.2

Our Contribution

Our work was motivated by the observation that the existing schemes [8], [9] did not take into account the dynamically changing topology of VANET and, consequently, kept either the transmission power of a vehicle or the QoS related parameters fixed. Indeed, the major contribution of this work is a new scheme for joint adaptation of transmission power at the PHY layer and of the CW size at the MAC layer, according to local vehicle density and network condition, respectively. The proposed scheme adapts transmission power dynamically based on estimated local traffic/vehicle density. For estimating local vehicle density, we use a different approach than the one proposed by Artimy [8]; as it turns out, our traffic density approximation is more accurate than the one in [8], resulting in a more appropriate transmission range. In addition, we prioritize messages according to their urgency by

2

incorporating IEEE 802.11e EDCA for timely propagation of high priority messages towards the destination region. We assume that the vehicles start running the proposed scheme as soon as they receive traffic-related messages and note that issues related to incident detection, admission control, as well as security and privacy of transmitted messages and participating vehicles, which are important for ITS, are outside of the scope of this paper. The remainder of this work is organized as follows: In Section 2, we discuss the EDCA mechanism of 802.11e and its role in VANET, and we formally state our problem. In Section 3 we present the proposed scheme for joint dynamic adaptation of transmission power and CW values based on vehicle density estimation and network condition. In Section 4 we formally state the algorithm for joint power and CW adaptation followed by presentation of numerical results obtained from simulations in Section 5. Finally, Section 6 offers concluding remarks and directions for future investigations.

2 T HE EDCA M ECHANISM OF 802.11 E AND ITS R OLE IN VANET

IEEE

The IEEE 802.11 standard plays a major role in wireless networking. Due to their simplicity, scalability, flexibility, and cost effectiveness, wireless local area networks (WLAN) based on IEEE 802.11 are among the most widely deployed WLAN technologies. The fundamental access mechanism of IEEE 802.11 is applicable to VANET communications, which use IEEE 802.11p [5], a modified version of IEEE 802.11a. It is widely known that the baseline IEEE 802.11 standard does not provide for the service differentiation necessary for supporting QoS for time critical data such as voice traffic in WLAN [12], [13]. In order to address the issue of service differentiation the IEEE 802.11e standard [4] specifies the distributed contention-based channel access mechanism, referred to as EDCA. The EDCA is available in the ad-hoc mode where no infrastructure is available. The EDCA scheme relies on CSMA/CA along with a slotted Binary Exponential Backoff (BEB) mechanism for contention-based channel access [4] and supports MAC-level QoS and prioritization of different data/traffic by defining multiple Access Categories (ACs) with different CW and Arbitration Inter Frame Space (AIFS) values. According to [4] a station with QoS implements four access categories (ACs) and there is a set of EDCA parameters associated with each AC. These parameters include AIFS [AC] and CW with its minimum and maximum value CWmin [AC] and CWmax [AC], respectively. Each AC from every station independently starts a backoff timer after detecting that the channel is idle for an AIFS [AC] interval and competes with other ACs for channel access and


(a) TN of possible reachable neighbor vehicles.

3

(b) AN of reachable neighbor vehicles.

Fig. 1. Estimating the local vehicle density K on the road for a given vehicle for its transmission range (TR). the opportunity to transmit. For each AC, the backoff period is selected from a uniform distribution over [0, CW [AC]]. The CW size is initially assigned CWmin and doubles when transmission fails, up to CWmax. The CW size is re-initialized when the CW reached CWmax, and the process is repeated. We note that the smaller CWmin [AC], the shorter the channel access delay for the corresponding priority and hence the better the chance of the station to access the channel in a given traffic condition. When an application is admitted, it has a number of QoS parameters. If two or more backoff timers within the same station finish backoff at the same time, there will be a virtual collision which will be solved by the station’s internal scheduler. We note that MAC protocols for VANET have to consider different types of traffic messages as well as a rapidly changing network topology. For example, it is highly desirable for emergency messages related to traffic incidents on the roadway to have higher priority than other messages in order to get rapid channel access, and thus prioritization of different messages according to their urgency is an important requirement in VANET. As a consequence, incorporating EDCA in VANET enables better message differentiation and ensures that the high priority messages get transmission opportunities on a preferential basis. Furthermore, in congested traffic the network topology is very dense, while, as soon as the congested region is passed the network topology may become sparse again. In the context of broadcast transmission used for V2V communication, high transmission power in a region with high vehicle density results in high network load. Therefore, vehicle density in a given region is a useful metric for adapting the transmission power. We present a new method which combines the advantage of dynamic adaptation of transmission power at the PHY layer as a function of vehicle density

with dynamic adaptation of CW size in EDCA at the MAC layer to enhance the performance of V2V communications in VANET. The proposed approach ensures that propagation and dissemination of prioritized messages will occur with high throughput and low end-to-end delay.

3

T RANSMISSION P OWER A DAPTATION P RIORITIZATION OF M ESSAGES

AND

In this section, we describe how transmission range and transmission power are calculated based on local density of vehicles and network conditions, and how different messages are assigned different priorities based on their urgency. One of the starting points of our investigation was provided by the following expression derived in Artimy [8] for the transmission range (TR) based on the estimated local vehicle density √ { } L ln L T R = min L(1 − K), + αL (1) K where • α is a constant from traffic flow theory [8], • L is the length of the road segment over which the vehicle estimates its initial local vehicle density, and • K is the local vehicle density for a given vehicle, calculated as the ratio K = AN T N of the actual number (AN ) of vehicles on the road that are present within its transmission range to the total number (T N ) of vehicles that can be present on the road for current transmission range, travel speed and safety separation distance as shown schematically in Fig. 1. We note that the method used for estimating K in [8] is based solely on the vehicle’s movement and may not always give a good estimate of the local traffic density K. For instance, when a given vehicle moves


to the theoretical value obtained when the actual vehicle density K is used in equation (1) instead of its estimates.

Normalized transmission range (TR/L)

1.2 1

Theoretical Reference paper [8] Proposed scheme

3.1 Transmission Power Adaptation

0.8 0.6 0.4 0.2 0 0

4

0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 Local vehicle density (K)

1

Fig. 2. Comparison of normalized transmission range (T R/L) vs. local vehicle density K for the proposed scheme (that is running only Stage 1 of Algorithm 1) and for the method used in [8]. at low speed the method in [8] will estimate that the local vehicle density K is high, while when it moves at high speed it will estimate a low vehicle density K. In order to improve the accuracy of the local vehicle density estimate K we employ a different method which uses information obtained from the interaction of a given vehicle with other vehicles in the network. For this purpose, we note that in the DSRC standard a given vehicle exchanges its status with neighboring vehicles approximately 10 times each second [14], and individual vehicles can use this information to estimate the actual number of vehicles AN in their vicinity by using the 12-bit sequence number of the IEEE 802.11 MAC header. We note also note that this method does not introduce significant network overhead since it exploits the periodic message in DSRC enabled systems, and that a similar approach has been used in [9] in a different context where it has used to measure the collision. For example, consider that the current TR for a given vehicle is 600 m that is obtained from equation (1) with L = 1000 m and the vehicles on the same lane maintain average safety separation distance of 20 m. For a separated highway with two lanes of traffic in each direction, the T N for the given vehicle is calculated as T N ≈ [600/20] ∗ 2 ∗ 2 = 120, while the AN for the given vehicle is calculated based on the received information from its neighboring vehicles. Assuming for example that AN = 65 we calculate the estimated vehicle density K = AN/T N ≈ 0.54. Using this value of K the given vehicle will update its transmission range using equation (1). The transmission range TR obtained by using the outlined method for estimating K was compared to that in [8] and the results, plotted in Fig. 21 , show that our method provides a better estimates of the local vehicle density and, consequently, of the normalized transmission range TLR , which turns out to be closer 1. The simulation setup used in obtaining Fig. 2 is described in the supplementary electronic document file.

Once the transmission range is obtained using equation (1), we need to map it to an actual transmission power value. For this purpose, we use the lookup table captured in Table 1 containing the transmit power values corresponding to different transmission ranges. We note that the data in Table 1 was obtained by simulations of basic wireless propagation models for different VANET scenarios and a specific power value is assigned for a given transmission range interval to include urban, city, and rural environments. We also note that the lookup table approach is faster since no computations are required. TABLE 1 Lookup table for Transmission Power Corresponding to a Given Transmission Range Transmission Range (m) 0—9 10 — 49 50 — 100 100 — 125 126 — 149 150 — 209 210 — 299 300 — 349 350 — 379 380 — 449 450 — 549 550 — 649 650 — 749 750 — 849 850 — 929 930 — 970 971 — 1000 >1000

Transmission Power (dBm) -20 -12 -5 -3 1 4 6 10 12 14 17 20 24 27 29 31 32 N/A in DSRC

3.2 Prioritization of Messages As discussed in Section 2, the IEEE 802.11e EDCA has the service differentiation to provide QoS for different types of messages: voice traffic, video traffic, best effort traffic and background traffic [4]. To incorporate the EDCA mechanism in VANET we categorize the different messages according to their urgency and delay requirements [7] as listed in Table 2. TABLE 2 Message Priorities [7] Priority (traffic in EDCA) Priority 1: (Voice – AC(3)) Priority 2: (Video – AC(2)) Priority 3: (Best-effort – AC(1)) Priority 4: (Background – AC(0))

Message Types in VANET Accident messages, etc. Accident indication messages Warning related messages General messages

The different access categories in EDCA will have different QoS parameters associated with them. Table 3 gives the QoS parameters corresponding to the


ACs in 802.11e EDCA. The higher the access category number, the higher the channel access or transmission opportunity will be. That means the CWmin value for AC(3) will be the least among all ACs. The backoff counter drawn uniformly from [0, CW[AC]] will have an initial value of CWmin, implying that AC (3) will get the highest transmission opportunity over others. Moreover, high priority classes in turn use a shorter inter frame spacing (IFS) and a smaller CW size so that they will get preferential treatment over lower priority classes. Each vehicle will have four different queues, one for each priority class with a virtual collision handler to handle internal collisions. 3.3

Contention Window Size Adaptation

In order to support message differentiation for different types of messages listed in Table 2 the size of the CW in Table 3 should also be adapted taking into account the fact that vehicles that have higher priority messages should not get the chance to be greedy (i.e., get channel access most of the time) while higher priority messages should not be waiting for a long time for the opportunity to transmit. TABLE 3 Priority Specific Parameters [4] AC 0 1 2 3

CWmin CWmin CWmin (CWmin+1)/2 1 (CWmin+1)/4-1

CWmax CWmax CWmax CWmin (CWmin+1)/2

AIFS 2 1 1 1

We note that a vehicle attempting to get transmission opportunity must wait for the channel to remain idle for the duration of the AIFS before starting its back-off timer. We also note that holding channel access for a long time for higher priority messages may result in a delay in message propagation which will not be able to notify and/or prevent incidents on the roadway such as congestion and traffic-jam buildups. Therefore, to mitigate these adverse effects the dynamic adaptation of QoS parameters, in particular CW sizes, for different access categories is essential since the backoff counter value is obtained uniformly from [0, CW [AC]] and the initial CW value is CWmin. In our proposed approach, the size of the CW may either increase or decrease, and CW adaptation is carried out by applying the well-known approach used in the IEEE 802.11 algorithms by which the size of window CW [AC] is varied by a factor of two. In other words, the window size is doubled if one has to increase the size, and is reduced by half if one has to decrease its size. The CW size will continue to increase until it reaches to maximum size of the window, CWmax [AC] after which it will be re-initialized to CWmin. Thus, in our proposed approach the window size fluctuates according to the network conditions observed by a vehicle while in conventional 802.11 technologies (including EDCA) the size of the window

5

1000

Throughput (Kbps)

800

600

400

200

0 0

Proposed scheme Reference paper [11] 50

100 150 200 Simulation Time (s)

250

300

Fig. 3. Throughput vs. simulation time for proposed scheme (i.e by keeping transmission power fixed and without running the power adaptation steps of Algorithm 1) and the method used in Chen et al. [11] for higher priority messages (AC(2) and AC(3)). remains fixed no matter what the network condition is. We note that the increase in CW [AC] values (for all ACs maintaining the hierarchy of CW [AC] values as in EDCA [4]) when the network is congested, will give less opportunity for all ACs to reduce network load because of broadcast and rebroadcast. Similarly the decrease in CW [AC] values (for all ACs maintaining hierarchy of CW values) when the network has less or no collision, will give higher opportunity for all ACs. In both cases the preferential treatment is preserved by hierarchical increments in corresponding CW [AC] values. The local state of the network can be determined as in [9] by using the record of sequence numbers corresponding to individual vehicles from which it receives messages. By using lost sequence numbers each vehicle can calculate the approximate percentage of lost frames sent by other vehicles to in a given period of time. Using these statistics and based on the local reception rate, each vehicle determines the local state of the network as suggested by Balon and Guo [9] and can use it to adapt the CW parameters in the EDCA mechanism. The adaptation of CW size according to network conditions results in high throughput and lower delay for high priority messages, while lower priority messages also get channel access but with lower preference over higher priority ones [10]. The throughput obtained by the proposed scheme using dynamic adaptation of CW values based on network condition was compared to the one reported by Chen et al. in [11] and the results, plotted in Fig. 32 , show that our scheme outperforms the one in [11].

4

T HE A LGORITHM

Based on the methods discussed in the previous section we present an algorithm which adapts both transmitted power and CW size and which should be 2. The simulation setup used to obtain Fig. 3 is described in the supplementary electronic document file.


run by individual vehicles periodically to ensure that proper updates of transmission power and CW [AC] values occur according to the local vehicle density and the network condition respectively. The algorithm is formally stated as Algorithm 1 and consists of two stages: transmit power adaptation and CW size adaptation. 4.1

Dynamic Adaptation of Transmission Power

Initially, individual vehicles start with an arbitrary transmission power and listen for information from other vehicles. Once a vehicle receives message packets from other vehicles, it starts to analyze the sequence numbers and to count the vehicles around its locale. In order to mitigate the adverse effects of high transmission power and to increase the duration of the communication link in case of low traffic density for inter-vehicle communication, each vehicle dynamically adapts its transmission power based on the estimated local vehicle density. The vehicle density within its transmission range is calculated using the method discussed in Subsection 3.1 which is based on observation of packets that are currently received and does not introduce significant network overhead to identify the neighbors of the vehicle [9]. Using the estimated vehicle density the algorithm calculates the transmission range using equation (1) and then sets up the corresponding transmission power using the look up Table 1. We note that maximum transmit power corresponding to maximum transmission range is selected when either the local vehicle density K is sparse, that is, lower than some application-dependent threshold value τ1 or when the vehicle needs to transmit priority 1 messages3 . The choice of threshold value τ1 plays a significant role in the implementation of the algorithm and is also adapted according to the local density observed by a given vehicle periodically based on the threshold history for the given vehicle. Moreover, since the algorithm is employed in a distributed manner threshold values for different vehicles might be different. 4.2

Dynamic Adaptation of CW Size

Dynamic adaptation of CW size causes changes in the back-off counter so that timely transmission of messages occurs according to the network conditions, namely the perceived collision rate and local vehicle density. The CW size adaptation is performed as discussed in Subsection 3.3 in response to network conditions estimated by analyzing the received sequence numbers at MAC layer as discussed in Section 3. The estimated collision rate is an indication of how 3. Priority 1 messages are transmitted in case of accidents or from emergency vehicles as shown Table 3. They have low delay requirements and should propagate on time in a single hop (if possible) within the maximum transmission range as all vehicles in the destination region seek emergency related messages so that they respond to the situation according to the message received.

6

Input: Maximum transmission range R value, traffic constant α from traffic flow theory [8] and threshold value τ1 , and default CW[AC] values ∀ ACs [4] and threshold value τ2 Output: Transmission power P corresponding to the calculated transmission range T R and adapted CW[AC] values ∀ ACs while (“messages are received” ) do Stage 1: Transmit power adaptation. foreach Time do Estimate the local vehicle density, K, based on the approach mentioned in Section 3 (B).; if K < τ1 or (Highest priority message) then Assign transmission range T R equal to maximum value R.; else Calculate TR using equation (1).; end Assign the suitable transmission power corresponding to calculated transmission range T R using look up Table 1.; end Stage 2: CW size adaptation. foreach Time do Estimate the collision rate based on the approach mentioned in Section 3.; if estimated collision rate > τ2 then Increase the corresponding CW [AC] values for all ACs using the approach mentioned in Section 3.3. else if estimated collision rate < τ2 then Decrease the corresponding CW [AC] values for all ACs using the approach mentioned in Section 3.3. else Maintain corresponding CW [AC].; end end end Algorithm 1: Joint Adaptation of Transmit Power and CW Size. congested the network is and how information flow from a vehicle should be controlled. The dynamic adaptation of the CW size is regulated by a threshold value τ2 . We note that, as it was the case with the power adaptation, proper choice of the threshold value τ2 in CW adaptation is important and affects the performance of the system, and that the threshold τ2 may also be adapted periodically based on the network conditions and on the threshold history for a given vehicle.

5

S IMULATIONS

AND

N UMERICAL R ESULTS

In order to illustrate the performance of the proposed joint dynamic adaptation scheme we have simulated Algorithm 1 and have compared it with the default


Value 150 12 - 30 m/s Poisson distribution 512 Bytes 10 packets per second Table 3 0.25 1000m (0.622 mile) 1 lane in each direction 0.5 0.5

In order to incorporate the EDCA mechanism in VANET using ns-2, we have mapped the suitable messages with the corresponding service differentiated EDCA access categories as listed in Table 2, and have assigned the QoS parameters for each AC once the message becomes available at the vehicle. We illustrate the performance of the proposed joint dynamic adaptation scheme for transmit power and contention window by comparing it with the default scheme in terms of the overall throughput and endto-end message delay in a scenario consisting of approximately 50% highest priority messages AC(3) and approximately 50% other types of messages AC(0)AC(2) in the network. This corresponds to the third simulation experiment as described in the supplementary electronic document file. The throughput variation is plotted Fig. 4, and the average end-to-end message delay is plotted in Fig. 5. From Fig. 4 we note that initially (up to approximately 48 s) the throughput value is similar for both schemes since the vehicles might not be able at first to detect their neighbors in order to estimate vehicle density and adjust their transmission power and CW size values. As soon as Algorithm 1 start adjusting the transmission power and CW size according to vehicle density and network condition the overall throughput of the proposed scheme is higher than that of the default scheme. Fig. 5 shows the average end-to-end delay for (a) all messages, (b) the highest priority – AC(3) only – messages, and (c) the other AC(0) – AC(2) categories (all message types except highest priority). In all cases there is no delay in the beginning stage (up to approximately 48 s) since all messages get transmission opportunity as soon as they are available at the vehicle. Delay at around 50 s is high since vehicles could not be able to adapt the power and CW values according to the vehicle density and network condition during the initial stage of the simulation.

Throughput (Kbps)

2000

1500

1000

500

0 0

50


250

300

Fig. 4. Overall throughput variation. 2 default scheme proposed scheme 1.5

Average Delay (s)

Parameter number of vehicles vehicle speed Data generation packet size generation rate default QoS values Traffic constant α segment L road configuration threshold τ1 threshold τ2

default scheme proposed scheme

1

0.5

0 0

50


250

300

(a) For all messages 2 default scheme proposed scheme 1.5

Average Delay (s)

TABLE 4 Simulation parameters

2500

1

0.5

0 0

50


250

300

(b) For AC(0), AC(1) and AC(2) 1 default scheme proposed scheme 0.8

Average Delay (s)

EDCA scheme which has fixed values for transmission power and QoS parameters [4]. The simulation setup is described in the supplementary electronic document file and its simulation parameters are summarized in Table 4.

7

0.6

0.4

0.2

0 0

50


250

(c) For AC(3)

Fig. 5. Average end-to-end message delay.

300


As simulation time increases, the delay decreases since individual vehicles adapt their transmission power and CW values dynamically according to vehicle density and network condition. As can be observed from Fig. 5 the average end-to-end delay is lower for the proposed scheme than for the default scheme in all cases.

6

C ONCLUSION

In this paper we presented a new scheme for reliable broadcast transmission in vehicular communication with joint dynamic adaptation of transmission power and CW size. The scheme incorporates the EDCA medium access mechanism of IEEE 802.11e in VANET to set priority for different messages according to their urgency, and consists of an algorithm by which individual vehicles dynamically adapt transmission powers according to the estimated local vehicle densities and adjust CW [AC] for all ACs based on data collision rate on the network. Performance of the proposed scheme is illustrated with numerical results obtained from simulations which show that better throughput is achieved with lower delay than when the default scheme is used.

ACKNOWLEDGMENTS The authors are grateful to the anonymous reviewers for their constructive comments on the paper.

R EFERENCES [1]

W. Chen, S. Cai, and T. Inc, “Ad Hoc Peer-to-peer Network Architecture for Vehicle Safety Communications,” IEEE Communications Magazine, vol. 43, no. 4, pp. 100–107, 2005. [2] O. Riva, T. Nadeem, C. Borcea, and L. Iftode, “Context-Aware Migratory Services in Ad-Hoc Networks,” IEEE Transactions on Mobile Computing, vol. 6, no. 12, pp. 1313–1328, Dec. 2007. [3] J. Nzounta, N. Rajgure, G. Wang, and C. Borcea, “VANET Routing on City Roads Using Real-Time Vehicular Traffic Information,” IEEE Transactions on Vehicular Technology, vol. 58, no. 7, pp. 3609–3626, September 2009. [4] Part 11: Wireless LAN Medium Access Control (MAC) and Physical Layer (PHY) specifications Amendment 8: Medium Access Control (MAC) Quality of Service Enhancements, IEEE Std. 802.11e, 2005. [5] D. Jiang and L. Delgrossi, “IEEE 802.11p: Towards an International Standard for Wireless Access in Vehicular Environments,” in IEEE Vehicular Technology Conference, 2008. VTC Spring 2008, Marina Bay, Singapore, May 2008, pp. 2036–2040. [6] R. Sengupta and Q. Xu, “DSRC for Safety Systems,” vol. 10, no. 4. California PATH – Partners for Advanced Transit and Highways, 2004, pp. 2–5. [7] C. Suthaputchakun and A. Ganz, “Priority Based Inter-Vehicle Communication in Vehicular Ad-Hoc Networks using IEEE 802.11e,” in IEEE 65th Vehicular Technologies Conference- VTC 2007-Spring, Apr. 2007. [8] M. M. Artimy, “Assignment of Dynamic Transmission Range Based on Estimation of Vehicle Density,” in IEEE Transactions on Intelligent Transportation Systems, vol. 8, no. 3, Sep. 2007, pp. 400–412. [9] N. Balon and J. Guo, “Increasing Broadcast Reliability in Vehicular Ad Hoc Networks,” in VANET ’06: Proceedings of the 3rd ACM international workshop on Vehicular ad hoc networks, Los Angeles, California, Sep. 2006, pp. 104–105. [10] D. B. Rawat, D. C. Popescu, and M. Song, “Performance Enhancement of EDCA Access Mechanism of IEEE 802.11e Wireless LAN,” in Proceedings 2008 IEEE Radio and Wireless Symposium – RWS 2008, Orlando, FL, Jan. 2008, pp. 507–510.

8

[11] X. Chen, H. Zhai, X. Tian, and Y. Fang, “Supporting QoS in IEEE 802.11e Wireless LANs,” IEEE Transactions on Wireless Communications, vol. 5, no. 8, pp. 2217 – 2227, Aug. 2006. [12] G.-M. Muntean and N. Cranley, “Resource Efficient qualityOriented Wireless Broadcasting of Adaptive Multimedia Content,” IEEE Transactions on Broadcasting, vol. 53, no. 1, pp. 362 – 368, Mar. 2007. [13] Y. Xiao, “Performance Analysis of Priority Schemes for IEEE 802.11 and IEEE 802.11e Wireless LANs,” IEEE Transactions on Wireless Communications, vol. 4, no. 4, pp. 1506 – 1515, Aug. 2005. [14] “Vehicle Safety Communications Project Task 3 Final Report: Identify Intellegent Vehicle Safety Applications Enabled by DSRC,” Vehicle Safety Communications Consortium consisting of, BMW, DaimlerChrysler, Ford, GM, Nissian, Toyota, and VW., Mar. 2005. Danda B. Rawat received the Bachelor’s Degree in Computer Engineering in 2002 and the Master’s Degree in Information and Communication Engineering in 2005 from Tribhuvan University, Kathmandu, Nepal. He received the PhD degree in Electrical and Computer Engineering from Old Dominion University in 2010. His research interests are in the areas of wireless communications and wireless cellular/ad-hoc networks. Dimitrie C. Popescu received the Engineering Diploma and M.S. degrees from the Polytechnic Institute of Bucharest, Romania, and the Ph.D. degree from Rutgers University, all in Electrical Engineering. His research interests are in the areas of wireless communications, digital signal processing, and control theory. Dr. Popescu is currently an Assistant Professor in the Department of Electrical and Computer Engineering, Old Dominion University. He is an associate editor for IEEE Communications Letters, he has served as technical program chair for the vehicular communications track of the IEEE VTC 2009 Fall, finance chair for the IEEE MSC 2008, and technical program committee member for the IEEE GLOBECOM, ICC, WCNC, and VTC conferences. Gongjun Yan received his Ph.D. in Computer Science from Old Dominion University in 2010, and is currently an Assistant Professor in the Department of Natural, Information, and Mathematical Sciences, Indiana University Kokomo. He has been working on the issues surrounding Vehicular Ad-Hoc Networks, Sensor Networks and Wireless Communication. His main research areas include security, privacy, routing, and healthcare. Dr. Yan applies mathematical analysis to model behavior of complex systems and integrates existing techniques to provide comprehensive solutions. Stephan Olariu received PhD degree in Computer Science from McGill University in 1986. He is currently a professor in the Computer Science Department, Old Dominion University, and is a world-renowned technologist in the areas of wireless networks, mobile multimedia systems, parallel and distributed systems, architectures and networks. He was invited and visited more than 120 universities and research institutes around the world lecturing on topics ranging from wireless networks and mobile computing, to biology-inspired algorithms and applications, to telemedicine, to wireless location systems, and security. Dr. Olariu is an Associate Editor of Networks and IEEE Transactions on Parallel and Distributed Systems and serves on the editorial board of Journal of Parallel and Distributed Computing, Journal of Ad hoc and Sensor Networks, and Parallel, Emergent and Distributed Systems.