Optimal Data Rate Selection for Vehicle Safety

Communications
Daniel Jiang, Qi Chen, Luca Delgrossi
Mercedes-Benz Research & Development North America, Inc.
850 Hansen Way, Palo Alto, CA 94304, U.S.A.
{Daniel.Jiang, Qi.Chen, Luca.Delgrossi}@Daimler.com

ABSTRACT terminal effect is amplified. Consequently, the efficiency of

This paper answers a simple but important question in VANET channel usage is important in managing the broadcast
research: what is the optimal data rate to be used in DSRC-based transmissions. In other words, efficient channel usage helps
vehicle safety communications? While it is generally accepted reduce the overall interference level and in turn impacts on the
that the default choice is 6 Mbps, this assumption is not rooted in broadcast reception performance.
strong technical considerations. This paper provides a systematic The data rate used in transmissions has a direct impact on the
evaluation of optimized data rates choices in a variety of level of interference introduced into the system. In general, a
scenarios. The answer found enables researchers to generally higher data rate results in shorter frame durations, thus producing
eliminate one dimension of complexity in relevant VANET less interference. On the other hand, a higher data rate requires
studies. Additionally, the methodology used in this paper is more power to transmit towards a certain intended range, thus
possibly as interesting as the conclusion. producing more interference. Therefore, the topic of selecting an
optimal data rate is highly relevant for vehicle safety
Categories and Subject Descriptors communications research.
C.2.1 [Network Architecture and Design]: Wireless In this paper, an innovative method is introduced to compare
Communication different data rate choices in a fair manner. We show that it is
possible to configure broadcast transmissions with different data
General Terms rates to produce exactly the same level of interferences in the
Performance, Design channel by manually adjusting the power used. In turn, the
performances of these transmissions are compared to identify the
Keywords optimal data rate.
Data Rate, Interference, IEEE 802.11, DSRC, Vehicle Safety The rest of this paper is organized as follows: Section 2 describes
Communications DSRC technology and the data rate options in details. Section 3
presents previous work. Section 4 discusses the proper
methodology for this study, including a description of a DSRC
channel load metric needed by the methodology. Section 5 lists
1. INTRODUCTION all the details of the simulation setup. Section 6 analyzes the
In traditional wireless systems, it is a common practice to use the results. Section 7 summarizes the findings.
lowest supported data rate for broadcast transmissions. This
simplistic approach is reasonable because such transmissions 2. DSRC OVERVIEW
generally contribute only a small fraction of the overall In 1999, the U.S. Federal Communication Commission allocated
communications activities on the channel. As such, efficiency of 75 MHz of Dedicated Short Range Communications (DSRC)
the channel usage is usually not a major concern for broadcast spectrum at 5.9 GHz to be used exclusively for vehicular
transmissions. In turn, the lowest data rate, which is the most communications. The primary goal is to enable public safety
robust for reception, is the preferred choice with the intent to applications that can save lives and improve traffic flow [10].
maximize the number of stations with successful receptions. Similar efforts are going on in Japan and Europe.
Vehicle safety communications, on the contrary, present a The radio technology for DSRC is being standardized as IEEE
significantly different pattern, consisting almost entirely of 802.11p Wireless Access for Vehicular Environments (WAVE)
pervasive broadcast transmissions among an unbounded [2], which is an amendment to the overall IEEE 802.11 standard
distribution of nodes [7]. Within such a system, the hidden [1]. IEEE 802.11p uses essentially the same PHY defined for
IEEE 802.11a, but operates in a 10 MHz wide channel instead of
20 MHz. This change in channel width is accomplished through a
Permission to make digital or hard copies of all or part of this work for simple scaling (i.e., doubling all relevant timing parameters) of
personal or classroom use is granted without fee provided that copies are the PHY. The purpose is to create a longer guard interval, which
not made or distributed for profit or commercial advantage and that is designed to address the increased RMS delay spread in the
copies bear this notice and the full citation on the first page. To copy
otherwise, or republish, to post on servers or to redistribute to lists,
vehicular environments [3].
requires prior specific permission and/or a fee.
VANET’08, September 15, 2008, San Francisco, California, USA.
Copyright 2008 ACM 978-1-60558-191-0/08/09...$5.00.

30
2.1 Modulation and Coding Rate channels halved all data rates in comparison to IEEE 802.11a; but
IEEE 802.11p uses Orthogonal Frequency Division Modulation the resulting lowest data rate of 3 Mbps was considered to
(OFDM), which divides the 10 MHz channel into 52 orthogonal “sound” too low for vehicle safety communications; the higher 6
sub-carriers, of which 48 are used to carry data. Sub-carriers are Mbps rate was picked as the default choice instead.
said to be “orthogonal” as they do not interfere with each other This study is an effort to use a rigorous technical evaluation to
even though they overlap in frequency. either justify the 6 Mbps assumption or to introduce a new
The following modulation schemes are available for IEEE scheme to select an optimal data rate for vehicle safety
802.11p to use on sub-carriers: Binary Phase Shift Keying communications under various conditions.
(BPSK), Quadrature Phase Shift Keying (QPSK), 16-point
Quadrature Amplitude Modulation (16-QAM), and 64-point 3. PREVIOUS WORK
Quadrature Amplitude Modulation (64-QAM). While the issue of data rate selection was touched in previous
wireless ad hoc network studies, the approach was generally
These modulation schemes work by varying periodic wave forms
oriented as follows: given a data rate choice, how does it affect
on the sub-carriers to carry information. Each of the schemes is
the optimization of the number of hops and transmission power
able to carry a varying number of bits per periodic wave form per
utilized towards achieving a certain level of end-to-end packet
sub-carrier. Only a fraction of the total modulated bits is used to
delivery rate [9].
carry the data. The rest are redundancies used to correct errors in
the reception process. This fraction is called “coding rate”. Therefore, a direct study on optimal data rate selection for vehicle
safety communications is generally not present in the literature.
Choosing a modulation scheme and a coding rate translates into a
We are only aware of the work by Xu et al, which briefly brought
data rate for the channel. The possible data rates for IEEE
up this subject while evaluating inter-vehicle communication
802.11p operating in a 10 MHz channel are listed in Table 1.
designs in [6].
Table 1: DSRC modulation, coding rate and data rate
That study compared overall system performances at different
data rates, based on IEEE 802.11a radio parameters. It was
channel (Mbps)

SINR threshold
Coded bits per

Coded bits per

OFDM symbol

OFDM symbol

Data rate for a

concluded that at 24 Mbps (i.e., 16-QAM with ½ coding rate), the

reception (dB)
periodic wave

10 MHz wide
Data bits per
Coding rate

system would experience the lowest probability of collision in the

Modulation

for frame

channel. However, we feel the question remains open for the

following reasons:
form

• Only one communications scenario, with low overall channel

BPSK 1 48 1/2 24 3 5 load, was evaluated.
BPSK 1 48 3/4 36 4.5 6 • The evaluation approach was rooted in control oriented
QPSK 2 96 1/2 48 6 8 perspective seeking near perfect communication reliabilities.

QPSK 2 96 3/4 72 9 11 • The simulator used was NS-2 version 2.26, which contained
simplifications and flaws in its IEEE 802.11 MAC and PHY
16-QAM 4 192 1/2 96 12 15 modeling [4].
16-QAM 4 192 3/4 144 18 20 • The RF signal propagation model used was deterministic
with no fading considered.
64-QAM 6 288 2/3 192 24 25
64-QAM 6 288 3/4 216 27 N/A 4. METHODOLOGY
Vehicle safety communications consist mostly of self contained
As shown in the table, transmissions at high data rates are more safety broadcast messages with no interactions and protocols at
efficient. However, they are more susceptible to reception errors the applications level. For this reason, overall system performance
due to interferences and noise. For this reason, a higher Signal to is closely related to individual broadcast messages reception
Interference/Noise Ratio (SINR) is required for a higher data rate. performance. The latter is often plotted as a “reception probability
This is indicated by the values in the last column of the table. vs. distance from the sender” curve.
These values have been obtained through empirical testing. We As discussed in the introduction, one of the key constraints in
configured the simulator used for this study accordingly. vehicle safety communications design is the interference level on
It has been generally assumed by the participants in the DSRC the channel. This interference is affected by the data rate selected
standardization process that the 6 Mbps data rate (instead of the as well as the corresponding transmission power that is
lowest 3 Mbps rate) would be the default for usage in vehicle necessarily different for each data rate choice.
safety communications. The historical origin1 of this assumption The methodology of this paper is therefore to compare
is surprisingly non-technical: the decision to use 10 MHz performances of broadcast transmissions using different data
rates, while ensuring that a comparable level of channel
load/interference level is maintained across various scenarios
1
This is based on the lead author’s direct knowledge of the DSRC constructed for comparison by adjusting the power used.
development process.

31
4.1 Communication Density Concept simulator, the expected number of carrier sensible event is 2.82
To follow the methodology described above, a fair and effective times the communication density computed in the manner above.
metric for measuring channel load is required. Such a metric is
proposed in [8] and is named “Communication Density.” 5. SIMULATIONS
The concept of communication density comes from a 5.1 Simulator Setup
mathematical model created as a probabilistic analysis tool for All simulations are run on NS-2 version 2.33, which incorporates
IEEE 802.11 DCF performance studies. The analysis shows that a a completely overhauled IEEE 802.11 simulation engine designed
radio’s broadcast performance depends only upon message size for correct and proper modeling of the MAC and PHY [4, 5].
and the number of carrier sensible events of such messages per Table 2 lists various IEEE 802.11p specific parameters settings in
unit of time at various points on the road, assuming a fading the simulator. Additionally, the SINR thresholds for frame
model similar to Rayleigh. reception are the ones listed in the table in Section 2. The
Rayleigh fading model is applied in all simulations.
This definition is independent on how a particular level of
communication density is created. In other words, a large number Table 2: Various parameter values used in the simulator
of vehicles transmitting messages at low power as illustrated in Code location of Parameter name Value
Scenario 1 in Figure 1 below could produce the same configuration
communication density as by a smaller number of vehicles
transmitting at a higher power as illustrated in Scenario 2. So long Mac/802_11Ext HeaderDuration_ 0.000040
if the communication densities are the same for the two scenarios,
Mac/802_11Ext SymbolDuration_ 0.000008
then the load on the channel is the same.
Mac/802_11Ext CWMin_ 15
Mac/802_11Ext CWMax_ 1023
Mac/802_11Ext SlotTime_ 0.000013
Scenario 1: High number of low powered transmissions
Mac/802_11Ext SIFS_ 0.000032
Phy/WirelessPhyExt freq_ 5.9e+9
Phy/WirelessPhyExt PreRXDuration_ 0.000040
Scenario 2: Low number of high powered transmissions
Phy/WirelessPhyExt noise_floor_ 1.26e-13
Phy/WirelessPhyExt CSThresh_ 2.512e-13

Scenario 3: Mixture of high and low powered transmissions

5.2 Environment Setup
Figure 1: Scenarios with equal communication density Mobile nodes are distributed pseudo-uniformly on a circle with a
2000 m circumference. Imagine that this is a road around the
More importantly, the communication density is additive in
equator of a greatly shrunken earth. From each node’s point of
nature. That is: communication densities produced by two groups
view, it is on a straight and flat road, and the RF signal
of vehicles can be simply summed up to give the total system
propagation calculation is based on the distance along the arc of
communication density. For example, Scenario 3 is just a
the circle. This setup eliminates boundary effects.
combination of the top half of the Scenario 1 and the bottom half
of Scenario 2. While it is a non-homogenous scenario and more As shown in Figure 2, three groups of nodes, namely reference
complicated, Scenario 3 produces a channel condition completely group 1, reference group 2 and study group, are intermixed on the
equivalent to Scenario 1 and 2. road. In all simulations, all nodes in the two reference groups are
configured to transmit messages at 6 Mbps data rate. The
A convenient way of computing communication density produced
transmission power are set to reach 200 and 400 m theoretic
by a group of vehicles is to calculate the product of the group’s
reception ranges at this data rate respectively for these two
vehicle density, messaging frequency, and transmission range.
groups.
This provides a channel load metric in the unit of transmissions
per unit of time. In other words, communication density could be
understood as: given a point on the road, the number of messages
transmitted per second for which this point is within the intended
range of reception. Figure 2: Three vehicle groups used in simulations
Such a unit is not the same as the number of carrier sensible The study group consists of nodes configured to transmit frames
events per unit of time but effectively serves the same purpose with various data rates and power settings in different simulations.
since they are generally proportional to each other. In fact, Their communication performances are the ones being compared
calculation shows that, with Rayleigh fading applied and to identify the optimal data rate in each vehicle safety
particular reception and carrier sense thresholds used in our communications scenario studied.

32
The reference groups are used to assure that adjusted data rate and Each scenario is in turn divided into three sub-scenarios. The first
power used by nodes in the study group combine to produce the simulation is configured with the study group transmitting at 6
same level of interference (i.e., communication density) in the Mbps data rate and using a power that corresponds to a 100, 300
channel and therefore can be fairly compared. This assurance is and 500 m range. For each sub-scenario, four additional
made by showing that communication performances of the simulations are then run in which nodes in the study group are
reference groups match each other across different simulations. configured with a different data rate. The transmission power used
by the study group is then manually adjusted in these four
5.3 Scenarios additional simulations so that reference groups’ performances still
Under the communication density concept, all vehicle safety match these produced in the first simulation of each sub-scenario.
communication scenarios can be generalized by two parameters: These four additional simulations study the performances of 3,
message size and total system communication density. For this 4.5, 9, and 12 Mbps data rates respectively. Data rates above 12
study, three values are picked for each parameter. Therefore, a Mbps are not presented in this paper because simulation results
total of nine scenarios are defined. clearly indicate that they would not matter in the conclusion.
The message size is the payload size of the frames transmitted To summarize, each sub-scenario consists of 5 simulations. Each
over the air in addition to the MAC and PHY overhead. The scenario contains 3 sub-scenarios, and there are 9 scenarios
values used for message size are 100, 200 and 500 Byte. It is not studied. So together there are 135 simulations run in total.
known how much security and other overhead is needed in
Table 3 shows the simulation matrix of scenarios of
vehicle safety communications, so these three values are picked to
communication density of 200. Scenarios of communication
cover the expected range. The 500 Byte value is purposely chosen
density of 400 and 800 are produced by doubling or quadrupling
as one clearly above the realistic level.
the messaging frequencies for all nodes.
The total system communication density level has values of 200,
400 to 800. We do not presume to know what eventual congestion 6. ANALYSIS
control target would be set to for vehicle safety communications.
Therefore these three widely ranged values are sampled. It should
6.1 Equivalent Interference Contribution
be pointed that 800 is also purposely chosen as one likely too high with Different Data Rates
to be realistic, and this point is further explained in section 6.1. The methodology of this work is based on the premise that it is
possible for transmission power to be adjusted so that the
In each scenario, each reference group is configured to contribute contribution to the communication density stays the same while
20% of the total system communication density. Thus, the study the data rate is changed. This premise is clearly validated in the
group is intentionally set to dominate the communications in the reference group performance results obtained through all the
channel. This is so that if the reference groups’ performances are simulations.
shown to match across different simulations, then the data rate
and power settings in the study group are contributing to the
channel load equally.
Table 3: All communication density 200 scenarios
Study Group
Reference

Reference
Group 1

Group 2

Communication
40 40 120
density produced
Share of total
communication 20% 20% 60%
density
Vehicle density Figure 3: Perfectly matched reference group performances
50 50 200
(cars/km)
Figure 3 shows one example: it contains the performance curves
Message frequency 6, 2, 1.2 of both reference groups across 5 sub-scenarios with the overall
2 1
(Hz) (one / scenario) system communication density being 200 and the message size
Transmission range being 100 Byte. The transmission power used for the study group
100, 300, 500 reaches to a range of 300 m when the 6 Mbps data rate is used.
based on 6 Mbps 200 400
(one / scenario)
data rate (m) The figure clearly shows two groups of mostly overlapping lines,
Data rate used in 3, 4.5, 6, 9, 12 corresponding to reference group 1 and 2 respectively. Within
6 6 each group, the lines overlap each other almost perfectly. This
TX (Mbps) (one / sub-scenario)
means the interference produced by the study group stays

33
completely the same even when their data rate and transmission
power are adjusted across different sub-scenarios.
Such near perfectly matched performances by the reference
groups are produced for most of the simulation scenarios. For
space reasons, they are not all shown in the paper.
At high communication density and message size cases, some of
the reference group performance curves do not completely
overlap any more. This is due to the over saturation of the channel
in these configurations. For example, a 500 Byte message takes
about 700 us to transmit at the 6 Mbps data rate. When the
communication density is 800, the number of carrier sensible
events experienced at any point on the road per second is 800 x
2.82 = 2256. Even if these events are perfectly lined up in time, it
would take more than 1.5 second. In short, the channel is loaded
in this case to the breaking point and accordingly it is not possible
to tune the transmission power of the study group with perfectly
matching results for the reference groups. Figure 5: Data rate comparison in medium stress scenario
In almost all cases, the 6 Mbps curve is on the top. Figure 5
illustrates the study group performances with the CD being 400,
message size being 200 Byte and study group transmission power
reaching 300 m with 6 Mbps. In this case, the solid black line,
representing the 6 Mbps data rate performance, is above all other
curves. In some cases, 4.5 and 9 Mbps data rates are relatively
close to the one of 6 Mbps, but the 6 Mbps is consistently the top
performer.
The 4.5 Mbps curve is shown to be better in the lowest stress case
of 100 Byte message size and 200 communication density, and
with shorter intended study group transmission range. This is
expected because when the channel is very empty, the impact of a
longer duration in interference is minimized. Correspondingly, the
9 Mbps curve becomes the top performer in the very high load
cases, which is explained with the reason applied in reverse. Both
are shown in the figures below.
Figure 4: Misaligned reference groups performances in a high
load scenario
Figure 4 shows the misalignment among reference group
performance curves. These results are produced with careful
manual adjustment of the power used in the study group. While
not perfect, they are still good enough to provide a reasonably
level field of comparison for different data rates.

6.2 Optimal Data Rate

As analyzed above, the reference group performance figures give
confidence that various data rate can be compared on a level
basis. The selection of the optimal data rates for vehicle safety
communications is now merely a matter of picking the top curve
in all the study group performance figures.
The study group performance figures for all 135 simulations are
presented in the appendix section at the end of this paper. The
figures are arranged into 3 tables with 9 figures in each table. Figure 6: 4.5 Mbps data rate wins in low channel load cases
Each table of figures present results from one of the
communication density level.

34
 In other words, we show that it is feasible to use a single value to
describe the channel load of very complex vehicle
communications scenarios, which consist of non-homogenous
configurations of vehicle density, message size, messaging
frequency, and transmission power.

8. ACKNOWLEDGMENTS
We would like to acknowledge Gaurav Bansal for his valuable
support to our vehicular communications research efforts in
general and this study in particular.

9. REFERENCES
[1] “IEEE Std. 802.11-2007, Part 11: Wireless LAN Medium
Access Control (MAC) and Physical Layer (PHY)
specifications,” IEEE Std. 802.11, 2007
[2] “IEEE P802.11p/D4.0, Draft Amendment for Wireless
Figure 7: 9 Mbps data rate wins in high channel load cases
Access in Vehicular Environments (WAVE),” March 2008
7. SUMMARY [3] D. Jiang and L. Delgrossi, “IEEE 802.11p: Towards an
In this paper, we identify the optimal data rate for various vehicle International Standard for Wireless Access in Vehicular
safety communications use-cases across a general spectrum of Environments,” in Proc. of IEEE Vehicular Technology
overall channel load. The commonly assumed default 6 Mbps Conference (VTC) Spring 2008
data rate turns out, interestingly, to be the best selection for
various intended ranges and safety message sizes in most cases [4] Q. Chen, F. Schmidt-Eisenlohr, D. Jiang, M. Torrent-
except when the channel is either lightly loaded or saturated. Moreno, L. Delgrossi, and H. Hartenstein, “Overhaul of
IEEE 802.11 modeling and simulation in NS-2,” in Proc. of
Even though this paper does not overturn the long held the 10th ACM/IEEE International Symposium on Modeling,
assumption on default DSRC data rate, it provides technical Analysis and Simulation of Wireless and Mobile Systems
insight to why the choice makes sense. More importantly, the (MSWiM), 2007
finding of this paper allows researchers to bypass the need of
evaluating different data rates in future studies of DSRC based [5] “Network Simulator ns-2,” http://www.isi.edu/nsnam/ns/
vehicular safety communications. As such, this work makes it [6] Q. Xu, T. Mak, J. Ko, and R. Sengupta, “Vehicle-to-Vehicle
possible to eliminate one dimension of complexity in such Safety Messaging in DSRC,” in Proc. of the First ACM
studies. International Workshop on Vehicular Ad Hoc Networks
Perhaps an even more important contribution from this paper is (VANET), 2004
the methodology introduced for such vehicular communications [7] D. Jiang, V. Taliwal, A. Meier, W. Holfelder, and R.
studies. Herrtwich, “Design of 5.9 GHz DSRC-based Vehicular
The communication density concept works effectively as a metric Safety Communication, ” IEEE Wireless Communications
for comparing interferences or measuring channel load. One magazine, October 2006, Vol. 13 No. 5
particular benefit is its ability to generalize the entire vehicle [8] D. Jiang, Q. Chen, and L. Delgrossi, “Communication
safety communications scenarios into two dimensions of Density: A Channel Load Metric for Vehicular
parameters, namely the message size and communication density. Communications Research,” in Proc. of Mobile Vehicular
Networks (MoVeNet) Workshop, 2007
Furthermore, the efforts in this work demonstrated that
communication density theory can be further refined to factor in [9] O. K. Tonguz and G. Ferrari, “Ad Hoc Wireless Networks: A
the message size as well. Message size matters in the form of Communication-Theoretic Perspective,” John Wiley & Sons,
carrier sensible event durations on the channel. We were Ltd., page 196-217, 2006.
successful in manually adjusting the power so that transmissions
[10] U.S. Federal Communications Commission, “FCC Report
of different frame durations (resulted from data rate changes)
and Order 03-324: Amendment of the Commission’s Rules
produce the same contribution to the communication density. This
Regarding Dedicated Short-Range Communication Services
demonstrates that a particular communication density value
in the 5.850-5.925 GHz Band,” December 17, 2003.
associated with a particular message size can be normalized and
mapped to such measures associated with a different message
size.

35
 Study Group Power range = 100m Study Group Power range = 300m Study Group Power range = 500m

36
 Study Group Power range = 100m Study Group Power range = 300m Study Group Power range = 500m

37
 Study Group Power range = 100m Study Group Power range = 300m Study Group Power range = 500m

38