University of California Transportation Center
UCTC- FR- 2010- 33
An Empirical Study of Inter- Vehicle Communication Performance
Using NS- 2
Jaeyoung Jung, Rex Chen, Wenlong Jin,
R. Jayakrishnan, and
Amelia C. Regan
University of California, Irvine
August 2010
1 AN EMPIRICAL STUDY OF INTER- VEHICLE
2 COMMUNICATION PERFORMANCE USING NS- 2
3
4 Jaeyoung Jung
5 Ph. D. Candidate
6 Institute of Transportation Studies
7 Department of Civil and Environmental Engineering
8 University of California, Irvine
9 Irvine, CA 92697- 3600 U. S. A
10 Phone: + 1– 949– 824– 5989 / FAX: + 1– 949– 824– 8385 / Email: jaeyounj@ uci. edu
11
12 Rex Chen
13 Ph. D. Candidate
14 Institute of Transportation Studies
15 Department of Computer Science
16 University of California, Irvine
17 Irvine, CA 92697- 3600 U. S. A
18 Phone: + 1– 949– 824– 5989 / FAX: + 1– 949– 824– 8385 / Email: rex@ uci. edu
19
20 Wenlong Jin
21 Assistant Professor
22 Institute of Transportation Studies
23 Department of Civil and Environmental Engineering
24 University of California, Irvine
25 Irvine, CA 92697- 3600 U. S. A
26 Phone: + 1– 949– 824– 1672 / FAX: + 1– 949– 824– 8385 / Email: wjin@ uci. edu
27
28 R. Jayakrishnan
29 Associate Professor
30 Institute of Transportation Studies
31 Department of Civil and Environmental Engineering
32 University of California, Irvine
33 Irvine, CA 92697- 3600 U. S. A
34 Phone: + 1– 949– 824– 2172 / FAX: + 1– 949– 824– 8385 / Email: rjayakri@ uci. edu
35
36 Amelia C. Regan
37 Professor
38 Institute of Transportation Studies
39 Department of Computer Science
40 University of California, Irvine
41 Irvine, CA 92697- 3600 U. S. A
42 Phone: + 1– 949- 824- 5156 / FAX: + 1– 949– 824– 4163 / Email: aregan@ uci. edu
43
44
45 Submitted for 17th ITS World Congress
46
Jung, Rex, Jin, Jayakrishnan, and Regan
2
1
2 ABSTRACT
3 In recent years, there has been increasing interest in inter- vehicle communications ( IVC)
4 based on wireless networks to collect and distribute traffic information in various Intelligent
5 Transportation Systems applications. In this paper, we study the performance of IVC under
6 various traffic and communication conditions by means of simulation analysis. We consider
7 impacts of shock waves, transportation network, traffic densities, transmission ranges, and
8 multiple information sources. We used a state- of- the- art communication network simulator
9 ns- 2 to measure the probability of success ( success rate) and message delivery ratio ( MDR)
10 for flooding- based IVC communication. For reasonable realism in the deployment scenario,
11 we assume that only a partial set of vehicles on the road are equipped with communication
12 devices, according to the market penetration rate. A Monte- Carlo simulation method is used,
13 with repeated random sampling of IVC- equipped vehicles. The results indicate how these
14 parameters can impact the performance of IVC communications. By comparing the flooding-15
based approach ( theoretical and simulation) and simulation results using AODV ( Ad Hoc On-
16 Demand Distance Vector), we conclude the importance of traffic environment and network
17 protocol in determining the MDR for IVC communication.
18
19
Jung, Rex, Jin, Jayakrishnan, and Regan
3
1
2 AN EMPIRICAL STUDY OF INTER- VEHICLE COMMUNICATION
3 PERFORMANCE USING NS- 2
4 INTRODUCTION
5 With increasing availability of wireless communication devices, Inter- Vehicle
6 Communications ( IVC) is an emerging technology that can help vehicles share or propagate
7 useful information for drivers for traffic congestion mitigation, safety warning, and traffic
8 management. The Federal Communication Commission ( FCC) of USA has allocated a
9 spectrum of 75 MHz in 5.9 GHz range for Dedicated Short Range Communications ( DSRC)
10 ( 1). To develop Intelligent Transportation Systems ( ITS) strategies based on DSRC and other
11 wireless communication technologies, the US Department of Transportation started the
12 Vehicle Infrastructure Integration ( VII) initiative among eight others ( USDOT, 2004). In a
13 VII system, vehicles equipped with communication units and road- side stations installed by
14 transportation authorities are able to exchange information with each other through inter-15
vehicle communication, including vehicle- to- vehicle ( V2V) and Vehicle- to- Infrastructure
16 ( V2I) communications.
17 
18 As early as in the 1990s, IVC has been used to help drivers respond more promptly to
19 emergencies on a road in the California PATH automatic highway project ( 2). The Autonet
20 project at University of California, Irvine developed concepts for IVC in the late 90s, which
21 were further studied in a National Science Foundation Project from 2003 ( 3). In 2002, the
22 CarTalk project in Europe studied Advanced Driver Assistance Systems based on IVC ( 4).
23 In recent years, various stakeholders have come together to address these short- term and
24 long- term challenges and initiative efforts have been formed, such as the Europe eSafety and
25 US IntelliDrive programs.
26
27 Every year, millions of traffic accidents occur worldwide with forty thousand fatalities in US
28 and Europe alike. A central theme for transportation planners is focused on increasing road
29 safety. The European Transport Policy set the goal to reduce road fatalities by 50% by the
30 year 2010 ( 5). Furthermore, US DOT’s Research and Innovative Technology Administration
31 ( RITA) has challenged the industry to reduce traffic crashes by 90% by 2030 ( 6). As a result,
32 safety related applications with localized information exchange have been an important
33 driving force for the development of IVC.
34
35 Since the concept of Carnet ( 7) and the project of Fleetnet ( 8) were introduced in 2000, an
36 IVC system has been studied as a special case of mobile ad hoc networks ( MANET) and
37 termed as vehicular ad hoc networks ( VANET). Thus, an IVC network could develop into a
38 vehicular network ( car to car communication) or “ Internet on the road” ( 8), a possible venue
39 for publishing advertisement and infotainment information.
40 
41 In an IVC network, communication nodes, i. e., vehicles equipped with communication units,
42 usually move at high speeds and are constantly entering and leaving roadway segments. In
43 transportation networks, the density of vehicles can vary dramatically due to driving
44 behaviors and restrictions in the network geometry. The network topologies for IVC are
45 highly dynamic ( 9, 10). The performance of IVC is affected by the underlying transportation
46 network structure and vehicular traffic dynamics as well as the wireless device and
47 communication protocols.
48
Jung, Rex, Jin, Jayakrishnan, and Regan
4
There are various performance measures to analyze the 1 effectiveness of communication
2 protocols which include: connectivity, capacity, throughput, delivery ratio, end- to- end delay,
3 and packet reception rate. In our study, we evaluate the performance of IVC by measuring the
4 probability of successful information propagation and packet delivery ratio in uniform and
5 shockwave traffic streams in unidirectional roads ( one- dimension) and uniform traffic for bi-6
directional roads ( two- dimension). We use uniform traffic to compare our simulation results
7 with a theoretical model and for consistency in the speed- density relationship. We consider
8 the impact of density, transmission range, routing protocol, market penetration rate of
9 equipped vehicles, and number of information sources on success rate and message delivery
10 ratio ( MDR). We define success rate as a probability of success for information to travel
11 beyond a certain location and message delivery ratio as the percentage of data packets
12 received by the receiver from those transmitted by the information source.
13
14 In many studies, communication nodes are assumed to follow a spatial Poisson distribution
15 on a plane or to move randomly and independently in a given area. However, in real traffic
16 the movement of, and positions of vehicles are not independent of each other. Therefore, the
17 aim of this study is to understand the fundamental properties of IVC under different traffic
18 and communication scenarios. Since we assume a certain level of market penetration rate of
19 equipped vehicles, the Monte Carlo method that randomly selects equipped vehicles via
20 Bernoulli trials is used. For network simulation, we use ns- 2 ( 11) with realistic
21 communication protocol stack based on IEEE 802.11 Medium Access Control with the
22 information propagated based on a flooding scheme.
23
24 RELATED WORK
25 The fundamental performance measures in mobile ad hoc networks include multi- hop
26 connectivity, information throughput and communication delay ( 12, 13, 14). Theoretical
27 analyses of capacity and throughput of mobile ad hoc networks have revealed that per- node
28 capacity drops dramatically with the increase in the number of nodes ( 15). This has profound
29 implications on the scalability of MANETs. Through theoretical ( 16, 17, 18, 19), simulation-30
based ( 20, 21), and field studies ( 22), it has been observed that multi- hop connectivity of an
31 IVC system is highly related to the distribution of vehicles on a road, transmission range of
32 wireless units, and market penetration rate of equipped vehicles.
33 
34 As routing protocols in wireless multi- hop ad hoc networks can significantly influence
35 communication reliability and reachability ( 23), various types of routing protocols such as
36 unicast, multicast, and broadcast have been studied to evaluate the feasibility and
37 performances of ad hoc network on rectangular areas with random waypoint mobility ( 24, 25).
38 Wang et al. ( 26) studied information throughput of inter- vehicle communication in a
39 unidirectional uniform traffic stream using AODV ( 27). Similarly, it is necessary to
40 investigate how information propagation in an IVC network is affected by vehicular traffic
41 dynamics.
42
43 The rest of the paper is organized as follows. First we introduce success rate and message
44 delivery ratio as the performance measure of our study. Then, we describe our simulation
45 environment and evaluate different mobility patterns and communication scenarios. We
46 conclude with insights on the impact of traffic dynamics and network parameters in the
47 performance of an IVC system.
48
Jung, Rex, Jin, Jayakrishnan, and Regan
5
1 SIMULATION ENVIRONMENT
2 THEORETICAL MODEL
3 We first assume that whether a vehicle is equipped with communication capability or not is a
4 random occurrence based on a simple market penetration ratio, and if node and are
5 within transmission range , the probability of propagating information is set to 1. Therefore,
6 the information propagation from sender to receiver in a traffic stream is a random process,
7 and the throughput and message delivery ratio at the receiver depends on the connectivity
8 between the sender and the receiver. We denote the end node probability for vehicle to be
9 the end of a communication chain starting from sender by and the probability for
10 information to propagate from node to node by . is independent of
11 vehicles outside , where and indicate vehicle location. and
12 are defined as upstream reach and downstream reach as the farthest vehicle within its
13 transmission range , from vehicle . Finally, given vehicle positions distributed according
14 to uniform or general traffic, the recursive model of multi- hop connectivity can be written as
15
16 ,
17
18 where,
19
20
21
22 .
23
24 Further details of the model can be seen in ( 28).
25 PERFORMANCE MEASURES
26 The approach to measure success rate and message delivery ratio from an information source
27 to an equipped vehicle at location is based on the Monte- Carlo method with randomly
28 repeated simulation by Bernoulli trials, which is similar to ( 26). For the Monte- Carlo
29 simulation, we generate the mobility patterns of vehicles as and carry out
30 randomly repeated simulations. In each experiment, we have independent variables
31 which correspond to vehicles on a given traffic stream. For the Bernoulli
32 trials, we generate a random number in and if , vehicle is IVC equipped.
33 
34 For measurement of success rate, we set the most upstream vehicle as an information source
35 in uniform traffic, while in shockwave traffic scenario an information source is set at the mid-36
point of two traffic streams with varying densities. The following notations describe the
37 success rate after experiments:
38
39 • : Information propagation distance in the simulation
40 • : Indicator function for message reception at location in the simulation
41
42
43
44 • : Success rate at location
45
Jung, Rex, Jin, Jayakrishnan, and Regan
6
1 , ( )
2
3 The message delivery ratio is defined as the number of received data packets by the receiver
4 divided by the number of transmitted packet by the sender. In flooding, an information source
5 transmits a message to all neighbors within its transmission range. Subsequently, the nearby
6 nodes then transmit the message to their neighbors and finally the message is propagated to
7 all nodes in network. Although the flooding based approach incurs some unnecessary
8 overhead and inefficiencies, it can quickly disseminate information which is especially useful
9 for emergency information propagation and does not require any routing table maintenance or
10 update in the communication design. The following notations describe the message delivery
11 ratio in our experiments:
12
13 • : Total number of data packets transmitted by a source
14 • : Total number of data packets received at a receiver from a source
15 • : Message Delivery Ratio at a vehicle from a source
16
17
18
19 MOBILITY MODELS
20 We consider two mobility models, uniform traffic and shockwave traffic. For the speed-21
density relationship, we use the well- known triangular fundamental diagram ( 29, 30).
22
23
24
25 where = 104 km/ h, = 150 veh/ km/ lane, and veh/ km/ lane
26
27 In uniform traffic, vehicles are equally spaced on the road and travel at the same speed. The
28 shockwave scenario is created by two traffic streams with varying densities ( hence, different
29 speeds according to the triangular relationship) that meet on a unidirectional road.
30 SIMULATION FRAMEWORK
31 We use the network simulator ns- 2, an open- source object- oriented discrete event simulator.
32 The ns- 2 tool is the most common tool used by computer networking researchers. According
33 to a survey conducted in 2005, ns- 2 is the simulator of choice used by 43% of all published
34 ACM research papers related to mobile ad hoc networks ( 31).
35
36 When a simulation is completed, ns- 2 generates a trace (*. tr) text file which is then analyzed
37 using a scripting language such as perl and awk. In our study, since every scenario must be
38 simulated repeatedly, we build a Monte- Carlo simulation framework, nsHelper, written in
39 C++. Figure 1 illustrates the sequence of steps in the simulation framework and how the
40 custom- build 2Helper tool facilitates the Monte- Carlo method and the mobility generation,
41 data collection, and gathering of statistics related to the performance measures. A sample
42 screenshot of the visualization output produced by ns- 2 is shown in Figure 2 for a two-43
dimensional arterial network with 16 intersections.
44
Jung, Rex, Jin, Jayakrishnan, and Regan
7
1
2 Figure 1. Simulation Framework Figure 2. ns- 2 simulation
3
4 SUCCESS RATE
5 In this section, we investigate the success rate for both uniform traffic and shockwave traffic
6 by setting one vehicle as an information source, which transmits a single message of 230
7 bytes and measuring how far the message travels along the traffic stream.
8 UNIFORM TRAFFIC
9 For uniform traffic, we simulate unidirectional uniform traffic stream moving in the same
10 direction with four lanes along a 20 km highway stretch. We set the information source at the
11 most upstream point. For four lanes, the traffic densities are = 20 veh/ km and = 56
12 veh/ km, which has 800 and 1200 vehicles traveling at free flow speed ( = 104 km/ h). We
13 use the Monte- Carlo method ( = 500 times) with different transmission ranges = 0.1, 0.2,
14 0.5, and 1km with 10% market penetration rate ( = 0.1) of randomly IVC- equipped vehicles
15 in the simulation.
16
17
18 3( a) = 20 veh/ km 3( b) = 56 veh/ km
19 Figure 3. Success Rate with Uniform Traffic Steam
20
21 Figure 3 shows the success rate of a receiver at different locations ( [ 0,10] km) from the
22 sender located at distance 0. The dashed lines indicate theoretical values from an analytical
23 model ( 15). First, we see that the simulation results are consistent with the analytical model
24 and as the distance from the information source increases, the success rate decreases.
25 Communication performance is strongly affected by vehicle density and transmission range.
26 In Figure 3( a), when R = 500m, the success rate at 3 km is almost zero, while the success rate
Jung, Rex, Jin, Jayakrishnan, and Regan
8
at 3 km is more than 0.3 and the message travels more than 10 km 1 according to Figure 3( b).
2 When the transmission range is low ( i. e. 100 or 200 meters), information cannot propagate
3 more than 1 km.
4
Transmission range ( ) Traffic density ( ) and MPR 10 % ( = 0.1)
= 20 veh/ km = 56 veh/ km
= 0.1 km 105.6 m 133 m
= 0.2 km 232.22 m 422.30 m
= 0.5 km 873.14 m 2799.66 m
= 1.0 km 3572.72 m > 20 km
5 Tabble 1. Average Information Propagation Distance
6
7 Table 1 illustrates the maximal value of average information propagation distances from the
8 information source with the specified transmission ranges and traffic densities. Note that the
9 average maximum information propagation distances are generally greater than the
10 transmission range. As the message propagation in IVC is multi- hop over multiple vehicles,
11 shorter transmission range and low traffic density negatively affects the travel distance in the
12 traffic stream.
13 SHOCKWAVE TRAFFIC
14 In this section, we examine success rate in shockwave traffic scenarios. Initially, we assume
15 that we have capacity flow with = 30 veh/ km/ lane for upstream to = 0 and congested
16 flow = 40 veh/ km/ lane for downstream. Using the speed- density relationship described
17 earlier, the corresponding speeds = 104 km/ h and = 71.5 km/ h are derived respectively.
18 At time = 0, a shockwave is created and moves backward at speed = - 26 km/ h. In the
19 simulation, we assume the traffic stream length to be more than 80 km with market
20 penetration rate 10 % ( = 0.1) and transmission range = 1 km. To simulate shockwave
21 traffic, we set information source at = - 10 km in the capacity flow, density = 30
22 veh/ km/ lane and speed = 104 km/ h.
23
24
25 4( a) Flooding 4( b) Theoretical
26 Figure 4. Success Rate with Shockwave Traffic Stream
27
28 Figure 4 shows the success rates in both forward and backward directions at four instants of
29 time: = 0, = 2.3, = 4.6, and = 9.9 minutes. In the simulation, the corresponding
30 locations of information source are - 10 km, - 6 km, - 2 km, and 4.3 km, and the locations of
31 shockwaves are 0 km, - 1 km, - 2 km, and - 4.3 km. We observe that success rate is symmetric
32 with respect to information source within the same traffic density. However, it is clear that
Jung, Rex, Jin, Jayakrishnan, and Regan
9
success rate depends on traffic density and changes dramatically 1 when meeting a different
2 traffic density. Comparing Figure 4( a) with 4( b), we see that the analytical and simulation
3 results are similar initially, but are significantly different as the distance from the information
4 source increases. For example, at location 60 km, the difference in success rates for the case
5 of = 0 is more than 10%. This is attributed to the wireless communication signal
6 interference in the simulation while the theoretical model assumes guaranteed message
7 delivery within transmission range. Further, the theoretical model assumes that messages are
8 directly delivered to the farthest IVC- equipped vehicle ( most forward within range) to
9 minimize the hop count.
10
11 MESSAGE DELIVERY RATIO
12 In this section, we evaluate the performance of inter- vehicle communication by measuring
13 the message delivery ratio for vehicular network in different traffic densities, number of
14 information sources, and two- dimensional road layouts. We set the communication
15 bandwidth to 1 Mbps and information source that transmits packets at periodic intervals ( 0.02
16 sec) with a fixed packet size ( 230 bytes/ packet) in the simulation time period ( 32) over M =
17 500 simulation runs.
18 IMPACT ON ROUTING PROTOCOL
19 In this experiment, a single information source is set and follows the same communication
20 scenario as ( 26) to compare our flooding- based method with AODV. AODV is a popular on-21
demand routing protocol to deliver messages in MANETs.
22
23
24 5( a) = 56 veh/ km 5( b) = 20 veh/ km
25 Figure 5. Message Delivery Ratio with = 500 m, = 0.1
26
27 Figure 5 presents message delivery ratio for two different traffic densities with = 500 m.
28 Similar to success rate, the message delivery ratio also decreases as the distance from the
29 information source increases. For low traffic density, there is no significant difference
30 between flooding, AODV, and theoretical model as shown in Figure 5( b). However, in high
31 traffic density, Figure 5( a), degradation of the flooding method is evident in comparison with
32 the other methods. The lower message delivery ratio in flooding for higher traffic density is
33 caused by the broadcast storm problem where redundant broadcasts cause wireless radio
34 contention and collision problems. Further, AODV performed better than the flooding
35 method as AODV establishes a shortest- path- based routing scheme ( routing table construct)
and then disseminate messages in the MANET. Consequently, we can see that the choice of 36 Message Delivery Ratio
Message Delivery Ratio
Jung, Rex, Jin, Jayakrishnan, and Regan
10
routing protocols can exhibit different performance measures 1 for the same mobility scenario
2 and transmission range.
3
4 IMPACT ON MULTIPLE INFORMATION SOURCES
5 This experiment evaluates the overall communication performance when multiple vehicles
6 are sending messages simultaneously. We place multiple information sources ( up to a
7 maximum of four) equally distributed over the same traffic scenario with Figure 5( a) and
8 measure the message delivery ratio. Figure 6 compares two different cases, single and four
9 information sources. From Figures 6( a) and 6( b), we see the impact of communication traffic
10 on delivery distance when multiple information sources are present in the network.
11
12
13 6( a) Single Information Source 6( b) Four Information Sources
14 Figure 6. Message Delivery Ratio with Multiple Sources
15 IMPACT ON TWO DIMENSIONAL NETWORKS
16 In this section, we construct a two- dimensional network ( 5 km x 5 km) with traffic flow in
17 both forward and opposite directions for uniform traffic to better understand communication
18 performance in the intersection junction of arterial road. A fixed value of = 250 m is used.
19 We designate the four longitudinal traffic flows to 30 veh/ km and vary the four latitudinal
20 traffic flows with 15 veh/ km and 60 veh/ km in separate experiments. In Figure 7, we observe
21 that with a 10% MPR, a density of 15 veh/ km can only propagate 1 km ( covering 3
22 intersections) and 60 veh/ km 5 km ( covering 12 intersections). This is due, in part that as
23 traffic flow meets at an intersection information can be propagated further. Hence, Figure
24 7( b) shows significant gains in message distance traveled by doubling the traffic density.
25
7( a) ρ = 15 veh/ km and 30 veh/ km
7( b) ρ = 60 veh/ km and 30 veh/ km
26 Figure 7. Two Dimensional Road Network
Message Delivery Ratio
Message Delivery Ratio
Jung, Rex, Jin, Jayakrishnan, and Regan
11
1
2 CONCLUSION
3 In this paper, we investigate and illustrate the impact of traffic stream and wireless
4 communication on the performance of inter- vehicle communications. We develop a
5 simulation framework with ns- 2 that generates different combinations of communication and
6 mobility scenarios and use the Monte- Carlo method to evaluate system wide performances.
7
8 To measure the performance of IVC, we consider success rate and message delivery ratio.
9 First, we measure success rate for both uniform traffic and shockwave traffic. The result
10 shows that both traffic density and transmission range are major contributing factors on the
11 communication performance. In shockwave traffic scenarios, the success rate changes
12 dramatically when it meets a different traffic density. By comparing it with analytical model,
13 simulation results are lower than theoretical values due to signal interference and inefficiency
14 of the flooding method. Then, we study message delivery ratio for different traffic densities,
15 transmission ranges, multiple information sources, and two dimensional road layouts. We
16 conclude that higher traffic densities and longer transmission range causes greater
17 interferences that lead to more packet drops. Both traffic and network can significantly
18 impact the performance in inter- vehicle communication.
19 
20 Systematic consideration of the requirements and constraints imposed by applications,
21 communication, and vehicular traffic flow are necessary for communication routing protocol
22 design. For example, a mobility model can describe information on vehicle headways, which
23 is useful since vehicles need to be within transmission range to communicate. For future
24 research, we plan to extend our simulation framework to complex traffic scenarios using
25 microscopic traffic simulator such as Paramics. However, a joint approach involving both
26 network and traffic simulator can create greater simulation challenges such as time-27
synchronization between the two simulators and ensuring compatibility and portability. Our
28 future plans include measuring the performance of IVC for bidirectional directions and delay-29
tolerant network schemes where vehicles “ store- carry- forward” messages ( 33). These issues,
30 along with other improvements at the lower levels of the communication protocol stack, will
31 be important future research questions related to the design of reliable, scalable, and efficient
32 routing protocols for vehicular networks.
33
34 ACKNOWLEDGEMENT
35 This research is supported in part by a grant from University of California Transportation
36 Center ( UCTC).
37
38 REFERENCES
39 1. “ FCC allocates spectrum in 5.9 Ghz range for Intelligent Transportation Systems uses” FCC.
40 http:// www. fcc. gov/ Bureaus/ Engineering_ Technology/ News_ Releases/ 1999/ nret9006. html.
41 Accessed Sep. 26, 2007.
42 2. Hedric, J. K, Tomizuka, M., and Varaiya, P. Control issues in automated highway systems. IEEE
43 Control Systems Magazine, Vol. 14, No. 6, 1994, pp. 21- 32.
44 3. Recker, W. W., Jin, W.- L., Yang, X., and Marca, J. Autonet: Inter- vehicle communication and
45 network vehicular traffic. International Journal of Vehicle Information and Communication
46 Systems, Vol. 1( 3/ 4), 2008, pp. 306- 319.
Jung, Rex, Jin, Jayakrishnan, and Regan
12
4. “ Safe and comfortable driving based upon inter- vehicle com 1 munication” CarTALK2000.
2 http:// www. cartalk2000. net/. Accessed November 11, 2003.
3 5. Commission of the European Communities. European transport policy for 2010: time to decide.
4 2001.
5 6. “ Advisory Memo to USDOT # 2” http:// www. its. dot. gov/ itspac/ itspac_ advmemo2. htm.
6 7. Morris, R., Jannotti, J., Kaashoek, F., Li, J., and Decouto, D. CarNet: a scalable ad hoc wireless
7 network system. Proceeding of the 9th workshop on ACM SIGOPS European workshop: beyond
8 the PC: new challenges for the operating system, 2000, pp. 61- 65.
9 8. Internet on the road. Fleetnet. http:// www. fleetnet. de/. Accessed November 11, 2003.
10 9. Rudack, M., Meincke, M., and Lott, M. On the dynamics of ad hoc networks for inter vehicle
11 communications ( IVC). The 2002 International Conference on Wireless Networks ( ICWN’ 02),
12 Las Vegas, NV, USA, 2002.
13 10. Blum, J. J., Eskandarian, A. and Hoffman, L. J. Challenges of intervehicle ad hoc networks. IEEE
14 Transportations on ITS, Vol. 5( 4), Dec. 2004.
15 11. Network Simulator, ns- 2. http:// www. isi. edu/ nsnam/ ns/. Accessed July 01, 2008.
16 12. Shannon, C. E. A mathematical theory of communication. The Bell System Technical Journal,
17 Vol. 27, July, October, 1948, pp. 379- 423, 623- 656.
18 13. Abramson, N. The Throughput of Packet Broadcasting Channels. Communications, IEEE
19 Transportations on [ legacy, pre- 1988], Vol. 25( 1), 1977, pp. 117- 128.
20 14. Tanenbaum, A. S. Computer Networks. Prentice Hall PTR, 2002.
21 15. Gupta, P. and Kumar, P. R. The capacity of wireless networks. IEEE Transportation on
22 Information Theory, 46( 2), March 2000, pp. 388- 404.
23 16. Jin, W.- L. and Recker, W. W. An analytical model of multihop connectivity of inter- vehicle
24 communication systems. Submitted to IEEE Transportations on Wireless Communications, 2005.
25 17. Jin, W.- L. and Recker, W. W. Instantaneous information propagation in a traffic stream through
26 inter- vehicle communication. Transportation Research Part B: Methodological, Vol. 40( 3), 2006,
27 pp. 230- 250.
28 18. Jin, W.- L. and Recker, W. W. A Monte Carlo Simulation Model of Instantaneous Inter- Vehicle
29 Communication. In Transportation Research Record: Journal of the Transportation Research
30 Board, No. 2000, Transportation Research Board of the National Academies, Washington, D. D.,
31 2007, pp. 8- 15.
32 19. Jin, W.- L., Wang, Z., and Lu, W. A new framework for simulating inter- vehicle communication.
33 Third China Annual Conference on Intelligent Transportation, Nanjing, China, 2007.
34 20. Yang, X. Assessment of A Self- Organizing Distributed Traffic Information System: Modeling
35 and Simulation. PhD thesis, University of California, Irvine, 2003.
36 21. Yang, X. and Recker, W. W. Simulation studies of information propagation in a self organizing
37 distributed traffic information system. Transportation Research Part C: Emerging Technologies,
38 Vol. 13( 5- 6), 2005, pp. 370- 290.
39 22. Harmon, T., Marca, J., Klefstad, R. and Martini, P. Design, implementation, and test of a wireless
40 peer- to- peer network for roadway incident exchange. International Journal of Vehicle
41 Information and Communication Systems, 2007.
42 23. Viswanath, K. and Obraczka , K. Modeling the Performance of Flooding in Wireless Multi- Hop
43 Ad Hoc Networks, Computer Communications, Vol. 29( 8), 2006, pp. 949- 956.
44 24. Broch J., Maltz, D. A., Johnson, D. B., Hu, Y.- C., and Jetcheva, J. A Performance Comparison of
45 Multi- Hop Wireless Ad Hoc Network Routing Protocols. Proceedings of the 4th annual
46 ACM/ IEEE international conference on Mobile computing and networking, 1998, pp. 85- 97.
47 25. Yi, Y., Gerla, M., and Kwon, T. J. Efficient Flooding in Ad hoc Networks: a Comparative
48 Performance Study. Communications, 2003. ICC ’ 03. IEEE International Conference on, Vol. 2,
49 2003, pp. 1059- 1063.
50 26. Wang, Z., Lu, W., and Jin, W.- L. A Study on Information Throughput of Inter- vehicle
51 Communications in a Unidirectional Traffic Stream. Communications and Mobile Computing,
52 2009. CMC ’ 09. WRI International Conference on, Vol. 2, 2009, pp. 396- 401.
53 27. AODV. http:// www. ietf. org/ rfc/ rfc3561. txt. Accessed Mar 13, 2008.
54 28. Jin, W.- L. Modeling connectivity of inter- vehicle communication networks along discrete traffic
55 streams. Under review for IEEE Transactions on Intelligent Transportation Systems.
Jung, Rex, Jin, Jayakrishnan, and Regan
13
29. Munjal, P. K., Hsu, Y. S., and Lawrence, R. L. Analysis and validation of 1 lane- drop effects of
2 multilane freeways. Transportation Research 5, 1971, pp. 257- 266.
3 30. Newell, G. F. A simplified theory of kinematic waves in highway traffic I: General theory. II:
4 Queuing at freeway bottlenecks. III: Multi- destination flows. Transportation Research Part B,
5 Vol. 27, 1993, pp. 281- 313.
6 31. S. Kurkowski, T. Camp, and M. Colagrosso. Manet simulation studies: the incredibles.
7 SIGMOBILE Mobile Computer Communication Review. 2005.
8 32. AhleHagh, H., Michalson, W. R., and Finkel, D. Statistical characteristics of wireless network
9 traffic and its impact on ad hoc network performance. Advanced Simulation Technologies
10 Conference, Orlando, USA, April 2003, pp. 66- 71.
11 33. Abdulla, M. and Simon, R. The Impact of the Mobility Model on Delay Tolerant Networking
12 Performance Analysis. Simulation Symposium, 2007. ANSS ’ 07. 40th Annual, 2007, pp. 177- 184.
20 0890- 8044/ 10/$ 25.00 © 2010 IEEE IEEE Network • January/ February 2010
very year, millions of traffic accidents occur world-wide,
resulting in tens of thousands of casualties
and billions of dollars in direct economic costs. For
many years now, transportation planners have been
pursuing an aggressive agenda to increase road safety through
intelligent transportation system ( ITS) initiatives. Further-more,
in 2001 the European Transport Policy set out a goal
to reduce road fatalities by 50 percent by the year 2010. Simi-larly,
in 2008 the U. S. Department of Transportation’s
( DOT’s) Research and Innovative Technology Administra-tion
challenged the industry to reduce 90 percent of traffic
crashes by 2030. In recent years various stakeholders have
come together to address these short- term and long- term
challenges, and initiative efforts have been formed such as
the U. S. IntelliDrive and European eSafety programs. A
novel communication system known as dedicated short- range
communication ( DSRC) has been proposed within the 5.8– 5.9
GHz frequency spectrum allocated for its use. Standard activ-ities
for the overall system architecture and communication
framework are coordinated by a variety of entities that
include the IEEE ( IEEE 802.11p, IEEE 1609 working group)
in the United States, and the Car 2 Car Communications
Consortium ( C2C- CC), European Telecommunications Stan-dards
Institute ( ETSI, TC ITS), and International Organiza-tion
for Standardization ( ISO, TC204/ WG16) in Europe and
other parts of the world.
To achieve the future road safety vision, time- sensitive,
safety- critical applications in vehicular communication net-works
are necessary. Broadcasting will play an important role
in disseminating safety messages to all nearby vehicles such as
look- ahead emergency warnings and information about unsafe
driving conditions. However, the lack of packet acknowledg-ment,
packet retransmission, and a medium reservation
scheme makes it difficult to achieve high broadcast reliability
and efficiency in dense vehicular networks due to wireless
contention and interferences.
The Routing Problem
The fundamental design consideration for routing protocols is
the network environment and whether it is a static or dynamic
network. Design in the underlying communication system is
complicated by requirements that satisfy multiple constraints
which include high reliability, efficiency, and scalability perfor-mance
measures.
A vehicular ad hoc network ( VANET) is a specific type of
mobile ad hoc network ( MANET) where dynamic routing pro-tocols
are necessary. A VANET operates in a self- organized
manner without permanent infrastructure and, similar to a
MANET, encounters two major routing issues, the broadcast
storm problem and the network disconnection problem. The
broadcast storm problem occurs when mobile nodes send mes-sages
by flooding, causing frequent link layer contention with
other nearby broadcasting nodes that result in high packet loss
due to collisions. Specifically, this phenomenon happens during
multihop relay and message broadcast. Multihop relay occurs in
MANETs in wireless mesh configurations and in VANETs
when there are no roadside stations nearby. For MANETs, mes-sage
broadcast occurs during route discovery or route mainte-nance,
such as route request hello messages. For VANETs, this
happens in periodic broadcast beacons of vehicle or traffic infor-mation.
Achieving high communication reliability and efficiency
is an essential requirement for safety- based ITS applications.
Furthermore, the network disconnection problem for VANETs
is more severe than for MANETs due to high mobility caused
by fast moving vehicles and the sparse traffic densities during
off- peak hours. This disconnection time ( on the order of a few
seconds to several minutes) makes MANET protocols such as
Ad Hoc On Demand Distance Vector unsuitable for VANETs.
Hence, new network designs to improve broadcast reliability
in dense networks and routing decisions in sparse networks are
necessary. In this article we review existing methods and design
considerations for vehicular communication networks. In partic-
EE
Rex Chen, Wen- Long Jin, and Amelia Regan, University of California, Irvine
Abstract
A primary goal of intelligent transportation systems is to improve road safety. The
ability of vehicles to communicate is a promising way to alleviate traffic accidents
by reducing the response time associated with human reaction to nearby drivers.
Vehicle mobility patterns caused by varying traffic dynamics and travel behavior
lead to considerable complexity in the efficiency and reliability of vehicular com-munication
networks. This causes two major routing issues: the broadcast storm
problem and the network disconnection problem. In this article we review broad-cast
communication in vehicular communication networks and mechanisms to allevi-ate
the broadcast storm problem. Moreover, we introduce vehicular safety
applications, discuss network design considerations, and characterize broadcast
protocols in vehicular networks.
Broadcasting Safety Information in
Vehicular Networks: Issues and Approaches
IEEE Network • January/ February 2010 21
ular, our discussion includes application requirements, commu-nication
systems, traffic characteristics, and routing protocols.
We conclude by summarizing the lessons learned, field experi-ments,
and future challenges of broadcasting in vehicular com-munication
networks. In the literature previous surveys and
tutorials on routing protocols for VANETs have been explored
by [ 1– 7]. This article is an extension from these related works as
it focuses on broadcast methods with an emphasis on the design
requirement of high reliability and efficiency for vehicular safety
applications by alleviating the broadcast storm problem.
Design Considerations
Safety Applications
Specific ITS applications govern the performance require-ments
in vehicular communication networks. During phase
one DSRC experiments, several road safety scenarios based
on cooperative intersection collision avoidance systems were
tested. These scenarios included traffic signal violation warn-ings,
stop sign alerts, and left turn signal assistance. According
to the U. S. Vehicle Safety Communications Consortium, a
comprehensive list of more than 75 application scenarios for
intelligent vehicle safety applications enabled by DSRC have
been identified [ 8]. Table 1 describes a list of safety applica-tions,
and their corresponding communication and traffic
parameters. In particular, safety applications at intersection
roads ( infrastructure- to- vehicle) and message exchange among
vehicles ( vehicle- to- vehicle) have the most promising safety
benefits in the near and mid- term future.
Message transmit mode can be triggered periodically or
event- driven. In the periodic case, preventive safety messages
are disseminated to keep drivers informed with details such as
forward and opposing vehicle speed, acceleration, and decel-eration
values. On the other hand, event- driven messages are
delivered occasionally as in the case of a sudden hard braking
vehicle from other nearby vehicles or emergency vehicles such
as ambulances. Moreover, many applications that send event-driven
messages are relevant for farther vehicles, allowing
upstream vehicles to undertake early countermeasures to pre-vent
severe catastrophes such as chain- reaction accidents.
In Table 1 the latency for safety requirements are approxi-mate
values proposed previously by several sources that
include previous research papers, automotive practitioner rec-ommendations,
and consortium reports. In addition, prelimi-nary
evaluation in field tests indicate the typical delay
requirement for many safety applications is between 100 and
500 ms, a lower bound value compared with human reaction
time. The delay factor for safety applications is important, and
the IEEE 802.11p specification has set a minimum allowable
latency of 100 ms for periodic message broadcast. In general,
near real- time information is essential as even non- safety traf-fic-
based applications require delay latencies in the range of
several seconds to a few minutes for many ITS applications to
be useful. The maximum communication range depends on
usefulness of the safety information to nearby vehicles for
both upstream and downstream traffic in the same direction
for highways, as well as opposing directions on arterial roads
and local streets. In situations where the maximum communi-cation
range does not reach the intended distance, multihop
communication is a useful mechanism.
Communication
In communication networks packet delivery can be unicast,
multicast, or broadcast. The behavior of multicast and broad-cast
systems are different, as the former sends a message to
multiple destinations based on specific group attributes, while
the latter sends a message to all recipients within its coverage
area. In vehicular communication networks, for example, a
group of taxi or courier vehicles in a metropolitan city may
only relay messages among their fleets. However, an ambu-lance
siren alert must notify all nearby vehicles to pull over
rapidly and safely. In recent years other forms of network
delivery have been proposed that include geocast and anycast.
In particular, for vehicular networks geocast, which is based
on geographic routing, has been studied extensively by taking
a form of greedy forwarding in relaying information to the
destination such as most forward within range ( MFR) or
nearest with forward progress.
Different from other wireless networks, packets in vehicular
networks are mostly autonomous and have specific temporal
and spatial relevance. Furthermore, the assumptions may
include knowledge of digital road layouts, location coordinates
( GPS), and in some cases the location of the destination node.
Performance metrics that are important include message deliv-ery
ratio, packet reception rates, packet error rates, and end-to-
end transmission delay. A comprehensive classification of
different automotive applications in DSRC and detailed per-formance
measures for VANETs is reviewed in [ 9].
Traffic
The mobility patterns of communication nodes in VANETs
are significantly different from those in conventional wireless
networks. Vehicles’ space- time trajectories are restricted by
paved roadways and drivers’ choices of origins, destinations,
departure times, and routes. The positions of vehicles are not
independent on a road due to car following or lane changing
rules. Densities of vehicles can vary dramatically along a com-munication
path due to driving behaviors and restrictions
caused by network geometry.
Previous studies have shown that the topological properties
and mobility models can have dramatic impact on network pro-tocol
performance. Two popular mobility models for vehicular
communication that generate movements at the microscopic
level include SUMO and VanetMobiSim, incorporating aspects
of the car following model developed by Stefan Krauss and the
TSIS- CORSIM traffic simulator. An in- depth survey and taxon-omy
of mobility models for VANETs is described in [ 10].
Furthermore, vehicle movements can be complicated by
other factors such as traffic signals and stop signs in arterial
roads and ramp meters on highways. Traffic simulators such
as TransModeler and Paramics that incorporate traffic flow
theory and traffic control systems can provide greater realism
in vehicle trajectories. Another approach to formulating the
topological properties and mobility model involves using real-istic
vehicular traces to account for other variables. Some
research work has adopted this method, using mobility trace
data from SUVnet ( taxi traces via GPS) and BTL/ NG- SIM
( vehicle traces via loop detectors).
Overview of Broadcasting Protocols in
Vehicular Networks
In this section we present a classification of broadcast proto-cols
based on methods to reduce the broadcast storm problem
for vehicular communication networks. Table 2 illustrates the
historical taxonomy of broadcast communication with a quali-tative
comparison of the communication methods, traffic char-acteristics,
network simulation environment, and mobility
model used in the protocol design and evaluation. In certain
cases the literature on broadcast protocol did not specify the
simulation environment, road topology, and mobility models
used in their evaluation. For these situations, we omit their
discussion and leave the table field entries blank.
22 IEEE Network • January/ February 2010
Communication Characteristics
In the MANET literature several suppression schemes have
been proposed to improve the overall reliability of the shared
communication channel. These schemes include probabilistic-based,
counter- based, distance- based, and location- based
methods. These schemes have been adopted in broadcasting
for vehicular communication networks along with new meth-ods
such as cluster- based and traffic- based methods. In loca-tion-
and position- based methods, messages are broadcast
based on the geographic area of the transmitting and receiving
vehicle locations. In distance and hop- based methods, mes-sages
are broadcasted by considering the neighboring distances
and hop count from the transmitting node. Cluster- based
Table 1. Vehicular safety applications: communication requirements and traffic information.
Safety application Communication type Traffic information Transmit
mode
Latency
( ms)
Communication
range ( m)
Intersection collision avoidance
Traffic signal violation
warning Infrastructure- to- vehicle Traffic signal status and
timing; pedestrian crossing Periodic ~ 100 ≤ 250
Left turn assistant Vehicle- to- infrastructure
Infrastructure- to- vehicle
Traffic signal status and timing;
vehicle position, speed, heading;
intersection road shape
Periodic ~ 100 ≤ 300
Stop sign movement
assistance
Vehicle- to- infrastructure
Infrastructure- to- vehicle Vehicle position, heading, speed Periodic ~ 100 ≤ 300
Intersection collision
warning Vehicle- to- vehicle Vehicle position, heading, speed;
turn signal status
Event-driven
~ 100 ≤ 300
Blind merge warning Infrastructure- to- vehicle Vehicle position, speed, heading Periodic ~ 100 ≤ 200
Pedestrian cross informa-tion
at designated
intersections
Infrastructure- to- vehicle Pedestrian detection and crossing Periodic ~ 100 ≤ 200
Information from other vehicles
Cooperative collision
warning Vehicle- to- vehicle Vehicle position, speed, heading,
acceleration Periodic ~ 100 ≤ 150
Emergency electronic
brake lights Vehicle- to- vehicle Vehicle position, heading, speed,
deceleration
Event-driven
~ 100 ≤ 300
Highway merge
assistant Vehicle- to- vehicle Vehicle position, heading, speed;
vehicles in merge path Periodic ~ 100 ≤ 250
Blind spot warning Vehicle- to- vehicle Vehicle position, heading, speed Periodic ~ 100 ≤ 150
Pre- crash sensing Vehicle- to- vehicle Safety sensor coordination on
seatbelts, airbags, pre- arming
Event-driven
~ 20 ≤ 50
Transit vehicle signal
priority Vehicle- to- vehicle Vehicle position, heading, speed Event-driven
~ 1000 ≤ 1000
Cooperative vehicle- high-way
automation systems
( platoon)
Vehicle- to- vehicle
Vehicle- to- infrastructure
Vehicle headway distance,
position, speed; coordinated
platoon maneuvers
Periodic ~ 20 ≤ 100
Cooperative adaptive
cruise control Vehicle- to- vehicle Vehicle headway distance,
vehicle cut- in Periodic ~ 100 ≤ 150
Public safety
Approaching emergency
vehicle warning Vehicle- to- vehicle Emergency vehicle right- of- way
yield
Event-driven
~ 1000 ≤ 1000
Post- crash warning Vehicle- to- infrastructure
Vehicle- to- vehicle
Disabled vehicle due to crash or
mechanical breakdown
Event-driven
~ 500 ≤ 300
Sign extension
In- vehicle signage Infrastructure- to- vehicle
Signage typically conveyed by
traffic signs ( e. g., school zone,
speed limit)
Periodic ~ 1000 ≤ 200
Curve speed warning Infrastructure- to- vehicle
Curve location, curve speed
limits, curvature, road surface
condition
Periodic ~ 1000 ≤ 200
Work zone wWarning Infrastructure- to- vehicle Distance to work zone, road
closure, reduced speed limit Periodic ~ 1000 ≤ 300
IEEE Network • January/ February 2010 23
methods broadcast messages to vehicle groups, for example, to
a platoon of vehicles with common paths. In probabilistic-based
methods, messages are broadcast with a given probabili-ty
p, and in many cases this probability is based on the
protocol’s backoff timer. For traffic- based methods, informa-tion
on traffic dynamics such as vehicle speed are incorporated
into the message broadcast decision. The predominant net-work
simulation used is the state- of- the- art open source ns- 2
simulator. A variety of mobility models are used for simulating
vehicle movements in highway and arterial roads.
Urban Multihop Broadcast ( UMB) and Ad Hoc and Multihop
Broadcast ( AMB) — In these techniques, preference on a broad-cast
relay and suppression scheme is utilized based on road
location or vehicle position. To reduce the multihop messaging,
UMB and AMB elect vehicles farthest away ( MFR) from the
information source as relay nodes. This location metric is com-puted
based on the black- burst method, which lets receivers
send black- burst signals proportional to their location from the
source. Furthermore, the AMB protocol is an enhancement to
UMB that does not require repeaters ( infrastructureless) when
vehicles may not be in the intersection to retransmit a message
by nominating the node closest to the intersection position as
the relay node for broadcasting instead.
Smart Broadcast ( SB), Position- Based Adaptive Broadcast ( PAB),
and Distributed Vehicular Broadcast ( DV- CAST) — SB and PAB
use a dynamic backoff timer for medium access control ( MAC)
contention window adjustment to improve the efficiency of
packet transmissions. SB’s backoff timer scheme is based on
the sender and receiver node distance, while PAB determines
the backoff timer based on vehicle position and vehicle speed.
DV- CAST uses local one- hop neighbor topology to make rout-ing
decisions. The protocol adjusts the backoff timer based on
the local traffic density, and computes forward and opposing
direction connectivity with periodic heartbeat messages. More-over,
DV- CAST is adaptive to the totally disconnected net-work
and can temporarily wait- and- hold a packet until the
vehicle hears heartbeat messages from other vehicles.
Multihop Vehicular Broadcast ( MHVB) — MHVB adjusts the
packet transmission interval with a position- based method.
The two proposed schemes for packet retransmissions in
MHVB include the location between sender and receiver, and
Table 2. Classification of broadcast protocols in vehicular networks.
Location-/
position-based
Distance/
hop-based
Cluster-based
Proba-bilistic-based
Network
simulator
Traffic-based
High-ways
Arterials/
local
streets
Data
aggrega-tion
Mobility model
Broadcast
protocols Communication characteristics Traffic characteristics
UMB, 2004 √ √ WS √ √
Negative exponential
( headways) and Gaussian
( speed)
TrafficView, 2004 ns- 2 √ √ √ √ Random waypoint model
MDDV, 2004 √ QualNet √ √ CORSIM and Atlanta road
traces
ODAM, 2004 √ √ ns- 2
OAPB/ DB, 2005 √ √ √ ns- 2 √
AMB, 2006 √ WS √ Negative exponential ( head-ways)
and Gaussian ( speed)
SB, 2006 √ √ Negative exponential ( head-ways)
MHVB, 2006 √ ns- 2 √ Microscopic traffic simulator
D- FPAV, 2006 √ ns- 2 √ DaimlerChrysler road traces
TRRS, 2007 √ √
REACT, 2007 √ ns- 2 √ √ Nagel and Schreckenberg
cellular automata
DV- CAST, 2007 √
FB, 2007 √
DBAMAC, 2007 √ ns- 2 √ IMPORTANT mobility tool
PAB, 2008 √ √ ns- 2 √ √ Road Design Manual
REAR, 2008 √ ns- 2 √ Manhattan model
CTR, 2009 √ √ ns- 2 √
24 IEEE Network • January/ February 2010
the traffic congestion level, which is determined by a multi-tude
of threshold values that include number of nearby vehi-cles,
number of vehicles in forward and opposing directions,
and vehicle speed. A subsequent improvement for MHVB was
later published that includes more efficient angular coverage
from sender to receiver and introduces a dynamic scheduling
algorithm that prioritizes received packets.
Mobility- Centric Data Dissemination Algorithm for Vehicular Net-works
( MDDV) — MDDV is a geo- cast protocol that defines
the destination region and trajectory- based routing based on
travel directions to deliver packets to the region. The MDDV
protocol runs a localized broadcast routing algorithm to con-tinuously
forward messages to the head node in the cluster
pack and moves closer to the intended destination. Results
from MDDV indicate that the routing protocol performance
depends on the market penetration rate of vehicle- to- vehicle
communication and road traffic density, which is affected by
the time of day with its realistic movement traces.
Fast Broadcast ( FB) and Cut- Through Rebroadcasting ( CTR) —
FB is a distance- based protocol that minimizes forwarding
hops when transmitting messages and contains two compo-nents,
the estimation and broadcast phases. In the estimation
phase the protocol adjusts the transmission range using heart-beat
messages to detect backward nodes. In the broadcast
phase it gives higher priority to vehicles that are farther away
from the source node to forward the broadcast message. CTR
also gives higher priority to rebroadcast alarm messages to
farther vehicles within transmission range but operating in a
multichannel environment.
Distributed Fair Transmit Power Assignment for Vehicular Ad
Hoc Network ( D- FPAV) — D- FPAV describes a scheme that
provides fairness in broadcasting heartbeat messages by
dynamically adjusting every node’s transmission power based
on distance to other neighboring nodes. The method enables
all nodes to share the channel capacity fairly. Although power
control and adjustment is well explored in wireless networks,
D- FPAV is unique as it investigates the problem in the con-text
of broadcasting in vehicular networks by using realistic
movement traces obtained from DaimlerChrysler on a Ger-man
highway.
Dynamic Backbone- Assisted MAC ( DBA- MAC) — DBA- MAC is
a cluster- based broadcast for message propagation based on
cross- layer intersection in the MAC. For a group of intercon-nected
vehicles, higher- priority nodes within the cluster are
considered backbone members and are able to broadcast mes-sages.
The process of choosing backbone nodes within the
cluster occurs periodically by selecting nodes that are farther
apart to minimize hop count.
Receipt Estimation Alarm Routing ( REAR) — In the REAR pro-tocol,
nodes that relay broadcast messages are selected
based on estimated message delivery ratio. This is computed
based on the received signal strength and packet reception
rates for packets that nodes receive, and this information is
exchanged with neighboring nodes using heartbeat broadcast
messages. Hence, nodes with higher message delivery ratios
are likely candidates to flood messages in the network while
the other nodes are kept silent to alleviate wireless con-tention
conflict.
TrafficView — The TrafficView protocol is a part of the
broader e- Road project with the goal of building a scalable
and reliable infrastructure for intervehicle communication
systems. In TrafficView, the message data contain informa-tion
on a list of vehicle IDs and the vehicle’s own position
and speed, as well as broadcast duration time. TrafficView
conserves bandwidth and deals with flow control of broadcast
messages by aggregating multiple data packets based on rela-tive
vehicle distance and message timestamp. For example,
two vehicles on the same highway lane traveling at similar
speeds are likely to have similar vehicle positions and vehicle
trajectories. Hence, when updated information on vehicle
positions is available, vehicle speeds may not be necessary,
which reduces packet size and results in lower packet trans-mission
delay ( less air time).
Time Reservation- Based Relay Node Selection ( TRRS) and Rout-ing
Protocol for Emergency Applications in Car- to- Car Networks
Using Trajectories ( REACT) — TRRS proposes a method where
nodes in the communication range choose their waiting time
based on a specified time window. The time window is deter-mined
by a distance that is inversely proportional to the previ-ous
relay node and reservation ratio of the time window. A
node with higher reservation ratio will have received duplicate
broadcast messages and incurred longer time window waiting
duration in the next transmission round. REACT gives more
influence on the forwarding trajectory and angle, and inte-grates
the position- based information with the time- division
multiple access 802.11 MAC.
Optimized Dissemination of Alarm Message ( ODAM) and Opti-mized
Adaptive Probabilistic Broadcast and Deterministic Broad-cast
( OAPB/ DB) — ODAM has a “ defertime” to broadcast
messages, computed based on the inverse proportional dis-tance
between receiver and source node. For ODAM, broad-cast
messages can only occur within the risk zone region,
determined with a dynamic multicast group based on vehicles’
proximity to the incident site. OAPB/ DB uses an adaptive
approach to rebroadcast emergency warning messages near
the incident zone. Nodes rebroadcast messages probabilistical-ly
within the region based on the delivery ratio, which is com-puted
based on local traffic density information.
Lessons Learned, Field Experiments, and
Future Challenges
Lessons Learned
An overview of broadcast protocols in vehicular communica-tion
networks has been introduced. Specifically, these proto-cols
address the broadcast storm problem by reducing packet
redundancy, wireless contention, and collisions in the network.
Although numerous design methods have been proposed,
each protocol has its limitations and assumptions that may
cause certain issues. For instance, the concept of node selec-tion
for multihop relay based on node distance ( MFR),
although reducing the total number of traveling hops, incurs a
reliability trade- off with lower packet reception rates due to
the loss in radio power from longer propagation distances.
Also, several broadcast protocols to modify the MAC with dif-ferent
priority schemes have been proposed. However, such
schemes may result in “ unfairness” in the overall system
where certain nodes have more packet transmission rounds
than others. Yet another shortcoming for some methods is the
assumption that GPS is readily available to provide location
position to neighboring vehicles. Hence, the feasibility of
these vehicular communication network applications will
depend largely on the technology adoption and market pene-tration
rates of vehicles equipped with capabilities, GPS
devices, or both.
IEEE Network • January/ February 2010 25
Field Experiments
In the past few years field trials have been conducted to fine-tune
the DSRC specification. Initial results indicate packet
error rates ( PERs) can be highly affected by urban canyons,
caused by radio signal degradation due to multipath fading
[ 11]. The vehicle height profile can also significantly impact
the transmission range for DSRC. Initial road test experi-ments
indicate 20 percent PER with about 150 messages/ s,
and the results are better for shorter ( 300 bytes) rather than
longer ( 1200 bytes) messages since longer packet length con-sumes
more air time. The phase one stage provides a strong
proof of concept for DSRC. However, VANETs still have
many issues to address, including external factors such as road
terrain conditions, vehicle types, and environmental factors.
Future Challenges
There remain many open issues and future challenges to
solve. The field of vehicular networks has not only fostered
academic research interest, but has motivated experts to
publish books to share knowledge, most recently in 2009
[ 12– 15] and 2010 [ 16, 17]. In the lower layers of the commu-nication
stack, novel channel access methods, priority access
with IEEE 802.11e, dynamic contention window and power
adjustment, and multiradio interfaces are just some of the
techniques that can improve vehicular communication by
optimizing the wireless channel load. This can be thought of
as a scalability problem and characterized by the “ communi-cation
density” metric for vehicular communications [ 18].
An empirical analysis using 802.11 wireless interfaces in the
ORBIT emulation testbed provides some insights on the
complexity of broadcasting in dense vehicular networks [ 19].
However, the communication parameters and how these
contribute to the overall system reliability and efficiency are
not yet well understood and need further analysis. More-over,
the design of vehicular communication networks needs
to be integrated with the safety and traffic- based application
requirements. For example, the communication system can
dynamically consider the latency requirement in Table 1 and
fine- tune its MAC contention window size to the desirable
performance measures ( e. g., highest delivery ratio, mini-mum
delay).
Initially, the requirements will be for vehicular safety
applications. Multihop broadcasting is useful to provide an
early countermeasure to prevent catastrophes such as chain-reaction
accidents for nearby and following vehicles in the
upstream. Subsequent enhancements will include real- time
traffic information and environmental applications that
reduce emissions in vehicle platoons by stabilizing traffic on
the road through adaptive cruise control. In other cases ITS
traffic applications may tolerate small delay and allow mes-sages
to be queued at intermediate relay points prior to
sending information to the intended destination when the
network is sparse. In such cases a delay- tolerant geocast pro-tocol
that sends messages on demand based on time factors
when near other vehicles or a traffic collection roadside sta-tion
is more appropriate. Finally, security in VANETs
remains a rich research area with many problems that need
to be addressed including vehicle anonymity, message
integrity, and authentication, traceability, and revocation of
malicious attackers.
Conclusion
In this article we classify and survey broadcast protocols
for vehicular communication networks. Vehicular net-works
have many safety- based applications where reliabili-ty
is of utmost importance. Reducing message flooding
serves as a fundamental method to alleviate the broadcast
storm problem and increase the reliability and efficiency
of disseminating safety messages to other vehicles. Future
research for network engineers and researchers should
incorporate traffic characteristics and application require-ments
into the communication system design. Traffic flow
dynamics, along with improvements in the communication
stack, will be important in designing reliable, efficient,
and scalable broadcast methods for vehicular communica-tion
networks.
References
[ 1] F. Li and Y. Wang, “ Routing in Vehicular Ad Hoc Networks: Survey,” IEEE
Vehic. Tech., 2007.
[ 2] K. Laberteaux and H. Hartenstein, “ A Tutorial Survey on Vehicular Ad Hoc
Networks,” IEEE Commun. Mag., 2008.
[ 3] ML Sichitiu and M. Kihl, “ Inter- Vehicle Communication Systems: A Survey,”
IEEE Commun. Surveys & Tutorials, 2008.
[ 4] Y. Toor, P. Muhlethaler, and A. Laouiti, “ Vehicle Ad Hoc Networks: Applications
and Related Technical Issues,” IEEE Commun. Surveys & Tutorials, 2008.
[ 5] T. Willke, P. Tientrakool, and N. Maxemchuk, “ A Survey of Inter- Vehicle
Communication Protocols and Their Applications,” IEEE Commun. Surveys &
Tutorials, 2009.
[ 6] J. Bernsen and D. Manivannan, “ Review: Unicast Routing Protocols for Vehic-ular
Ad Hoc Networks: A Critical Comparison and Classification,” Pervasive
Mobile Computing, 2009.
[ 7] K. Lee, U. Lee, and M. Gerla, “ Survey of Routing Protocols in Vehicular Ad
Hoc Networks,” chapter in Vehicular Ad- oc Networks: Developments and
Challenges, 2010.
[ 8] “ Vehicle Safety Communications Project Task 3 Final Report — Identify Intelli-gent
Vehicle Safety Applications Enabled by DSRC,” VSC Consortium: U. S.
DOT HS 809859, 2005.
[ 9] F. Bai et al., “ Towards Characterizing and Classifying Communication- Based
Automotive Applications from A Wireless Networking Perspective,” IEEE
Wksp. Automotive Networking and Apps., 2006.
[ 10] J. Haerri, F. Filali, and C. Bonnet, “ Mobility Models for Vehicular Ad Hoc
2008.
[ 11] M. Zhang and R. S. Wolff, “ DSRC and Proof of Concept Test: Final Report:
Proof of Concept Results of Findings Summary,” VII Consortium, FHWA- JPO-
09- 043, 2009.
[ 12] S. Olariu and M. Weigle, Eds., Vehicular Networks: From Theory to Prac-tice,
Chapman & Hall/ CRC, 2009.
[ 13] H. Guo, Automotive Informatics and Communicative Systems: Principles in
Vehicular Networks and Data Exchange, Info. Sci. Reference, 2009.
[ 14] H. Moustafa and Y. Zhang, Eds., Vehicular Networks: Techniques, Stan-dards,
and Applications, Auerbach, 2009.
[ 15] H. Huang and Y. Chen, Eds., Telematics Communication Technologies and
Vehicular Networks: Wireless Architectures and Applications, Info. Sci. Ref-erence,
2009.
[ 16] M. Watfa, Ed., Advances in Vehicular Ad Hoc Networks: Developments
and Challenges, IGI Global, 2010.
[ 17] H. Hartenstein and K. Laberteaux, Eds., VANET — Vehicular Applications
and Inter- Networking Technologies, Wiley, 2010.
[ 18] D. Jiang, Q. Chen, and L. Delgrossi, “ Communication Density: A Channel
Load Metric for Vehicular Communications Research,” IEEE Int’l. Conf.
Mobile Ad Hoc and Sensor Sys., 2007.
[ 19] K. Ramachandran et al., “ Experimental Analysis of Broadcast Reliability in
Dense Vehicular Networks,” IEEE Vehic. Tech., 2007.
Biographies
REX CHEN ( rex@ uci. edu) is a Ph. D. candidate in computer science ( networked
systems) at the University of California, Irvine ( UCI). His research interests include
vehicular ad hoc networks, wireless network security, and peer- to- peer networks.
He was with Qualcomm from 2003 to 2005.
WEN- LONG JIN ( wjin@ uci. edu) received a Ph. D. in applied mathematics from the
University of California, Davis in 2003. He is a professor of civil and environ-mental
engineering at UCI. His research interests include intervehicle communica-tions,
traffic flow theory, and transportation network analysis. He was previously
a professor in automation at the University of Science and Technology of China
and a post- doctoral researcher at the UCI Institute of Transportation Studies.
AMELIA REGAN ( aregan@ uci. edu) received a Ph. D. in civil ( transportation systems)
engineering from the University of Texas at Austin in 1997. She is a professor of
computer science at UCI. Her recent research interests include security in
VANETs, short- and long- term pricing of network assets, and resource allocation
techniques for large- scale network improvement. She was previously an opera-tions
research analyst with United Parcel Service and the Association of Ameri-can
Railroads.
Multi- Hop Broadcasting in Vehicular Ad Hoc
Networks with Shockwave Traffic
Rex Chen, Wenlong Jin, Amelia Regan
University of California, Irvine
{ rex, wjin, aregan}@ uci. edu
Abstract- A primary goal of intelligent transportation systems
( ITS) is to improve road safety. The ability for vehicles to
communicate is a promising way to alleviate traffic accidents by
reducing the response time associated with human reaction to
nearby drivers. In addition the limitations of standard driving
can be overcome by providing drivers with instantaneous
information about complications up ahead. Shockwaves, induced
by vehicle speed differentials, are a typical mobility pattern that
occurs with the formation and propagation of vehicle queues and
increase the probability of traffic incidents. These induce sudden
braking and increase the occurrence of traffic incidents. In this
paper, we investigate safety applications in highways with
shockwave mobility and different lane configurations in vehicular
ad hoc networks ( VANET). We evaluate the performance of
multi- hop broadcast communication using the ns- 2 simulator with
vehicles following a shockwave mobility pattern in fully- connected
traffic streams. We propose mechanism to improve broadcast
reliability using dynamic transmission range that leverages our
understanding of fundamental traffic flow relationships.
I. INTRODUCTION
Every year, millions of traffic accidents occur worldwide,
resulting in tens of thousands of casualties and billions of
dollars in direct economic costs. For many years now,
transportation planners have been pursuing an aggressive
agenda to increase road safety through the ITS initiative such
as the U. S. IntelliDrive and Europe eSafety projects. With the
widespread adoption of wireless communication devices,
vehicular communication is becoming an essential and
emerging technology to allow vehicles to share or propagate
useful information for drivers such as traffic congestion alerts,
safety warnings, and traffic management suggestions. In the
United States, in particular, the Federal Communication
Commission ( FCC) has allocated a spectrum of 75 MHz in
5.9 GHz for Dedicated Short Range Communications
( DSRC), a technology for the ITS to improve road safety and
complementary traffic information with standardization efforts
described in IEEE 802.11p.
. Due to the time- sensitive, safety- critical applications in
VANET, broadcasting will play an important role in vehicular
communication to disseminate messages such as look- ahead
emergency warning and information about unsafe driving
conditions. However, the lack of packet acknowledgement and
packet re- transmission makes it difficult to achieve high
broadcast reliability due to wireless contention and
interferences in the medium. Unlike unicast, the optional
RTS/ CTS handshake to prevent the hidden terminal problem in
802.11 cannot be used for broadcast since the RTS/ CTS
exchange would cause even more packet flooding and
exacerbate the broadcast storm problem. The motivation for
our work derives from previous studies that suggest the
importance of examining the impacts of mobility patterns and
transportation network configurations on vehicular
communications. The work by [ 1] suggests these factors can
significantly impact multi- hop connectivity with vehicular
communications in both uniform and non- uniform traffic
streams. As such, we explore the impacts of network
environment on highways with different lane configurations
and mobility patterns on the performance of multi- hop
broadcasting.
In VANET, maintaining high connectivity and high
broadcast reliability is difficult, especially in dense networks
and with non- homogeneous vehicle mobility. In this paper, we
propose a mechanism to dynamically control the
communication range for vehicles by adjusting the
transmission power to mitigate the effects of broadcast storm.
Specifically, our safety- application scenario relates to
shockwave on highways, a common phenomenon that occurs
every day along with the formation and propagation of traffic
queues. A shockwave separates two traffic streams with
different traffic densities and speed, derived according to the
fundamental traffic flow relationships. When the first vehicle
in the following traffic stream meets the last vehicle of the
leading traffic stream, it senses the danger and immediately
sends a broadcast message to inform all nearby vehicles
( within a few kilometers away) of an upcoming shockwave and
caution the vehicles to reduce speeds. The information
propagation is relayed from one vehicle to the next, inspired by
the need for multi- hop broadcast [ 2]. Previous work in wireless
multi- hop networks [ 3] shows the benefits of dynamic
transmission power control ( which results in a dynamic
transmission range) as a way to increase network capacity at
the same time as reducing power consumption.
The contribution of this paper is a simulation- based
approach for a better understanding on the performance of
multi- hop broadcasting under shockwave mobility on highway
with different lane configurations. Efficiency in packet
reception is achieved by reducing packet collisions caused by
overhearing broadcast packets through transmission range
adjustment based on vehicle speed variation. Further, we
compare the performance of static and dynamic minimum
transmission range for different lane configurations on the
highway with free flow and congested traffic densities.
II. RELATED WORKS
The work by [ 4] uses a dynamic transmission- range
assignment ( DTRA) algorithm that employs transmission
power control based on the relationship between connectivity
and traffic density characteristics. Their approach uses an
analytical traffic flow model to derive and estimate local
density coupled with the RoadSim vehicle traffic simulator to
measure the performance of the communication system on
several road configurations. Further, the paper provides
simulation results identifying the minimum transmission range
for different traffic densities in non- homogeneous traffic that
does not require any message exchange with neighboring
vehicles. The focus of their work and the DTRA algorithm is to
maintain a high level of connectivity in vehicular networks by
estimating the local vehicle density and local traffic conditions
( free flow versus congested traffic). In the communication
model, they assume that two vehicles can communicate if their
Euclidean distance is less than or equal to the shorter
transmission range between the two vehicles. However,
communication issues associated with radio interface such as
contention in the shared transmission window, hidden
terminals, and other errors were not considered in their study.
The work by [ 5] uses simulation traces to derive an
empirical model that provides the broadcast reception rate
probability. Parameter optimizations and their empirical model
formulation include inspiration from Jiang et al. [ 6] that define
channel load in vehicular communication by the product of
traffic density, packet generation rate, and transmission range.
The simulation scenario is a circular road but their results
consider single- hop broadcast only with vehicles all having the
same transmission range.
The work by [ 7] evaluates the performance metrics of
delivery ratio and delay for broadcasting safety beacon
messages with varying packet transmission interval and data
packet sizes. The simulation methodology is similar to our
environment, but their study is based on a fixed transmission
range and does not consider multi- hop broadcasting.
The work by [ 8] proposes the distributed fair power
adjustment for vehicular networks ( D- FPAV) algorithm that
dynamically adjusts each vehicle’s transmission power ( and
hence transmission range) to prevent packet collisions. The
optimization focuses on fairness of each communicating
vehicle to receive and send safety information rather than
network capacity and connectivity. Fairness in their adaptive
transmit power scheme is validated through simulation results
on highway scenarios with different radio propagation models.
The work by [ 9] proposes a multi- hop broadcast protocol
called Fast Broadcast that reduces the time to propagate a
message and reduces the total number of hops to cover a
portion of the road. The scheme estimates forward and
backward transmission ranges, computed using two rounds of
transmission ranges ( current- turn and last- turn). However, their
scheme requires message exchange between vehicles in the
specific area- of- interest to determine vehicle spacing and make
transmission range adjustments accordingly.
The work by [ 10] uses simulation traces to present a
broadcast protocol for intermittent connectivity in highway and
urban traffic scenarios that improves reliability and efficiency
by reducing redundant retransmissions. It uses periodic beacon
messages to acquire neighboring vehicle locations and
piggyback acknowledgments for reception.
In the MANET and VANET literature, previous proposed
methods that avoid broadcast storm problem include hop- based,
location- based, cluster- based, probabilistic- based, and traffic-based
suppression schemes such as [ 11] and [ 12]. Our method
to improve broadcast reliability integrates the vehicular
communication system with traffic flow by dynamically
adjusting transmission range based on traffic density and
vehicle speed characteristics. Further, our study on multi- hop
broadcast extends the potential application use cases. Single-hop
broadcast are useful for high locality and very time
sensitive applications such as crash imminent collision.
However, it does not provide safety applications that stretch
several miles for look- ahead warning to alert the downstream
traffic for advance speed reduction. Finally, multi- hop
broadcast communication may also have environmental
applications that reduce emission in vehicle platoons by
stabilizing traffic on the road through cooperative cruise
control systems.
III. DESIGN
A. Traffic Scenarios
Our traffic scenario includes two traffic streams with each
traffic stream stretching five kilometers and one kilometer
apart with uninterrupted traffic flow. Market penetration rate
( MPR) of equipped vehicle with communication device is
100% and vehicles are uniformly distributed according to their
traffic density. Since shockwaves are caused by variation in
speed differentials, the two traffic streams have different traffic
density with the leading traffic stream’s density greater than
the following traffic stream. It is generally accepted that, for
uninterrupted traffic flow, there is a density- speed relationship
[ 13]. In our simulation, we assume the so- called triangular
fundamental diagram [ 14] [ 15] with density ρ and speed V.
( ) ( ) ,
, 0
( )
j
f c
f c j
c
j c
V
V
V r r r
r r r
r r
r r r
r
£ £
£ £
-
-
 
=   
( 1)
We assume the conditions in which the free flow speed Vf =
104 km/ h ( 64.6 mph), a reasonable value for highway speed
limit. The jam density is ρj = 150 veh/ km [ 16], and critical
density ρ c = 0.2 ρj. Further, we assume density ρ1 = 90 veh/ km
and ρ2 = 30 veh/ km for the two traffic streams with vehicle
spacing 11.1 meters and 33.3 meters. Based on these
assumptions for triangular fundamental diagram and the
formulation in ( 1), a lane consists of 600 vehicles with leading
traffic stream vehicles traveling at 17.4 km/ h ( 10.8 mph),
following traffic stream vehicles at free flow speed. The
backward shockwave speed is - 26 km/ h ( 16 mph). Specifically,
our traffic scenario is relevant to a typical shockwave
encounter on a highway where vehicles in the downstream are
congested while the upstream vehicles are un- congested. The
distance between vehicles on neighboring lanes is set to 3.65
meters according to the highway capacity manual. The
shockwave pattern in the simulation is based on the speed-density
relationship and parameters described above, and is
created using MatLab and ported onto ns- 2 mobility file.
Figure 1 shows the trajectory of shockwave traffic in our
scenario with each line representing vehicle’s movement for a
specific location and time instant. Moreover, the figure
illustrates backward shockwave point propagation as vehicle
reduces their speed with the congestion traffic ahead from 64.6
mph to 10.8 mph.
Figure 1. Trajectory of Shockwave Traffic
B. Simulation Environment
We use ns- 2.33 network simulator to evaluate
communication performance with the mobility model
according to section 3- A. In the simulation, all nodes are
configured to flood all un- heard messages to follow the multi-hop
broadcasting behavior. To evaluate the impact of varying
communication range and transmission power adjustment, we
use the deterministic two- ray ground propagation for radio
model. For higher fidelity with realistic vehicle- to- vehicle
communications, we set configuration values according to the
IEEE 802.11p draft standard. For security protection, we
assign packet size to 382 bytes with 200 bytes of data payload,
128 bytes for a certificate, and 54 bytes for a signature similar
to [ 5]. The main parameters used in the ns- 2 simulation are
presented in Table 1. The simulation ran on a 2.3 GHz quad-core
with 8 GB RAM and the multi- core processors provide
speed up in the Monte Carlo simulation.
Information source is the first vehicle of the following traffic
stream that after 41 seconds detects the upcoming shockwave
ahead and broadcast a shockwave alert message once in both
upstream and downstream directions. For multiple- lane
situations, we assume that the first vehicle ( information source)
originates from lane one. Sending the shockwave message alert
to downstream vehicles on the same direction can be beneficial
as those vehicles can later relay messages in the opposing
direction of the highway for non- instantaneous forwarding.
TABLE I COMMUNICATION CONFIGURATION
Parameters Values
Antenna height 1.5 m
Antenna gain 1 dB
RxTh - 95 dBm
CSTh - 99 dBm
CPTh 4 dB
Data rate 3 Mbps
Frequency 5.9 GHz
Packet size 382 bytes
Minimum contention window 15 slots
Number of messages send 1
Tx range ( meters)
Corresponding power ( dBm)
37, 18.5
- 15.8, - 21.8
C. Transmission Range Adjustment
In our simulation, we use minimum transmission range
( MinTR) which is computed based on the spacing distance
between a leading and following vehicle. Since the MPR is
100%, the communication equipped vehicles are fully
connected. We compare the results with fixed MinTR, derived
using the value from following traffic density ρ2 and dynamic
minimum transmission range values for each traffic density ρ1
and ρ2. Note however the actual MinTR shown in Table 1 and
used in our simulation is a few meters more to compensate for
multiple lanes and flexibility that messages send by vehicle on
lane one can be heard by vehicles one vehicle distance away
for all lanes.
IV. SIMULATION RESULTS
A. Discussion
For statistical reliability and to avoid correlation in the
results, a Monte Carlo approach of 500 runs ( with varying seed
in ns- 2) for each scenario with different highway lanes is
computed. Additional scripts were used to compute parse the
raw output and compute performance measures of the collected
data. In particular, we evaluate two performance metrics for
multi- hop broadcasting, message delivery ratio ( MDR) and
packet reception rate ( PRR). MDR is measured at the
application level and defined by the probability of the message
send by the information source to travel a certain distance
along the traffic stream. PRR is measured in the MAC level
and defined as the probability of packet reception for a given
distance, measured in 100 meter segments. In the figures,
performance measure starts at the information source where the
first shockwave transition occurs ( kilometer distance zero).
Figure 2 and Figure 3 shows the MDR and PRR for fixed
transmission range for all vehicles, MinTR= 37. Figure 4 and
Figure 5 shows the MDR and PRR for dynamic transmission
ranges where vehicles in traffic density ρ1 are assigned
MinTR= 18.5 and vehicles in ρ2 with MinTR= 37. Difference in
the two traffic streams are attributed to the congested and free-flow
traffic patterns. In the MDR measure, as the number of
lanes increases for free flow traffic, the MDR also improves as
shown in Figure 2 and Figure 4. Further, the result for two
lanes is particularly low since it endures communication
interferences from vehicles in the adjacent lane and its traffic
density is least among all the multi- lane scenarios. In the case
of congested traffic with fixed transmission range, the MDR
achieves 100% with three or more lanes as it can fully reach
the 5 km distances. However, in congested traffic with
dynamic transmission range, only the one- lane scenario has
guaranteed reliability as indicated in Figure 4. This is because
for one lane case with MinTR, there is no contention in
wireless medium and no interferences from other vehicles
farther away in the forward and backward directions as well as
adjacent lanes.
Contrary to MDR, the PRR shows opposite effect where
more lanes result in lower packet reception rate. Further,
Figure 3 illustrates that in all cases of fixed transmission range,
there is a downward spike in PRR from the information source
to its nearby downstream traffic. This is triggered by the
transition from free flow and the increase in overall vehicle
density in the congested traffic stream.
B. Impact of Lane Configuration
In our highway traffic scenario, the number of lanes affects
the communication densities. This can be observed in both
MDR and PRR results. As we describe earlier, the one lane
scenario with MinTR is a special case that has the best results
for all figures and lane configurations except in the forward
direction in Figure 2. For the multi- level scenarios, the more
lanes the higher the application level delivery probability.
However, it comes at a tradeoff where greater traffic densities
cause more collisions in the MAC level and results with lower
packet reception. For the two lane scenario, the multi- hop
broadcast message propagates only about half the entire 5 km
in the direction of the free- flow traffic and its packet reception
rates has higher volatility due to less overall received packets
in comparison with three, four, or five lanes. Finally, the
dynamic MinTR adjustment for two lanes in the direction of the
congested traffic causes it to reach only about 1 km in distance.
C. Impact of Transmission Range
Although the transmission range adjustment for dynamic
MinTR results in lower MDR, it can improve PRR. The
analytical model proposed by [ 17] describes the relationship
between application and communication level delivery ratios
and its formulation shown in ( 2).
Papp( N) = P ( at least 1 successful tx in N tries)
= 1- P ( all fail in N tries) = 1-( 1- Pcom) N
( 2)
The DSRC standard requires that the packet generation rate
for safety messages are triggered every 100 milliseconds.
Hence, the MDR delivery ratio can quickly be compensated in
the case when multiple N messages are sent. Hence, the
tradeoff of lower MDR to compensate for higher PRR with
dynamic transmission range is desirable. Real field
experiments by the USDOT RITA VII project on the
communication performance also suggest the desire for low
packet error rate as a design consideration for DSRC [ 18]. It is
valid that it may be difficult to compute the absolute MinTR for
different free- flow traffic densities since the vehicle speed
would be the same. In fact, the DTRA algorithm suggests using
maximum transmission range ( MaxTR) since the less traffic
density with free flow will have less impact on wireless
medium contention and interferences. Our result on free- flow
traffic is the critical density ( ρ c = 0.2 ρj). Intuitively, for free-flow
traffic, if the transmission range was rather set to MaxTR,
the results should indicate the farthest distance travel with
highest MDR and lowest PRR possible.
V. CONCLUSION AND FUTURE WORK
In this paper, we study the performance of multi- hop
broadcasting on the highway traveling in one direction. We
suggest a mechanism to improve multi- hop broadcast
reliability and efficiency with dynamic transmission ranges
based on our understanding of fundamental traffic flow
relationships. In particular, we show the benefits of employing
dynamic transmission ranges on the highway with shockwave
mobility that inter- mixes free flow and congested flow traffic.
Using ns- 2 simulator, we evaluate the performance measure of
message delivery ratio and packet reception rates. In addition,
we show that lane configurations can have a major impact on
the performance measures.
Future work can incorporate complex traffic and network
characteristics for greater realism in shockwave mobility with
non- homogeneous stop- and- go traffic pattern to describe heavy
congestion. Moreover, message generation rate for sending
messages multiple times or from multiple information sources
are possible and can further clog the communication medium.
Studies on dynamic contention window for broadcasting have
been proposed by [ 19] and the metric of contention window
adjustment and its formulation can incorporate traffic flow
dynamics. Analytical methods to model the wireless contention
and communication reliability and efficiency for safety- based
DSRC systems have been studied recently by [ 20] [ 21]. Further,
theoretical analysis on the results and relationship for delivery
ratio in the application and communication level would be
helpful for understanding the factors that impact the
performance metrics in VANET. These methodologies can be
beneficial in the routing protocol design for VANET.
VI. ACKNOWLEDGMENT
This research is supported in part by a grant from University
of California Transportation Center. We would like to thank
Hao Yang for assistance in the shockwave mobility and to
Jaeyoung Jung and Dr. Jay Jayakrishnan for their discussions.
REFERENCES
[ 1] W. Jin and W. Recker, " An Analytical Model of Multihop Connectivity of
Inter- Vehicle Communication Systems," IEEE Transactions on Wireless
Communications, 2008.
[ 2] S. Schnaufer and H. F, " Vehicular Ad- Hoc Networks: Single- Hop Broadcast
is not enough," Proceedings of 3rd International Workshop on Intelligent
Transportation, 2006.
[ 3] J. Gomez and A. Campbell, " Variable- Range Transmission Power Control in
Wireless Ad Hoc Networks," IEEE Transactions on Mobile Computing, 2007.
[ 4] M. Artimy, " Local Density Estimation and Dynamic Transmission- Range
Assignment in Vehicular Ad Hoc Networks," IEEE Transactions on
Intelligent Transportation Systems, 2007.
[ 5] M. Killat and H. Hartenstein, " An Empirical Model for Probability of Packet
Reception in Vehicular Ad Hoc Networks," EURASIP Journal on Wireless
Communications and Networking, 2009.
[ 6] D. Jiang, Q. Chen, and L. Delgrossi, " Communication density: a channel load
metric for vehicular communications research," Proceedings of IEEE
Internatonal Conference on Mobile Adhoc and Sensor Systems, 2007.
[ 7] S. Yousefi, M. Fathy, and A. Benslimane, " Performance of beacon safety
message dissemination in Vehicular Ad hoc NETworks ( VANETs)," Journal
of Zhejiang University SCIENCE A, 2007.
[ 8] M. Torrent- Moreno, P. Santi, and H. Hartenstein, " Distributed fair transmit
power adjustment for vehicular ad hoc networks," IEEE Communications
Society on Sensor and Ad Hoc Communications and Networks, 2006.
[ 9] C. Palazzi, S. Ferretti, M. Roccetti, G. Pau, and M. Gerla, " How do you
quickly choreograph inter- vehicular communications? A fast vehicle- to-vehicle
multi- hop broadcast algorithm, explained," IEEE CCNC International
Workshop on Networking Issues in Multimedia Entertainment, 2007.
[ 10] F. Ros, P. Ruiz, and I. Stojmenovic, " Reliable and Efficient Broadcasting in
Vehicular Ad Hoc Networks," IEEE Vehicular Technology Conference, 2009.
[ 11] O. Tonguz, N. Wisitpongphan, J. Parikh, F. Bai, P. Mudaliget, and V.
Sadekart," On the Broadcast Storm Problem in Ad Hoc Networks,"
International Conference on Broadband Communications, Networks and
Systems, 2006.
[ 12] G. Korkmaz, E. Ekici, and F. Ozguner, " An Efficient Fully Ad- Hoc Multi-
Hop Broadcast Protocol for Inter- Vehicular Communication," IEEE
International Conference on Communications, 2006.
[ 13] F. L. Hall, B. Allen, and M. Gunter, " Empirical analysis of freeway flow-density
relationships," Transportation Research Part A, vol. 20, 1986.
[ 14] P. K. Munjal, Y. S. Hsu, R. L. Lawrence, “ Analysis and validation of lane- drop
effects of multilane freeways.” Transportation Research, 1971.
[ 15] G. F. Newell “ A simplified theory of kinematic waves in highway traffic I:
General theory. II: Queuing at freeway bottlenecks. III: Multi- destination
flows.” Transportation Research Part B, 1993.
[ 16] J. Del Castillo and F. Benitez, " On the functional form of the speed- density
relationship— I: General theory," Transportation Research B, 1995.
[ 17] F. Bai, T. Elbatt, and G. Hollan, " Towards characterizing and classifying
communication- based automotive applications from a wireless networking
perspective," Proceeding of the IEEE Workshop on Automotive Networking
and Applications, 2006.
[ 18] S. Andrews and M. Cops, " USDOT: Vehicle Infrastructure Integration Proof
of Concept Results and Findings ( FHWA- JPO- 09- 043)," Final Report, 2009.
[ 19] N. Balan, J. Guo, " Increasing Broadcast Reliability in Vehicular Ad Hoc
Networks," Proceedings of the Third ACM International Workshop on
Vehicular Ad Hoc Networks, 2006.
[ 20] X. Ma, X. Chen, and H. Refai, " Performance and Reliability of DSRC
Vehicular Safety Communication: A Formal Analysis," EURASIP Journal On
Wireless Communications and Networking, 2009.
[ 21] J. He, Z. Tang, T. O'Farrell, and T. Chen, " Performance analysis of DSRC
priority mechanism for road safety applications in vehicular networks,"
Wireless Communications and Mobile Computing, 2009.
1
Abstract— Inter- vehicle communication is a promising way
to share and disseminate real- time and nearby safety
information on the road. However, several pressing open
questions require solutions in order to achieve high reliability
and efficiency with these systems. Further, previous studies
have shown that the mobility model can significantly influence
the communication performance in vehicular networks. In this
paper, we analyze communication in stop- and- go waves and
propose a method to optimize an important network
parameter, the transmission range, based on traffic stability
measures. Our findings suggest a transmission range
adjustment scheme that achieves high reliability by considering
network coverage and packet reception rates.
I. INTRODUCTION
In recent years, computing systems and communication
capabilities have become more affordable, powerful, and
accessible. For example, the proliferation of smart phone
computing devices has enabled more people to stay
connected to the Internet over longer time spans. Similarly,
this trend is now expanding to vehicles. The global
positioning system ( GPS) that integrates computing and
satellite communication has resulted in millions of vehicle
drivers with real- time road navigation information in the
United States. Advanced telematic systems will only
continue to grow and facilitate drivers with better and more
accurate real- time traffic and safety information.
Dedicated Short Range Communication ( DSRC) is a
technology based on 802.11p that operates using 75 MHz of
spectrum band in the 5.9 GHz range, and is specifically
designed for automotive use in road safety and
complementary traffic information. Due to the time-sensitive,
safety- critical applications in VANET,
broadcasting will play an important role in vehicular
communication to disseminate messages about unsafe
driving conditions to immediate nearby vehicles ( one- hop)
and other vehicles in the vicinity ( multi- hop). However,
there are several challenges to broadcast packets reliably.
First, broadcast lacks acknowledgement ( ACK) packets from
the receiver. As a result, there is no retransmission of
dropped packets. Due to this lack of MAC- layer recovery,
the contention window size for broadcast is often held
constant ( fixed). This differs from unicast which adjusts the
contention window size based on a binary exponential back-off
scheme, depending on the packet failure probability. In
addition, reservation schemes used in unicast such as
RTS/ CTS exchange cannot be efficiently used for broadcast
since the nature of disseminating packets would exacerbate
the broadcast storm problem with the additional RTS/ CTS
control packet exchanges. Inherently, communicating
devices should adapt based on the dynamic vehicular
network.
One of the most important factors that impacts network
reliability is the interference level which is highly dependent
on the transmission range for each communicating node. In
this paper, we carefully study stop- and- go movement and
incorporate an understanding of traffic waves onto the
network design for one- hop periodic broadcast. Stop- and- go
movement, a phenomenon that arises from a combination of
shockwave and rarefaction waves, can occur in highways,
especially during peak hours or when road incidents occur.
Through analytical and simulation- based studies, we
illustrate the coverage and packet reception rates
performance measures for different traffic dynamics. Taking
into consideration both reliability and interference
minimization, we compare the performance for various
transmission range adjustment schemes relative to the traffic
stability.
II. RELATED WORKS
Our work is motivated by [ 1] which provides a first study
to obtain the analytical lower- bound for the minimum
transmission range in non- homogeneous distribution of
vehicles in congested densities. Following this initial work,
[ 2] uses a dynamic transmission- range assignment ( DTRA)
algorithm that employs transmission power control based on
the relationship between connectivity and traffic density
characteristics. Their approach is based on an analytical
traffic flow model to estimate local density and derive
vehicle trajectories using RoadSim to measure the
performance of the communication system on several road
configurations. The focus of their work and the DTRA
algorithm is to adjust the transmission range by estimating
local vehicle density and local traffic conditions ( free flow
versus congested traffic) without any prior message
exchange with neighboring vehicles. In their work, the
minimum transmission range is defined as an average
maximum value of vehicle spacing for multi- lane case and
Dynamic Transmission Range in Inter- Vehicle Communication
with Stop- and- Go Traffic
Rex Chen, Hao Yang, Wen- Long Jin, Amelia Regan
{ rex, hyang5, wjin, aregan}@ uci. edu
University of California, Irvine
2
the widest gap among vehicles for single- lane scenario.
Further, to compensate for the non- homogeneous
distribution of vehicles on a single- lane, the transmission
range is increased by an additional constant that is
proportional to length of the road of interest. Although their
work achieves the goal of maintaining high connectivity, the
communication issues such as collision due to the hidden
and exposed terminal problems were not evaluated. An
optimal adjustment in transmission range would improve
communication by reducing wireless transmission collisions.
Our work extends the dynamic transmission range by
analyzing traffic dynamics on the road and incorporating
traffic stability information as a relative measure to increase
transmission range.
The work by [ 3] proposes the distributed fair power
adjustment for vehicular networks ( D- FPAV) algorithm that
dynamically adjusts each vehicle’s transmission power to
prevent packet collisions. The optimization focuses on
fairness of each communicating vehicle to receive and send
safety information rather than network capacity, connectivity
or coverage. Fairness in their adaptive transmission power
scheme is validated through simulation results on a highway
with different radio propagation models.
The work by [ 4] proposes an analytical model to evaluate
the performance and reliability of safety- related services in
DSRC systems on highways. The model considers several
design metrics which include different safety- message
priorities, the hidden terminal problem, transmission range,
and contention window back- off mechanisms. From their
analytical model, channel throughput, transmission delay,
and packet reception rates were computed. The findings
suggest that delay requirements can be met but high
reliability cannot. The work by [ 5] provides extensive
simulations to study the performance of one- hop broadcast
beacon safety messages. Communication parameters used in
the performance measures include transmission range,
packet transmission interval, and message payload size.
The work by [ 6], [ 7] proposes an analytical model for
connectivity in non- uniform traffic stream based on the
Lighthill- Whitham- Richards ( LWR) traffic flow model. The
instantaneous connectivity factor is based the multi- hop
broadcast communication and with different market
penetration rates of DSRC- equipped vehicles. Further,
connectivity can be computed as the traffic pattern evolves
in a time- dependent manner. Theoretical results on the
propagation distance for different transmission range values
are shown for non- uniform traffic. The work by [ 8] proposes
an analytical method to approximate connectivity for
vehicular communication in highway under different traffic
conditions as factors such as traffic density and vehicle
velocity parameters can significant influence the
performance of connectivity. Finally, [ 9] proposes to
improve communication reliability with dynamic
transmission range by incorporating fundamental traffic flow
relationship. The work is focused on shockwave mobility
pattern for multi- hop broadcast communication which is
different from this paper.
III. TRAFFIC BEHAVIOR AND MODELING
This section describes the traffic scenario, vehicle
movements and trajectories, and methodology to precisely
compute vehicle locations and traffic stability in detail.
A. Traffic Scenarios
Our traffic scenario is a non- uniform congested traffic
stream that covers a three kilometer unidirectional, one- lane
highway network. We assume a critical density ρ c = 0.2 ρj
and a jam density of 150 veh/ km. Further, we assume that
every vehicle is DSRC- enabled ( 100% market penetration
rate). Initially, the vehicles are randomly distributed within
the three kilometer road segment with a condition that the
distance between any two DSRC- enabled communications
device is minimally 6.66 meters based on jam density value.
Due to the non- uniform distribution of vehicles, there are
instances of the road segment where the spacing between the
forward and rear vehicle can be greater than the average
vehicle spacing of the entire traffic stream for a given traffic
density.
B. Car- Following Model
In traffic flow theory, various microscopic traffic models
have been proposed such as Gibbs, General Motors, Pipes or
the K- S car following models. In our traffic network,
vehicles movement is based on Newell’s car- following
model for its simplicity. Furthermore, the accuracy of
Newell’s car- following model [ 10] has been compared with
other microscopic car- following models [ 11], and have
subsequently been verified with real highway results [ 12],
[ 13].
The following formulation ( 1) describes Newell’s car-following
model in a congested road:
􀜺 􀯡 􁈺 􀝐 􀵅 􀟬 􁈻 􀵌 􀜺 􀯡 􀬿 􀬵 􁈺 􀝐 􁈻 􀵆 􀝀 ( 1)
where Xn and Xn- 1 are the following and leading vehicles’
locations, respectively, d is the jam spacing of vehicle Xn,
and τ is the time displacement of vehicle Xn. From the NG-SIM
data [ 14], d and τ are set to 6.66 meters and 1 second,
respectively. Hence, the nth vehicle trajectory will follow
the trajectory of the ( n- 1) st vehicle as described in ( 1) for all
vehicles on a congested road.
C. Vehicle Trajectories
Vehicle trajectories of stop- and- go waves for different
congested traffic densities ( from ρ = 0.2ρj to ρ = 0.9ρj) of two
minutes of driving time are computed in Figure 1. Increasing
traffic density not only increases the number of vehicles on
the road, but decreases vehicle speed which reduces spacing
between vehicles. From the vehicle trajectories, we observe
that all stop- and- go waves propagate backward as shown in
Figures 1( a) to 1( h). As shown in those figures, as traffic
density increases, more stop- and- go waves are created.
However, when the traffic pattern is denser ( ρ > 0.5ρj), these
narrower stop- and- go waves start to merge into wider ones
as shown in Figures 1( e) to 1( h).
D. Traffic
Using New
of each indiv
knowing the
variance ( CV)
can be compu
vehicle distri
according to
spacing decre
Figure 2 illus
for different
convergences
densities. For
value converg
veh/ km, it too
traffic stream
most of the tr
points where
traffic densiti
since these h
the formation
Figure 2. E
Ra
Dynamics
well’s car- follo
vidual vehicle
precise vehi
V) of spacing fo
uted. Initially
ibution. In la
o Newell’s c
eases until it
strates an exa
t traffic den
s much faster
r example, wh
ges within 10
ok up to 40
m to reach “ st
raffic stream i
the stop- and- g
ies have a low
higher traffic
n of large gaps
Evolution of t
andom Initial D
Figu
owing model
e on the road
icle locations
for all vehicles
the CV is hig
ter time step
car- following
t convergence
ample of CV a
nsities. In co
for higher den
hen ρ > 60 veh
0 seconds or l
seconds to c
tationary.” A
s smooth exce
go wave occu
wer CV value
streams have
s between veh
the Coefficien
Distribution o
ure 1. Vehicl
in III- B, the l
can be deriv
s, the coeffic
s in the traffic
gh due to the r
s, as vehicles
model, the
es to a fixed
adjustment ( s
omparison, t
nsity than with
h/ km ( 0.5ρj) ,
less but with
onverge and
t the stability
ept for a few s
urs. In general,
e to start ( tim
less space to
hicles.
nt of Variance
f Vehicles
3
e Trajectories
location
ved. By
cient of
c stream
random
s move
CV of
d value.
pacing)
the CV
h lower
the CV
ρ = 30
for the
y point,
specific
, higher
me t= 0)
o allow
with
tra
co
A
bro
sin
oc
pe
sam
int
for
bro
ad
B
dy
Th
en
ne
tra
fol
co
ve
tra
as
C
co
s under Varyin
This section d
ansmission
mmunication
A. Broadcast
In vehicular
oadcast storm
ngle- hop perio
curs due to
riodic messa
me sender) is
terval is short
r safety applic
oadcast which
daptive cruise c
B. Transmiss
Our propose
ynamically by
he increase in
nsure a desirab
twork for a
ansmission ra
llowing rule:
􀜶 􀜴 􀯔 􀯗 􀯝
where n is t
efficient of v
hicle spacing
affic becomes 􀜶 􀜴 􀯔 􀯩 􀯚 _ 􀯦 􀯣 .
C. Coverage
In this sectio
mmunication
ng Densities
IV. NETW
describes the
range adjus
reliability.
ting
networks, tw
m problem ar
odic messages
message floo
aging ( consec
s problematic
t. In this wor
cations on hig
h include pre
control applic
sion Range Ad
ed scheme a
y taking traff
n transmissio
ble coverage
specific t
ange ( 􀜶 􀜴 􀯔 􀯗 􀯝 )
􁈺 􀝊 􁈻 􀵌 􁈺 1 􀵅 􀝊 the order of
variance ( CV)
g over the en
uniform, CV
Model
on, we descr
coverage in t
WORK DESIGN
mechanism in
stment for
wo scenarios
re multi- hop
s. In the form
oding while i
cutive transm
c when the p
rk, we evalua
ghways with s
e- crash sensin
cations.
djustment
djusts the tr
fic stability in
on range is r
value for all
traffic patter
can be com
􀗛 􀜥 􀜸 􁈻 􀗛 􀜶 􀜴 􀯔 􀯩
magnitude f
V) and 􀜶 􀜴 􀯔 􀯩 􀯚 _ ntire traffic s
V is zero and 􀜶
ribe the mod
the vehicular
n broadcasting
improving
that lead to
event- driven
mer case, the
in the latter
missions from
packet time e
ate communic
ingle- hop per
g and cooper
ransmission r
nto considera
relative to C
nodes in the
rn. The adju
mputed using
􀯚 _ 􀯦 􀯣 ( 2)
for increasing 􀯦 􀯣 is the ave
stream. When 􀜴 􀯔 􀯗 􀯝 is the
del for measu
network. Sup
g and
the
o the
n and
issue
case,
m the
lapse
cation
riodic
rative
range
ation.
CV to
road
usted
g the
g the
erage
n the
same
uring
ppose
4
n vehicles travel in a road defined as 􀝒 􀬵 , 􀝒 􀬶 , 􀚮 , 􀝒 􀯡 , and the
positions for all n vehicles are defined as 􀝔 􀬵 , 􀝔 􀬶 , 􀚮 􀝔 􀯡 .
Further, assume that 􀝒 􀬵 is the leading vehicle of the traffic
stream 1,2, 􀚮 , 􀝊 an 􀵆 d 1 􀝒 . 􀯜 􀬾 L 􀬵 e t isth et hetr afnoslmloiwssiinogn vraenhgicel eo ff ovre h 􀝒 ic 􀯜 , l e 􀗊 i 􀝅 b 􀵌 e
denoted as 􀜴 􀯜 . Then the upstream and downstream coverage
is defined by the following definition:
􀜥 􀯜 , 􀯨 􀯣 􀯦 􀯧 􀯥 􀯘 􀯔 􀯠 􀵌 􁈼 10/ 2 􀗌 􀸫 􀝔 􀯜 􀵆 􀝔 􀯝 􀸫 􀵑 􀜴 􀯜 ,, 􀗊 􀝋 􀝐 􀝆 􀝄 􀵌 􀝁 􀝎 1 􀝓 ,2 􀝅 􀝏 , 􀝁 􀚮 , 􀝅 􀵆 1
( 3)
􀜥 􀯜 , 􀯗 􀯢 􀯪 􀯡 􀯦 􀯧 􀯥 􀯘 􀯔 􀯠 􀵌 􁈼 10/ 2 􀗌 | 􀝔 􀯜 􀵆 􀝔 􀯞 | 􀵑 􀜴 􀯜 ,, 􀝋 􀗊 􀝐 􀝄 􀝇 􀝁 􀵌 􀝎 􀝓 􀝅 􀝅 􀵅 􀝏 􀝁 1, 􀚮 , 􀝊
( 4)
The coverage of each vehicle i is defined in terms of the
Euclidean distance to the nearest upstream and downstream
vehicles in the traffic stream:
􀜥 􀯜 􀵌 􀜥 􀯜 , 􀯨 􀯣 􀯦 􀯧 􀯥 􀯘 􀯔 􀯠 􀵅 􀜥 􀯜 , 􀯗 􀯢 􀯪 􀯡 􀯦 􀯧 􀯥 􀯘 􀯔 􀯠 ( 5)
The total coverage 􀜥 of this vehicular network is denoted by:
􀜥 􀵌 Σ 􀯜 􀯡 􀭀 􀬵 􀜥 􀯜 / 􀝊 ( 6)
D. Results and Discussion
Here, we illustrate the effects of traffic dynamics that
range and density ( from critical to jam density) on
transmission range adjustment and coverage value defined
earlier in sections IV- B and IV- C. Tables 1 and 2 provide
details of the simulation runs of the analytical model for
coverage with different transmission range adjustments. For
higher fidelity in the results, the simulation was run 100
times with randomized traffic locations ( with minimum 6.66
meters apart) for all vehicles and the average results are
presented.
Table 1 shows the actual transmission range value
increases according to equation ( 2). This adjustment value
can be observed to be highly related by traffic stability.
Comparing the two traffic patterns, we observe that the
actual transmission range adjustment is greater in the initial
randomized traffic. This is due to the fact that the coefficient
of variance value is lower for stationary traffic using
Newell’s car- following model. Also, the transmission range
differences between initial randomized and stationary traffic
is less apparent in higher traffic densities.
As observed in Table 2, the increase in coverage is most
apparent from 􀜶 􀜴 􀯔 􀯗 􀯝 􁈺 0 􁈻 to 􀜶 􀜴 􀯔 􀯗 􀯝 􁈺 1 􁈻 except when the traffic
density is high such as ρ = 0.9 ρj and the traffic is near
stationary to begin with. In order to achieve a 95% percentile
in coverage in most cases, a transmission range adjustment
of 􀜶 􀜴 􀯔 􀯗 􀯝 􁈺 2 􁈻 and 􀜶 􀜴 􀯔 􀯗 􀯝 􁈺 3 􁈻 is necessary for initial
randomized traffic and stationary traffic.
We can see the impact of stop- and- go waves on traffic
stability in the converged traffic scenario. In the 􀜶 􀜴 􀯔 􀯗 􀯝 􁈺 0 􁈻
and 􀜶 􀜴 􀯔 􀯗 􀯝 􁈺 1 􁈻 values, the coverage increase is consistent
with higher traffic density. In addition, the coverage for a
few traffic densities stay the same, in ρ = 0.2 ρj with 􀜶 􀜴 􀯔 􀯗 􀯝 􁈺 1 􁈻 and thereafter, and in ρ = 0.3 ρj and ρ = 0.4 ρj with 􀜶 􀜴 􀯔 􀯗 􀯝 􁈺 2 􁈻 and thereafter. When traffic density increases, the
ratio between 􀜶 􀜴 􀯔 􀯩 􀯚 _ 􀯦 􀯣 and “ go” pattern spacing of the stop-and-
go wave is greater and a larger transmission range
adjustment of 􀜶 􀜴 􀯔 􀯗 􀯝 􁈺 3 􁈻 is necessary to achieve a coverage
value that approach 1.
V. SIMULATION ANALYSIS
A. Simulation Environment
We use the ns- 2.33 network simulator to evaluate
communication performance with the mobility model
described in section III- C. For higher fidelity, we set
configuration values according to the IEEE 802.11p standard
draft and the main parameters used in the ns- 2 simulation are
presented in Table 3. To measure reliability of single- hop
periodic broadcast, all nodes in the highway broadcast safety
messages at 100 ms intervals for a duration of two seconds
( an upper bound on human reaction time). The packet size is
set to 382 bytes with 200 bytes of data payload, 128 bytes
for a certificate, and 54 bytes for a signature, similar to [ 15].
The preferred data rate of 6 Mbps for vehicular safety
applications is used which has the greatest benefit in overall
reliability ( in terms of packet reception rates) as confirmed
by [ 16]. The simulation ran on a 2.3 GHz quad- core machine
with 8 GB RAM and the multi- core processors provide
speed up in the Monte Carlo simulations.
TABLE 3 COMMUNICATION CONFIGURATIONS
Parameters Values
Antenna height 1.5 m
Antenna gain 1 dB
RxTh - 95 dBm
CSTh - 99 dBm
CPTh 4 dB
Data rate 6 Mbps
Frequency 5.9 GHz
Packet size 382 bytes
Transmission criteria Single- hop periodic for
all nodes in network
Message transmission interval 100 ms
Contention window size 15 slots ( fixed)
Slot time 16 μs
Tx range ( meters) See table 1
B. Results and Discussion
For statistical reliability and to avoid correlation in
the results, 100 independent runs ( with varying seeds in ns-
2) for each scenario are computed. Additional scripts were
used to parse the raw output and compute performance
measures. In particular, we evaluate the performance metric
of packet reception rates ( PRR) for all nodes. PRR is
measured in the MAC level and is defined as the probability
of receiving a packet sent within transmission distance.
5
TABLE 1. TRANSMISSION RANGE ADJUSTMENT ( IN METERS)
Initial Traffic ( randomized) Stationary Traffic ( after convergence)
density ( veh/ km) 􀜶 􀜴 􀯔 􀯗 􀯝 􁈺 0 􁈻 􀜶 􀜴 􀯔 􀯗 􀯝 􁈺 1 􁈻 􀜶 􀜴 􀯔 􀯗 􀯝 􁈺 2 􁈻 􀜶 􀜴 􀯔 􀯗 􀯝 􁈺 3 􁈻 􀜶 􀜴 􀯔 􀯗 􀯝 􁈺 0 􁈻 􀜶 􀜴 􀯔 􀯗 􀯝 􁈺 1 􁈻 􀜶 􀜴 􀯔 􀯗 􀯝 􁈺 2 􁈻 􀜶 􀜴 􀯔 􀯗 􀯝 􁈺 3 􁈻
ρ = 0.2ρj ( 30) 33.333 60.556 87.779 115.001 33.333 40.282 47.231 54.180
ρ = 0.3ρj ( 45) 22.222 39.569 56.916 74.263 22.222 34.377 46.531 58.686
ρ = 0.4ρj ( 60) 16.667 29.286 41.905 54.525 16.667 27.710 38.753 49.795
ρ = 0.5ρj ( 75) 13.333 23.300 33.266 43.232 13.333 22.756 32.180 41.603
ρ = 0.6ρj ( 90) 11.111 19.067 27.022 34.977 11.111 18.874 26.637 34.400
ρ = 0.7ρj ( 105) 9.524 15.794 22.064 28.334 9.524 15.733 21.941 28.150
ρ = 0.8ρj ( 120) 8.333 13.027 17.721 22.415 8.333 13.006 17.679 22.351
ρ = 0.9ρj ( 135) 7.407 10.463 13.519 16.575 7.407 10.461 13.514 16.567
TABLE 2. NETWORK COVERAGE
Initial Traffic ( randomized) Stationary Traffic ( after convergence)
density ( veh/ km) 􀜶 􀜴 􀯔 􀯗 􀯝 􁈺 0 􁈻 􀜶 􀜴 􀯔 􀯗 􀯝 􁈺 1 􁈻 􀜶 􀜴 􀯔 􀯗 􀯝 􁈺 2 􁈻 􀜶 􀜴 􀯔 􀯗 􀯝 􁈺 3 􁈻 􀜶 􀜴 􀯔 􀯗 􀯝 􁈺 0 􁈻 􀜶 􀜴 􀯔 􀯗 􀯝 􁈺 1 􁈻 􀜶 􀜴 􀯔 􀯗 􀯝 􁈺 2 􁈻 􀜶 􀜴 􀯔 􀯗 􀯝 􁈺 3 􁈻
ρ = 0.2ρj ( 30) 0.644 0.900 0.944 0.978 0.122 0.989 0.989 0.989
ρ = 0.3ρj ( 45) 0.607 0.852 0.941 0.970 0.474 0.644 0.993 0.993
ρ = 0.4ρj ( 60) 0.633 0.861 0.956 0.967 0.594 0.783 0.994 0.994
ρ = 0.5ρj ( 75) 0.689 0.862 0.951 0.978 0.667 0.813 0.889 0.996
ρ = 0.6ρj ( 90) 0.733 0.863 0.948 0.974 0.719 0.841 0.922 0.967
ρ = 0.7ρj ( 105) 0.737 0.863 0.937 0.962 0.737 0.863 0.937 0.959
ρ = 0.8ρj ( 120) 0.819 0.903 0.944 0.972 0.819 0.897 0.939 0.967
ρ = 0.9ρj ( 135) 0.904 0.943 0.960 0.983 0.904 0.941 0.958 0.978
To calculate the probability of packet reception with the
corresponding transmission range adjustment, our analysis
on reliability is based on a weighted packet reception rate
that multiplies the PRR and coverage. Figures 3 and 4
illustrate the performance measures for initial traffic and
stationary traffic which exhibit the stop- and- go waves. For
both Figures 3 and 4, a 70% packet reception rate with
coverage is achieved in the optimal case.
In Figure 3, the packet reception rate with coverage is
consistence with a higher transmission range adjustment.
Further, 􀜶 􀜴 􀯔 􀯗 􀯝 􁈺 2 􁈻 and 􀜶 􀜴 􀯔 􀯗 􀯝 􁈺 3 􁈻 have similar results for all
traffic densities. Actual selection of 􀜶 􀜴 􀯔 􀯗 􀯝 􁈺 2 􁈻 and 􀜶 􀜴 􀯔 􀯗 􀯝 􁈺 3 􁈻
is dependent on the network design criteria and whether
higher reliability or higher coverage is more important.
6
Figure 3. PRR with Coverage for Initial Randomized Traffic
Figure 4 indicates a large difference in packet reception
rate with coverage. For small and large traffic densities, 􀜶 􀜴 􀯔 􀯗 􀯝 􁈺 2 􁈻 performed better, while moderate congested
traffic, 􀜶 􀜴 􀯔 􀯗 􀯝 􁈺 3 􁈻 showed better results. This is because there
are more stop- and- go patterns in the moderate congested
traffic, as previously shown in Figures 1( d) and 1( e).
Figure 4. PRR with Coverage for Stationary Traffic
VI. CONCLUSION
Deploying successful large scale VANETs hinges on the
ability of these systems to guarantee message delivery. In
this work, we examine the performance of broadcast
communication and seek to improve its reliability with
dynamic transmission range adjustment. In particular, we
analyze traffic dynamics as a result of stop- and- go waves for
varying traffic densities.
Longer transmission range allows for more receiving
nodes but at the expense of higher interference. Our
evaluation of dynamic transmission range adjustment
includes an analytical study of coverage and simulation
study of packet reception rates using ns- 2. Based on our
observation, we see that the near optimal transmission range
adjustment with traffic stability consideration is near two to
three times the coefficient of variance. Moreover, a stop-and-
go traffic pattern can impact the transmission range
adjustment decision, depending on traffic density.
For future work, mixed traffic can be considered with
different vehicle types, time displacement values, and multi-lane
highway scenarios. To study how traffic should inform
network design in large scale vehicular networks,
macroscopic traffic model can be used. In addition, a multi-layer
networking model that involves both the upper
( application) and lower ( network) layers for wireless
broadcast should be investigated and designed for future
inter- vehicle communication systems.
VII. ACKNOWLEDGMENT
This research is supported in part by a grant from
University of California Transportation Center.
REFERENCES
[ 1] M. Artimy, W. Robertson, W. J. Phillips, “ Minimum transmission
range in vehicular ad hoc networks over uninterrupted highways,” 9th
International IEEE Conference on Intelligent Transportation Systems,
2006.
[ 2] M. Artimy, " Local Density Estimation and Dynamic Transmission-
Range Assignment in Vehicular Ad Hoc Networks," IEEE
Transactions on Intelligent Transportation Systems, 2007.
[ 3] M. Torrent- Moreno, J. Mittag, P. Santi, and H. Hartenstein, " Vehicle-to-
Vehicle Communication: Fair Transmit Power Control for Safety-
Critical Information," IEEE Transactions on Vehicular Technology,
2009.
[ 4] Ma, X. Chen, and H. Refai, " Performance and Reliability of DSRC
Vehicular Safety Communication: A Formal Analysis," EURASIP
Journal On Wireless Communications and Networking, 2009.
[ 5] Y. Saleh, F. Mahmood, B. Abderrahim. " Performance of beacon
safety message dissemination in Vehicular Ad hoc NETworks
( VANETs)," Journal of Zhejiang University SCIENCE A, 2007.
[ 6] W. L. Jin and W. Recker, " An analytical model of multihop
connectivity of intervehicle communication systems." IEEE
Transactions on Wireless Communications, 2010.
[ 7] W.- L. Jin and W. Recker, “ Instantaneous connectivity of one-dimensional
inter- vehicle communication networks with general
traffic conditions” Proceedings of Transportation Research Board
Annual Meeting, 2009.
[ 8] S. Ukkusuri, L. Du, " Geometric connectivity of vehicular ad hoc
networks: Analytical characterization ," Transportation Research Part
C: Emerging Technologies, volume 16, Issue 5, pp. 615- 634, 2008.
[ 9] R. Chen, W. L. Jin, A. Regan, " Multi- Hop Broadcasting in Vehicular
Ad Hoc Networks with Shockwave Traffic," 2nd IEEE Intelligent
Vehicular Communications System Workshop, 2010.
[ 10] G. F. Newell, " A simplified car- following theory: a lower order
model," Transportation Research Part B 36, pp. 195- 205. 2002
[ 11] E. Brockfeld, R. D. Kühne, P. Wagner, " Calibration and Validation of
Microscopic Traffic Flow Models," Transportation Research Record:
Journal of the Transportation Research Board, No. 1876, pp. 62– 70,
2004.
[ 12] A. Soyoung, M. J. Cassidy, J. Laval, “ Verification of a simplified car-following
theory,” Transportation Research Part B 38, pp. 431– 440,
2004.
[ 13] H. Yeo, A. Skabardonis, " Understanding Stop- and- go Traffic in View
of Asymmetric Traffic Theory," Transportation and Traffic Theory,
2009 ( book chapter).
[ 14] NG- SIM website:
http:// ops. fhwa. dot. gov/ trafficanalysistools/ ngsim. htm
[ 15] M. Killat, H. Hartenstein, " An Empirical Model for Probability of
Packet Reception in Vehicular Ad Hoc Networks," EURASIP
Journal on Wireless Communications and Networking, 2009.
[ 16] D. Jiang, Q. Chen, L. Delgrossi, " Optimal Data Rate Selection for
Vehicle Safety Communications," ACM International Workshop
on Vehicular Ad Hoc Networks ( VANET), 2008.