1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47

A Framework of a V2X Communication System for Enhancing Vehicle and
Pedestrian Safety at Un-signalized Intersections
Ziqiang Zeng, Ph.D.
Research Associate in Smart Transportation Application and Research Lab, Department of Civil
and Environmental Engineering, University of Washington
101 More Hall, Seattle, WA, 98195
Tel: (206) 303-0165; Fax: (206) 543-5965
Email: ziqiang@uw.edu
John Ash, Ph.D. Candidate
Smart Transportation Application and Research Lab, Department of Civil and Environmental
Engineering, University of Washington
101 More Hall, Seattle, WA, 98195
Tel: (414) 467-9555; Fax: (206) 543-5965
Email: jeash@uw.edu
Ziyuan Pu, Ph.D. Candidate
Smart Transportation Application and Research Lab, Department of Civil and Environmental
Engineering, University of Washington
101 More Hall, Seattle, WA, 98195
Tel: (206) 617-0494 Fax: (206) 543-5965
Email: wbzhu@uw.edu
Yifan Zhuang, Ph.D. Student
Smart Transportation Application and Research Lab, Department of Civil and Environmental
Engineering, University of Washington
101 More Hall, Seattle, WA, 98195
Tel: (206) 519-9298; Fax: (206) 543-5965
Email: ker27@uw.edu

Yinhai Wang, Ph.D. (Corresponding Author)
Professor in Smart Transportation Application and Research Lab, Department of Civil and
Environmental Engineering, University of Washington
121F More Hall, Seattle, WA, 98195
Tel: (206) 616-2696; Fax: (206) 543-5965
Email: yinhai@uw.edu

Total number of words: 5661 (text) +1750 (2 table, 12 figures) = 7411
Submitted for presentation only at the TRB 97th Annual Meeting
Submission Date: August 1, 2017

 Zeng et al.

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24

2

ABSTRACT
Connected vehicle (CV) technology presents great potential to improve the safety of our
roadway system by increasing driver situational awareness and reducing or eliminating crashes
through vehicle-to-vehicle (V2V), vehicle-to-infrastructure (V2I), and vehicle-to-device systems
(V2X) communications. There have been several studies focusing on safety and operational
benefits of V2V and V2I communications. However, the safety benefits of V2X communications
were not sufficiently explored and discussed in the current published materials. This paper
focuses on developing a framework of a V2X communication system for enhancing vehicle and
pedestrian safety at un-signalized intersections. A comprehensive review of the literature has
been made to investigate existing V2X safety applications. A cost-effective, solar-energy driven,
small, and lightweight communication node device, called the smart road sticker (SRS), is
developed to communicate with connected vehicles via LoRa and dedicated short range
communications (DSRC), and with pedestrians and unconnected vehicle through cell phones and
other mobile devices via Bluetooth. A mobile application that allows pedestrians, bicyclists, and
drivers of unconnected vehicles to communicate with the SRS device and vice versa is also
designed. A crash prediction algorithm is developed to identify unsafe conditions and determine
appropriate CV based safety countermeasures to be presented to system users. Finally, a CV
simulation test bed is established in VISSIM to evaluate the safety benefits of the proposed
methodology under various traffic and landscape conditions. The simulation results indicate that
the number of conflicts increases when the penetration rate of connected devices decreases.
Keywords: Connected vehicle; V2X communication system; Vehicle and Pedestrian Safety;
Intersections.

 Zeng et al.

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46

3

INTRODUCTION
In 2014, there were 4884 pedestrians killed in traffic crashes in United States. In addition,
about 65000 pedestrians were estimated to be injured. In other words, a pedestrian is killed
approximately every two hours and injured every eight minutes in the United States (1). These
statistics indicate the importance of targeting pedestrians in development of new traffic safetyrelated technologies. In a connected environment, road users (drivers and pedestrians/bicyclists)
will receive warnings that inform them of potential hazards through visual displays, audible
warning messages, etc. These warnings will increase roadway users’ awareness of hazards and
other dangerous situations they may not otherwise be aware of. As a result, drivers for example,
can be alerted of imminent crash situations such as head-on collisions or rear-end crashes.
Furthermore, when going through an intersection, roadway users can be alerted about another
system user who has failed to observe the right of way so that these roadway users can take
necessary actions to avoid potential crashes.
Using connected vehicle (CV) data and technology to improve traffic operations has been
the topic of many studies over the past five years. Some examples include Lee and Park (2), He
et al. (3), He et al. (4), Christofa et al. (5), Hu et al. (6), Feng et al. (7), Guler et al. (8), and Lee et
al. (9). There are also a number of research projects that aim at evaluating the effectiveness of a
number of CV safety applications. For instance, the effectiveness of six vehicle-to-vehicle (V2V)
safety applications are currently being tested in Ann Arbor, MI. Those applications are forward
collision warning, emergency electronic brake light, blind spot/lane change warning, intersection
movement assist, and left turn assist. The University of Michigan is performing this study and
has a fleet of 3000 connected vehicles. In addition, the USDOT has sponsored and performed
research in the area of V2V communications for safety. The research focus has been on finding
out if vehicle based safety application using V2V communications are effective and if they have
the perceived benefits. A report published in 2013 described pre-crash scenarios that might be
addressed by V2V communications (10).
Most of the current studies consider safety applications of CV technologies in V2V
communications and their application in real-world situations (11-13). Therefore, the probability
of collisions occurring and near-collision events are reduced significantly. In fact, several types
of warning messages such as forward collision warning, lane change warning, do-not-pass
warning, and control loss warning keep the drivers aware of crash prone situations. Although
V2V communication provides effective tools to reduce the number and severity of the crashes, it
cannot address all possible types of crashes. Thus, V2I communications that make use of
roadside equipment such as signal controllers are considered as another means to improve safety
via an application of CV technologies. For example, Maile et al. (14) developed a cooperative
intersection collision avoidance system to warn drivers about violation of stop signs or traffic
signals. Moreover, the information from signalized crosswalks can be sent to approaching
vehicles as a warning message to make the drivers aware about the presence of pedestrians in the
crosswalk.
There have also been some studies considering the vehicle-to-pedestrian (V2P)
communication to issue alerts between vehicles and vulnerable road users when potential
collisions are possible to arise. Liu et al. (15) a dedicated short-range communication/long-term
evolution/WiFi-based vehicular system is developed to support the V2V and V2P
communication for the safety of vehicles and pedestrians. Borges et al. (16) used the stereo
vision to detect pedestrians in the proposed active pedestrian safety system that combined the
sensing, situation analysis, decision making, and vehicle control. The General Motors

 Zeng et al.

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46

4

researchers applied the peer-to-peer standard WiFi Direct to collaborate with other sensor-based
object detection and driver alert systems on production vehicles to detect pedestrians and
bicyclists carrying smart-phones equipped with WiFi Direct (17). Wang et al. (18) proposed an
Android application WalkSafe that uses the back camera of the mobile phone to detect vehicles
and possible dangerous situation and alert the user using sound and vibration. Dhondge et al.
(19) proposed a smartphone-based Car2X communication system with the WiFi beacon stuffed
to alert collisions for pedestrians. While these studies have made significant contributions to
promote the development of V2P or V2X, a lot of issues are still unsolved. Typical problems
include the following aspects: (1) the association time of WiFi Direct users takes only around 1
s, which is actually limited in the practical traffic environment that the driving speed should not
be larger than 20 km/h (); (2) the DSRC applied for the V2X communication requires the
pedestrians’ mobile devices to be installed with the IEEE 802.11p enabled module; (3) it is
difficult for the user to put the phone with Android application WalkSafe in the suitable position
to capture the transportation conditions while walking; (4) the users’ walking states are not
considered in the Car2X communication system.
As mentioned above, most of the ongoing research activities focus on how V2V and V2I
communications may increase safety for drivers. V2X impacts on road user safety are typically
not sufficiently explored and discussed in these existing studies while they can offer a great
potential for further improvement in the safety of mixed-use roadways. To address the gaps, the
proposed research project will focus on developing a communication node (CN) device that
facilitates communications between connected vehicles, pedestrians, bicyclists, and unconnected vehicles (as long as they have a smart phone and the mobile app). The devices can be
placed on spots or links that are known to have safety problems. The design of the CN devices is
a challenging task and needs to meet time-to-collision constraints. Such devices will allow
communications between CVs and road users without CV devices to connect via regular mobile
electronic devices. A mobile app that enables communication with the CV system is developed
to facilitate communications and displays of safety messages. An algorithm is also built in the
app to evaluate safety condition and identify response messages to communicate. Finally, a CV
simulation test bed is established in VISSIM to evaluate the safety benefits of the proposed
methodology under various traffic and landscape conditions.
DEVELOPMENT OF CN DEVICE AND MOBILE APPLICATION
An initial literature search has shown that researchers have investigated V2X
communications and developed devices and methodologies to communicate information on
possible safety issues with pedestrians to drivers of motor vehicles (20-21). Although these
papers have made initial strides in the V2X arena, they focus on the safety issue from one
perspective only, namely, that of the motorist. To date, little work has been done to address
safety issues, particularly conflicts between motorized and non-motorized users, in a connected
vehicle environment from the perspective of the non-motorized user. We feel providing safety
information to the non-motorized users is just as important as providing similar information to
drivers as both parties may ultimately be able to take preventative actions to avoid conflicts.

Smart Road Sticker Design for Bluetooth Detection
A solar-energy driven, small, and lightweight CN device is developed, as shown in
Figure 1, that could be installed along the roadside. This sensor would receive messages from

 Zeng et al.

1
2
3
4
5
6
7
8
9
10
11

12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28

5

connected vehicles and infrastructure (such as signal controller cabinets) and transmit pertinent
safety information/warning messages to users’ cellular devices via Bluetooth or another suitable
cellular communication protocol. This would allow for system users, especially non-motorized
users who do not have access to DSRC devices, to send/receive critical safety information and
greatly increase the potential number of “connected” users across all modes on a given facility.
Besides allowing for communication between vehicles, infrastructure, and non-motorized users,
the sensors would be able to transmit information from one sensor to another. By developing a
small, and relatively inexpensive sensor, multiple sensors could be easily installed in critical
areas to overcome issues of transmission range and allow users (regardless of mode) to be
informed of safety issues/potential conflicts well in advance of their arrival to the site of the
potential conflict.

FIGURE 1 Illustration of the CN device application in a vision restricted street scenario
The CN device developed in this section is named smart road sticker (SRS). There are
four components in the SRS, which are the central processing module, long range
communication module, Bluetooth detection module, and the power module. The central
processing module uses the ATmega 328P chip for processing the Bluetooth device information,
and the long-range communication module uses LoRa techniques to achieve both long distance
communication and low power consumption. The Bluetooth detection module uses the HC-05
Bluetooth module at the master mode to search surrounding Bluetooth devices positively. The
power module is a high-capacity Lithium-ion battery designed to support continuous work. The
exterior and inner structure of the SRS are shown in Figure 2 (a) and (b). The detection time for
one interval is 5 seconds and the maximum number of devices able to be detected at once is 10.
During the detection interval, the sticker will send detection results when the number of devices
detected reaches the maximum number. Otherwise, the sticker will still send data at the end of
the detection interval.

 Zeng et al.

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26

6

(a) Exterior of the SRS
(b) Inner structure of the SRS
FIGURE 2 The design of SRS
Software Development
Once the sensor is designed to receive information from CVs and CV infrastructure has
been designed, the next step will be to develop a mobile device application for use of receiving
information from the sensors and presenting pertinent safety information to users who have the
application installed on their phones or other mobile devices. Such information will be
comprised of warnings about potential conflicts that may allow users to take corrective or
evasive actions to prevent the conflicts from ever happening. Again, the goal of developing the
mobile application and the CN device is to allow for transmission of safety information to a
larger percentage of the traveling population, regardless of their mode, as at least in the present
time, non-motorized users do not have DSRC-capable devices. Existing research, such as
Roodell (22), has addressed communication of safety information in a connected vehicle
environment between DSRC and cellular phones, but primarily from the perspective of the driver
and not in terms of presenting warnings/information to pedestrians and bicyclists. As
aforementioned, cyber security and privacy are important issues to address when designing the
mobile application.
The software can be categorized into two types mobile phone application and PC
application. The mobile phone application (STAR Detection App) is based on the Android
platform and can send beacons at a specific rate. The PC application will listen to the LoRa
receiver port and filter the data received. Only Bluetooth MAC address registered in the system
will be displayed. The interfaces of Android and PC applications are shown in Figure 3. In order
to guarantee the cyber security and privacy, Bluetooth MAC address will be encrypted in the
mobile and PC applications as they are technically personally identifying information.

 Zeng et al.

7

1
2
3
4
5
6
7
8
9
10
11
12

Experiment Results
Experiments were done on several perspectives including basic function verification and
detection distance measurement. The basic function of the sticker is detecting mobile devices with
the STAR Detection App installed. Once a device is detected, the SRS will send data to the receiver
and PC via LoRa. The coded MAC address of the detected device will then be displayed on the
screen. Ultimately, an experiment was conducted near the Burke Gilman Trail (a multi-use path
for bicyclists and pedestrians) on the University of Washington campus in Seattle. The detection
results and experimental scenario (in which a pedestrian is walking past the SRS) are shown in
Figure 4.

13
14
15
16
17

FIGURE 4 Detection result and experiment scene
The testers opened and ran the mobile application at different distances from the SRS. The
detection results are shown in Table 1. The results indicate that the detection error ratio increases
dramatically when the distance from the SRS exceeds 8 m. As a result, the maximum effective
distance of detection is set to be 8 m.

18

FIGURE 3 Interface of Android and PC applications

TABLE 1 The detection results at different distances from the SRS
Distance
2m
4m
6m

Total Number
20
20
20

Detection Number
19
18
18

Error Ratio
0.05
0.10
0.10

 Zeng et al.

8

8m
10m

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19

15
8

0.25
0.60

CRASH PREDICTION ALGORITHM
To estimate the risk of collision between pedestrians and vehicles, there is a need to develop
a framework that makes the proper connection between devices. This framework should use
information such as current position and speed from the smartphones and onboard units in the
vehicles to estimate the collision risk and warn both pedestrians and drivers to take the appropriate
action within a reasonable time. In this research, we consider the time to collision as the surrogate
measure of risky situations for potential crossing and rear-end conflicts. A general framework is
developed based on the assumption that the smartphone devices and onboard units provide
accurate data about the position and speed of the pedestrians and vehicles. First, the means of
conflict point estimation and the approach to find it will be discussed. Then, the time to collision
will be estimated, and finally the proposed algorithm will be presented.
Computations to Find the Conflict Point
To find the conflict point between two vehicles or a vehicle and a pedestrian, we use the
information collected from the onboard units in the vehicles and the smartphones. The longitude
and latitude positions and the direction of the movements are used to estimate the conflict point.
Figure 5 shows the position and the direction of the movements of the pedestrian and a vehicle.
The position of the conflict point can be defined based on Eqs. (1) and (2).

20
21
22

FIGURE 5 Finding the collision point
𝑥𝑐 =
𝑦𝑐 =

23
24
25
26
27

20
20

(𝑦𝑣 − 𝑦𝑝 ) − (𝑥𝑣 tan(𝜃𝑣 ) − 𝑥𝑝 tan(𝜃𝑝 ))
tan(𝜃𝑝 ) − tan(𝜃𝑣 )
(𝑥𝑣 − 𝑥𝑝 ) − (𝑦𝑣 cot(𝜃𝑣 ) − 𝑦𝑝 cot(𝜃𝑝 ))
cot(𝜃𝑝 ) − cot(𝜃𝑣 )

(1)
(2)

Computations for Time to Collision
Finding the conflict points could be helpful to find the frequency of the conflict situations,
but it does not show the severity of the potential conflict. Thus, time to collision is introduced to

 Zeng et al.

1
2
3
4
5
6
7
8
9

9

measure the severity of the conflict situation. Time to Collision (TTC) indicates the time at which
two vehicles or a vehicle and a pedestrian would collide if they continue their movement with their
current speed and path. Lower values of TTC indicate more severe possible conflicts. If TTC drops
below a given threshold value, the safety risk gets to a critical condition and the users should take
actions to prevent the crash. The value of TTC can be calculated based on the speed and the
distance to conflict point for each of the users. Eq. (3) shows the measurement of TTC for crossing
conflicts. However, this formulation is not appropriate for rear-end collisions. Thus Eq. (4) is
introduced for rear-end cases. Two rear-end and crossing conflicts are shown in Figure 6 (a) and
(b) respectively.
𝑑2
𝑑1
𝑑2
𝑑1 +𝑙1 +𝑤2
𝑖𝑓
<
<
𝑣
𝑣1
𝑣2
𝑣1
(3)
Crossing TTC= {𝑑21
𝑑2
𝑑1
𝑑2 +𝑙2 +𝑤1
𝑖𝑓
<
<
𝑣
2
𝑣
𝑣
1

Rear-End TTC =
10
11
12
13
14
15
16

17
18
19
20
21
22
23

𝑥1 −𝑥2 −𝑙1
𝑣2 −𝑣1

1

2

𝑖𝑓 𝑣2 > 𝑣1

(4)

where
𝑥1 = Location of the vehicle 1.
𝑥2 = Location of the vehicle 2.
𝑣1 = Speed of vehicle 1.
𝑣2 = Speed of vehicle 2.
𝑑1 = Distance to the conflict point from the front of vehicle 1.
𝑑2 = Distance to the conflict point from the front of vehicle 2.

FIGURE 6 Rear-end and crossing conflicts
We can also calculate the threshold for critical situations. The critical time for a vehicle is
the time that a driver needs to perceive the warning message (𝑡𝑝 ) and react to it by activating the
brake (𝑡𝑏𝑟 ). In addition, we can consider a safe distance between vehicles and pedestrians which
should not be violated (𝑡𝑠𝑎𝑓𝑒 ). Eq. (1) indicates the minimum time (𝛿) that a driver or a pedestrian
needs to react to avoid the collision.
𝛿 = 𝑡𝑝 + 𝑡𝑏𝑟 + 𝑡𝑠𝑎𝑓𝑒
(1)

 Zeng et al.

1
2
3
4
5
6
7
8
9
10
11
12
13

14
15
16
17
18
19
20
21

22
23
24
25
26
27
28
29

10

We assume that at a critical time, the user’s reaction with braking is at the maximum
deceleration rate, as this is an emergency situation. To be more realistic, the maximum deceleration
is considered to be less when the vehicle is moving with higher speed because decelerating harshly
is not safe for the passengers inside the vehicle. Thus, a linear relationship between the current
speed and maximum deceleration is assumed. Eq. (2) shows this relationship. In this equation 𝑎0
(which takes a negative value) indicates the maximum deceleration rate that a vehicle can reach
when its speed is near zero. The value for 𝑏 is positive as the rate of maximum deceleration
increases when speed increases. However, for pedestrians, it is assumed that the maximum
deceleration rate is a constant value equal to 𝑎𝑝 . As such, the braking time for vehicles and
pedestrians can be calculated with Eqs. (3) and (4), respectively. However, since pedestrians travel
at lower speeds, we can assume that they stop instantaneously. It should be noted that the speed of
pedestrians is assumed to maximum in order to consider the worst case. The unit of the time, speed,
𝑚
𝑚
and deceleration rates in all proposed equations are, 𝑠, 𝑠 , and 𝑠2 , respectively.
(2)
𝑀𝑎𝑥𝐷 = 𝑎0 + 𝑏𝑣
1
𝑏
𝑣
𝑡𝑏𝑟
=   ln(1 + 𝑣) 
(3)
𝑏
𝑎0
𝑝
𝑣𝑚𝑎𝑥
𝑝
(4)
𝑡𝑏𝑟
= 
 
𝑎𝑝
The safe time ( 𝑡𝑠𝑎𝑓𝑒 ) for vehicles and pedestrians can be calculated based on Eqs.
(5) and
(6). We also define a variable, 𝑇𝑇𝐶𝑃, to measure the
remaining time for a vehicle or a pedestrian to get to the conflict point if it keeps its current speed
and path. Eq.
(7) shows this value, where 𝑑 is the remaining
distance. Assuming 𝛿𝑣 and 𝛿𝑝 are the critical times for the vehicle and pedestrian respectively, the
critical time that considers both can be introduced as 𝛿𝑐𝑟𝑖𝑡 = 𝑚𝑎𝑥{𝛿𝑣 , 𝛿𝑝 }. When TTC is equal to
𝛿𝑐𝑟𝑖𝑡 , either the vehicle, the pedestrian, or both of them should reduce their speed in order to
prevent a collision.
𝑑𝑠𝑎𝑓𝑒
𝑣
(5)
𝑡𝑠𝑎𝑓𝑒
=
𝑣
𝑑𝑠𝑎𝑓𝑒
𝑝
𝑡𝑠𝑎𝑓𝑒
= 𝑝
(6)
𝑣𝑚𝑎𝑥
𝑑
(7)
𝑇𝑇𝐶𝑃 =
𝑣
Proposed Crash Prediction Algorithm
The developed algorithm to predict safety risk and prevent collisions is described in this part.
This framework considers V2V and V2X communications as the vehicles are connected to each
other and are connected to pedestrians and bicyclists as well. The objective of the algorithm is to
warn the proper user in and give them sufficient time to avoid the collision. Figure 7 shows the
procedure of the algorithm as to how it identifies risky conditions and sends warning messages to
users.

 Zeng et al.

1
2
3
4
5
6
7
8
9

11

FIGUR 7 Algorithm for preventing the collision between vehicles and pedestrians
In the proposed algorithm, it is assumed that at each time step the onboard unit in the vehicle
(device 𝑖) collects the information (position and speed) of the surrounding devices including the
other vehicle’s onboard units and smartphones within a certain range of distance that DSRC
technology covers. However, since the smartphones can be connected to vehicles through the
cellular network, this range will be accordingly less. Device 𝑗 among all surrounding devices will
be selected to analyze the risk condition between device 𝑖 and device 𝑗. The analysis for the current

 Zeng et al.

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45

12

time step will be finished when all the surrounding devices in the covered area are selected and
analyzed. As the interaction between vehicles and pedestrians is different than the interaction
between vehicles and other vehicles, user 𝑗 must be defined as a pedestrian or a driver.
If device 𝑗 is a pedestrian, two conditions are possible; either the pedestrian is in the road or
on the sidewalk (potentially walking along or standing still on the sidewalk). When the pedestrian
is in the road and the driver 𝑖 gets to the critical distance (𝑇𝑇𝐶𝑃𝑖 ≤ 𝛿𝑖 ), the warning message will
be sent to the corresponding driver to brake and yield to the pedestrian. However, there may be
some cases that the driver does not perform well or does not pay attention to the warnings. Thus,
these cases are considered in this algorithm. In such situations, the warning message continues to
be sent to the driver while the warning messages will be sent to pedestrian to stop or take the proper
action as well. When the risky condition is resolved, the warning messages stop being sent. On the
other hand, if the pedestrian is in the sidewalk and there is a risky condition when 𝑇𝑇𝐶 ≤ 𝛿𝑐𝑟𝑖𝑡 ,
the warning message will be sent to the pedestrian to stop moving forward and stay on the sidewalk
until the risky condition is resolved.
When device 𝑗 is an onboard unit in the vehicle, the interaction can be different. In this case,
when a risk of collision is detected as 𝑇𝑇𝐶 ≤ 𝛿𝑐𝑟𝑖𝑡 , we decide to send the warning message to the
vehicle that has the higher value for time to reach the collision point. In other words, the closer
vehicle will pass the conflict point while the farther away one will yield to it. It is assumed that the
vehicles start to reduce their speed with the maximum deceleration rate as described in the previous
section when the warning message is sent to them. The value of the new speed can be estimated
with equation
(8). In this equation, 𝑣𝑡 is the current
speed of the vehicle and 𝑣𝑡+1 is the estimated speed for the next time step, where the length of the
time step is ∆𝑡.
(8)
𝑣𝑡+1 = 𝑣𝑡 + 𝑀𝑎𝑥𝐷. ∆𝑡
SIMULATION EVALUATION APPROACH
To investigate the effectiveness of the proposed safety performance algorithm, there is a need to
go through a simulation approach. Simulation provides the opportunity to evaluate the ability to
reduce the number of collision and pre-collision situations based on different settings. In this
study, the VISSIM microscopic simulator is used because it provides the required tools to
simulate the CV condition and corresponding interactions between vehicles and pedestrians. The
simulation test bed is created in VISSIM and different scenarios are tested within it. The
procedures for creating the test bed and the tested scenarios are explained as below.
VISSIM-based Simulation
VISSIM provides the possibility to export the results of the simulation as inputs of the
Surrogate Safety Assessment Model (SSAM). VISSIM also provides the opportunity to take
control of the vehicles and pedestrians with an external logic through the COM interface. The
latter feature of this software is desirable for the purposes of this study because V2X
communication provides more information to the drivers which could be used to enhance the
reaction of drivers and pedestrians in unsafe situations.
The earlier versions of VISSIM (e.g., VISSIM 5.00 and prior) used to introduce a readyto-use Car2X module for driving behavior which automatically makes the equipped vehicles
aware about unsafe situations and has them react =accordingly. However, the later versions of
VISSIM (e.g., VISSIM 6.00 and after) do not offer this module anymore. Moreover, using the
predefined modules could be limiting. The VISSIM COM interface provides the opportunity to

 Zeng et al.

1
2
3
4

5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24

13

identify the traffic condition and control the behavior of drivers and pedestrians in an online
environment. For example, when a vehicle receives a warning message about the possibility of a
collision, it will reduce its speed to keep maintain a safe distance. The structure of implementing
the proposed safety performance algorithm in VISSIM is described in Figure 8.

FIGURE 8 The structure of simulation with VISSIM.
As shown in Figure 8, the network and the required inputs such as the percentages of
vehicles equipped with onboard units and the percentages of pedestrians equipped with
smartphones are predefined. Using the COM interface, the proposed crash prediction algorithm
is implemented in order to detect risky situations by using information from vehicles and
pedestrians that are connected to each other. Then, time to collision can be estimated and used as
a basis on which to change the behavior of users by changing their speeds. After each simulation
run is complete, the trajectory file output of VISSIM is inserted into SSAM and the number of
conflicts between vehicles and pedestrians, as well as their severities, are found.
Simulation Test Bed
As aforementioned, for this project, the simulation test bed is created in VISSIM. As the
literature review section shows, most of the collisions between vehicles and pedestrians occur in
urban areas. The most frequent cases are those in which a vehicle is turning and a pedestrian is
crossing the road. However, the most severe cases occur when a vehicle is going straight and a
pedestrian is crossing the road. Simulating the behavior of drivers and pedestrian in an urban
intersection covers most of these cases. Thus, a network with a four-leg intersection with both
vehicle and pedestrian movements is considered as the test site for investigating the V2V and

 Zeng et al.

1
2
3
4
5
6
7
8
9
10

11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35

14

V2P communications and the corresponding safety performance. Figure 9 shows the studied
network containing an intersection.
This intersection has a symmetric shape with four legs. Each direction has one lane with a leftturn pocket. The studied intersection can be signalized or un-signalized. In addition, pedestrians
can move through the intersection in all directions. When a vehicle is going straight, turning left,
or turning right, there is potential for the vehicle to collide with pedestrians that are crossing the
road. In this network, it is assumed that the speed limit for vehicles is 50 km/hr and the
pedestrians are moving with a maximum speed 5 km/hr. In this study, we work with an unsignalized intersection in order to consider the high-risk scenarios for cases when the users are
connected and not connected.

FIGURE 9 Simulation test bed.
Simulation Scenarios
The intersection for this case study is un-signalized to increase the number of potential
collisions between vehicles and pedestrians. Such a situation helps better display the
performance of the proposed algorithm. In this case study, pedestrians can move across the
intersection in all directions. Vehicles can go straight, turn right, and turn left. It is expected that
the proposed algorithm addresses most of the possible collisions between vehicles and
pedestrians, as well as those between vehicles and other vehicles.
To compare the performance of the proposed algorithm under different conditions, there is
a need to define several scenarios. Generally, a combination of two sets of scenarios considering
the volume of vehicles and pedestrians and the penetration rate of the connection between them
can indicate the performance of the algorithm in different situations.
Increasing the volume increases the number of movements in the intersection. As a result,
the number of interactions between vehicles and pedestrians increases. In this study, three
volume levels are considered 100 veh/hr/lane, 200 veh/hr/lane, and 300 veh/hr/lane. It is
assumed that the volume for pedestrians is 15 percent of the vehicular volume.
The second important factor in the performance of the proposed algorithm is the
penetration rate of the CVs in the network. If there are less communication between vehicles to
vehicles and vehicles to pedestrians, higher probability of collision is expected. In this study four
penetration rates will be considered as follows: 100, 75, 50, and 25 percent of the total number of
vehicles and pedestrians.
SIMULATION RESULTS AND DISCUSSIONS

 Zeng et al.

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45

15

The results from the simulation study based on the evaluation approach described above is
presented. First, we describe the parameters that are fixed in the network. Then, the criteria used
to define a situation as risky are explained, and finally, the results and their supporting discussion
are provided.
Defining the Input Parameters
For the purpose of the analysis, the network is loaded with different volume levels, in
different simulation runs, as explained before. In the studied intersection, the rate of turning and
through movements are defined before running the simulation. In the case study, 70 percent of
vehicles for each movement go straight, 15 percent of vehicles turn right, and 15 percent turn
left. The volume for pedestrian is 15 percent of the volume for vehicles.
In this case study, the users communicate with each other every 0.1 seconds as the
connected environment provides this capability. It should be noted that smaller time steps for
analyzing the risk condition gives more accurate results. In addition, each of the equipped
vehicles and pedestrians look for surrounding users within a range of 100 meters. If they find an
equipped vehicle, the connection happens and each device calculates the estimated TTC to define
the risk of crash occurrence. When TTC drops to a value less than the predefined threshold,
warning messages will be sent to the users.
To test each of the described scenarios, several simulations were implemented in order to
address the randomness in each. Each scenario is repeated 10 times, which takes 5 minutes.
Recall that each of the users needs to measure the risk of a crash with their surrounding vehicles
within the range of 100 meters. Therefore, several calculations necessary for this risk assessment
are carried out in the underlying COM interface. As a result, the run time increases significantly
for higher volumes, higher demands, and longer study durations. Based on the experimental
results, five minutes is set to be the maximal simulation time for each simulation.
Selection of Threshold Value for Conflicts
Based on previous studies, the threshold for considering a conflict with respect to TTC is
suggested to be 1.5 s (23-24). This value indicates that when the time to collision drops to less
than 1.5 second, there is a high chance of a collision between a vehicle and a pedestrian. In
general, smaller values of TTC indicate a higher risk of collision. Ultimately, when the value of
TTC is equal to zero, we can say a collision has occurred.
VISSIM provides a trajectory file as an output of each simulation run which can be
imported into the SSAM software. Given the predefined TTC threshold, SSAM will detect the
number of conflicts and the safety measures corresponding to that conflict. Based on this
approach we analyzed the performance of the proposed crash prediction algorithm.
Results for Safety Measures
This section provides the results and analysis of the case study based on the predefined
parameters and the safety measures. As aforementioned, TTC is considered as the safety
indicator. Table 2 shows the number of conflicts for each of the simulated scenarios. Based on
the results, the number of conflicts increases when the volume in the network increases. In
addition, increasing the penetration rate of connected devices decreases the number of conflicts.
TABLE 2 The number of conflicts in different scenarios

 Zeng et al.

1
2
3
4
5
6
7
8
9

10
11
12
13
14
15
16

17

16

Volume / Penetration Rate

25%

50%

75%

100%

100 veh/hr/lane

10

4

2

0

200 veh/hr/lane
300 veh/hr/lane

22
71

22
46

15
36

3
11

Three types of collisions between vehicles and pedestrians are possible in the simulation
and they include those in which a pedestrian is crossing the road and a vehicle is going straight,
turning right, or turning left. Figure 10 shows the movements considered in the conflict scenarios
in the case study. When the penetration rate is low, a conflict between a vehicle and pedestrian
most commonly occurs when the vehicles are going straight and pedestrians cross the road. In
addition, the conflicts for left turn movements are more frequent than conflicts for right turn
movements. However, when the penetration rates increase, the total number of conflicts as well
as the number of conflicts with through movements decrease.

FIGURE 10 The number of conflicts based on the movements and penetration rates.
In addition, Figure 11 shows increasing the volume increases the number of conflicts for
all movements. Specifically, the conflicts corresponding to through movements increase
substantially in number. The number of right turn and left turn conflicts also increase, but the
rate of increase for right turn conflicts is higher than that for left turn conflicts.

 Zeng et al.

1
2
3
4
5
6

7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32

17

FIGURE 11 The number of conflicts based on the movements and volume.
Figure 12 shows the decreasing rate of the number of conflicts based on the penetration
rate for all of the different volume levels considered in the simulations. When the volume is
higher, a simultaneous increase in the penetration rate of connected devices has a larger effect on
decreasing the number of conflicts.

FIGURE 12 The relationship between the number of conflicts with penetration rates for
different volumes
CONCLUSION
This research focuses on using connected vehicle information to identify locations prone to
conflicts between motorized and non-motorized users to improve traffic safety on mixed-use
roadway networks. With this information in hand, transportation system users can be alerted of
potential conflicts they may be involved in, prior to their occurrence, allowing these users to take
preventative actions, perhaps by making evasive maneuvers. A literature review was made to
investigate existing V2X safety applications. These methods to allow a sensor to communicate
via both dedicated short range communications (DSRC) and via Bluetooth, WiFi or another
communication protocol commonly used by mobile devices were also investigated as well as
existing surrogate safety measures and their use in conflict- and safety-prediction algorithms.
While there have been several studies focusing on safety and operational benefits of V2V and
V2I communications, the safety benefits of V2X communications were not found in published
materials so far. In this study, a cost-effective, solar-energy driven, small, and lightweight
communication node device, called the smart road sticker (SRS), was developed to communicate
with connected vehicles via LoRa and DSRC, and with pedestrians, bicyclists, and unconnected
vehicle through cell phones and other mobile devices via Bluetooth. A mobile application that
allows pedestrians, bicyclists, and drivers of unconnected vehicles to communicate with the SRS
device and vice versa was also designed. A crash prediction algorithm was developed to identify
unsafe conditions and determine appropriate CV based safety countermeasures to be presented to
system users. Finally, a connected vehicle simulation test bed was established in VISSIM to
evaluate the safety benefits of the proposed methodology under various traffic and landscape
conditions. The simulation network was tested under different scenarios including different

 Zeng et al.

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46

18

penetration rates and volumes. Each case was tested several times and the results are compared.
Based on the findings of this study, the number of conflicts increases when the penetration rate
of connected devices decreases. In addition, increasing the volume has a direct relationship with
increasing the number of conflicts. In the simulation study, the number of conflicts between
through-moving vehicles and pedestrians crossing the road is the highest. This is due to higher
volumes of vehicles that go straight (70 percent) as compared to making turning movements.
Increasing the penetration rate of connected devices reduces the number of through-moving
conflicts significantly. Moreover, the number of conflicts involving left turns and right turns
decreases as well. Generally, the number of conflicts decreases more rapidly when the
penetration rate of connected devices increases among higher volumes.
ACKNOWLEDGEMENT
This research was supported by the project of the Pacific Northwest Transportation Consortium.
REFERENCES
1. National Highway Traffic Safety Administration (NHTSA), 2014. Traffic Safety Facts 2014:
A Compilation of Motor Vehicle Crash Data from the Fatality Analysis Reporting System and
the General Estimates System. National Center for Statistics and Analysis, U.S. Department of
Transportation.
2. Lee, J., Park, B., 2012. Development and evaluation of a cooperative vehicle intersection
control algorithm under the connected vehicles environment. IEEE Transactions on Intelligent
Transportation Systems, 13 (1), 81-90.
3. He, Q., Head, K. L., Ding, J., 2012. PAMSCOD: Platoon-based arterial multi-modal signal
control with online data. Transportation Research Part C 20, 164-184.
4. He, Q., Head, K. L., Ding, J., 2014. Multi-modal traffic signal control with priority, signal
actuation and coordination. Transportation Research Part C 46, 65-82.
5. Christofa, E., Argote, J., Skabardonis, A., 2013. Arterial queue spillback detection and signal
control based on connected vehicle technology. Transportation Research Record: Journal of
the Transportation Research Board 2356, 61-70.
6. Hu, J., Park, B.B., Lee, Y., 2015. Coordinated transit signal priority supporting transit
progression under connected vehicle technology. Transportation Research Part C 55, 393408.
7. Feng, Y., Head, K.L., Khoshmagham, S., Zamanipour, M., 2015. A real-time adaptive signal
control in a connected vehicle environment. Transportation Research Part C 55, 460-473.
8. Guler, S.I., Menendez, M., Meier, L., 2014. Using connected vehicle technology to improve
the efficiency of intersections. Transportation Research Part C 46, 121-131.
9. Lee, J., Park, B.B., Malakorn, K., So, J., 2013. Sustainability assessments of cooperative
vehicle intersection control at an urban corridor. Transportation Research Part C 32, 193-206.
10. Najm, W.G., Ranganathan, R., Srinivasan, G., Toma, J.S., Swanson, E., Burgett, A., 2013.
Description of Light-Vehicle Pre-Crash Scenarios for Safety Applications Based On Vehicleto-Vehicle Communications. US Department of Transportation. DOT HS 811 731,
Washington DC.
11. Ahmed-Zaid, F., Bai, F., Bai, S., Basnayake, C., Bellur, B., others, 2011. Vehicle Safety
Communications—Applications: Second Annual Report. US Department of Transportation.
DOT HS 811 492A, Washington DC.

 Zeng et al.

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39

19

12. Kandarpa, R., Chenzaie, M., Dorfman, M., Anderson, J., Marousek, J., Schworer, I., Beal, J.,
Anderson, C., Weil, T., Perry, F., 2009. Vehicle Infrastructure Integration Proof-of-Concept
Executive Summary—Infrastructure. US Department of Transportation, Washington DC.
13. Dion, F., Oh, J.S., Robinson, R., 2011. Virtual testbed for assessing probe vehicle data in
intellidrive systems. IEEE Transactions on Intelligent Transportation Systems 12, 635-644.
14. Maile, M., Neale, V., Ahmed-Zaid, F., Basnyake, C., Caminiti, L., Doerzaph, Z., Kass, S.,
Kiefer, R., Losh, M., Lundberg, J., 2008. Cooperative Intersection Collision Avoidance
System Limited to Stop Sign and Traffic Signal Violations (CICAS-V). US Department of
Transportation, Washington DC.
15. Liu, Z.S., Liu, Z.Y., Meng, Z., Yang, X.Y., Pu, L., Zhang, L., 2016. Implementation and
performance measurement of a V2X communication system for vehicle and pedestrian
safety. International Journal of Distributed Sensor Networks 12 (9), 1-14.
16. Borges, P.V.K., Zlot, R., Tews, A., 2013. Integrating off-board cameras and vehicle on-board
localization for pedestrian safety. IEEE Transactions on Intelligent Transportation 14 (2), 720730.
17. General Motors, 2012. Developing Wireless Pedestrian Detection Technology,
http://media.gm.com/media/us/en/gm/news.detail.html/content/Pages/news/us/en/2012/Jul/07
26_pedestrian.html.
18. Wang, T.Y., Cardone, G., Corradi, A., Torresani, L., Campbell, A.T., 2012. WalkSafe: a
pedestrian safety app for mobile phone users who walk and talk while crossing roads. In:
Proceedings of the 12th workshop on mobile computing systems and applications
(HotMobile’12), San Diego, CA, 28–29 February 2012, p.5. New York: ACM.
19. Dhondge, K., Song, S., Choi, B.Y., Park, H., 2014. WiFiHonk: smartphone-based beacon
stuffed WiFi car2X-communication system for vulnerable road user safety. IEEE 79th
vehicular technology conference (VTC spring), Seoul, South Korea, 18-21 May, pp.1-5.
20. Sugimoto, C., Nakamura, Y., Hashimoto, T., 2008. Prototype of pedestrian-to-vehicle
communication system for the prevention of pedestrian accidents using both 3G wireless and
WLAN communication. In: Wireless Pervasive Computing, 2008. ISWPC 2008. 3rd
International Symposium on. pp. 764-767.
21. Hisaka, S., Kamijo, S., 2011. On-board wireless sensor for collision avoidance: Vehicle and
pedestrian detection at intersection. Intelligent Transportation Systems (ITSC), 2011 14th
International IEEE Conference on, pp.198 (205), 5-7.
22. Roodell, B.D., 2009. Vehicle Driver Message Alert System Using DSRC/WAVE and
Bluetooth. Masters’ Thesis. University of Minnesota.
23. Gettman, D., Pu, L., Sayed, T., Shelby, S.G., 2008. Surrogate safety assessment model and
validation: Final report. PublicAtion No. FHWA-HRt-08-051. US Department of
Transportation, Washington DC.
24. Sayed, T., Brown, G., Navin, F., 1994. Simulation of traffic conflicts at unsignalized
intersections with TSC-Sim. Accid. Anal. Prev. 26, 593-607.