Вы находитесь на странице: 1из 38

Reliable Congestion Control Mechanism for Safety Applications in Urban VANETs

Journal Pre-proof

Reliable Congestion Control Mechanism for Safety Applications in


Urban VANETs

Wenfeng Li, Wuli Song, Qiang Lu, Chao Yue

PII: S1570-8705(19)30723-1
DOI: https://doi.org/10.1016/j.adhoc.2019.102033
Reference: ADHOC 102033

To appear in: Ad Hoc Networks

Received date: 25 July 2019


Revised date: 7 October 2019
Accepted date: 28 October 2019

Please cite this article as: Wenfeng Li, Wuli Song, Qiang Lu, Chao Yue, Reliable Congestion
Control Mechanism for Safety Applications in Urban VANETs, Ad Hoc Networks (2019), doi:
https://doi.org/10.1016/j.adhoc.2019.102033

This is a PDF file of an article that has undergone enhancements after acceptance, such as the addition
of a cover page and metadata, and formatting for readability, but it is not yet the definitive version of
record. This version will undergo additional copyediting, typesetting and review before it is published
in its final form, but we are providing this version to give early visibility of the article. Please note that,
during the production process, errors may be discovered which could affect the content, and all legal
disclaimers that apply to the journal pertain.

© 2019 Published by Elsevier B.V.


Reliable Congestion Control Mechanism for Safety
Applications in Urban VANETs

Wenfeng Lia,∗, Wuli Songa , Qiang Lua , Chao Yuea


a College of Medical Information Engineering, Shandong First Medical University &
Shandong Academy of Medical Sciences, Taian, China.

Abstract

In vehicular ad hoc networks (VANETs), cooperation between vehicles is


needed for proper performance of safety applications. Beacons are periodically
broadcasted by vehicles to inform their neighbors of their information, such
that periodic beacons are one of the building blocks that enable the operation
of safety applications. Safety applications have their own strict quality of service
(QoS) requirements, and they require reliable and timely data communication
within networks. When many vehicles are trying to broadcast within a con-
fined area at the same time, frequent channel contention and congestion may
occur. Therefore, channel congestion control mechanisms are key for reliable
and efficient operation of safety applications.
In this paper, an adaptive beacon generation rate (ABGR) congestion control
mechanism is proposed to reduce channel congestion and contention. ABGR
can dynamically adjust the beacon generation rate according to varying levels
of vehicles’ density to ensure reliable and timely delivery of beacons. Inspired by
the correlation between traffic density and vehicle speed, a dynamic application-
level reliability assessment scheme (T-Pro) is proposed to evaluate the reliability
of various safety applications at varying density. Finally, the application-level
reliability of three safety applications is evaluated using the ABGR mechanism.

∗ Corresponding author. Address:College of Medical Information Engineering, Shandong

First Medical University & Shandong Academy of Medical Sciences, No. 619 Greet Wall Road,
Taian, 271016, China. Tel: (+86)13853803177
Email addresses: 13853803177@163.com (Wenfeng Li), songswl@163.com (Wuli Song),
luqiang271016@163.com (Qiang Lu), yuechao_71@163.com@163.com (Chao Yue)

Preprint submitted to Journal of LATEX Templates November 8, 2019


Numerical results show that the performance of ABGR is far higher than that of
the current DSRC. The application-level reliability of three safety applications
exceeds 99%, even if density reaches 450 vehicles/(km ∗ 3lanes).
Keywords:
safety application reliability, VANET, congestion control mechanism, quality
of service(QoS)

1. Introduction

Intelligent Transportation Systems (ITS), i.e., the integration of telecommu-


nication and information technologies into transportation systems, were devel-
oped to address critical issues such as passenger safety and traffic congestion[1].
5 The vehicular ad hoc network (VANET) forms the cornerstone of modern ITS.
It consists of mobile vehicles connected by wireless communication in an ad
hoc manner without central control while moving along roads. Dedicated Short
Range Communication (DSRC) is considered the most promising wireless ac-
cess technology to support vehicle-to-vehicle (V2V) and vehicle-to-infrastructure
10 (V2I) information exchange[2]. Vehicles generally disseminate two types of in-
formation related to event-driven messages and beacons (also called BSM Basic
Safety Messages in the US or CAM Cooperative Awareness Messages in Europe).
Beacons are periodically broadcast by vehicles to inform their neighbors of infor-
mation such as position, direction and speed. On the other hand, event-driven
15 messages are multi-hop distributed in a geographic area with higher priority
in the case of emergencies, such as car collisions, accidents and road surface
collapse[3]. These messages assist vehicles in expanding their awareness range
beyond the line of sight as well as avoiding potential dangers.
Depending on reliable and timely beacon exchange via V2V or V2I communi-
20 cation, many safety applications such as car collision avoidance and slow vehicle
indication have been designed for VANET, revolutionizing the quality of expe-
rience for drivers and passengers. Research[4] shows that 60% of car collisions
can be avoided if safety applications were provide a warning to drivers on time

2
before car collisions. But safety applications timely warning relies on reliable
25 and timely disseminating of beacons between vehicles. Therefore, disseminating
beacons reliably and in a timely manner is crucial for proper operation of safety
applications.
However, reliably disseminating beacons to all surrounding vehicles in a
timely manner in VANET remains a challenge[5]. First, IEEE 802.11p pro-
30 tocol, which is based on Carrier Sense Multiple Access with Collision Avoid-
ance (CSMA/CA), is employed as the PHY and MAC layer protocol of DSRC.
The broadcast procedure of the IEEE 802.11p MAC protocol follows the basic
medium access protocol of the distributed coordination function (DCF) without
three functions (acknowledgment, request to send (RTS)/clear to send (CTS),
35 and retransmission). All vehicles employ the shared channel to broadcast their
beacons through a contention manner referred to as the back-off mechanism.
The broadcasting mechanism in dense networks may lead to frequent contention
and packet collisions in transmission among neighboring vehicles[6].
In addition, the DSRC standard provides a control channel (CCH) as a
40 shared channel to transmit beacons. Its limited bandwidth (10 M Hz) brings
limited transmit capacity. In dense traffic conditions as the V2V deployment
scales up, the resultant channel load on CCH increases and leads to channel
congestion[7]. Meanwhile, dense vehicles will inevitably increase the number of
hidden terminals, which aggravates channel congestion and contention. Such
45 congestion might have devastating consequences for network performance; in
particular, data transmission reliability will become poorer in congested net-
works. Safety application reliability may be degraded, possibly endangering
the safety of road users. Therefore, reducing channel congestion and providing
access to the harmonized and fairness channel among vehicles is necessary.
50 Traditionally, reliability in the context of VANET broadcast services is de-
fined as the ability for all intended mobile nodes in an area of interest of a
network to receive broadcast messages[4][5]. However, this reliability does not
reflect the reliability of the safety application due to the memory-less property of
the safety application[8]. In addition, the quality of service (QoS) requirements

3
55 for each safety application are different. Although some QoS metric is adopted
in [9][10][11] to evaluate the reliability of safety applications, it is difficult to
evaluate whether all safety applications’ QoS requirements are satisfied.
To address the aforementioned channel congestion issues, an adaptive beacon
generating rate (ABGR) congestion control mechanism based on vehicle density
60 is proposed in this paper. The main idea of the ABGR mechanism is that the
beacon generation rate of the vehicle is adaptively adjusted according to the
vehicle density and congestion threshold to reduce the probability of packet
collision, thereby effectively controlling channel congestion and improving the
reliability of the VANET safety application. The congestion threshold is derived
65 from the analysis result of the effect of vehicle density on QoS requirements for
the safety application. Then, the QoS requirements for the three typical safety
applications that are believed to have the most stringent QoS requirements are
specified and discussed. Based on the correlation between vehicle density and
speed in traffic theory, a dynamic application-level reliability assessment scheme
70 (T-Pro) is proposed. Metrics for QoS requirements of safety applications need to
change not only for different safety applications but also for the same application
as the density changes. The T-Pro scheme will dynamically adjust parameters in
metric with changing density. The main contributions of this paper are detailed
below:

75 • An adaptive beacon generation rate (ABGR) congestion control mech-


anism based on vehicle density is proposed in this paper. The ABGR
mechanism can automatically adjust the vehicle beacon generation rate
according to the change in vehicle density. It can effectively reduce chan-
nel congestion and packet collisions in dense networks, improving system
80 reliability.

• Based on the correlation between road traffic and vehicle speed, a dy-
namic application-level reliability assessment scheme(T-Pro) is proposed
to evaluate application-level reliability of various safety applications with
different traffic densities.

4
85 The rest of this paper is organized as follows. Section II presents related
work within the field. Section III illustrates the features of the traffic scenario,
and an application-level evaluation model is built up in this section. Section
IV introduces the proposed ABGR channel congestion control mechanism and
T-Pro application-level reliability assessment scheme. The numerical results
90 and corresponding performance evaluations are discussed in section V. Finally,
section VI provides the conclusion of this paper.

2. Related Work

The development of an optimal congestion control scheme faces many chal-


lenges. Congestion control mechanisms for the VANET have been studied in
95 many studies. They usually dynamically adjust one or more transmission pa-
rameters, i.e., beacon generating rate, transmission power, and parameters in
the CSMA/CA protocol.
Tonguz et al.[12] proposed an adaptive protocol, called Distributed Vehicular
Broadcast (DV-CAST) , in which a neighbor detection mechanism estimates the
100 local topology using a periodic beacon. DV-CAST is able to disseminate beacons
in both dense and sparse networks using different methods. It suffers from
the use of static intervals for sending beacons, regardless of vehicles’ density.
Sommer et al. [13] proposed the Adaptive Traffic Beaconing (ATB) protocol;
the beacon interval is adapted dynamically according to the channel quality
105 and message priority. The basic idea is to implement a network load control
to address scalability issues of data dissemination in VANETs. Nevertheless,
this protocol did not have mechanisms that address the problem of prioritizing
messages when vehicles try to transmit at the same time. In[14] , the authors
proposed the distributed fair power adjustment (D-FPAV) protocol. D-FPAV
110 controls the transmission power of vehicles by keeping the beacon traffic under
a maximum beacon load (MBL). The transmission range of nodes is reduced to
avoid congestion.
The authors in [15] studied the trade-off between broadcasting efficiency and

5
reliability, and obtained the optimal packet transmission probability correspond-
115 ing to the vehicle density. Then, congestion control approach was proposed to
achieve the maximum broadcasting efficiency. However, the investigated com-
munication scenario, where all vehicles are randomly positioned in a single lane
modeled as a one-dimensional (1-D) spatial network, is not a general case. Gani
et al.[16] proposed different channel congestion control algorithms based on
120 adaptation of beacon length, transmission rate, and a combination of both of
these two parameters. They determined that a joint rate-length congestion con-
trol algorithm offered more flexibility than algorithms with separate parameters,
by examining the trade-off between beacon length and transmission rate. But,
the length of the beacon is limited. Too long or too short beacons will have a
125 negative impact on the system.
Bai et al.[17] introduced a schedule scheme referred to as Context Awareness
Beacon Scheduling (CABS) to address the congestion issue in VANETs. CABS
re-divides the time slots and dynamically schedules the beacons on each time
slot, which is similar to TDMA transmission. Channel congestion is addressed
130 through tuning the beacon frequency in this scheme. In [18], the authors intro-
duced a novel multichannel TDMA-based MAC protocol specifically proposed
for a VANET scenario. This protocol reduces transmission collisions on the CCH
by assigning disjoint sets of time slots to vehicles that are moving in opposite
directions. However, in the protocol, each node must transmit a beacon during
135 its time slot only even if the node has no data to include in the high-priority
safety applications field.
The authors in[19] proposed a new control protocol that integrates conges-
tion and awareness control processes. This protocol dynamically adapts the
transmission frequency and power of beacons of each vehicle to guarantee that
140 its application requirements are satisfied while controlling the generated channel
load. Zemouri et al. [20] presented P&A-A, a new congestion control protocol
for channel congestion where beacon transmit rate and power are adapted suc-
cessively, relying on an altruistic short-term prediction algorithm that estimates
the vehicular density around a given vehicle within the next short while. P&A-

6
145 A guarantees strict beacon requirements and satisfies the level of awareness
required for the operation of most critical VANET applications.
To make an accurate assessment of the reliability of VANET and safety ap-
plications, the appropriate assessment metrics are indispensable. The package
delivery ratio (PDR) and transmission delay were adopted as reliability met-
150 rics in the early research work. Yousefi et al. [21] introduced two new metrics,
called beacon rate and effective range, to evaluate the performance and relia-
bility of single-hop beacon broadcasting in VANET. Ma et al. [5] summarized
four reliability metrics including package acceptance rate (PRR), packet deliv-
ery ratio (PDR), packet delivery probability (PDP) and effective range(ER).
155 Nevertheless, Bai et al.[8] noted that the communication reliability (MAC-level
reliability) is different from the reliability of safety applications (application-
level reliability). Thus, the reliability metrics aforementioned are unsuitable for
evaluating application-level reliability. They proposed a new metric referred to
as T-window reliability in their work. Afterwards, An et al.[9] defined aware-
160 ness probability as an application-level reliability metric. Awareness probability
with different parameters can be used to evaluate reliability of various safety
applications, but there is no explicit or uniform specification for setting the
appropriate parameters for each safety application.

3. Problem Statement

165 3.1. Scenario Analysis

The performance and reliability of real-world radio networks are influenced


by many factors such as concurrent transmission collision, interference from
hidden terminals, channel fading and pass loss. To generate a simplified yet
reasonable analytical model, several scenario assumptions are made as follows.
170 Various types of roads, such as highways, expressways and elevated roads,
are used in urban and rural environments to create a public road network. A
typical multi-lane straight highway scenario, which is taken from the work in
[22], is described in Fig. 1(a). It is believed that this scenario is one of the most

7
(a) A Typical Highway Scenario Tagged
Vehicle

R R R R

Potential Hidden Potential Hidden


Transmission Area of Tagged Vehicle
Terminal Area Terminal Area

Tagged Vehicle Vehicle in transmission Vehicle in hidden terminal


area of tagged vehicle area of tagged vehicle
Ttransmission area of Hidden terminal area of
tagged vehicle tagged vehicle
(b) An abstraction of the highway scenario

Figure 1: Road topology abstraction: 2-D to 1-D

demanding scenarios for the MAC method[23]. Each vehicle traveling along
175 the highway is equipped with DSRC, GPS devices and other necessary sensors.
Since the distance between parallel lanes can be ignored given the long length
of the road, the scenario in Fig.1(a) is abstracted into a one-dimensional vehicle
network as in Fig. 1(b). This abstraction is affordable when the network size is
very large and mobile nodes are placed with a certain finite network density[24].
180 A collection of mobile nodes representing vehicles is randomly placed on the
line that is assumed to follow a Poisson point process with vehicle density β
(in vehicles per meter). This assumption is a good approximation for networks
where transmitters and/or receivers are located in or move around randomly
over a large area[5][25]. Thus, the probability P (i; l) of finding i vehicles in
185 length of l is given by
i
(βl) e−βl
P (i, l) = (1)
i!

8
All vehicles on the line have the identical transmission/receiving range and
carrier sensing range, which is denoted as R. Thus, the average number of nodes
on the line within the transmission range of a tagged node (the node located on
origin O) is 2βR. The potential hidden terminal area of the tagged vehicle in
190 broadcast communication drops in the range of the blue area.
Since the Poisson arrival process is a suitable approximation of message
arrivals in packet-data networks[26], the packet arrivals at each node follow
a Poisson process with a rate λ (in packets per second). Vehicle mobility is
not considered in this paper because it has been proved and verified in [27]
195 that high node mobility has a very minor impact on the reliability of VANET
communication. To analyze the impact of DSRC channel fading in highway
scenarios, the Nakagami-m model[28] is employed in this paper.

3.2. Reliability Evolution Model

In this part, we briefly review the reliability analytical model to evaluate


200 the reliability of safety applications in VANET on the basis of our previous
work[22][29]. Broadcast procedure of IEEE 802.11p MAC follows the basic
medium access protocol of DCF, which is a random access scheme based on
CSMA/CA protocol. In fact, the IEEE 802.11p channel access procedure of
a vehicle in broadcast mode is a type of competitive progress called backoff
205 progress[29]. Since there is no retransmission attempt in DCF for broadcast
service, the size of the backoff window maintains an initial backoff window size
of W0 . Therefore, the backoff progress in each vehicle can be characterized by
a one-dimensional discrete time Markov chain(DTMC), as shown in Fig.2. The
state of DTMC is represented by k, where k denotes the value of the backoff
210 counter and takes values (0, 1, . . ., W0 − 1).
Let τ be the probability that a vehicle transmits a packet in the backoff
process. Since any transmission occurs only when the backoff counter value is
equal to zero, it is obtained from solutions to the one-dimensional DTMC as:

2p0
τ= (2)
W0 + 1

9
1/W0

1 1 1 ... 1 1
0 1 2 W0-2 W0-1

Figure 2: Markov chain model for backoff progress in broadcast

where p0 is the probability that there is a packet ready to transmit at the MAC
215 layer in each vehicle.
Now, we take tagged vehicle O in Fig.1 as an example. Denote pb as the
probability that the channel is sensed busy by vehicle O. Therefore, the proba-
bility pb is calculated as[22]


X (2βR)i −2βR
pb = 1 − (1 − τ )i e = 1 − e−2βRτ (3)
i=0
i!

Obviously, the solution pb depends on p0 , which can be derived by using an


220 iterative algorithm as Table I. Here, Q(z) is a probability generating function
(PGF), as detailed in [22][29].

Table 1: Algorithm for calculating p0

1) Let p0 = 0.
2) Substituting p0 = 0 into Eq.(2) and Eq.(3) to yield pb .
3) Calculate Q(z).
4) Let service rate µ = 1/Q0 (1);
5) If λ < µ, p0 = 1 − λ/µ; otherwise, p0 = 0.
6) If p0 converges with the previous value, then algorithm is stopped;
otherwise, go to step 2.

In the scenario shown in Fig. 1, nodes U and V are the receiving vehicles
within transmission range of node O, and nodes A and B are placed on the
hidden terminal area of node O. When vehicles within transmission range of
225 node O are receiving packets sent by O, transmission may be disturbed by the
hidden terminal, concurrent transmission collisions and channel fading. We first
consider the probability of vehicle U successfully receiving packets from vehicle

10
O. Assuming that the distance between node U and O is x(0 < x < R).

1. The interference of the hidden terminal to the transmission of vehicle O


230 is divided into two parts: first, the probability that no hidden terminal
vehicle is in the transmitting state can be calculated as:

PH1 (x) = 1 − xβλT (1 − pb /2) (4)

Second, the probability that none of hidden terminal vehicles start trans-
mitting during the whole time period of vehicle O transmitting packet can
be calculated as:

0
(βx) −xβλTp
PH2 (x) = e = e−xβλTp (5)
0!

235 where Tp is the average time to transmit the full packet.

2. When node O broadcasts a packet in a slot, a concurrent transmission


collision will occur if any node within range[−(R − x), R ](e.g. node V )
transmits in the same slot. The mean number of nodes transmitting in
area [−(R − x), R] can be calculated as[29]:

nP = βRτ + βτ (R − x) − β 2 τ λT (R − x)2 (1 − pb /2)/2 (6)

240 Therefore, the probability that no concurrent packet collision occurs in


area [−(R − x), R] can be written as[29]:

0
(nP ) −nP P
PC (x) = e = e−n (7)
0!

3. Considering a N akagami − m distribution as a model to analyze the im-


pact of channel fading and path loss, the expected probability of node U
successfully receiving a packet from node O is written as[29]:

m Z 1/R2
(x2 m) 2
PF (x) = 1 − z m−1 e−x mz
dz (8)
Γ(m) 0

11
245 where m is a function of distance x:


 3, x < 50m


m(x) = 1.5 50m ≤ x < 150m (9)



 1 x ≥ 150m

From the above, the probability that node U receives the broadcast message
from tagged node O is

Ps (x) = PH1 (x)PH2 (x)PC (x)PF (x) (10)

3.3. Reliability Performance Metrics

1. Packet Reception Ratio (PRR): PRR is calculated as a percentage of nodes


250 that successfully receive a packet from the sender among the total number
of nodes that are within specified range of the sender. Based on the node
reception probability Ps (x), PRR can be written as:
Rx Z x
0
βPs (t)dt 1
P RR(x) = = Ps (t)dt; x≤R (11)
βx x 0

2. Awareness Probability(AP): AP is defined as the probability of a vehicle


successfully receiving at least k packets from the sender in tolerance time
255 window Ta , which is written as:
 
bTa /tc
X k
PA (x, n, Ta ) =   Ps (x)k (1 − Ps (x))bTa /tc−k (12)
k=n n

where n and k are the number of packets from the sender and the number
received successfully by the receiver in tolerance time Ta , respectively.
x is the distance between the sender and receiver, and t is the packet
generation interval.

12
260 4. Proposed Congestion Control Approach

In this section, an adaptive beacon generation rate (ABGR) channel con-


gestion control mechanism is first introduced. Based on the vehicle density,
the ABGR mechanism adjusts the beacon generation rate to reduce the inter-
ference of channel conflicts and improve system reliability. Then, a dynamic
265 reliability evaluation approach(T-Pro) is proposed based on re-consideration of
the safety application’s operating characteristics and the relationship between
vehicle speed and density.

4.1. ABGR channel congestion control mechanism

The operation of safety applications relies on each vehicle periodically broad-


270 casting its safety information to other vehicles in the context of VANET. These
pieces of safety information including beacons and emergency messages are
transmitted through shared channels of DSRC. Therefore, only through con-
tention can each node access the shared channel to broadcast beacons. But this
contention-based communication mechanism has limitations: aggravated chan-
275 nel collisions and increased failure probability of consecutive transmission as a
result of contention. In addition, each node sends beacons periodically, and the
DSRC protocol adopts a fixed-size contention window. So, channel contention
will increase, and the shared channel is easily congested when a large number
of vehicles send messages at the same time. Thereby, the reliability of data
280 transmission will become poorer in DSRC communication. To guarantee oper-
ation reliability of the DSRC system, the effective channel congestion control
mechanism should be employed.
The main idea of the channel congestion control mechanism is to reduce the
load on shared communication channels and to coordinate fair channel access
285 among vehicles in the context of VANET. According to the measured real-time
channel congestion condition, the congestion control algorithms usually adopt
a method of dynamically adjusting one or more transmission parameters such
as beacon generation rate, data transmission rate, power and range.

13
Table 2: Beacon generation rate control algorithm

Initialization :
Given the maximum beacon generation rate λmax , minimum beacon
generation rate λmin , density impact factor δ, density threshold βthreshold .
The rth time period:
1) Calculate the current vehicles’ density β according to beacon
received in this time period;
2) If β ≤ βthreshold , then current beacon generation rate λ = λmax ;
otherwise λ = λmax − b(β − βthreshold )/δc;
3) If λ < λmin , then λ = λmin ;
4) If λ > λmax , then λ = λmax ;
5) Output λ as the current beacon generation rate in next time
period. Goto next time period

It is noted that a higher beacon generation rate is crucial to improving relia-


290 bility of communication in VANET. However, the higher beacon generation rate
is maintained when vehicle density increases and the number of beacons to be
transmitted per unit time increases dramatically, which can easily lead to chan-
nel congestion. At the same time, the fixed size of the contention window also
aggravates channel competition and increases channel collision. It will greatly
295 reduce the probability of successful beacon transmission and decrease the reli-
ability of DSRC communication. Undoubtedly, reducing the beacon generating
rate is the most direct way to mitigate the intense channel contention and colli-
sions among vehicles when the network becomes congested[30]. Therefore, this
approach is employed in the proposed ABGR algorithm; the detailed steps are
300 shown in Table 2.
ABGR algorithm can adjust the beacon generation rate dynamically ac-
cording to different levels of network density. First, the vehicle identification
information or DSRC device address information is encapsulated in beacons.
Each of the surrounding vehicles can be identified by a vehicle after it receives
305 beacons from surrounding vehicles. In this way, the onboard density calculation
module can count the number of surrounding vehicles by identifying the beacon
that is received during the unit time period. Then, the current vehicles’ density
can be obtained by a vehicle. Second, the current vehicles’ density is compared

14
with the density threshold. If the current vehicles’ density is lower than the
310 density threshold, the beacon generation rate is set to a pre-defined maximum
generation rate; otherwise, the beacon generation rate will automatically reduce
(increase) when the current vehicle’s density increases (decreases). It is noted
that the beacon generation rate will not decrease when it is reduced to a cer-
tain value. At this moment, the density of vehicles has become so large that
315 road is congested, and vehicles can only crawl forward ro stop. In such road
environment, the minimum beacon generation rate is sufficient for communica-
tion among vehicles. Meanwhile, too low beacon generation rate will affect the
reliability of vehicular safety applications.

4.2. Traffic Flow Model of Speed-Density Relationship

320 In traffic flow theory, the characteristics of uninterrupted traffic flow can
be described by three parameters: vehicle density K, speed V and traffic flow
q. Under uninterrupted flow conditions, three parameters are all related to
each other and can be express as: q = KV . In the study of traffic flow, the
speed-density relationship is the basic relationship among three parameters,
325 from which the velocity speed-flow and flow-density relationships can be derived.
The speed-density relationship is not difficult to observe in the real world. Var-
ious forms of speed-density models have been proposed, such as Greenshields
model, Greenberg model, and Multi-segment model[31]. Artimy et al. [32] have
employed speed-density models in the research on VANET. Similarly, a Multi-
330 segment speed-density model[31] is adopted as follows.
Empirical evidence and traffic simulations show that vehicles can travel as
fast as they want (free-flow speed) when density is lower than a certain density
(free-flow density). In this phase, traffic flow is “sparse” and vehicles maintain
free-flow speed without inference from other vehicles. Thus, there is no rela-
335 tionship between speed and density; at a density above free-flow density, traffic
flow becomes “dense” and drivers begin to decrease their speed gradually with
increasing density; then, the relationship between speed and density emerges.
When the density increases to a certain density (critical density), the value of

15
traffic flow, which is the number of vehicles passing the same observation point
340 in a unit time, reaches its peak at this moment. Afterward, traffic flow is re-
duced gradually once the critical density is exceeded. If density increases more,
traffic would reach the jam state, and in the worst case, all vehicles would have
to stop completely [33]. In traffic theory, each phase is dealt with in a different
manner, while the relationship between speed and density can be expressed by:


 K
mVf + (1 − m)Vf Kf Kf ≤ K < K1


V = q0 /K K = K1 (13)



 mV (1 − K )
f Kj K1 < K ≤ Kj

345 where V is the instantaneous speed, Vf is the free-flow speed, K is the in-
stantaneous density,Kf is free-flow density, q0 is the maximum traffic flow,
m(0 ≤ m ≤ 1) is a constant, Kj should be the maximum number of vehi-
cles parked on the road per unit length, which is called as theoretical maximum
density. But in practice, drivers tends to stop driving when the distance be-
350 tween the vehicles reduces to a certain distance to avoid collisions. This leads to
the difference between the theoretical maximum density and the actual block-
ing density. Usually, the theoretical maximum density is replaced by the actual
blocking density as Kj in Eq.(13) , which is equal to 2/3 of the theoretical
maximum density[31]. K1 denotes critical density, which happens to be the
355 maximum value point of traffic flow on the lower curve, and K1 = Kj /2[34].
Since our research is more concerned about the scenario with larger density, the
numerical interval (K1 < K ≤ Kj ) is considered in this paper.

4.3. Dynamic Reliability Evaluation Scheme (T-Pro)

The aforementioned reliability metric AP, which is a function of distance and


360 time, is equivalent to the probability of successfully receiving at least n packets
in the tolerance time window Ta within a specific distance. Specifically, for each
given time t0 , if enough packets are received during time interval [t0 − Ta , t0 ],
the receiver can reliably predict the sender movement based on these received
packets. On the one hand, tolerance time window Ta , which is the key parameter

16
Dinit
L Dsafe

A B

Figure 3: The scenario for calculating beacon reception time period

365 of the definition for AP, is directly related to the requirements of the safety
applications and should be determined based on speed, traffic density, and driver
reaction time. On the other hand, although the headway will be shortened due
to increased density, speed decreases and braking distance becomes shorter,
which results in more time for drivers to avoid collisions.
370 It is assumed that vehicle A following vehicle B drives at the same speed v
in the same lane as shown in Fig. 3. To avoid a rear-end collision, the distance
between the two vehicles must be greater than a specified distance, referred to
as the minimum safety distance Dsaf e . Once the headway between A and B is
less than the minimum safety distance, A does not have enough time to avoid
375 a collision with B even though the safety application generates a warning. The
minimum safety distance Dsaf e can be calculated as follows.
If B initiates an emergency brake with acceleration a and broadcasts a bea-
con to inform vehicle A, the braking distance of B is

DB = v 2 /2a (14)

v represents the vehicle speed in m/s; a is their acceleration in m/s2 . Vehicle A


380 also initiates an emergency brake with the same acceleration a after it receives a
beacon from B; then, the total distance of A traveled from the receiving beacon
to the point of stopping is

DA = v ∗ treact + v 2 /2a (15)

17
where treact denotes the reaction time of the driver. Obviously, a collision can be
avoided as long as the driving distance of A is less than the sum of the braking
385 distance of B and the minimum safety distance. Thus, we have:

DA + L ≤ DB + Dsaf e (16)

Substitute Eq.(14) and (15) into Eq.(16), the minimum safety distance is:

Dsaf e = v ∗ treact + L (17)

where L is the mean length of vehicles. Drivers usually maintain a headway,


which is greater than the minimum safety distance. The safety application in
vehicle A can reliably provide a warning as long as A receives a sufficient number
390 of new beacon from B during the time period before A enters the minimum
safety distance of B. This time period can be expressed by:

Dinit − Dsaf e Dinit − L


tc = = − treact (18)
v v

where Dinit denotes the headway between two vehicles. It is determined by


density which is closely related to the speed. Thus, it can be derived from Eq.
(13) as:
1 1
Dinit = = v (19)
K (1 − mV f
)Kj

395 Assuming the free-flow speed Vf is 120km/h, the average length of vehicles
is 5m, so the theoretical maximum density is 200 vehicles/km/lane, and Kj in
Eq.(19) is equal to 2/3 of theoretical maximum density. Then, beacon reception
time period tc at larger density can be calculated by Eq. (18) as depicted
in Fig. 4. The X-axis shows the different speed below 30km/h. It can be
400 observed that the time period tc is rapidly increasing with decreasing density.
The reason for this phenomenon is that there is a linear relationship between
density and speed in the context of larger density. The speed reduction is due to
the increasing density, which will inevitably lead to shortened headway between

18
5 
With react time
4.5 Without react time
4

3.5

3
WF (s)

2.5

1.5

0.5

0 
0 5 10 15 20 25 30
Speed(km/h)

Figure 4: The beacon reception time period with different density

vehicles. However, the length of the braking distance is proportional to the


405 square of the speed, and the proportion of the braking distance to headway will
become smaller with increasing density. Thus, vehicle B has a longer time to
receive beacons from A.
It is obvious that the change in density and speed has a significant impact
on time from the example above, whereas the current reliability related QoS
410 requirements for the safety application are set without considering it; e.g., the
tolerance time window Ta in AP is set to a fixed value. Therefore, taking
speed, density and time into account, a dynamic reliability evaluation scheme
(T-pro) is proposed to estimate the Application-level reliability of the safety
applications. The main idea of T-pro scheme is that the tolerance time window
415 Ta in AP can be prolonged appropriately with the change in density and speed
in the context of larger density. In this way, the number of beacons transmitted
within the time window can still be maintained even if the beacon generation
rate is reduced, so the QoS reliability requirements of the safety application in
the context of larger density can be satisfied.

19
420 5. NUMERICAL RESULTS

5.1. Simulation Description

In this section, we set the most stringent test conditions for simulations,
including the highest density of vehicles, farthest intended distance of region
of interest (ROI) for the applications, and the maximum length of a beacon.
425 The highway scenario is set as Fig.1 with three lanes in each direction. Each
vehicle in the network is equipped with DSRC capability with the following com-
munication parameters as listed in Table 3. Communication range (transmis-
sion/carrier sensing) is R = 500m. Beacons are generated with maximum rate
λmax = 10packets/s and the maximum length of beacon P Lmax =400bytes). In
430 our simulation, We conduct simulations using NS2 simulation tool and MAT-
LAB on a straight road with length of 5000m. The simulation covers the main
physical and MAC behaviors of IEEE 802.11 broadcast communication. Each
simulation round lasts 30 seconds.
Table 3: Parameters for Communication in DSRC

Parameter Value
Signal Bandwidth 10M Hz
Channel Data Rate 27M bps
Slot Time σ 16µs
DIFS 64 µs
SIFS 32 µs
Propagation delay δ 1µs
OFDM Symbol Duration 8µs
Size of CW W0 15
Transmission Power 20 dbm
Transmission/Receive Range 500m
Carrier Sensing Range 500m

To evaluate whether the DSRC protocol and ABGR mechanism can meet the
435 QoS requirements of safety applications in VANET, three safety applications,
cooperative collision warning (CCW), slow vehicle indication (SVI), and rear-
end chain collision warning (RCW), are considered in this paper. According to

20
the range of region of interest (ROI), all safety applications can be classified into
large range, medium range and short range[35]. CCW, SVI and RCW belong to
440 three categories of application. Meanwhile, they are believed to have the most
stringent QoS requirements among all safety applications.
For DSRC-based safety applications, they are counted as being reliable as
long as at least one or several packets are successfully received from a neighbor
vehicle within a given duration time; the receiver vehicle should be able to
445 predict and update the neighbor vehicle information accurately enough for safety
application processing. The metric AP can be applied to evaluate whether the
current DSRC can meet the reliability requirements of safety applications by
setting its range and probability, as well as on parameters T and n[9][29].
CCW enables vehicles to predict and send notifications regarding the poten-
450 tial risk of collisions via information exchange and warns driver to take correct
action in a timely way to avoid risk. The ROI of CCW has a large range, and
the number of required packets n = 1 in a tolerance time window Ta =1s at
the edge of ROI is enough, as one beacon provides full information regarding
the accident ahead. Thus, the QoS requirements of CCW might be satisfied
455 when the probability that a vehicle within 400 meters of the sender successfully
receives at least one packet in the tolerance time window is larger than 99.0%,
i.e., PA (400, 1, 1) ≥ 0.99. AP is greater than or equal to 99% is reasonably high
to guarantee reliability[9].
SVI enables vehicles to provide alerts to drivers about the potential hazard
460 if a slow vehicle ahead is detected based on the beacon received. It requires the
vehicle to receive more beacons from a slow vehicle in Ta to predict the move-
ment of the slow vehicle. Then, within ROI 100 meters, the AP requirement
for SVI is that a vehicle successfully receives at least 3 packets in the tolerance
time window Ta =1s with a probability larger than 99.5%.
465 RCW can advise the driver of an approaching vehicle to make a proper
maneuver to avoid the crash when a possible rear-end collision is detected. The
ROI of the RCW is the shortest since the distance between the two vehicles is
already very close when a possible risk of rear-end collision in front is detected.

21
As many beacons as possible that are received in Ta are needed to maintain an
470 up-to-date picture of the neighborhood. Thus, the AP requirement for SVI is
that a vehicle within 400 meters of the sender successfully receives at least 5
packets in Ta =1s is larger than 99.9%.

5.2. Reliability Analysis of Current DSRC

In this section, the reliability of the current DSRC will be discussed. The
475 reliability evaluation metric PRR from Eq. 11 has been depicted, and the
impact of different parameters in Eq. 11 on the PRR has been revealed in the
previous research [29]. Among these parameters, vehicle density degrades PRR
significantly. The increase in density aggravates channel competition, which
leads to a decrease in reliability PRR. Similarly, the distance between sender and
480 receiver also has a greater impact on PRR. In addition, lower beacon generation
rates can effectively improve PRR with larger density. Other parameters, larger
content window sizes, higher data transmission rates and shorter beacon lengths
will improve PRR to a minor extent. Research [29] shows that the PRRs are
too low to meet reliability requirements for most VANET safety applications
485 from a MAC-level perspective.
Fig. 5 shows the reliability of DSRC for the CCW application communica-
tion under setting conditions in the application-level. In Fig.5, we can see the
metric AP for CCW, the probability that a receiver within 400 meters from the
sender successfully receives at least one packet in tolerance time Ta should be
490 larger than 99%, can still meet the QoS requirement as the distance between
the sender and receiver is less than 300m even if density reaches 450veh/km.
However, the AP is greatly reduced when the communication distance exceeds
300m. In particular, the greater the density, the faster the AP drops. The AP
is already as low as 0.988 when vehicle density is 300veh/km, and the com-
495 munication distance is 400m, which fails to meet the reliability requirement for
CCW application.
The AP probability for the SVI application is depicted as Fig. 6. The AP
with 3 beacons during Ta within 100m is used for the reliability evaluation

22
1 

0.99
Packet Reception Ratio˄PRR˅

0.98

0.97

0.96 Density=200 vehicles/km


Density=250 vehicles/km
Density=300 vehicles/km
0.95 Density=350 vehicles/km
Density=400 vehicles/km
Density=450 vehicles/km
0.94 
0 50 100 150 200 250 300 350 400
Distance of vehicles to the source node

Figure 5: The reliability of DSRC for CCW with the most stringent parameters under
different vehicle densities

1 

0.998
Packet Reception Ratio˄PRR˅

0.996

0.994

0.992

Density=200 vehicles/km
0.99 Density=250 vehicles/km
Density=300 vehicles/km
Density=350 vehicles/km
0.988
Density=400 vehicles/km
Density=450 vehicles/km
0.986 
10 20 30 40 50 60 70 80 90 100
Distance of vehicles to the source node

Figure 6: The reliability of DSRC for SVI with the most stringent parameters under
different vehicle densities

of the SVI application, and its probability is set to larger than 0.995. It is
500 observed from Fig.6 that the AP probability degrades smoothly within 100m.
The QoS requirement of SVI that all vehicles within the intended range receive
the broadcast message from the sender can be met under most densities, unless

23
1

0.99

0.98

0.97
Packet Reception Ratio˄PRR˅

0.96

0.95 Density=200 vehicles/km


Density=250 vehicles/km
0.94
Density=300 vehicles/km
0.93 Density=350 vehicles/km
Density=400 vehicles/km
0.92 Density=450 vehicles/km

0.91

0.9
5 10 15 20 25 30 35 40 45 50
Distance of vehicles to the source node

Figure 7: The reliability of DSRC for RCW with the most stringent parameters under
different vehicle densities

vehicle density is above 450veh/km.


The AP is also applied to evaluate the QoS requirement of the RCW appli-
505 cation, which requires that a vehicle within 50 meters of the sender successfully
receiving 5 beacons have a probability larger than 99.9%. In Fig.4, we can
observe that the AP probabilities are lower than 0.999 when vehicle density is
250veh/km. This indicates that it is difficult for the current DSRC to meet the
QoS requirement of RCW application in most densities.
510 In summary, the current DSRC is able to meet the QoS requirements of the
above three safety applications with low vehicle density, although the MAC-level
packet reception ratios (PRR) in these scenarios are very low. However, DSRC
fails to meet the QoS requirements of safety applications for greater vehicle
density, which is often encountered in reality.

515 5.3. Reliability Analysis of ABGR Mechanism

In this section, we evaluate the QoS requirements of three safety applications


based on the ABGR mechanism using the T-pro scheme and perform a com-
parison with LIMERIC algorithm, DSRC with fixed 10Hz transmissions (no

24
congestion control present) for safety applications. The same communication
520 parameters are adopted as listed in Table 3.
As detailed in [36], LIMERIC is a distributed and adaptive linear rate control
algorithm where each vehicle adapts its message rate in a way such that the total
channel load converges to a specified target.
The ABGR beacon generation rate control algorithm as shown in Table 2
525 can be expressed as:


 λmax β ≤ βthreshold

 j k
β−βthreshould
λ(β) = λmax − δ βthreshould < β < βblock (20)



 λmin β ≥ βblock

where λ(β) represents the instant beacon generation rate at instant density
β. βblock represents the actual blocking density, which is equal to 2/3 of the
theoretical maximum density.
According to the numerical results in the previous section, the QoS require-
530 ments of all three safety applications are satisfied when density is no more than
200 veh/km. Thus, the parameter density threshold βthreshold can be set to
200, and the density impact factor δ is set to 50. In our simulation scenario,
the theoretical maximum density is equal to the sum of theoretical maximum
density on 3 lanes, i.e., 600 vehicles/km. Thus, βblock is 400 vehicles/km.
535 Given the maximum beacon generation rate λmax as 10 packets/s, Eq. (20) is
modified as follows:


 10 β ≤ 200

 j k
β−200
λ(β) = 10 − 50 200 < β < 400 (21)



 6 β ≥ 400

In Fig. 8-10, the comparisons of the reliability of three safety applications


for three schemes, respectively, are depicted. For convenience, the reliability at
540 the edge of the ROI of safety applications with different densities are shown in

25
1 
The Awareness Probability for CCW
0.99

0.98

0.97

0.96

0.95 Original
LIMERIC
ABGR

0.94 
200 250 300 350 400 450
The vehicle density(vehicles/km)

Figure 8: Comparison of the reliability for CCW using ABGR, LIMERIC and DSRC

1 
The Awareness Probability for SVI

0.998

0.996

0.994

0.992

Original
0.99
LIMERIC
ABGR

0.988 
200 250 300 350 400 450
The vehicle density(vehicles/km)

Figure 9: Comparison of the reliability for SVI using ABGR, LIMERIC and DSRC

the figures. In Fig. 8, the evaluation distance between receiver and sender is
400m, which is the furthest distance for the ROI of the CCW application. In
Fig. 9 and Fig. 10, this distance, the furthest distance for the ROI of the SVI
and RCW application, is 100m and 50m, respectively. It is obvious that the

26
1 

The Awareness Probability for RCW 0.99

0.98

0.97

0.96

0.95

0.94

0.93

0.92

0.91

0.9 
200 250 300 350 400 450
The vehicle density(vehicles/km)

Figure 10: Comparison of the reliability for RCW using ABGR, LIMERIC and DSRC

545 AP for the three safety applications degrades significantly and far below their
respective QoS requirements with increasing density. Increasing density and a
fixed beacon generation rate causes the number of beacons to be transmitted to
increase significantly and aggravates channel contention and interference from
hidden terminals. Then, it affects the reliability of data transmission.
550 We observe that the AP using LIMERIC algorithm for three safety applica-
tions are higher than current DSRC, but it still cannot meet QoS requirements
when the density is too high, respectively. The ABGR mechanism enables the
beacon generation rate to automatically decrease with the increase in vehicle
density. The number of beacons to be transmitted by all the vehicles in the
555 communication range does not increase significantly. This means that the inter-
ference from channel competition and hidden terminals will not be aggravated
by increasing density. Meanwhile, the tolerance window will be prolonged ap-
propriately with increasing vehicle density in the T-Pro scheme. This ensures
that the number of beacons transmitted within the time window can still be
560 maintained at a larger value. Thereby, the probability that several beacons are
successfully received by a receiver in a time window will be be improved. In Fig.

27
8 and Fig. 9, we can see that the application-level reliability ABGR provided
is larger than the DSRC provided, and it can meet the QoS requirements of
CCW and SVI applications. Although the application-level reliability ABGR
565 provided fails to meet the QoS requirements of the RCW application in Fig. 10,
it is greatly improved with respect to approximating the QoS requirements of
the RCW application.

6. Conclusion

Designing a congestion control mechanism to make the network safer and


570 more reliable is one of the challenges in VANET. In this paper, we first analyze
the influence of the change of communication environment parameters on DSRC
communication reliability. The parameters include data transmission rate, bea-
con generation rate, contention window size, and beacon length. Based on the
analysis result, the ABGR mechanism is proposed to improve the reliability
575 of the VANETs safety application. ABGR can automatically adjust the bea-
con generation rate according to varying density. Thus, it can reduce channel
congestion and contention even at larger densities, ensuring reliable and timely
delivery of beacons in the network. Compared with LIMERIC algorithm and
current DSRC, the ABGR significantly improves the reliability of three safety
580 applications. The numerical results validate that ABGR can achieve up to 4.5%,
1%, 10% performance improvements over current DSRC in the AP reliability of
three safety applications, respectively.

ACKNOWLEDGMENT

This Work was supported by the Project of Shandong Province Higher Edu-
585 cational Science and Technology Program, P. R. China(Grant Nos. J18KA351),
the Natural Science Foundation of Shandong Province(Grant Nos. ZR2016AL04,
ZR2016FL05, ZR2017MF039 and ZR2012AM021) and the High-Level Training
Project of Taishan Medical University(No.2015GCC07).

28
References

590 [1] R. Ghebleh, A Comparative Classification of Information Dissemination


Approaches in Vehicular Ad hoc Networks from Distinctive Viewpoints: A
Survey, Computer Networks 131 (2018) 15–37.

[2] J. B. Kenney, Dedicated Short-Range Communications (DSRC) Standards


in the United States, Proceedings of the IEEE 99 (7) (2011) 1162–1182.

595 [3] K. N. Qureshi, A. H. Abdullah, O. Kaiwartya, S. Iqbal, R. A. Butt,


F. Bashir, A Dynamic Congestion Control Scheme for safety applications in
vehicular ad hoc networks, Computers & Electrical Engineering 72 (2018)
774–788.

[4] R. Oliveira, C. Montez, A. Boukerche, M. S. Wangham, Reliable data dis-


600 semination protocol for VANET traffic safety applications, Ad Hoc Net-
works 63 (2017) 30–44.

[5] X. Ma, J. Zhang, T. Wu, Reliability Analysis of One-Hop Safety-Critical


Broadcast Services in VANETs, IEEE Transactions on Vehicular Technol-
ogy 60 (8) (2011) 3933–3946.

605 [6] S. Panichpapiboon, W. Pattara-Atikom, A review of information dissem-


ination protocols for vehicular ad hoc networks, IEEE Communications
Surveys & Tutorials 14 (3) (2012) 784–798.

[7] Ahmad, Syed Amaar and Hajisami, Abolfazl and Krishnan, Hariharan and
Ahmed-Zaid, Farid and Moradi-Pari, Ehsan, V2V System Congestion Con-
610 trol Validation and Performance, IEEE Transactions on Vehicular Technol-
ogy 68 (3) (2019) 2102–2110.

[8] F. Bai, H. Krishnan, Reliability Analysis of DSRC Wireless Communica-


tion for Vehicle Safety Applications, in: IEEE Intelligent Transportation
Systems Conference, 2006, pp. 355–362.

29
615 [9] N. An, T. Gaugel, H. Hartenstein, VANET: Is 95% probability of packet
reception safe?, in: International Conference on ITS Telecommunications,
2011, pp. 113–119.

[10] C. B. Math, H. Li, S. H. D. Groot, I. Niemegeers, V2X Application-


Reliability Analysis of Data-Rate and Message-Rate Congestion Control
620 Algorithms, IEEE Communications Letters 21 (6) (2017) 1285–1288.

[11] W. F. Li, M. A. Xiaomin, J. Wu, K. S. Trivedi, X. L. Huang, Q. Liu,


Analytical Model and Performance evaluation of Long Term Evolution for
vehicle Safety Services, IEEE Transactions on Vehicular Technology 66 (3)
(2017) 1926–1939.

625 [12] O. K. Tonguz, N. Wisitpongphan, F. Bai, DV-CAST: A distributed ve-


hicular broadcast protocol for vehicular ad hoc networks, IEEE Wireless
Communications 17 (2) (2010) 47–57.

[13] C. Sommer, O. K. Tonguz, F. Dressler, Traffic information systems: effi-


cient message dissemination via adaptive beaconing, Communications Mag-
630 azine IEEE 49 (5) (2011) 173–179.

[14] M. Torrent-Moreno, J. Mittag, P. Santi, H. Hartenstein, Vehicle-to-Vehicle


Communication: Fair Transmit Power Control for Safety-Critical Informa-
tion, IEEE Transactions on Vehicular Technology 58 (7) (2009) 3684–3703.

[15] F. Ye, R. Yim, S. Roy, J. Zhang, Efficiency and Reliability of One-Hop


635 Broadcasting in Vehicular Ad Hoc Networks, IEEE Journal on Selected
Areas in Communications 29 (1) (2011) 151–160.

[16] S. M. O. Gani and Y. P. Fallah and G. Bansal and T. Shimizu, A Study


of the Effectiveness of Message Content, Length, and Rate Control for Im-
proving Map Accuracy in Automated Driving Systems, IEEE Transactions
640 on Intelligent Transportation Systems 20 (2) (2019) 405–420.

30
[17] S. Bai, J. Oh, J. I. Jung, Context awareness beacon scheduling scheme
for congestion control in vehicle to vehicle safety communication, Ad Hoc
Networks 11 (7) (2013) 2049–2058.

[18] H. A. Omar, W. Zhuang, L. Li, VeMAC: A TDMA-Based MAC Protocol for


645 Reliable Broadcast in VANETs, IEEE Transactions on Mobile Computing
12 (9) (2013) 1724–1736.

[19] M. Sepulcre, J. Gozalvez, O. Altintas, H. Kremo, Integration of congestion


and awareness control in vehicular networks, Ad Hoc Networks 37 (P1)
(2016) 29–43.

650 [20] S. Zemouri, S. Djahel, J. Murphy, An Altruistic Prediction-Based Con-


gestion Control for Strict Beaconing Requirements in Urban VANETs,
IEEE Transactions on Systems, Man, and Cybernetics: Systems (2018)
1–16doi:10.1109/TSMC.2017.2759341.

[21] SalehYousefi, MahmoodFathy, Metrics for performance evaluation of safety


655 applications in vehicular ad hoc networks, Transport 23 (4) (2008) 291–298.

[22] W. Li, J. Wu, X. Ma, Z. Zhang, On reliability requirement for BSM broad-
cast for safety applications in DSRC system, in: Intelligent Vehicles Sym-
posium Proceedings, 2014, pp. 946–950.

[23] A. A. Gómez, C. F. Mecklenbräuker, Dependability of Decentralized Con-


660 gestion Control for Varying VANET Density, IEEE Transactions on Vehic-
ular Technology 65 (11) (2016) 9153–9167.

[24] X. Chen, H. H. Refai, X. Ma, A Quantitative Approach to Evaluate DSRC


Highway Inter-Vehicle Safety Communication, in: IEEE GLOBECOM
2007 - IEEE Global Telecommunications Conference, 2007, pp. 151–155.

665 [25] J. G. Andrews, R. K. Ganti, M. Haenggi, N. Jindal, A primer on spa-


tial modeling and analysis in wireless networks, Communications Magazine
IEEE 48 (11) (2010) 156–163.

31
[26] K. Trivedi, Probability & statistics with reliability, queuing and computer
science applications, John Wiley & Sons, 2008.

670 [27] X. Ma, X. Chen, H. H. Refai, On the Broadcast Packet Reception Rates
in One-Dimensional MANETs, in: Global Telecommunications Conference,
2008. IEEE GLOBECOM, 2008, pp. 1–5.

[28] M. Nakagami, The m-distributiona general formula of intensity distribution


of rapid fading, in: Statistical methods in radio wave propagation, Elsevier,
675 1960, pp. 3–36.

[29] X. Ma, X. Yin, M. Wilson, K. S. Trivedi, MAC and application-level broad-


cast reliability in vanets with channel fading, in: 2013 International Confer-
ence on Computing, Networking and Communications (ICNC), 2013, pp.
756–761. doi:10.1109/ICCNC.2013.6504183.

680 [30] X. Shen, X. Cheng, R. Zhang, B. Jiao, Y. Yang, Distributed Congestion


Control Approaches for the IEEE 802.11p Vehicular Networks, IEEE In-
telligent Transportation Systems Magazine 5 (4) (2013) 50–61.

[31] X. Chen, Research on Key Technologies of Expressway Safety Early-


warning System Based on Traffic Flow Theory, Ph.D. thesis, School of
685 Transportation, Jilin University (2011).

[32] M. M. Artimy, W. Robertson, W. J. Phillips, Assignment of dynamic trans-


mission range based on estimation of vehicle density, in: International
Workshop on Vehicular Ad Hoc Networks, Vanet 2005, Cologne, Germany,
September, 2005, pp. 40–48.

690 [33] M. Treiber, A. Kesting, Traffic flow dynamics, Traffic Flow Dynamics:
Data, Models and Simulation, Springer-Verlag Berlin Heidelberg, 2013.

[34] F. L. Hall, V. Hurdle, J. H. Banks, Synthesis of recent work on the nature


of speed-flow and flow-occupancy (or density) relationships on freeways,
1993.

32
695 [35] F. Bai, H. Krishnan, T. Elbatt, G. Holland, Towards characterising and
classifying communication-based automotive applications from a wireless
networking perspective, International Journal of Vehicle Autonomous Sys-
tems 10 (3) (2008) 165–197.

[36] J. B. Kenney, G. Bansal, C. E. Rohrs, LIMERIC: a linear message rate con-


700 trol algorithm for vehicular DSRC systems, in: Proceedings of the Eighth
ACM international workshop on Vehicular inter-networking, ACM, 2011,
pp. 21–30.

33
Wenfeng Li has completed his B.S. degree in Computer Software from
705 Shandong University of Science and Technology and M.E. degree in Computer
Application from Jiangxi University of Science and Technology in 1997 and 2007,
respectively. He got the Ph.D. degree at Tongji University, Shanghai, China in
2017. Now he is a lecture in the College of Medical Information Engineering,
Shandong First Medical University & Shandong Academy of Medical Sciences,
710 Taian, China. His research interests include VANET and LTE.

Wuli Song was born in Wenshang, Shandong, P.R. China, in 1978. He re-
ceived his masters degree from Shandong University of Science and Technology,
P.R. China. Now he is studying at China University of Mining and Technology;
715 he is an associate professor of Shandong First Medical University & Shandong
Academy of Medical Sciences. His research interest include computational in-
telligence, information security and big data analysis.

34
Qiang Lu received the B.S. degree from Qingdao University of Science
720 and Technology, Qingdao, China, Master degree from Shandong University of
Science and Technology, Qingdao, China, the Ph.D. degree in control theory
and control engineering from Tongji University, Shanghai, China. He is cur-
rently a association professor in College of Medical Information Engineering,
Shandong First Medical University & Shandong Academy of Medical Sciences,
725 Taian, China(e-mail: luqiang271016@163.com). His current research interests
include intelligent robot control and neural networks.

35
Chao Yue was born in Shandong, China, in 1971. He is a reviewer for
Mathematical Reviews on AMS. He received the Ph.D. degree in mathematics
730 from the Shanghai University, Shanghai, China, in 2015. From September 2015
to July 2016, he was a Visiting Scholar in the Department of mathematics,
Tongji University, Shanghai, China. He is now an Associate Professor with the
College of Medical Information Engineering, Shandong First Medical University
& Shandong Academy of Medical Sciences.

36
735 Conflict of interest statement

We declare that we have no conflict of interest.

37

Вам также может понравиться