Vehicle-to-vehicle (V2V) and vehicle-to-infrastructure (V2I) communication in a heterogeneous wireless network – Performance evaluation

Abstract

Connected Vehicle Technology (CVT) requires wireless data transmission between vehicles (V2V), and vehicle-to-infrastructure (V2I). Evaluating the performance of different network options for V2V and V2I communication that ensure optimal utilization of resources is a prerequisite when designing and developing robust wireless networks for CVT applications. Though dedicated short range communication (DSRC) has been considered as the primary communication option for CVT safety applications, the use of other wireless technologies (e.g., Wi-Fi, LTE, WiMAX) allow longer range communications and throughput requirements that could not be supported by DSRC alone. Further, the use of other wireless technology potentially reduces the need for costly DSRC infrastructure. In this research, the authors evaluated the performance of Het-Net consisting of Wi-Fi, DSRC and LTE technologies for V2V and V2I communications. An application layer handoff method was developed to enable Het-Net communication for two CVT applications: traffic data collection, and forward collision warning. The handoff method ensures the optimal utilization of available communication options (i.e., eliminate the need of using multiple communication options at the same time) and corresponding backhaul communication infrastructure depending on the connected vehicle application requirements. Field studies conducted in this research demonstrated that the use of Het-Net broadened the range and coverage of V2V and V2I communications. The use of the application layer handoff technique to maintain seamless connectivity for CVT applications was also successfully demonstrated and can be adopted in future Het-Net supported connected vehicle applications. A long handoff time was observed when the application switches from LTE to Wi-Fi. The delay is largely due to the time required to activate the 802.11 link and the time required for the vehicle to associate with the RSU (i.e., access point). Modifying the application to implement a soft handoff where a new network is seamlessly connected before breaking from the existing network can greatly reduce (or eliminate) the interruption of network service observed by the application. However, the use of a Het-Net did not compromise the performance of the traffic data collection application as this application does not require very low latency, unlike connected vehicle safety applications. Field tests revealed that the handoff between networks in Het-Net required several seconds (i.e., higher than 200 ms required for safety applications). Thus, Het-Net could not be used to support safety applications that require communication latency less than 200 ms. However, Het-Net could provide additional/supplementary connectivity for safety applications to warn vehicles upstream to take proactive actions to avoid problem locations. To validate and establish the findings from field tests that included a limited number of connected vehicles, ns-3 simulation experiments with a larger number of connected vehicles were conducted involving a DSRC and LTE Het-Net scenario. The latency and packet delivery error trend obtained from ns-3 simulation were found to be similar to the field experiment results.

Full text

Vehicle-to-vehicle (V2V) and vehicle-to-infrastructure (V2I)
communication in a heterogeneous wireless network –
Performance evaluation
q
Kakan Chandra Dey
a,
⇑
, Anjan Rayamajhi
b,1
, Mashrur Chowdhury
c,2
, Parth Bhavsar
d,3
,
James Martin
e,4
a
Glenn Department of Civil Engineering and Department of Automotive Engineering, 216 Lowry Hall, Clemson University, SC 29634, United States
b
Department of Electrical and Computer Engineering, 314 EIB, Clemson University, SC 29634, United States
c
Eugene Douglas Mays Professor of Transportation, Glenn Department of Civil Engineering and Department of Automotive Engineering, 216 Lowry Hall,
Clemson University, SC 29634, United States
d
Civil and Environmental Engineering, 327 Rowan Hall, Rowan University, Glassboro, NJ 08028, United States
e
School of Computing, Clemson University, 211 McAdams Hall, Clemson, SC 29634, United States
article info
Article history:
Received 27 June 2015
Received in revised form 24 March 2016
Accepted 25 March 2016
Keywords:
Connected vehicle technology
Handoff
Het-Net
Collision warning message
DSRC
V2V and V2I
abstract
Connected Vehicle Technology (CVT) requires wireless data transmission between vehicles
(V2V), and vehicle-to-infrastructure (V2I). Evaluating the performance of different network
options for V2V and V2I communication that ensure optimal utilization of resources is a
prerequisite when designing and developing robust wireless networks for CVT applica-
tions. Though dedicated short range communication (DSRC) has been considered as the
primary communication option for CVT safety applications, the use of other wireless tech-
nologies (e.g., Wi-Fi, LTE, WiMAX) allow longer range communications and throughput
requirements that could not be supported by DSRC alone. Further, the use of other wireless
technology potentially reduces the need for costly DSRC infrastructure. In this research, the
authors evaluated the performance of Het-Net consisting of Wi-Fi, DSRC and LTE technolo-
gies for V2V and V2I communications. An application layer handoff method was developed
to enable Het-Net communication for two CVT applications: traffic data collection, and
forward collision warning. The handoff method ensures the optimal utilization of available
communication options (i.e., eliminate the need of using multiple communication options
at the same time) and corresponding backhaul communication infrastructure depending
on the connected vehicle application requirements. Field studies conducted in this research
demonstrated that the use of Het-Net broadened the range and coverage of V2V and V2I
communications. The use of the application layer handoff technique to maintain seamless
connectivity for CVT applications was also successfully demonstrated and can be adopted
in future Het-Net supported connected vehicle applications. A long handoff time was
observed when the application switches from LTE to Wi-Fi. The delay is largely due to
the time required to activate the 802.11 link and the time required for the vehicle to
associate with the RSU (i.e., access point). Modifying the application to implement a soft
http://dx.doi.org/10.1016/j.trc.2016.03.008
0968-090X/Published by Elsevier Ltd.
q
This article belongs to the Virtual Special Issue on CA Vehicles.
⇑
Corresponding author. Tel.: +1 (313) 523 0865; fax: +1 (864) 656 2670.
E-mail addresses: kdey@clemson.edu (K.C. Dey), arayama@g.clemson.edu (A. Rayamajhi), mac@clemson.edu (M. Chowdhury), bhavsar@rowan.edu
(P. Bhavsar), jim.martin@cs.clemson.edu (J. Martin).
1
Tel.: +1 (864) 633 8659; fax: +1 (864) 656 0145.
2
Tel.: +1 (864) 656 3313; fax: +1 (864) 656 2670.
3
Tel.: +1 (856) 256 5399; fax: +1 (856) 256 5242.
4
Tel.: +1 (864) 656 4529; fax: +1 (864) 656 0145.
Transportation Research Part C 68 (2016) 168–184
Contents lists available at ScienceDirect
Transportation Research Part C
journal homepage: www.elsevier.com/locate/trc

handoff where a new network is seamlessly connected before breaking from the existing
network can greatly reduce (or eliminate) the interruption of network service observed
by the application. However, the use of a Het-Net did not compromise the performance
of the traffic data collection application as this application does not require very low
latency, unlike connected vehicle safety applications. Field tests revealed that the handoff
between networks in Het-Net required several seconds (i.e., higher than 200 ms required
for safety applications). Thus, Het-Net could not be used to support safety applications that
require communication latency less than 200 ms. However, Het-Net could provide
additional/supplementary connectivity for safety applications to warn vehicles upstream
to take proactive actions to avoid problem locations. To validate and establish the findings
from field tests that included a limited number of connected vehicles, ns-3 simulation
experiments with a larger number of connected vehicles were conducted involving a
DSRC and LTE Het-Net scenario. The latency and packet delivery error trend obtained from
ns-3 simulation were found to be similar to the field experiment results.
Published by Elsevier Ltd.
1. Introduction
A robust wireless communication network is the foundation for connected transportation systems. Reliable and seamless
vehicle-to-vehicle (V2V) and vehicle-to-infrastructure (V2I) data communication is the critical component of Connected
Vehicle Technology (CVT) applications. Though there are several communication technologies/options available, such as
Wi-Fi, WiMAX, LTE, and DSRC, not all can support low latency, accuracy, and the reliability of data transmission required
for CVT safety applications (RITA, 2015a). While Dedicated Short-Range Communication (DSRC) provides low latency, fast
network connectivity, highly secure and high-speed communication for safety applications, reliance on DSRC only may prove
detrimental to diverse CVT applications. As a result, research efforts for wireless technologies that can enhance V2V and V2I
communications for diverse applications have been undertaken. The wireless communication research community has been
exploring combinations of DSRC with Wi-Fi, WiMAX and LTE communication technologies to provide a robust next-
generation communication network for connected vehicles (Dar et al., 2010). Moreover, it is expected that DSRC roadside
units will be installed at key locations such as intersections and interchanges. Thus, the limited coverage of DSRC (approx-
imately 300 m) and integration of existing Wi-Fi, WiMAX, and LTE network creates a heterogeneous wireless network
(Het-Net) for CVT applications.
For continuous connectivity, the shift from one communication network to another relies on the successful handoffs
between the networks that ensure optimal utilization of available communication options (i.e., eliminate the need of using
multiple communication options at the same time) and corresponding backhaul communication infrastructure requirements
depending on the connected vehicle application requirements. In this study, the authors investigated the potential of a
Het-Net to provide connectivity for V2V and V2I communications with optimal network resource allocation based on the
connected vehicle application requirements. The objectives of this research were to evaluate the performance of Het-Net
for (i) V2I communications for collecting traffic data, and (ii) V2V communications for a collision warning application. For
field experiments, the authors utilized the Clemson University Wi-Fi network, and the National Science Foundation (NSF)
sponsored Science Wireless Network (SciWiNet) infrastructure at Clemson, South Carolina that supports WiMAX, 3G and
LTE, and DSRC infrastructure installed in test vehicles and roadsides. SciWiNet project supports a mobile virtual network
operator (MVNO) for academic research communities (Martin et al., 2015). In additional, an ns-3 simulation experiment
was conducted to evaluate Het-Net performance when there are larger numbers of vehicles within close proximity, and
validate field test findings.
2. Previous studies
Various wireless technologies have been used to support the data transfer requirements of diverse intelligent transporta-
tion system (ITS) applications (Ma et al., 2009a). The selection of wireless communication option relies on the accessibility
and feasibility of a wired communication option and wireless communication options (Wi-Fi, WiMAX, LTE), and data transfer
requirement of particular applications (Ma et al., 2009a; Goodwin, 2003; Dey et al., 2015a; Haseman et al., 2010; Zhou et al.,
2013). While existing ITS applications are infrastructure-based (i.e., installed at roadside locations), the next major deploy-
ment of wireless technologies within the transportation grid is the high speed wireless communication between moving
vehicles and transportation infrastructures (V2V and V2I). V2V and V2I communication supported CVT safety applications
require fast acquisition, fast message delivery (i.e., low latency) with high reliability, and the highest security and privacy
standards (Watfa, 2010). To meet these particular requirements of CVT applications, the Federal Communications Commis-
sion (FCC) assigned 75 MHz bandwidth between 5.850 GHz and 5.925 GHz frequency in the US, which is known as Dedicated
Short-Range Communication (DRSC). DSRC based V2V communication has been developed, evaluated and demonstrated
under the supervision of the USDOT and the auto-manufacturers coalition for multiple safety applications (Lukuc, 2012).
K.C. Dey et al. / Transportation Research Part C 68 (2016) 168–184 169

However, DSRC is insufficient for supporting the breadth of CVT applications outlined by the current connected vehicle
reference implementation architecture (CVRIA) (RITA, 2015b). While DSRC is the primary option for CVT safety applications,
other wireless technologies are also accessible for CVT applications. These technologies can supplement availability, cover-
age, and peak data rate requirements that cannot be supported by DSRC alone (RITA, 2015c). The heterogeneous communi-
cation network (Het-Net) can resolve the issues of range and peak data rate.
2.1. Data communication for safety application
Vehicular Ad-hoc Network (VANET) is the most studied data delivery methods in vehicular communication (Yousefi et al.,
2006). VANET entails the creation of an ad-hoc based communication mesh network of highly dynamic vehicular nodes.
Packets being forwarded in the mesh nodes (i.e., vehicles) need to know the structure of the mesh network, the direction
to forward the data traffic based on the vehicle location and route to the destination nodes/vehicles. Multi-hop routing pro-
tocols for VANET provide different options that could be used for CVT applications based on their specific latency require-
ments (Saleet et al., 2011; Al-Rabayah and Malaney, 2012; Tee and Lee, 2010; Jarupan and Ekici, 2010; Sahu et al., 2013;
Dey et al., 2015b). Also, several studies reviewed the VANET protocols for CVT applications (Darwish and Bakar, 2015;
Dey et al., 2015b). Table 1 summarizes several potential protocols for CVT applications with low and medium latency char-
acteristics. When there are limited numbers of vehicles in a roadway traffic network, a reliable connectivity through vehic-
ular nodes depend on the availability of backup protocols that could be activated in cases when the multi-hop protocols do
not locate a route from a source node to a destination node. Delay tolerant networking protocols were developed to enable
data communication between vehicles without steady connection. For example, Vehicular Delay Tolerant Network (VDTN)
routing protocols implement ‘‘Store and Forward” methods for reliable packet delivery in the case of any lost connection in
the connected vehicle network (Little and Agarwal, 2005), but such approaches are not efficient in the case of a time sensitive
safety critical warning message delivery scenario for a collision avoidance application. Although VDTN may work for some
applications, such as traffic monitoring and traffic data collection, but for latency sensitive applications, such as collision
warnings, a communication network with very low latency is always required.
2.2. Data communication for traffic data collection application
Real time traffic data collection is one of the major functions of transportation agencies to inform travelers and provide
reliable transportation services. Historically infrastructure based sensors (e.g., inductive loop detector, video cameras, infra-
red sensors, blue-tooth) have been extensively used for traffic data collection (Leduc, 2008; Chowdhury and Wang, 2007).
The coverage of the infrastructure based traffic data collection system is very limited and resource intensive to cover all high-
ways. Most transportation agencies deploy roadway sensors to monitor traffic condition on major corridors, however, a
growing number of agencies have been using services from private traffic data providers to report real-time traffic speeds
and travel times for large transportation network (Herrera et al., 2010). The transportation data service providers aggregate
real time traffic data collected from multiple sources, including mobile phones, freight fleets, and GPS devices, that covers a
broader highway network in comparison to limited roadway sensor coverage (INRIX-Iowa DOT, 2015). Such as mobile phone
MAC address has been used using Wi-Fi, Bluetooth technologies to predict pedestrian and cyclist movement (Abedi et al.,
2015). Seo and Kusakabe (2015) proposed a traffic state estimation method, which uses probe vehicle position and spacing
data. Roncoli et al. (2015) developed an optimal control problem that could be used to minimize traffic congestion using
connected vehicle data. Several vehicle models, with multiple wireless communication functionalities (Wi-Fi, LTE, DSRC),
are expected to be introduced in the market over the next few years (GM, 2015), that would provide new set of traffic data
through V2I and V2V integration. These new data sources will improve the reliability of traffic operations, such as traffic
condition assessment and prediction, and response to traffic incidents based on the assessment, (e.g., routing of vehicles
and closing of lanes based on these assessments and predictions) (Ma et al., 2009b; Ma et al., 2012), and reduce public trans-
portation agencies reliance on private service providers. VANET based data communication techniques and algorithms that
are common in V2V communication could be used for traffic monitoring data collection and incident detection (Baiocchi
et al., 2015; Darwish and Bakar, 2015). Argote-Cabañero et al. (2015) proposed a traffic operation measure of effectiveness
(MOE) estimation method using limited connected vehicle data. Piccoli et al. (2015) investigated the sampling requirement
for first order and higher order traffic parameters. Thus, a motivating argument for CVT for traffic condition assessment
applications is that even a small penetration of CVT-enabled vehicles that provide traffic sensing information through V2I
and V2V will improve the accuracy and robustness of the traffic management system, with a significantly reduction in
real-time traffic management costs (Darwish and Bakar, 2015; Herrera et al., 2010).
2.3. Data communication in Het-Net for handoff between communication options
The differences in the characteristics of various wireless communication technologies, such as data rates, coverage radius,
and security requirements do not prevent their access via mobile nodes for continuous connectivity (Magagula and Chan,
2008). The routing of data traffic during the movement of mobile nodes is a major concern, however, because of the wide-
spread adoption of mobile devices such as laptop computers and smart phones (Giovanardi and Mazzini, 1997). One of the
big drawbacks in the use of Het-Nets is the handoff duration, which requires the devices to switch between networks. The
170 K.C. Dey et al. / Transportation Research Part C 68 (2016) 168–184

time necessary for the handoff between networks in turn hinders the reliability and speed of the Het-Net (Magagula and
Chan, 2008). The mobile IP (MIPv6) idea is one concept for managing this handoff, in which a mobile user associated with
a home agent uses the home agent to re-route the data connection. A modified version of the MIPv6 handoff protocol, known
as the PMIPv6, uses a handoff scheme that communicates with the next wireless access point by measuring the mobile
node’s signal strength (Giovanardi and Mazzini, 1997). Several handoff methodologies have been developed, which are char-
acterized by modifications to the existing firmware structure (Kong et al., 2008; Kim et al., 2010; Ning et al., 2014; Bargh
et al., 2008; Baojiang, 2011; Huang et al., 2013). For example, in the creation of a software defined network based implemen-
tation of the Link layer handoff scheme, Izard et al. (2014) developed a smoother transition from one network to another in a
Het-Net scenario. However, the use of the application layer handoff scheme (Leiner et al., 1985) to improve Het-Net perfor-
mance in CVT applications has yet to be explored. The switching from one network technology to other is decided by the
application (discussed more in Section 3). Due to the distinctly separate WAVE stack and IP stack, application layer handoff
deemed to be the most effective solution for Het-Net scenarios considered in this study as also concluded in Uzcategui and
Acosta-Marum (2009). A National Highway Traffic Safety Administration (NHTSA) study suggests that DSRC will be inade-
quate for supporting the safety applications for traffic corridors with high traffic density (i.e., dense roadway infrastructure
such as interchanges) (Harding et al., 2014). Thus, a Het-Net including DSRC is particularly important as it will expand the
application horizon of connected vehicles by allowing seamless V2V and V2I communication (USDOT, 2015; 3GPP, 2015).
3. Method
Fig. 1 illustrates the research method adopted in this study to determine the efficacy of Het-Net in supporting CVT appli-
cations. Field tests and simulation experiments using ns-3 were conducted to answer the research questions and learn about
integration challenges of Het-Net for CVT applications that ensures the optimal utilization of available communication
options (i.e., eliminate the need of using multiple communication options at the same time) and corresponding backhaul
communication infrastructure depending on the application requirements. To evaluate the performance of a Het-Net, case
study 1 was conducted for a CVT supported traffic data collection application. To evaluate the performance of Het-Net for
a CVT safety application, a forward collision warning application for V2V was conducted in the second case study. Based
on field tests and simulation findings, the usability of Het-Net for CVT applications was determined using multiple perfor-
mance measures. The field tests were conducted with an application layer handoff process developed in this study to test
seamless communication between vehicles in a Het-Net system subjected to a hard handoff (Fig. 2). For the purposes of this
study, the data flow between roadside unit (RSU)–on board unit (OBU), and OBU–OBU do not require an uninterrupted con-
nection simply because of the dynamic nature of vehicular networks. Therefore, applications use burst of packet exchanges
through DSRC and through LTE or Wi-Fi, and the term ‘‘handoff” is used to define the process of switching from one network
technology to another. This application layer handoff is effective as it does not require modification to existing hardware and
firmware of connected vehicle equipment, such as OBU and RSU. The technique was designed to detect hard handoff con-
trolled by underlying Link and IP layers. The handoff was controlled by giving priority to DSRC or Wi-Fi over LTE in a
Het-Net environment. The application recorded information about the handoff, such as the interface used before and after
the handoff, packet loss during handoff, and the time necessary to complete the handoff. A comprehensive description of
case studies are presented in the Sections 3.1.1 and 3.1.2 respectively.
3.1. Case study 1: Traffic data collection using vehicle-to-infrastructure (V2I) communication
In order to strategically understand and test the idea of supplementing DSRC with Het-Net for a traffic data collection
application, tests were performed in two Het-Net scenarios as described below.
3.1.1. Wi-Fi and LTE Het-Net infrastructure for traffic data collection
To evaluate the efficacy of V2I data communication using Wi-Fi and LTE Het-Net infrastructure for a traffic data collection
application, a client-to-server packet transfer scheme was developed, in which the client (i.e., vehicle) sends data containing
GPS coordinates, time stamps, and signal strengths of different network interfaces (Wi-Fi or LTE) to a server (i.e., traffic
Table 1
Latency of different VANET protocols.
VANET protocols Network delay
Intersection-based Geographical Routing Protocol (IGRP) (Saleet et al., 2011) Low
PROMPT (Jarupan and Ekici, 2010)
Back-Bone-Assisted Hop Greedy (BAHG) (Sahu et al., 2013)
Vehicle-Assisted Data Delivery (VADD) (Zhao and Cao, 2008)
Greedy Perimeter Stateless Routing (GPSR) (Karp and Kung, 2000) Medium
Hybrid Location based Ad-Hoc Routing (HLAR) (Al-Rabayah and Malaney, 2012)
Junction-based Adaptive Reactive Routing (JARR) (Tee and Lee, 2010)
Static Node-assisted Adaptive Data Dissemination Protocol (SADV) (Ding and Xiao, 2010)
K.C. Dey et al. / Transportation Research Part C 68 (2016) 168–184 171

management center, TMC). The test system utilized in this study consists of a RSU with Wi-Fi and LTE support along the test
roadway section and in-vehicle OBUs with Wi-Fi and LTE support. The test was conducted on an access route to a parking lot
with flat topology and light to medium roadside foliage in Clemson, South Carolina. As shown in Fig. 3, the vehicle began
from Wi-Fi and LTE availability zone and was connected to the data collection server (i.e., TMC) via Wi-Fi. As the vehicle
moved out of Wi-Fi coverage, the connection went through handoff to connect to the server (i.e., TMC) via LTE. In Wi-Fi
and LTE enabled Het-Net scenario, Internet Protocol (IP) was used to enable vehicle to RSU, and vehicle to TMC data com-
munication (Fig. 4). This was a hard-handoff, where the connection was broken before attempting to reconnect. Data was
collected during the time between successive packets during the handoff to determine hard handoff duration.
3.1.2. DSRC and LTE Het-Net infrastructure for traffic data collection
The authors used a client-to-server packet transfer model to evaluate the applicability of DSRC and LTE Het-Net infras-
tructure for traffic data collection. The test system is comprised of OBU and a RSU with DSRC enabled radios on a Linux
machine. A RSU resides by roadside at a predefined location whereas cars equipped with an OBU are moving up and down
along the stretch of roadway and passes the RSU location in each trip. To enable Het-Net functionality, both OBU and RSU are
equipped with LTE support, which allows the application layer handoff between the DSRC and LTE network. Tests were
conducted on a section of the Perimeter Road in Clemson, South Carolina, which has a flat topology and light roadside foliage.
In DSRC and LTE enabled Het-Net, the Wireless Access in Vehicular Environments (WAVE) protocols were used to enable
communication between the OBU and the RSU, and the Internet Protocol (IP) was used to enable communication between
a remote server/TMC–RSU, and OBU–TMC (Fig. 4). Table 2 summarizes the technical specifications of the DSRC RSU and OBU
unit used in this study. The parameters in Table 2 were adopted from equipment specifications of OBU and RSU acquired
from an USDOT approved DSRC equipment provider. The power level of 23 dBm at 64 QAM is the maximum transmission
power and modulation of the equipment used in the field tests. The experimental setup of the application is presented in
Fig. 5. The Het-Net RSU platform of the traffic data collection application functioned within the laptop connected to a RSU.
As shown in Fig. 5, a vehicle with the capability to use either DSRC or LTE network for V2I communication could receive
beacon-like messages from a nearby RSU as soon as it reaches the RSU’s DSRC coverage area. Upon receiving a beacon from
the RSU, the application would then initiate a handoff to DSRC from LTE and use a backhaul network to store the data in a
remote server/TMC. The backhaul of collected data from the RSU to TMC can be transferred either through wireless or wired
connection depending upon the availability. The results from the tests of Het-Net handoff behavior and the performance in
terms of message delivery latency for the traffic data collection application are presented in Section 4.
3.2. Case study 2: Forward collision warning application support using DSRC and LTE within vehicle-to-vehicle (V2V)
communication
V2V communication enabled safety applications use data collected from nearby connected vehicles, such as vehicle
speed, vehicle location, and heading direction, in the determination of any crash risk of the subject vehicle. A safety
application-forward collision warning was selected for field experiment. This application warns vehicles at risk of rear
end collision to take corrective measures due to a suddenly stopped vehicle in a travel lane. The stopped vehicle broadcasts
Can CVT applications be
supported by Het-Net?
What are different
Het-Net network technologies ?
Traffic data collection
(CVT mobility application)
Forward collision warning
(CVT safety application)
How field test could be used to evaluate
Het-Net for CVT applications ?
Case Study 1 Case Study 2
Research
Questions
Research
Findings
Vehicle to Infrastructure
(V2I): Data logging
application
Vehicle to Vehicle (V2V):
Safety message broadcast
application
Is Het-Net suitable for traffic
management applications?
Is Het-Net suitable for forward
collision warning application?
Research Method
Conduct ns-3 simulation to validate field experiment findings
Fig. 1. Evaluation of the Het-Net for CVT Applications-Research Method.
172 K.C. Dey et al. / Transportation Research Part C 68 (2016) 168–184

collision warning message to vehicles within or soon to be within the area about potential safety hazards using V2V connec-
tivity. As with early deployment of connectivity in new vehicle models, there will be a very limited number of vehicles with
DSRC connectivity; thus the possibility of an intermittent broken link and the lack of DSRC enabled vehicles will not allow
continued multi-hop communication. Considering this early deployment scenario, in this study, the authors adopted a single
hop message delivery between vehicles (i.e., OBUs). However, in future research, multi-hop communication such as GeoNet-
working (ETSI, 2014; Cadzow et al., 2012) for time sensitive safety message delivery could be explored to support potential
CVT applications.
An experimental setup was developed with three vehicles traveling in the same direction and same lane, where two
vehicles, one with DSRC support (Vehicle 2) and third with LTE support (Vehicle 3) receive a safety warning message from
a downstream vehicle (Vehicle 1) with DSRC and LTE support that had just stopped abruptly due to a road hazard. Vehicle 2
was traveling within DSRC reach of Vehicle 1 while Vehicle 3 was traveling towards Vehicle 1 but away from the DSRC reach
of Vehicle 1 and Vehicle 2 (Fig. 6). Tests were conducted on a section of Perimeter Road with a flat topology and light road-
side foliage in Clemson, South Carolina. The safety warning message packets including GPS location of the stopped vehicle
and speed were broadcasted at the rate of 10 messages per second (i.e., at 100 ms interval between consecutive broadcast)
by Vehicle 1 upon its abrupt halts at a predefined location on the test route (Fig. 6). During each test run, a series of events
NO
Form Collision
Avoidance Packet
Send Message
through DSRC
End Application
YES
Send Message
through LTE or Wi-Fi
Start Application
Get Speed and GPS
Location
Is Collision Detected?
YES
Form Traffic
Monitoring Packet
Attempt to receive
Beacon from RSU
Is Beacon Received?
NO
Fig. 2. A high-level view of the application layer handoff procedure developed in the study.
K.C. Dey et al. / Transportation Research Part C 68 (2016) 168–184 173

occurs: ➊– Vehicle 1 stops abruptly, ➋– Vehicle 1 starts sending safety warning message packets, ➌– Vehicle 2 receives the
warning message through DSRC, and ➍– Vehicle 3 receives the warning message via LTE. Field tests were conducted to
estimate the time required for the warning messages to reach the two following vehicles using the DSRC and LTE networks
Fig. 3. V2I field test experiment setup using Wi-Fi and LTE networks for traffic monitoring.
Fig. 4. Handoff process between DSRC and LTE or Wi-Fi.
174 K.C. Dey et al. / Transportation Research Part C 68 (2016) 168–184

Table 2
Physical layer specifications of DSRC equipment.
Properties RSU specifications OBU specifications
Frequency 5.85–5.925 GHz 5.85–5.925 GHz
DSRC radio 23 dBm @ 5.9 GHz standard 24 dBm @ 5.9 GHz standard
Channels 10 MHz 10 MHz
Modulation BPSK rate 1/2 BPSK rate 1/2
Fig. 5. Het-Net infrastructure setup using DSRC and LTE networks for traffic data collection.
Fig. 6. Safety application communication support using Het-Net infrastructure.
K.C. Dey et al. / Transportation Research Part C 68 (2016) 168–184 175

respectively, where the times were measured between events ➋and ➌for Vehicle 2 using DSRC, events ➋and ➍for Vehicle 3
using LTE. To observe the influence of vehicle speed on message delivery time, tests were repeated with vehicles traveling at
three different speeds – 20 mph, 35 mph and 50 mph.
To evaluate the message delivery performance when more than two vehicles were traveling toward a stopped vehicle in a
given lane, a second set of experiments were conducted with four vehicles (three vehicles with DSRC support and one vehicle
with LTE support traveling towards an incident location). In addition, to supplement our field study, which is the primary
focus of this paper, simulation experiments in a Network Simulator (Network Simulator Version 3 or ns-3) with more
vehicles were conducted. Fig. 7 shows the setup of a simulation study conducted to determine the effect of a larger number
of connected vehicles than observed/utilized in our field study and their effects on latency of message delivery. Random 2D
Walk Mobility Model was used to create the effect of moving vehicles (Network Simulator-3, 2015). A RSU capable of DSRC
communication was placed at (0, 0) coordinate in an 800 ft 800 ft simulation area as shown in Fig. 7. A LTE transmitter that
can reach all vehicles in the simulation area was placed at (400 ft, 400 ft) coordinate, so that the vehicles moving inside the
simulation region have both DSRC and LTE communication capabilities. Propagation loss model based on Benin et al. (2012)
was added in ns-3 model. An application for the ns-3 simulation was developed that was able to broadcast messages to
vehicles using both DSRC and LTE networks. The simulation tests comprised of up to 40 vehicles that can move within
the simulation grid rendering a simplified simulation model for this study. All the simulated vehicles were equipped with
DSRC and LTE radios. Table 3 shows the ns-3 simulation parameters used in the simulation. Standard LTE simulation scheme
was used that was already available for ns-3 to test LTE throughput and latency.
4. Results and discussion
4.1. Case study 1 test results
4.1.1. Wi-Fi and LTE Het-Net infrastructure for traffic data collection
Two data transfer protocols-UDP and TCP were used for the V2I based traffic data collection tests. The developed test
application was able to detect the breakage of one communication link with one communication network and was designed
to restart a new socket as soon as possible with other available network technologies without breaking the connection
(Fig. 2). The vehicle speed data packet arrival latency during handoff between Wi-Fi and LTE was calculated. The average
delay recorded between successive packets arrival at the TMC is used to calculate the handoff time. Table 4 shows the aver-
age delay between successive packets arriving at the TMC during handoff. In both TCP and UDP handoff scenarios, the time
between arrivals of successive vehicle speed data packets were relatively longer. The longest delay observed was 28 s using
the TCP mode. The delay includes the time required to activate the 802.11 link and the time required for the vehicle to asso-
ciate with the RSU (i.e., access point). However, the traffic data collection application does not require vehicle dynamics data
as frequently as safety applications (3GPP, 2015). A study found that a data collection frequency with 40 s intervals assessed
the average traffic condition with a 95% degree of accuracy (Benin et al., 2012). Thus, transportation agencies can collect
traffic data for a larger highway network irrespective of a 25 s handoff delay (UDP connection) and a 28 s handoff delay
(TCP connection) observed in the field experiment with sufficient accuracy. Furthermore, the collection of traffic data at
longer time intervals reduce substantial investment required to install data storage and processing infrastructure in compar-
ison to the investment need for traffic data collection at a shorter time interval (Liu et al., 2006).
According to the Safe, Accountable, Flexible, Efficient Transportation Equity Act: A Legacy for Users (SAFETEA-LU) requirement,
state DOTs in the US must provide travel time information along major corridors to travelers with 85% accuracy (FHWA,
2013). Using existing technologies (i.e., loop detectors) to achieve the required level of accuracy is challenging, because of
estimation errors, frequent malfunction, and higher maintenance and installation costs of such devices (FHWA, 2006).
Several recent vehicle models are already equipped with Wi-Fi and LTE capabilities (Buss, 2014; Uzcategui and Acosta-
Marum, 2009). Thus, in the future connected transportation systems, the Het-Net infrastructure can be used to support
traffic data collection requirements with a sufficient accuracy.
4.1.2. DSRC and LTE Het-Net infrastructure for traffic data collection
This phase of the research entailed the use of DSRC and LTE in Het-Net for the purpose of collecting and transferring
vehicle speed and location data to a TMC. It was determined that when the vehicles moved into RSU’s DSRC coverage areas,
handoff occurs from LTE to DSRC, and the vehicle speed data was transmitted to TMC through DSRC. The coverage of speed
data collected by vehicles with only DSRC support is shown in Fig. 8(a), the coverage of speed data collected by DSRC and LTE
support together is shown in Fig. 8(b), and the speed profile indicating the collected speeds of the study corridor using both
DSRC and LTE is shown in Fig. 8(c). It is evident that DSRC support alone (without the use of LTE) was insufficient for obtain-
ing a complete speed profile of the study corridor.
In terms of average vehicle speed data packet delivery, the observed latency in the field tests during DSRC and LTE handoff
was 5.75 s, which was much smaller than the Wi-Fi and LTE handoff time. Thus, a Het-Net of DSRC and LTE is more reliable
for the traffic data collection application compared to the Wi-Fi and LTE Het-Net. It can be concluded that two Het-Net
options considered in this study can be implemented for traffic data collection satisfying the SAFETEA-LU requirement with
sufficient reliability.
176 K.C. Dey et al. / Transportation Research Part C 68 (2016) 168–184

4.2. Case study 2 test results: forward collision warning application support using DSRC and LTE within vehicle-to-vehicle (V2V)
communication
The time required for delivering emergency safety message from an abruptly stopped vehicle (Vehicle 1) to two following
vehicles, one with DSRC support (Vehicle 2) and the other with LTE support (Vehicle 3) was calculated from field tests. As
discussed earlier (shown in Fig. 3), the message delivery time of the vehicle using the DSRC fell between events ➋and ➌, and
the message delivery time of the vehicle using LTE fell between events ➋and ➍. The first few safety warning messages
received by Vehicle 2 after Vehicle 1 began broadcasting warning messages under repeated tests at different speeds are
shown in Fig. 9. The small circles marked with ①in Fig. 9 are the first safety message packets received by Vehicle 2 in
different test runs. The distance from Vehicle 1, where Vehicle 2 received the first warning message packets through DSRC
are shown in Table 5, Column 2. Also shown is how far Vehicle 2 would travel applying the brake after receiving the warning
message (Table 5, Column 3) with the comfortable deceleration rate of 11.2 ft/s
2
(AASHTO, 2011). As can be seen, Vehicle 2
has a sufficient time/distance to safely stop upon receiving the safety warning message packets from Vehicle 1 using DSRC.
The distance at which the vehicles received the first safety message from the stopped vehicle varied due to the dynamic
nature of vehicle movements and an increase in packet error rate at high speeds.
Fig. 7. ns-3 Simulation set up.
Table 3
ns-3 Simulation parameters.
DSRC simulation parameters LTE simulation parameters
Propagation loss model Three log distance Propagation loss model Friis spectrum
Fading Nakagami-m Mac Scheduler Proportional fairness MAC
Tx Power Level 23 dBm UL Bandwidth 25 MHz
Freq 5.9 GHz DL Bandwidth 25 MHz
Channel BW 10 MHz Tx Power (UE) 10 dBm
Packet Size 250 Bytes (both case) Tx Power (enodeB) 30 dbm
Table 4
Average delay between packets at the TMC during Wi-Fi and LTE handoff.
Het-Net type Average delay between packets during handoff (s)
Wi-Fi and LTE (TCP connection) 28
Wi-Fi and LTE (UDP connection) 25
K.C. Dey et al. / Transportation Research Part C 68 (2016) 168–184 177

4.2.1. Impact of vehicle speed on message delivery latency
The average time recorded for a warning message to reach Vehicles 2 and 3 (at different vehicle speeds – 20, 35, and
50 mph is illustrated in Figs. 10 and 11 respectively. The message delivery latency for Vehicle 2 (using DSRC) increased with
vehicle speed, with a maximum of 125 ms at 50 mph and a minimum of 88 ms at 20 mph (Fig. 10), both of which satisfy the
Fig. 8. (a) Data points using only DSRC coverage. (b) Data points using DSRC and LTE coverage. (c) Speed profile using DSRC and LTE data points.
Fig. 9. Delivery of forward collision warning packets to Vehicle 2 at different speed.
178 K.C. Dey et al. / Transportation Research Part C 68 (2016) 168–184

200 ms latency requirement for CVT safety applications (Xu et al., 2003). The vehicle with LTE supports (Vehicle 3), which
was outside of the DSRC range of Vehicle 1, received the warning message with a longer delay compared to the latency
requirement for CVT safety applications (Fig. 11). However, Vehicle 3 had a greater margin of reaction time to avoid the
incident location because it received the warning message further upstream. Similar to DSRC, using the LTE safety warning
message packet delivery time increased with the vehicle speed. Additional research to evaluate the suitability of LTE for
safety application warning message delivery must be done because of the highly dynamic nature of delays (i.e. congestion
and concurrent background flow) tied to LTE networks (Garcia et al., 2014). Complementary communication options using
LTE developed from such research will benefit more vehicles upstream by providing advanced warnings when those vehicles
are out of the DSRC coverage area.
In four vehicles field-test scenario, the different geographic locations indicating where the vehicles received the first
safety warning message packets are shown in Fig. 12. These test results were similar to tests with two vehicles scenario.
Here, all three vehicles with DSRC connectivity received warning message packets within a closer proximity of the stopped
vehicle (i.e., within DSRC range) while vehicles with LTE connectivity received the same information much further upstream.
4.2.2. ns-3 simulation results
Four test runs with different number of connected vehicles (i.e., 10, 20, 30 and 40) in the ns-3 simulation were conducted
to determine the impact of number of connected vehicles on message delivery latency for using the DSRC or LTE network.
Table 5
Stoppage distance for Vehicle 2.
Vehicle speed (mph) Distance from Vehicle 1 when 1st safety
warning message packet received (ft)
Distance required to stop at
11.2 ft/s
2
deceleration rate (ft)
20 272 (test run 1) 38
336 (test run 2)
308 (test run 3)
35 260 (test run 1) 118
340 (test run 2)
237 (test run 3)
50 355 (test run 1) 240
304 (test run 2)
282 (test run 3)
70
90
110
130
150
170
190
210
230
250
15 20 25 30 35 40 45 50 55
Message delivery time
(milliseconds)
Vehicle Speed (mph)
Latency requirements for CVT Safety
applications: 200 milliseconds
Fig. 10. Average time required to deliver an emergency safety message to Vehicle 2 using DSRC.
190
690
1190
1690
2190
2690
15 20 25 30 35 40 45 50 55
Message delivery time
(milliseconds)
Vehicle Speed (mph)
Latency requirements for CVT Safety
applications: 200 milliseconds
Fig. 11. Average time required to deliver an emergency safety message to Vehicle 3 using LTE.
K.C. Dey et al. / Transportation Research Part C 68 (2016) 168–184 179

Fig. 13 shows the plot of average latency in seconds with a 95% confidence interval (CI) for four simulation scenarios with
different number of connected vehicles that use DSRC and LTE networks. An analysis of variance of average latency was
conducted to determine the statistical significance of increase in average latency for using the DSRC or LTE due to an increase
in number of connected vehicles. It was found that the latency increase was statistically significant with a greater number of
connected vehicles compared to scenarios with less number of connected vehicles as greater number of vehicles increase
network congestion and cause higher latency in the DSRC and LTE network. Another four test runs were conducted for four
different connected vehicle speeds (i.e., 20, 35, 50 and 70 mph) to determine the impact of vehicle speed on the DSRC and LTE
latency. Fig. 14 shows average latency in seconds with 95% CI for four different vehicle speed scenarios. An analysis of
variance of average latency revealed that the latency increase was statistically significant with a higher vehicle speed
compared to scenarios with lower vehicle speed in the DSRC or LTE network. Table 6 compares the packet delivery latency
at different speeds between simulation and field tests for both DSRC and LTE networks. It can be concluded that packet deliv-
ery latency trend in simulation is similar to what was observed in field tests. The delays in LTE system depend on network
congestion level, user demands and environmental parameters, not only on the physical link state at any timestamp. The
subscribers’ priorities, availability of resource blocks, traffic volume and packet congestion play a major role in the LTE delay.
Depending on the number of users within a network at any given time and associated congestion level of the system;
network performance varies (Mir and Filali, 2014). In the ns-3 simulation conducted in this research, vehicles equipped with
LTE were moving in the simulated network received broadcasted message through LTE networks (as shown Fig. 7). The
authors evaluated the variation in latency due to a change in vehicle speed, and number of connected vehicles. As shown
in Figs. 11, 13 and 14, LTE latency is dependent on the speed of the vehicles as well as number of consecutive connections
with other vehicular and user applications. The LTE latency observed in ns-3 simulation in this research (Table 6) is similar as
reported in a simulation study performed for a city area by Mir and Filali (2014). Moreover, an increase in the number of
connected vehicles in an area causes packet error rate to also increase (Fig. 15). As more vehicles attempt to transmit
Fig. 12. Location of collision avoidance packet arrival in different vehicles.
180 K.C. Dey et al. / Transportation Research Part C 68 (2016) 168–184

# of CVs Lower Limit Upper Limit Lower Limit Upper Limit
10 35.47 35.54 1204.87 1205.23
20 50.66 50.70 1349.39 1350.62
30 66.63 66.66 1485.34 1485.64
40 87.80 87.84 1741.75 1742.11
DSRC LTE
Average Latency (95% CI), milliseconds
Fig. 13. Average latency of DSRC and LTE for different number of connected vehicles in the simulated network.
Lower Limit Upper Limit Lower Limit Upper Limit
20 89.35 89.39 1304.85 1305.08
35 93.79 93.84 1319.76 1320.21
50 96.10 96.16 1374.75 1375.43
70 101.47 101.54 1402.30 1402.87
Average Latency (95% CI), milliseconds
DSRC LTE
Vehi cle
Speed
(mph)
Fig. 14. Average latency at different speeds using DSRC and LTE in the simulated network.
Table 6
Comparison of latencies from ns-3 simulation and field tests.
Speed (mph) DSRC latency (ms) LTE latency (ms)
Field test Simulation Field test Simulation
20 88 89 2555 1305
35 107 94 2851 1320
50 125 96 2977 1375
K.C. Dey et al. / Transportation Research Part C 68 (2016) 168–184 181

messages to the RSU, wireless channel congestion causes large number of packet loss (i.e. higher packet error rate). This
illustrates a typical traffic data collection application where large numbers of OBUs attempt to register their data with
the RSU. However, due to the lack of proper scheduling of data from different OBUs near a RSU and no requirement of
acknowledgements, the packet loss rate was significantly high. A higher packet error rate is a significant concern for safety
applications if congestion results into a significant delay in receiving warning messages.
5. Contribution of this research
To overcome the limited range of DSRC, the authors demonstrated that LTE and Wi-Fi, in combination with DSRC, was an
effective solution in the traffic data collection with reliability that ensures the optimal utilization of alternate communica-
tion options and corresponding backhaul communication infrastructure. Case studies performed in this research demon-
strate the use of Wi-Fi and LTE to widen the range of vehicular communication, which is not possible with DSRC alone.
The use of the application handoff technique to maintain seamless connectivity for CVT applications was also evaluated. This
study is provides a unique perspective on the use of Het-Net for seamless CVT applications.This study demonstrates that,
although there is a long handoff time of about 25 s using UDP mode between Wi-Fi and LTE, and approximately six seconds
between DSRC and LTE, traffic data collection related applications could be supported by Het-Net with significant accuracy.
However, these long handoff times between network options are not suitable for collision warning message delivery for an
imminent crash that would require a much lower latency of less than 200 ms. In future work, different Het-Net service
models that are more appropriate for safety-oriented vehicular applications could be explored.
6. Conclusions
In this study, the authors demonstrated the potential of V2V and V2I in a Het-Net environment with Wi-Fi, DSRC and LTE
that guarantee the optimal utilization of available communication options and minimize the corresponding backhaul
communication infrastructure requirements while considering connected vehicle application requirements. Existing CVT
architecture, such as CVRIA, suggests that seamless communication for V2V and V2I are required to utilize the full potential
of CVT applications. The performance of a CVT application using Het-Net depends on the availability of multiple wireless
communication options, acceptable communication latency, data security, and reliability of timely message delivery. For
a broad range of CVT applications (i.e., safety, mobility, environmental), a viable communication option should include
V2V and V2I capable of utilizing a Het-Net optimally without losing connectivity while moving from one communication
network to another. It will require a successful handoff from one wireless network to another in a Het-Net environment.
However, these different networks have not been designed for the seamless message transfer from one network to another
for moving nodes/vehicles. Consequently these networks must be reconfigurable for developing such a robust synchronized
Het-Net. The research detailed here is the first attempt in designing and evaluating such a network for CVT applications.
A long handoff time was observed due to the time required to activate the 802.11 link and the time required for the vehi-
cle to associate with the RSU (i.e., access point) in LTE and Wi-Fi Het-Net scenario. Field test results revealed that Het-Nets
did not compromise the performance of the traffic data collection application studied in this research. Unlike safety appli-
cations, very low latency (200 ms) is not required for traffic data collection. Het-Nets provide additional connectivity beyond
DSRC range to collect traffic data for a larger traffic network, which is required for many CVT applications. The handoff
between networks (Wi-Fi to LTE, DSRC to LTE and vice-versa) require several seconds to establish a connection and resume
the data transfer, which means that Het-Net could not be used to support time sensitive safety applications. Field tests
revealed that the message delivery time during the handoff was much higher than the CVT safety application latency
requirement of 200 ms. However, Het-Net could provide supplementary connectivity for CVT safety applications to warn
vehicles upstream about any safety hazardous conditions downstream, so that they can take proactive actions to avoid
0.9
0.91
0.92
0.93
0.94
0.95
0.96
0.97
0.98
0.99
1
5 10 25 50
Packet error rate
Numbers of Connected Vehicles
Fig. 15. Packet error rate while transmitting from OBU to RSU.
182 K.C. Dey et al. / Transportation Research Part C 68 (2016) 168–184

the problem locations. ns-3 simulation experiments with a larger number of connected vehicles, compared to the field tests,
were conducted for a DSRC and LTE Het-Net scenario to complement and validate the findings from field tests that included a
limited number of connected vehicles. Results from these ns-3 simulations were similar to the field experiment results. This
study is the first of its kind to evaluate the performance of Het-Net for a seamless V2I and V2V communication for CVT
applications. This application layer handoff method for Het-Net communication developed in this study can be adopted
in future Het-Net-supported connected vehicle applications. Examinations of the feasibility and performance of Het-Net
for a traffic data collection application and a collision warning application were the focus of this research.
Acknowledgements
The LTE service provided through the SciWiNet project was used as the platform for this research. Sponsored by the
National Science Foundation, SciWiNet was developed to support a mobile virtual network operator (MVNO) for academic
research communities. The authors also wish to acknowledge the efforts of Ms. Aniqa Chowdhury and Mr. Godfrey Kimball
for reviewing the draft.
References
3GPP, 2015. The Mobile Broadband Standard. <www.3gpp.org> (accessed on July 27, 2015).
AASHTO, 2011. A Policy on Geometric Design of Highways and Streets, sixth ed. American Association of State Highway and Transportation Officials,
Washington, DC.
Abedi, N., Bhaskar, A., Chung, E., Miska, M., 2015. Assessment of antenna characteristic effects on pedestrian and cyclists travel-time estimation based on
Bluetooth and WiFi MAC addresses. Transport. Res. Part C: Emer. Technol. 60, 124–141.
Al-Rabayah, M., Malaney, R., 2012. A new scalable hybrid routing protocol for VANETs. IEEE Trans. Veh. Technol. 61 (6), 2625–2635.
Argote-Cabañero, J., Christofa, E., Skabardonis, A., 2015. Connected vehicle penetration rate for estimation of arterial measures of effectiveness. Transport.
Res. Part C: Emer. Technol. 60, 298–312.
Baiocchi, A., Cuomo, F., De Felice, M., Fusco, G., 2015. Vehicular ad-hoc networks sampling protocols for traffic monitoring and incident detection in
intelligent transportation systems. Transport. Res. Part C: Emer. Technol. 56, 177–194.
Baojiang, W., 2011. An efficient fast handoff scheme with network mobility in heterogeneous networks. In: 6th International ICST Conference on
Communications and Networking in China, CHINACOM 2011.
Bargh, M., Hulsebosch, B., Eertink, H., Heijenk, G., Idserda, J., Laganier, J., Prasad, A.R., Zugenmaier, A., 2008. Reducing handover latency in future IP-based
wireless networks: proxy mobile IPv6 with simultaneous binding. In: International Symposium on World of Wireless, Mobile and Multimedia
Networks, WoWMoM 2008.
Benin, J., Nowatkowski, M., Owen, H., 2012. Vehicular network simulation propagation loss model parameter standardization in ns-3 and beyond. In:
Southeastcon, 2012 Proceedings of IEEE, pp. 1–5.
Buss, D., 2014. GM’s New OnStar 4G LTE Wi-Fi Stakes Lead in Connectivity. <http://www.forbes.com/sites/dalebuss/2014/05/12/gms-new-onstar-4g-lte-
system-stakes-it-to-new-lead-in-connectivity/> (accessed on June 26, 2015).
Cadzow, S., Hoefs, W., Kargl, F., Roy, R., Sill, S., Whyte, W., 2012. EU-US Standards Harmonization Task Group Report: (Document HTG1&3-3), USDOT.
Chowdhury, M., Wang, K.C., 2007. Distributed Intelligent Traffic Sensor Network. Transport Science and Technology. Elsevier, Amsterdam, Netherlands, ISBN
# 0–08-044707-4.
Dar, K., Bakhouya, M., Gaber, J., Wack, M., Lorenz, P., 2010. Wireless communication technologies for ITS applications [topics in automotive networking].
IEEE Commun. Mag. 48 (5), 156–162.
Darwish, T., Bakar, K.A., 2015. Traffic density estimation in vehicular ad hoc networks: a review. Ad Hoc Netw. 24, 337–351.
Dey, K.C., Mishra, A., Chowdhury, M., 2015a. Potential of intelligent transportation systems in mitigating adverse weather impacts on road mobility: a
review. IEEE Trans. Intell. Transport. Syst. 16 (3), 1107–1119. http://dx.doi.org/10.1109/TITS.2014.2371455.
Dey, K.C., Yan, L., Wang, X., Wang, Y., Shen, H., Chowdhury, M., Yu, L., Qiu, C., Soundararaj, V., 2015b. A review of communication, driver characteristics, and
controls aspects of cooperative adaptive cruise control (CACC). IEEE Trans. Intell. Transport. Syst. http://dx.doi.org/10.1109/TITS.2015.2483063.
Ding, Y., Xiao, L., 2010. SADV: static-node-assisted adaptive data dissemination in vehicular networks. IEEE Trans. Veh. Technol. 59 (5), 2445–2455.
ETSI, TCITS, 2014. Intelligent Transport Systems (ITS); Vehicular Communications; GeoNetworking; Part 4: Geographical Addressing and Forwarding for
Point-to-point and Point-to-multipoint Communications; Sub-part 1: Media-Independent Functionality, ETSI EN 302 636-4-1 V1.2.1.
FHWA, 2006. Traffic Detector Handbook, third ed., vol. II. Publication Number: FHWA-HRT-06-139, US Department of Transportation.
FHWA, 2013. Real-Time System Management Information Program Fact Sheet. US Department of Transportation. <http://ops.fhwa.dot.gov/1201/factsheet/>
(accessed on June 26, 2015).
Garcia, J., Alfredsson, S., Brunstrom, A., 2014. A measurement based study of TCP protocol efficiency in cellular networks. In: WINMeE 2014.
Giovanardi, A., Mazzini, G., 1997. Transparent mobile IP: an approach and implementation. In: IEEE Global Telecommunications Conference, GLOBECOM’97,
vol. 3, pp. 1861–1865.
GM, 2015. Cadillac to Introduce Advanced ‘Intelligent and Connected’ Vehicle Technologies on Select 2017 Models. <http://media.cadillac.com/media/us/
en/cadillac/news.detail.html/content/Pages/news/us/en/2014/Sep/0907-its-overview.html> (accessed on June 26, 2015).
Goodwin, L.C., 2003. Best Practices for Road Weather Management. Final Project Report FHWA-OP-03-081.
Harding, J., Powell, G., Yoon, R., Fikentscher, J., Doyle, C., Sade, D., Lukuc, M., Simons, J., Wang, J., 2014. Vehicle-to-Vehicle Communications: Readiness of V2V
Technology for Application. Report No. DOT HS 812 014. National Highway Traffic Safety Administration, Washington, DC.
Haseman, R.J., Wasson, J.S., Bullock, D.M., 2010. Real-time measurement of travel time delay in work zones and evaluation metrics using Bluetooth probe
tracking. Transport. Res. Rec.: J. Transport. Res. Board 2169 (1), 40–53.
Herrera, J.C., Work, D.B., Herring, R., Ban, X.J., Jacobson, Q., Bayen, A.M., 2010. Evaluation of traffic data obtained via GPS-enabled mobile phones: the mobile
century field experiment. Transport. Res. Part C: Emer. Technol. 18 (4), 568–583.
Huang, K.-L., Chi, K.-H., Wang, J.-T., Tseng, C.-C., 2013. A fast authentication scheme for WiMAX–WLAN vertical handover. Wirel. Pers. Commun. 71 (1), 555–
575.
INRIX-Iowa DOT, 2015. Iowa Department of Transportation uses INRIX’s Big Data and Analytics Solutions to Make Smarter Transportation Decisions and
Eliminate Bottlenecks. <http://inrix.com/iowa-dot/> (accessed on November 07, 2015).
Izard, R., Hodges, A., Liu, J., Martin, J., Wang, K.C., Xu, K., 2014. An openflow testbed for the evaluation of vertical handover decision algorithms in
heterogeneous wireless networks. In: Proceedings of TRIDENT.
Jarupan, B., Ekici, E., 2010. PROMPT: a cross-layer position-based communication protocol for delay-aware vehicular access networks. Ad Hoc Netw. 8 (5),
489–505.
Karp, B., Kung, H.T., 2000. GPSR: Greedy perimeter stateless routing for wireless networks. In: Proc. MobiCom, pp. 243–254.
K.C. Dey et al. / Transportation Research Part C 68 (2016) 168–184 183

Kim, M.S., Lee, S., Cypher, D., Golmie, N., 2010. Fast handover latency analysis in proxy mobile IPv6. In: IEEE Global Telecommunications Conference,
GLOBECOM 2010.
Kong, K.S., Lee, W., Han, Y.H., Shin, M.K., You, H., 2008. Mobility management for all-IP mobile networks: mobile IPv6 vs. proxy mobile IPv6. IEEE Trans.
Wirel. Commun. 15 (2), 36–45.
Leduc, G., 2008. Road traffic data: collection methods and applications. Work. Pap. Energy Transp. Clim. Change 1, 55.
Leiner, B.M., Cole, R., Postel, J., Mills, D., 1985. The DARPA internet protocol suite. IEEE Commun. Mag. 23 (3), 29–34.
Little, T.D., Agarwal, A., 2005. An information propagation scheme for VANETs. In: Proc. IEEE Intell. Transp. Syst., pp. 155–160.
Liu, K., Yamamoto, T., Morikawa, T., 2006. An analysis of the cost efficiency of probe vehicle data at different transmission frequencies. Int. J. ITS Res. 4 (1),
21–28.
Lukuc, M., 2012. NHTSA Research, Light Vehicle Driver Acceptance Clinics. <http://www.safercar.gov/v2v/index.html> (accessed on July 27, 2015).
Ma, Y., Zhou, Y., Chowdhury, M., Wang, K.C., Fries, R.N., 2009a. A framework for performance evaluation of communication alternatives for intelligent
transportation systems. J. Intell. Transport. Syst. 13 (3), 111–126.
Ma, Y., Chowdhury, M., Sadek, A., Jeihani, M., 2009b. Real-time highway traffic condition assessment framework using vehicle–infrastructure integration
(VII) with artificial intelligence (AI). IEEE Trans. Intell. Transport. Syst. 10 (4), 615–627.
Ma, Y., Chowdhury, M., Sadek, A., Jeihani, M., 2012. Integrated traffic and communication performance evaluation of an intelligent vehicle infrastructure
integration (VII) system for online travel-time prediction. IEEE Trans. Intell. Transport. Syst. 13 (3), 1369–1382.
Magagula, L., Chan, H.A., 2008. Optimized handover delay in proxy mobile IPv6 using IEEE802.21 MIH services. In: Southern Africa Telecommunication
Networks and Applications Conference.
Martin, J., Wang, K.C., Seskar, I., 2015. SciWiNet: A Science Wireless Network for the Research Community. <http://sciwinet.org/#nsf> (accessed on July 27,
2015).
Mir, Z.H., Filali, F., 2014. LTE and IEEE 802.11 p for vehicular networking: a performance evaluation. EURASIP J. Wirel. Commun. Netw. 2014 (1), 1–15.
Network Simulator-3 (ns-3), 2015. <https://www.nsnam.org/> (accessed on September 26, 2015).
Ning, Z., Song, Q., Liu, Y., Wang, F., Wu, X., 2014. Markov-based vertical handoff decision algorithms in heterogeneous wireless networks. Comput. Electr.
Eng. 40 (2), 456–472.
Piccoli, B., Han, K., Friesz, T.L., Yao, T., Tang, J., 2015. Second-order models and traffic data from mobile sensors. Transport. Res. Part C: Emer. Technol. 52, 32–
56.
RITA, 2015a. Overview of Dedicated Short Range Communications (DSRC) Technology. <http://www.its.dot.gov/DSRC/#sthash.pegMwDx3.dpuf> (accessed
on July 27, 2014).
RITA, 2015b. Communications Data Delivery System Analysis. <http://www.its.dot.gov/connected_vehicle/communications_system_analysis.htm>
(accessed on July 27, 2015).
RITA, 2015c. ITS Research Fact Sheets, DSRC: The Future of Safer Driving. <http://www.its.dot.gov/factsheets/dsrc_factsheet.htm#sthash.ONoFSgkG.dpuf>
(accessed on July 27, 2015).
Roncoli, C., Papageorgiou, M., Papamichail, I., 2015. Traffic flow optimisation in presence of vehicle automation and communication systems – Part II:
Optimal control for multi-lane motorways. Transport. Res. Part C: Emer. Technol. 57, 260–275.
Sahu, P.K., Wu, E., Sahoo, J., Gerla, M., 2013. BAHG: back-bone assisted hop greedy routing for VANET’s city environments. IEEE Trans. Intell Transport. Syst.
14 (1), 199–213.
Saleet, H., Langar, R., Naik, K., Boutaba, R., Nayak, A., Goel, N., 2011. Intersection-based geographical routing protocol for VANETs: a proposal and analysis.
IEEE Trans. Veh. Technol. 60 (9), 4560–4574.
Seo, T., Kusakabe, T., 2015. Probe vehicle-based traffic state estimation method with spacing information and conservation law. Transport. Res. Part C: Emer.
Technol. 59, 391–403.
Tee, C., Lee, A., 2010. A novel routing protocol—junction based adaptive reactive routing (JARR) for VANET in city environments. In: Proc. EW, pp. 1–6.
USDOT, 2015. Connected Vehicle Reference Implementation Architecture. <http://www.iteris.com/cvria/html/applications/applications.html> (accessed on
July 27, 2015).
Uzcategui, R., Acosta-Marum, G., 2009. WAVE: a tutorial. IEEE Commun. Mag. 47 (5), 126–133.
Watfa, M., 2010. Advances in Vehicular Ad-Hoc Networks: Developments and Challenges. Intelligent Transport Systems. IGI Global.
Xu, Q., Segupta, R., Jiang, D., Chrysler, D., 2003. Design and analysis of highway safety communication protocol in 5.9 GHz dedicated short range
communication spectrum. In: The 57th IEEE Semiannual Vehicular Technology Conference, vol. 4, pp. 2451–2455.
Yousefi, S., Mousavi, M.S., Fathy, M., 2006. Vehicular ad hoc networks (VANETs): challenges and perspectives. In Proc. IEEE 6th ITS, pp. 761–766.
Zhao, J., Cao, G., 2008. VADD: vehicle-assisted data delivery in vehicular ad hoc networks. IEEE Trans. Veh. Technol. 57 (3), 1910–1922.
Zhou, Y., Chowdhury, M., Wang, K.C., Dey, K.C., 2013. Evaluation of wireless communication performance between adjacent nodes for intelligent
transportation systems applications. In: Transportation Research Board 92nd Annual Meeting (No. 13-3942).
184 K.C. Dey et al. / Transportation Research Part C 68 (2016) 168–184