Using DSRC Road-Side Unit Data to Derive Braking Behavior

Rahul Sengupta, Tania Banerjee, Yashaswi Karnati, Sanjay Ranka and Anand Rangarajan

University of Florida, U.S.A.

ﬂ

Keywords:

Connected Vehicle, DSRC, CV2X, Road-Side Unit, Braking, Trafﬁc Safety.

Abstract:

With the increasing deployment of Connected Vehicle Technology (via DSRC/CV2X), public trafﬁc authori-

ties are presented with a potential treasure trove of valuable data for analysis. However, several practical limi-

tations exist that pose unique challenges in this publicly-collected data such as lack of vehicle re-identiﬁcation

due to privacy measures, sparsity of data, limited range of transmission, noise in recorded trajectories etc. In

this paper, we analyze trajectories for braking behaviors of Connected Vehicle Road-side Unit (RSU) data.

The dataset consists of trajectories collected from a dense urban grid consisting of 25 intersections along 4

high-volume arterials, for a period of 1 year. We begin by providing a brief description of the data collection

and processing modalities. We then present a tool to perform exploratory analytics on the data with a focus on

anomalous trajectories with hard-braking events. We show the beneﬁts of such a tool for public trafﬁc author-

ities to gain insights into the performance and safety aspects of urban arterials, and to guide policy decisions.

1 INTRODUCTION

The recent advancements in the Intelligent Trans-

portation System (ITS) have led to the extensive

deployment of various real-time sensing and data

collection systems. Public trafﬁc agencies col-

lect various types of data, including loop detec-

tor data (Haas et al., 2001; Lamas et al., 2015),

GPS, video, BlueTooth, CV2X (Cellular Vehicle-to-

Everything), DSRC (Dedicated Short Range Com-

munications) (Kenney, 2011), (Wolf et al., 2014).

Such publicly-collected data has the potential for use

in gaining insights into the state of corridors and

real-time adaptive arterial trafﬁc signal optimization

(Wang et al., 2022; Zhao et al., 2012). However, it

is resource-intensive to collect, collate, store several

terabytes of data, and then perform analytics to gain

insights.

To address this challenge, this paper

describes a

tool that uses the data broadcast by V2X (Vehicle-

to-everything) enabled vehicles to analyze anomalous

trajectory events, speciﬁcally unexpected braking.

The V2X communication systems, like DSRC (Dedi-

cated Short Range Communications) and CV2X (Cel-

The work was supported in part by NSF CNS 1922782

and by the Florida Department of Transportation (FDOT).

The opinions, ﬁndings and conclusions expressed in this

publication are those of the authors and not necessarily

those of NSF or FDOT.

lular Vehicle-to-Everything), are becoming increas-

ingly widespread in urban trafﬁc networks. These

communication systems usually consist of two types

of components: On-board Units (OBUs), which are

installed and interfaced with vehicles, and Road-side

Units (RSUs) which are usually installed at intersec-

tions. OBUs and RSUs communicate with each other

via a series of predeﬁned messages (Kenney, 2011).

Importantly, OBUs transmit their host vehicles’ data

including location, speed, heading to RSUs.

While there is signiﬁcant work (Liu et al., 2022)

that uses trip-level probe vehicle data via GPS (Global

Positioning Satellite), such data is usually available

in real-time to auto guidance platforms (such as An-

droid Auto

, Apple CarPlay

, HERE Technologies

WEJO

etc.) and ride-sharing companies (such as

Uber

, Lyft

etc.). This proprietary data is not pub-

licly accessible. However, public trafﬁc authorities

may have access to GPS trajectory traces transmitted

by OBUs to RSUs, when the vehicles are within a

couple of hundred meters of the RSUs near intersec-

tions.

Interactions of vehicle trajectories within a signal-

www.android.com/auto/

www.apple.com/ios/carplay/

www.here.com/

www.wejo.com/

www.uber.com/

www.lyft.com/

420

Sengupta, R., Banerjee, T., Karnati, Y., Ranka, S. and Rangarajan, A.

Using DSRC Road-Side Unit Data to Derive Braking Behavior.

DOI: 10.5220/0012025300003479

In Proceedings of the 9th International Conference on Vehicle Technology and Intelligent Transport Systems (VEHITS 2023), pages 420-427

ISBN: 978-989-758-652-1; ISSN: 2184-495X

 2023 by SCITEPRESS – Science and Technology Publications, Lda. Under CC license (CC BY-NC-ND 4.0)

ized intersection is a complex phenomenon(Banerjee

et al., 2022), inﬂuenced by factors such as signal tim-

ing plan, road geometry, surrounding driver behav-

iors, pedestrian behaviors etc. Intersections are also

critical points of conﬂict along the roadways and ac-

count for one-quarter of all trafﬁc fatalities as well as

one-half of all trafﬁc injuries (FHWA, ; FDOT, ).

Vehicle trajectories near intersections (including

anomalous braking behavior) can be detected by

video data with automated object segmentation and

tracking algorithms (Huang et al., 2020; Banerjee

et al., 2022). These algorithms detect and track the

trajectories of various road users. Based on these,

dangerous braking can be detected. However, these

algorithms are greatly limited by the deployed video

camera instrumentation and have a very limited ﬁeld

of view: ﬁsh-eye lens cameras can only capture the

inner intersection region and regular cameras can gen-

erally capture only one approach of the intersection.

GPS-based trajectories have been used to detect

dangerous interactions. For example, in (Li et al.,

2021), important bus-driving events (using GPS tra-

jectory data) were extracted and used for surrogate

safety measures for pedestrians and bicycles.

In this paper, we focus on utilizing the

anonymized GPS data from the Connected Vehicles

infrastructure to develop algorithms to detect braking

incidents which are often representative of the danger-

ous events that happen near intersections. The GPS

data was collected at the RSUs for the Connected Ve-

hicles such as transit buses and ofﬁcial University ve-

hicles that broadcast their coordinates, in terms of lat-

itude and longitude, to the RSUs. An added advantage

with this data over video is the larger tracking region

over which vehicles can be tracked.

The contributions in this paper may be summa-

rized as follows:

1. We created a system for collecting and storing

data in real-time from 25 intersections in the

vicinity of University of Florida, by connecting to

the RSUs. The collected data was then stored on

the cloud in data buckets.

2. We developed a novel system to analyze and vi-

sualize location coordinate data transmitted by

OBUs to the RSUs, to determine braking behavior

which is an important marker for trafﬁc intersec-

tion safety.

3. In this work, we lay the groundwork for effec-

tively using DSRC (and in future, CV2X) data for

analyzing braking behaviors by local trafﬁc au-

thorities based on the RSU data likely available to

them. This is despite the practical challenges as-

sociated with such data including the lack of ve-

hicle re-identiﬁcation (due to privacy concerns),

limited range restricted to the proximity of the

RSU and low OBU penetration rates.

This work is intended to serve as a foundation

for future applications, particularly as the number of

Connected Vehicles grows and their presence at inter-

sections becomes more widespread. The paper is or-

ganized as follows: Section 2 describes our data col-

lection framework and methodology for determining

the braking behavior. Section 3 presents our results

of analyzing data collected for over a year for brak-

ing behavior at an intersection-level as well as at the

network-wide system-level. Finally, we conclude in

Section 4.

2 DATA COLLECTION AND

PROCESSING

In this section, we describe the data source, the data

location, and the data collection and storage pipeline,

the data processing, and our methodology for detect-

ing braking behavior.

2.1 Data Source

The data source comprises of the RSUs at various in-

tersections. The collected data is composed of incom-

ing and outgoing data with respect to an RSU. The in-

coming data are DSRC messages sent by the OBUs,

and by neighboring RSUs. On the other hand, the

outgoing DSRC messages are sent by the RSU to the

neighboring RSUs, and all OBUs in its range.

DSRC, or Dedicated Short-Range Communica-

tions (Biddlestone et al., 2012), is a radio communica-

tion technology that follows the IEEE 802.11 ”Wi-Fi”

standards and enables secure one-way and two-way

communication between vehicles and the road trafﬁc

infrastructure. The range of DSRC communication

can vary from 100-1000m based on the topography

and line-of-sight. This technology enables vehicles to

share information with other Connected Vehicles and

to send messages about road conditions and safety is-

sues. In addition to relaying information between ve-

hicles and road trafﬁc infrastructure, DSRC technol-

ogy also enables vehicles to communicate with RSUs

installed at trafﬁc intersections. This communication

capability can be utilized for signal timing manage-

ment and optimization (Hsu and Shih, 2015; Mandava

et al., 2009). For the purpose of this work, we gather

the messages sent from vehicles to the RSUs. These

messages are formatted in a speciﬁc way and include

various ﬁelds. Our main focus is on the GPS coordi-

nates ﬁelds within these messages.

Using DSRC Road-Side Unit Data to Derive Braking Behavior

421

GPS, or Global Positioning System (Hofmann-

Wellenhof et al., 2012), is a radio navigation system

that uses satellites to send low-energy radio signals

to Earth, which can be picked up by GPS receivers

commonly found in smartphones, for example. When

GPS receivers are installed on vehicles, these sensors

provide an estimate of the vehicle’s location, head-

ing, and speed. It should be noted that GPS signals

can become inaccurate in areas with tall buildings due

to scattering effects, in which case we would need to

post-process the information to dampen the noise.

GPS data has been widely utilized in trafﬁc ﬂow

analysis, particularly for congestion analysis (Yong-

chuan et al., 2011; D’Andrea and Marcelloni, 2017;

Kan et al., 2019; Wang et al., 2016; Sun et al., 2019).

Connected vehicles have GPS systems built into the

OBUs which broadcast their GPS trajectory at a 10Hz

resolution to RSUs via Onboard Units (OBUs) by

embedding the latitude and longitude within a Basic

Safety Message (BSM). Figure 1 shows the structure

of a DSRC BSM message.

Figure 1: Fields in a Basic Safety Message (BSM) from

SAE J2935.

This study focuses on the structure of Basic Safety

Messages (BSMs) registered by OBUs as they pass

near intersections equipped with RSUs. The BSM

format, described in J2935 (DSRC, 2020), allows the

RSU to capture various properties of the vehicle, in-

cluding its location, speed, and more. However, to

maintain privacy, the vehicle is assigned a temporary

ID that changes every few minutes, so the count of

such IDs does not reﬂect the actual number of unique

vehicles.

The type of data that can be captured by the OBU

depends on its capability to connect to the vehicle.

If the OBU is directly interfaced with the vehicle, it

can record ﬁelds such as transmission, heading, an-

gle, acceleration and braking. However, for this study,

the vehicles were ﬁtted with external portable OBUs.

Hence, only the vehicle’s ID, time stamp, latitude and

longitude, and speed (id, secMark, lat, long, speed

ﬁelds in the BSM message structure) were captured.

Thus, the data of interest was obtained from

RSUs. Although the collected data included incom-

ing messages from OBUs and neighboring RSUs, this

paper focuses on the messages received from OBUs

in the form of BSMs. The OBUs generally transmit

BSMs at a rate of 10 messages per second. The BSMs

gathered include GPS coordinates of the vehicles in

terms of latitude and longitude, and they are exten-

sively utilized in this study.

2.2 Data Location

Figure 2: Intersections ﬁtted with RSUs are shown here.

The intersections surround the vast campus of the Univer-

sity of Florida. The four roads forming a Trapezium have

25 intersections that are ﬁtted with RSUs.

In this section, we provide an overview of the region

from which the data is collected. The region is the

Trapezium surrounding the University of Florida in

Gainesville, as shown in the Figure 2. The Trapez-

ium is formed by four major roads: 34

Street, West

University Avenue, 13

Street, and Archer Road,

forming its Western, Northern, Eastern, and Southern

boundaries, respectively. These roads are among the

busiest in Gainesville, with high levels of pedestrian

trafﬁc.

The data for this study was collected from 25 sig-

nalized intersections, primarily from vehicles owned

by the City of Gainesville and the University of

Florida, including transit buses and service vehicles.

The data collection took place from April 2021 to July

2022 and involved 50-60 vehicles. Although the lim-

ited deployment of OBUs at the time of the study re-

sulted in sparse data, the methodology developed in

this paper provides a foundation for future connected

vehicle applications as their deployment increases.

The distribution of unique identiﬁers in our mes-

sage logs is depicted in Figure 3. Our logs contained

approximately 200, 000 unique identiﬁers for Con-

nected Vehicles. It’s noteworthy that while a single

physical vehicle may have multiple identiﬁers for pri-

vacy purposes, the relative number of unique identi-

ﬁers is indicative of the relative number of actual ve-

VEHITS 2023 - 9th International Conference on Vehicle Technology and Intelligent Transport Systems

422

Figure 3: Counts of unique identiﬁers of Connected Vehi-

cles tracked between the months of April 2021 to July 2022,

across various intersections.

Figure 4: Various messages exchanged between the users

and the RSUs. The data collection software resides in the

server, which in our case is housed at the City of Gainesville

Laboratory.

hicles at the intersection, especially considering that

the speed limits at these intersections are similar. As

such, we can conclude, for example, that intersections

7353 and 7354 have a higher presence of OBUs com-

pared to intersection 5057.

2.3 Data Collection

This section details the data collection process used

in our study. The data was collected from Road Side

Units (RSUs) in the region surrounding the Univer-

sity of Florida, Gainesville. There are two methods

for connecting to the RSUs: through the RSU web ap-

plication (provided by the Original Equipment Man-

ufacturer) or using Java WebSocket APIs. We chose

the latter method for data collection.

To receive messages from the RSUs, a Java appli-

cation was written to subscribe to the messages us-

ing websocket. In the process of collecting data from

the RSUs, the Java application sent the subscription

command and executed the ”!wait” command to keep

the websocket connection alive and receive messages

from the RSUs.

Once a message was received, it was written to the

standard output and forwarded to a Python (version

3.7) interpreter through the Python subprocess PIPE.

In addition, the control message was also written to

the standard output for the Python interpreter to mon-

itor the connection status.

The Python interpreter, upon receiving the mes-

sages, ﬁrst extracted the byte-encoded XML message

from the raw message wave and decoded it. The syn-

tax and semantics of the XML message can be found

in the SAE J2735 standard (Kenney, 2011). In ad-

dition to the BSM, we received other types of mes-

sages from the RSUs such as PSM (Personal Safety

Message), SPAT (Signal Phase And Timing Mes-

sage), TIM (Traveler Information Message), MAP

(Map Data Message), SSM (Signal Status Message),

and SRM (Signal Request Message). Figure 4 shows

the various DSRC message senders and recipients in

a CV2X infrastructure. The data collector and proces-

sor codes reside in the Server.

Finally, the messages received from the RSUs

were stored in an Amazon Web Services (AWS) S3

bucket

. AWS S3 provides high data durability and

availability, close to 100%, and the data was orga-

nized by city, year, month, day, and intersection prop-

erties for quick access. All data stored in AWS S3 is

encrypted by default, and the S3 API endpoints sup-

port Secure Sockets Layer/Transport Layer Security

for encrypting data in transit.

2.4 Data Processing

The initial stage of the data processing pipeline in-

volves downloading the RSU logs from AWS S3

bucket and using the data parsing module to read

the logs and establish a spatio-temporal trajectory

database. This is achieved by using Python XML-

parsing library LXML 4.9.0

and Pandas 1.4

, to

parse the BSM messages. A trajectory database is

built including all available tracks, which is stored

as dictionaries indexed by vehicle IDs and containing

time-stamped latitude and longitude of the vehicles’

paths.

However, the stored trajectories often contain

noise due to factors such as surrounding buildings

and atmospheric conditions. To address this, trajec-

tory smoothening algorithms are developed using ve-

hicle motion data, based on windowed moving aver-

ages and interpolation techniques to ﬁll any gaps in

the data. The goal of these algorithms is to provide

a smoother representation of the raw, noisy trajecto-

ries. The original data is obtained at one decisecond

resolution. We smoothen the data by using a moving

average with a window of ﬁve deciseconds.

aws.amazon.com/

www.lxml.de

pandas.pydata.org/

Using DSRC Road-Side Unit Data to Derive Braking Behavior

423

Figure 5: Counts of unique identiﬁers of Connected

Vehicles tracked showing braking behavior of concern

(MILD/HARD/EXTREME) between the months of April

2021 to July 2022, across various intersections. Compared

to Figure 3, we can see that there is a correlation between

the intersections with most vehicle identiﬁer detections, and

the intersections having more braking events of concern.

2.5 Methodology for Detecting Braking

Behavior

A near-miss trafﬁc event can be described as an event

where a signiﬁcantly higher risk is involved in the

interactions between road users than in an ordinary

case. One of the key indicators of a near-miss is un-

expected sudden braking by a vehicle. The sudden

braking may be in response to another road user (such

as another vehicle or pedestrian) coming too close. It

could also be in response to a static hazard such as a

pothole. While both cases are undesirable, they can be

distinguished by their frequency of occurrence. Sup-

pose multiple vehicles show a similar braking pattern

in the same region, around the same time, which is

sustained for at least several minutes. In that case, it

can be reasonably inferred that the braking is in re-

sponse to a persistent static hazard. However, if such

braking only occurs sparsely and irregularly, it may

be of concern. In this analysis, we focus on detecting

braking behavior that may be of concern.

We obtain braking behavior from GPS (Global Po-

sitioning System) trajectory data from Basic Safety

Messages (BSMs), by looking at the deceleration pro-

ﬁles. Deceleration is found by taking the second

difference of the smoothened trajectory, as the vehi-

cles were not digitally interfaced with the OBU and

were thus not able to capture actual braking. Numpy

1.23.0

Python library is used to compute the sec-

ond difference (i.e. second derivative). Haversine for-

mula

is used to convert the latitude/longitude coor-

dinates to meters.

We characterize braking behavior based on decel-

eration thresholds (Baldanzini et al., 2016; Staputis

and

Zuraulis, 2023). In this analysis, we use the ac-

celeration due to gravity “g” (9.8 m/s

) as a unit, and

mark:

www.numpy.org

rosettacode.org/wiki/Haversine formula

• 0g to -0.35g as SAFE Braking

• -0.35g to -0.47g as MILD Braking

• -0.47g to -0.62g as HARD Braking

• Beyond –0.62g is EXTREME Braking

We only present results for MILD, HARD, and EX-

TREME braking in this paper. SAFE braking is not

of concern.

3 EXPERIMENTS

In this section, we present the outcomes of imple-

menting our approach on a per-intersection basis and

also take a comprehensive view that encompasses all

the intersections.

3.1 Intersection Braking Behavior

Figure 5 shows all MILD/HARD/EXTREME com-

bined events per intersection. We break up this in-

formation based on braking severity per instance in

Figure 6. Note that it is possible that a single iden-

tiﬁer may contain multiple braking instances. An in-

stance here lasts for 1 decisecond, which is the time

resolution of Basic Safety Messages. Thus, this plot

also captures the duration (in deciseconds) of the class

of braking behavior seen. Also, in the plots shown,

blank rectangles indicate that exactly zero such brak-

ing instances were found in the given time bucket.

3.1.1 Temporal Behavior

Our study on braking events includes an analysis of

the occurrences based on the time of the day and the

day of the week. This analysis is represented in the

form of three plots, each one for a speciﬁc level of

severity - mild, hard, and extreme braking events.

The plots in Figures 7, 8, 9 and 10 depict that the

majority of the braking events of concern occur dur-

ing the weekdays and during the mid-day hours. This

could be due to several reasons such as the presence of

a higher number of pedestrians during the lunch hour,

but also the increased presence of Connected Vehicles

on the road at that time.

3.1.2 Spatial Behavior

In order to better understand the braking behavior

of vehicles, we segment the data based on the dis-

tance from the intersection where the braking event

occurred. To do this, we divide the distance into three

categories: 0-20 meters, 20-40 meters, and 40 meters

VEHITS 2023 - 9th International Conference on Vehicle Technology and Intelligent Transport Systems

424

Figure 6: Counts of unique identiﬁers of Connected Vehicles tracked showing braking behavior of concern

(MILD/HARD/EXTREME) between the months of April 2021 to July 2022, across various intersections. Compared to

Figure 3, we can see that the intersections with most vehicle identiﬁer detections, also have more braking events of concern. .

Figure 7: Heatmap showing the ﬂow of all Connected Ve-

hicles across the week, system-wide.

Figure 8: Heatmap showing MILD braking events of all

Connected Vehicles across the week, system-wide.

and above. The ﬁrst category, 0-20 meters, is partic-

ularly important as it captures the braking events that

occur near the intersection. This is likely to indicate

braking in response to pedestrian crossings, which are

common near intersections. On the other hand, the

next category of 20-40 meters likely indicates brak-

ing due to a vehicle’s response to last-minute lane

changes, as vehicles prepare to make a turn at the in-

tersection.

As an illustration of the analysis, Figure 11 shows

the different levels of braking behavior against the

distance from the intersection for intersection 5050.

From the plot, it can be observed that most of the

braking events of concern occurred beyond 40 meters

from the intersection, which is likely due to a variety

of factors such as road conditions, vehicle speed, and

driver behavior.

Figure 9: Heatmap showing HARD braking events of all

Connected Vehicles across the week, system-wide.

Figure 10: Heatmap showing EXTREME braking events of

all Connected Vehicles across the week, system-wide.

Figure 11: Plots showing the severity of braking against the

distance from the center of the nearest intersection.

3.1.3 Speed Characteristics

Our next analysis involves examining the relationship

between braking behavior and vehicle speed. To do

Using DSRC Road-Side Unit Data to Derive Braking Behavior

425

Figure 12: Plots showing the severity of braking against the

speed at which the braking took place.

this, we plot the braking events against the speed of

the vehicle at the time of the event. We divide the ve-

hicle speeds into three segments: 0-20 mph (0-32.2

kmph), 20-40 mph (32.2-64.4 kmph), and 40 mph

(64.4+ kmph) and above.

As shown in Figure 12, for intersection 5050, we

observe that a signiﬁcant number of hard and ex-

treme braking events occur at relatively low speeds.

This information is useful in understanding the sever-

ity of braking behavior and its relationship with the

speed of the vehicle. It could indicate that drivers are

more likely to brake harshly when they are traveling at

lower speeds, which could be due to a variety of fac-

tors, such as sudden trafﬁc changes, road conditions,

or distractions. This observation also may be due to

the fact that most vehicles approach the intersection

at lower speeds. The future development of the dash-

board will allow for combined derived metrics such

as such as events normalized by vehicle ﬂow (based

on nearby loop detector data) to explore such insights.

These ﬁndings can help inform future studies on brak-

ing behavior and road safety.

3.2 System-Wide Braking Behavior

We also developed an interactive visualization tool to

provide a more in-depth analysis of braking behav-

ior. This tool, which runs in a web browser, presents

the braking events of concern (i.e., MILD, HARD,

and EXTREME events) on a map in the form of a

heatmap, providing a bird’s-eye view of the distribu-

tion of such events. The user can interact with the

tool to view speciﬁc regions and zoom in to inspect

particular intersections or stretches of roads.

This tool makes it easier to identify hotspots

where braking events of concern are more prevalent.

As an example, Figures 13 and 14 show the overall

view of the braking behavior and a zoomed-in view

of an intersection approach, respectively. From the

visualization, one can see that some stretches of West

University Avenue and Archer Road have fewer brak-

ing events of concern compared to other areas.

Figure 13: Interactive web browser-based application con-

taining a plot of all braking events of concern.

Figure 14: Zoomed-in view of an intersection approach in

the interactive application. It allows the user to inspect in-

tersection approaches. Here the East-bound approach of the

intersection of Archer Road and 34

Street is shown.

4 CONCLUSION AND FUTURE

WORK

The recent advancements in ITS have led to the exten-

sive deployment of various real-time sensing and data

collection systems for public trafﬁc agencies. Our

work focuses on using the Connected Vehicle (CV)

data available through V2X (Vehicle-to-everything)

communication systems like DSRC and CV2X. The

data collected by these technologies is useful for

studying the braking dynamics of vehicles as they ap-

proach, enter, and exit intersections.

We presented a methodology for analyzing brak-

ing behavior in connected vehicle data. The study

utilized a spatio-temporal trajectory database created

from RSU logs and BSM messages, which were

parsed using XML-parsing libraries and Python. The

trajectories were smoothened using windowed mov-

ing averages and interpolation.

We presented a study, for which we collected data

in real-time from 25 intersections surrounding the

University of Florida and stored the data on the cloud.

The data was processed ofﬂine and braking behavior

was analyzed on a per-intersection basis as well as

from an overall perspective, taking into account the

time-of-day and day-of-week, distance from the inter-

section, and speed at which the vehicle was traveling.

The results of the analysis were presented in various

plots and an interactive visualization tool was devel-

oped to allow the user to get a better understanding of

VEHITS 2023 - 9th International Conference on Vehicle Technology and Intelligent Transport Systems

426

braking behavior in the network. Overall, the method-

ology provides a valuable tool for analyzing braking

behavior and identifying potential hotspots for safety

concerns.

During our investigation into the real-world dy-

namics and potential of data analytics in an urban traf-

ﬁc grid, we encountered the challenge of sparse data

caused by the limited deployment of Connected Ve-

hicle (V2X) technology. Nevertheless, this research

sheds light on the enormous potential of utilizing

Connected Vehicle data to optimize trafﬁc ﬂow and

improve road safety, opening up a promising future

for transportation and trafﬁc management.

REFERENCES

Baldanzini, N., Huertas Leyva, P., Savino, G., and Pierini,

M. (2016). Rider behavioral patterns in braking

manoeuvres. Transportation Research Procedia,

14:4374–4383.

Banerjee, T., Chen, K., Almaraz, A., Sengupta, R., Kar-

nati, Y., Grame, B., Posadas, E., Poddar, S., Schenck,

R., Dilmore, J., Srinivasan, S., Rangarajan, A., and

Ranka, S. (2022). A modern intersection data ana-

lytics system for pedestrian and vehicular safety. In

2022 IEEE 25th International Conference on Intel-

ligent Transportation Systems (ITSC), pages 3117–

3124.

Biddlestone, S., Redmill, K., Miucic, R., and Ozguner,

(2012). An integrated 802.11p wave dsrc and vehicle

trafﬁc simulator with experimentally validated urban

(los and nlos) propagation models. IEEE Transactions

on Intelligent Transportation Systems, 13(4):1792–

1802.

D’Andrea, E. and Marcelloni, F. (2017). Detection of traf-

ﬁc congestion and incidents from gps trace analysis.

Expert Systems with Applications, 73:43–56.

DSRC (2020). DSRC Technical Committee On-Board Sys-

tem Requirements for V2V Safety Communications.

FDOT. Intersection operations and safety, https://www.fdot

.gov/trafﬁc/trafﬁcservices/intersection-operations.sh

tm.

FHWA. Intersection safety, https://highways.dot.gov/safet

y/intersection-safety.

Haas, C., Mahmassani, H., Khoury, J., Haynes, M., Rioux,

T., and Logman, H. (2001). Evaluation of automatic

vehicle identiﬁcation for san antonio’s transguide for

incident detection and advanced traveler information

systems.

Hofmann-Wellenhof, B., Lichtenegger, H., and Collins, J.

(2012). Global positioning system: theory and prac-

tice. Springer Science & Business Media.

Hsu, M. and Shih, A. (2015). A trafﬁc signal control mech-

anism in a connected vehicle environment. In 2015

IEEE International Conference on Consumer Elec-

tronics - Taiwan, pages 256–257.

Huang, X., He, P., Rangarajan, A., and Ranka, S. (2020). In-

telligent intersection: Two-stream convolutional net-

works for real-time near-accident detection in trafﬁc

video. ACM Transactions on Spatial Algorithms and

Systems (TSAS), 6(2):1–28.

Kan, Z., Tang, L., Kwan, M.-P., Ren, C., Liu, D., and Li,

Q. (2019). Trafﬁc congestion analysis at the turn level

using taxis’ gps trajectory data. Computers, Environ-

ment and Urban Systems, 74:229–243.

Kenney, J. B. (2011). Dedicated short-range communica-

tions (dsrc) standards in the united states. Proceedings

of the IEEE, 99(7):1162–1182.

Lamas, J., Castro-Castro, P.-M., Dapena, A., and Vazquez-

Araujo, F. (2015). Vehicle classiﬁcation using the dis-

crete fourier transform with trafﬁc inductive sensors.

Sensors, 15.

Li, P., Abdel-Aty, M., and Yuan, J. (2021). Using bus crit-

ical driving events as surrogate safety measures for

pedestrian and bicycle crashes based on gps trajectory

data. Accident Analysis & Prevention, 150:105924.

Liu, Y., Jia, R., Ye, J., and Qu, X. (2022). How ma-

chine learning informs ride-hailing services: A sur-

vey. Communications in Transportation Research,

2:100075.

Mandava, S., Boriboonsomsin, K., and Barth, M. (2009).

Arterial velocity planning based on trafﬁc signal in-

formation under light trafﬁc conditions. In 2009 12th

International IEEE Conference on Intelligent Trans-

portation Systems, pages 1–6.

Staputis, A. and

Zuraulis, V. (2023). Research of efﬁciency

of anti-lock braking system during emergency corner-

ing manoeuver. In TRANSBALTICA XIII: Transporta-

tion Science and Technology: Proceedings of the 13th

International Conference TRANSBALTICA, Septem-

ber 15-16, 2022, Vilnius, Lithuania, pages 224–235.

Springer.

Sun, S., Chen, J., and Sun, J. (2019). Trafﬁc conges-

tion prediction based on gps trajectory data. In-

ternational Journal of Distributed Sensor Networks,

15(5):1550147719847440.

Wang, X., Sanner, S., and Abdulhai, B. (2022). A criti-

cal review of trafﬁc signal control and a novel uniﬁed

view of reinforcement learning and model predictive

control approaches for adaptive trafﬁc signal control.

arXiv preprint arXiv:2211.14426.

Wang, Y., Cao, J., Li, W., and Gu, T. (2016). Mining traf-

ﬁc congestion correlation between road segments on

gps trajectories. In 2016 IEEE International Confer-

ence on Smart Computing (SMARTCOMP), pages 1–

8. IEEE.

Wolf, J., Bachman, W., Oliveira, M., Auld, J., Mohamma-

dian, A. K., Vovsha, P., and Zmud, J. (2014). Apply-

ing gps data to understand travel behavior, volume ii:

guidelines. Technical report.

Yong-chuan, Z., Xiao-qing, Z., Zhen-ting, C., et al. (2011).

Trafﬁc congestion detection based on gps ﬂoating-car

data. Procedia Engineering, 15:5541–5546.

Zhao, D., Dai, Y., and Zhang, Z. (2012). Computational

Intelligence in Urban Trafﬁc Signal Control: A Sur-

vey. IEEE Transactions on Systems, Man, and Cyber-

netics, Part C (Applications and Reviews), 42(4):485–

494.

Using DSRC Road-Side Unit Data to Derive Braking Behavior

427