Article citation information:
Le, K.G. Enhancing
traffic safety: a comprehensive approach through real-time data and intelligent
transportation systems. Scientific Journal of Silesian University of Technology. Series
Transport. 2024, 122,
129-149. ISSN: 0209-3324. DOI: https://doi.org/10.20858/sjsutst.2024.122.8.
Khanh Giang LE[1]
ENHANCING TRAFFIC SAFETY: A COMPREHENSIVE APPROACH THROUGH REAL-TIME
DATA AND INTELLIGENT TRANSPORTATION SYSTEMS
Summary. Today,
traffic accidents are still a difficult and urgent problem for many countries
around the world. Traffic accidents on highways are often more serious than
accidents on urban roads. Therefore, disseminating emergency information and
creating immediate connections with road users is key to rescuing passengers
and reducing congestion. Thus, this study applies data fusion and data mining
techniques to analyze travel time and valuable information about traffic
accidents based on the real-time data collected from On-Board Unit installed in
vehicles. The results show that this important information is the vital
database to analyze traffic conditions and safety factors, thereby developing a
smart traffic information platform. This result enables traffic managers to
provide real-time traffic information or forecasts of congestion and traffic
accidents to road users. This helps limit congestion and serious accidents on
the Highway.
Keywords: traffic
accident, highway, big data, data mining, intelligent transportation system
1.
INTRODUCTION
The
implementation of eTag sensors on the Freeway in Taiwan serves a dual purpose:
facilitating toll calculations and generating valuable traffic-related data.
This data encompasses essential metrics such as travel times, traffic volumes,
speeds, and the tolls necessary to traverse different segments between ramps.
The authority and road users can readily access and reference this information,
as it is promptly stored on servers [1].
In
the event of an accident on the Freeway, the conventional reporting mechanism
involves telephonic communication with the traffic control center. The
accident's location is determined based on milestones and verified through
traffic cameras before dispatching rescue teams. Subsequently, the traffic
control center disseminates accident information to road users through
broadcasting and online platforms [2]. Despite this established process, this
study proposes leveraging smart technologies to streamline the processing and
dissemination of traffic information.
One
notable limitation of the existing eTag system is its inability to access
real-time information about road conditions if the distance between two eTag
sensor gantries is too extensive. This limitation often hampers road users'
ability to make informed decisions, such as changing routes or adjusting their
driving speed. The eTag system may not capture comprehensive traffic
information for a particular section between sensor gantries, creating an
information gap [3].
To address
this gap, the study suggests integrating the capabilities of the On-Board Unit
(OBU) with the eTag system. When a vehicle equipped with the OBU passes through
an eTag sensor gantry, data transmission occurs for comparative analysis,
ensuring data integrity and mitigating the information gap issue. Beyond
managing the vehicle fleet, the OBU contributes real-time information about the
vehicle and road conditions. This data is relayed in real-time to the traffic
information platform specifically developed for this study.
The
integration of real-time information from the OBU into both the eTag database
and the larger big data repository, which includes data from both the eTag and
the OBU, enables comprehensive data mining and visualized analysis. The study
envisions employing various models and data analyses to extract valuable
traffic information and safety factors. The ultimate goal is to utilize
real-time cloud computing to enhance the completeness of the traffic
information service, as illustrated in Fig. 1.
Fig. 1. Establishment of study
foundation
By
integrating data from eTag and the OBU, this study seeks to optimize the
potential for data mining techniques. The collected and analyzed data about the
Freeway will be instrumental in establishing relevant models, and the big data
analysis model will further uncover useful traffic information. The overarching
aim is to advance the capabilities of the traffic information service through a
more thorough integration of eTag and OBU data.
The
remainders of the paper are arranged as follows. Section 2 shows literature
review. Section 3 presents study data and methods. Section 4 illustrates data
analysis and discussions. In Section 5, conclusions are displayed.
2.
LITERATURE REVIEW
Passive delays in the delivery of
emergency or accident information can prevent neighboring road users from
obtaining real-time updates. Real-time communication can be achieved by using
the OBU and mobile devices such as Vehicle Information and
Communication System (VICS) in Japan to send information to
neighboring vehicles quickly. Numerous research works suggest that trip time
can be predicted using eTag data. In order to enhance data integrity and match
forecasts with factual circumstances, several researches integrate vehicle
detector or OBU data for thorough examination [4]. Certain researches use the
OBU to gather more accurate data for fleet management and monitoring.
Predictions and data analysis are rarely integrated into actual applications
inside the domestic traffic information infrastructure, nevertheless.
Various factors
contribute to traffic safety, including individuals, vehicles, and the
surrounding environment [5]. The OBU has the capability to monitor both
individuals and their vehicles. Additionally, the author aim to utilize data
from road sensors for analyzing and predicting vehicle flow. However, in the
event of an accident, most road users typically receive information passively
from the police or broadcasting. Predicting and preventing factors that could
lead to accidents and understanding road conditions are crucial.
Providing road
users with reminders before emergencies can significantly reduce the severity
and occurrence of accidents. Beyond understanding accident causes, the
knowledge gained can be employed to develop preventive strategies for accident
occurrence. The accident forewarning acts as a behavioural indicator between
the core indicator and performance indicator, aiming to enhance traffic safety.
Utilizing the accident forewarning as an intermediary indicator can improve the
accuracy of assessments and help dissect the complex causes of accidents [6,
7]. The following section will delve into the development of the accident forewarning
and the associated challenges encountered thus far.
There are a few
previous studies that have addressed this issue, specifically as follows:
Swiftly and accurately identifying sections on the Freeway prone to accidents
[8]-[10]. Quickly and precisely evaluating the impact of strategies aimed at
enhancing traffic safety, as demonstrated by [11]. Serving as the foundation
for assessing safe driving behaviors [12]. Developing driveway security devices
and measures for traffic safety [13]-[15]. Allowing the insurance market to
implement price discrimination based on driver safety [16, 17]. Conducting
further exploration into driving behaviors, vehicle designs, and their
interactions with the road environment.
Currently, the
traffic information service platform lacks complete integration of eTag data
analysis and application in Taiwan. Also, it does not facilitate the
transmission of emergency information. The strength of this study lies in
combining eTag data with the OBU data to construct a comprehensive Big Data
database. This integrated data undergoes classification using data mining
techniques, with relevant models applied to extract valuable traffic
information. The study envisions an advanced traffic platform offering a
holistic traffic information service, covering general road conditions and
accidents. Additionally, the application of accident forewarning is expected to
improve and expand with ongoing advancements in computing power and wireless
communications technology. The accuracy of information analysis relies on a
multitude of data sources, emphasizing the importance of the precision of these
relevant sources for accurate information assessments.
In summarizing
the literature on traffic information and safety considerations, the focal
points in traffic information services are the accuracy of information
transmission. Nevertheless, the integrity of data also plays a significant role
in influencing analysis and prediction outcomes. Challenges arise when the
distance between sensors is extensive, hindering the acquisition of information
on a Freeway section and resulting in insufficient data integrity.
In this
investigation, the objective is for the OBU to gather data on vehicle flow,
integrating eTag data. Subsequently, these data will undergo conversion,
filtering, compensation, and fusion through various data processing procedures,
including data mining techniques. The aim is to predict travel time, extract
valuable traffic information, and establish a comprehensive set of analysis
procedures, methods, and mechanisms. Subsequently, the traffic information
platform can provide real-time updates to road users, enabling them to avoid
congested Freeway sections, reduce travel time, and enhance traffic speed.
Additionally, this information can be referenced and applied by traffic
management authorities.
The main tasks of
this study encompass: First, providing real-time travel information and
predictions. Second, offering accident prediction information. Third, supplying
emergency relief information. Final, integrating information services for
safety and rescue within the smart traffic information platform.
3.
DATA AND METHOD
3.1.
Data Collection
The Traffic Data Collection Support
System (TDCS) on the Freeway gathers eTag data, with the original data being
publicly accessible. There are six file types, detailed in Tab. 1, with the M06
file type specifying the original data. The data containing the initial path
tends to be the most comprehensive, boasting the largest data volume. Daily, an
average of 4 million data entries are processed, necessitating a daily file
size of about 1GB or potentially more. Given the continuous 24-hour collection
of vehicle flow data on the Freeway, the system may accumulate over 1 million
data entries per hour, presenting a challenge to existing data processing
methods concerning current data analysis techniques, efficiencies, and software
and hardware capabilities.
Besides eTag data, this study also
uses data from OBU for analysis. The OBU has the capability to gather detailed
information about the vehicle, including basic data, operational status, and
precise location. The data formats are specified in Tab. 2. When merging data
from the OBU and eTag, challenges such as inaccuracies, discrepancies, noises,
fragmentation, or irrelevant information may appear. Consequently, there may be
a need for data processing and normalization.
The integration and extension of
functions of the OBU applied in this study is presented in Fig. 2. This
research uses the data to build the big database. During the combination of
data from the OBU and eTag, issues such as incorrectness, inconsistency,
noises, fragmentation or irrelevance of data may arise. Thus, data processing
and normalization may be required.
Tab. 1
The file kinds of eTag data
|
No |
File name |
Description |
|
1 |
TDCS_M03A_YYYYMMDD_hhmmss.csv |
Traffic volume |
|
2 |
TDCS_M04A_YYYYMMDD_hhmmss.csv |
Average travel time |
|
3 |
TDCS_M05A_YYYYMMDD_hhmmss.csv |
Average driving speed |
|
4 |
TDCS_M06A_YYYYMMDD_hhmmss.csv |
OD raw data (daily) |
|
5 |
TDCS_M07A_YYYYMMDD_hhmmss.csv |
OD average length (daily) |
|
6 |
TDCS_M08A_YYYYMMDD_hhmmss.csv |
OD average traffic volume
(daily) |
OD: Origin-Destination.
Tab. 2 Data
formats from the OBU
|
Vehicle route data |
||||||
|
Date: 20160802000000 ~
20160802164759 |
||||||
|
Vehicle
No.: CAR2 |
||||||
|
Vehicle
No. |
OBU |
Driver |
GPS time |
Longitude |
Latitude |
Location |
|
CAR2 |
6860422853 |
DR-2 |
2016/08/02 00:02:33 |
120.2429 |
22.8794 |
Dongfang Rd., Hunei District, Kaohsiung City |
|
CAR2 |
6860422853 |
DR-2 |
2016/08/02 00:03:01 |
120.2429 |
22.8794 |
Dongfang Rd., Hunei District, Kaohsiung City |
|
CAR2 |
6860422853 |
DR-2 |
2016/08/02 00:05:01 |
120.2429 |
22.8794 |
Dongfang Rd., Hunei District, Kaohsiung City |
|
CAR2 |
6860422853 |
DR-2 |
2016/08/02 00:07:01 |
120.2429 |
22.8794 |
Dongfang Rd., Hunei District, Kaohsiung City |
|
CAR2 |
6860422853 |
DR-2 |
2016/08/02 00:07:33 |
120.2429 |
22.8794 |
Dongfang Rd., Hunei District, Kaohsiung City |
|
CAR2 |
6860422853 |
DR-2 |
2016/08/02 00:09:01 |
120.2429 |
22.8794 |
Dongfang Rd., Hunei District, Kaohsiung City |
3.2. Method
The Intelligent
Transport System (ITS) serves as an application designed for the coordination
of people, roadways, and vehicles. Its purpose is to provide instantaneous
information, thereby improving the security, efficiency, and convenience of the
transport system while mitigating the environmental impact of traffic. Cloud
technologies are employed in storing, disseminating, and processing the
substantial volume of data derived from traffic information. The positive
outcomes can only be achieved through comprehensive data processing and
strategic managing, as illustrated in Fig. 3.
Fig. 2. Functions of the OBU
Fig. 3. The traffic information platform diagram
3.2.1. Data
Pre-processing
As we delve into
the intricate process of thoroughly analyzing information from ITS, it becomes
evident that key stages such as data collection, information transmission,
integration, and disclosure necessitate the establishment of supporting
mechanisms. In this study, the compliance analysis model takes center stage,
meticulously screening and analyzing conditional data, gradually shaping them
into valuable information. This intricate process is distilled into five
essential steps: data filtering, conversion, compensation, fusion, and
extension, as shown in Fig. 4.
Fig. 4. Data preprocessing
process
v Data filtering: The data
may contain abnormal or extreme values, possibly due to equipment
abnormalities, specific driving behaviors, or inherent vehicle properties.
Excessive presence of such data may impact the accuracy of subsequent
algorithms. Therefore, it is essential to filter these data first.
v Data conversion: Initial
frequencies and content of data collected by each device may vary.
Consequently, data conversions can be performed to ensure uniform usability in
subsequent applications. To identify missing data, data collection over a
specific time interval is necessary to confirm any gaps.
v Data compensation: Data may
be missing due to device abnormalities post-collection, or defects may emerge
after data filtering. In such cases, the algorithm, prediction model, and
historical data can be employed to compensate for any missing data on the
device at any given time.
v Data fusion: When a section
of the Freeway lacks data, data extension will be conducted for compensation.
When data from the OBU at a specific time becomes representative for the
Freeway section, it will be fused with the prediction data. According to the
weighted fusion method outlined by the Institute of Transportation under the
Ministry of Transportation and Communications (MOTC) [18], the OBU-collected
data, along with instantaneous speed and predicted traffic speed, will be
fused. Both values are used because the OBU data may be less representative
when the number of vehicles is lower. The weighted method helps account for the
properties of vehicle detection (VD) data. When there is sufficient data from
the OBU, the data will be closer to true values, and their weights may approach
100 percent, indicating that the prediction value can be completely spared.
Data fusion will be carried out through model construction, and the general
formula is shown as follows:
Where:
v Data extension: Data
extension may occur within a section on the Freeway or between sections on the Freeway.
Within a Freeway section, data extension may be conducted to expand the data
range when it does not align with the section's length. This extension is
seamless, and potential errors are accounted for. Additionally, data extension
can occur between Freeway sections to extend data from a fully conforming
section to one without any data. It is assumed that vehicle flow may exhibit a
directional extension, enabling data from the upstream Freeway section to be
extended for predicting data in the downstream section of the Freeway.
3.2.2. Traffic Information
Platform Development
This research
primarily explores the realm of Big Data, focusing on the development of a
real-time dynamic information system dedicated to the instantaneous computation
of Big Data. As the volume of historical traffic data rises and traffic systems
grow in complexity, it becomes crucial for the platform to possess the capacity
to manage Big Data effectively. Consequently, the central emphasis of this
study lies in devising methods to process, analyze, synthesize, harness,
utilize, disseminate, and store large datasets derived from real-time traffic
information swiftly and efficiently. The goal is to construct a comprehensive
traffic information platform tailored for highways, serving both general and
emergency purposes. The architectural layout of the platform is illustrated in
Fig. 5.
Fig. 5. Traffic information
platform
This study aims
to create a smart traffic information platform using Information and
Communications Technology (ICT) and Data Mining for real-time decision-making.
The system analyzes eTag and OBU data to generate Big Data, offering insights
like vehicle flow, travel time, trip count, vehicle types, and more.
Additionally, it manages traffic information, predicts congestion, and provides
visualization of results.
The platform
primarily offers two types of traffic information services. The first type is a
smart and secure navigation service primarily catering to road users, while the
second type is an emergency information service providing advanced warnings to
road users. To meet the evolving traffic needs of the Freeway in the future,
this research establishes a prediction model and develops the platform with
functions including: intelligent and safe navigation service and emergency
information service. The platform will be equipped with the following
functions:
(1) Smart and
secure navigation service:
a. To inquire about road conditions in
real-time.
b. To identify easily congested sections on
the Freeway, issue forewarning about congestion intervals, and predict travel
times on those sections.
c. To provide suggested routes and real-time
guidance to road users.
d. To offer historical data about vehicle
flow information to relevant authorities for analysis and applications.
(2) Emergency
information service:
a. To provide rescue services after
accidents occur.
b. To issue emergency alerts and push
notifications to road users to control of potential accident risks.
c. To offer suggested routes and guidance to
enhance smooth traffic and reduce accidents. As the Freeway section is a closed
road with no alternative routes for vehicle evacuation after an accident, this
may lead to a serious blockage and unforeseen impacts on road capacity.
The information
platform will be capable of forecasting easily congested sections, planning
alternative trips during congestion, and reporting accidents on Freeway
sections to help road users make informed decisions. For instance, road users
may choose alternative routes based on comparisons of travel time, average
speed, and congestion conditions.
4. DATA ANALYSIS AND DISCUSSION
4.1.
Occurrence of Accidents
Central Taiwan is
served by four Freeways, specifically Freeways No. 1, 3, 4, and 6. According to
Tab. 3, there were a total of 3,727 accidents in 2016, with 2,270 occurring on
Freeway No. 1. Accidents on Freeway No. 1 constituted 61 percent of all
accidents on Central Taiwan's Freeways in 2016, surpassing half (50 percent) of
the total accidents on these Freeways. Consequently, Freeway No. 1 is chosen as
the focus of this study. While the central section of Freeway No. 1 spans 158
kilometers, accidents are not uniformly distributed; instead, they are
concentrated in specific sections. This study targets segments with a
relatively higher accident frequency. The research target is set from the
Taichung system to the Puyan system (kilometer range 165-207), is shown in Tab.
4, as it encompasses 1,386 accidents, representing more than half (50 percent)
of the total accidents on the central section of Freeway No. 1.
Tab. 3 The
number of accidents on Freeways in 2016
|
Locations |
Directions |
|||||
|
North |
West |
East |
North-South |
South |
Total |
|
|
Freeway No. 3 |
589 |
0 |
2 |
3 |
562 |
1156 |
|
Freeway No. 4 |
0 |
45 |
29 |
0 |
0 |
74 |
|
Freeway No. 6 |
0 |
156 |
71 |
0 |
0 |
227 |
|
Freeway No. 1 |
1043 |
0 |
0 |
0 |
1227 |
2270 |
|
Total |
1632 |
201 |
102 |
3 |
1789 |
3727 |
4.1.1.
Variable Screening
The initial screening of variables is conducted through two approaches.
Initially, the author refers to literature to examine various accident types,
considering the impact of congestion, and utilizes variables such as backup
(induced or not induced) and the number of driveways occupied for analysis.
Subsequently, the author delves into the original data. Finally, a preliminary
screening of variables related to accident occurrence is performed, as depicted
in Fig. 6.
Tab. 4 The
number of accidents on Freeway No. 1 (from Mileage 99 to 257k)
|
Interchanges |
Mileage |
North |
South |
Total |
|
Hsinchu system-Toufen |
99-110k |
35 |
13 |
48 |
|
Toufen-Touwu |
110-125k |
31 |
33 |
64 |
|
Touwu-Miaoli |
125-132k |
22 |
8 |
30 |
|
Miaoli-Tongluo |
132-140k |
29 |
18 |
47 |
|
Tongluo-Sanyi |
140-150k |
26 |
39 |
65 |
|
Sanyi-Houli |
150-160k |
35 |
48 |
83 |
|
Houli-Taichung system |
160-165k |
13 |
53 |
66 |
|
|
165-168k |
46 |
98 |
144 |
|
Fongyuan-Daya |
168-174k |
39 |
282 |
321 |
|
Daya-Taichung |
174-178k |
77 |
106 |
183 |
|
Taichung-Nantun |
178-181k |
26 |
58 |
84 |
|
Nantun-Wang Tian |
181-189k |
54 |
34 |
88 |
|
Wang Tian-Changhua system |
189-192k |
13 |
23 |
36 |
|
Changhua system-Changhua |
192-198k |
168 |
80 |
248 |
|
Changhua-Puyan system |
198-207k |
109 |
173 |
282 |
|
Puyan system-Yuanlin |
207-211k |
79 |
31 |
110 |
|
Yuanlin-Beidou |
211-220k |
97 |
39 |
136 |
|
Beidou-Hsilo |
220-230k |
65 |
29 |
94 |
|
Hsilo-Huwei |
230-235k |
33 |
22 |
55 |
|
Huwei-Dounan |
235-240k |
16 |
12 |
28 |
|
Dounan-Yunlin system |
240-243k |
5 |
8 |
13 |
|
Yunlin system-Dalin |
243-250k |
21 |
18 |
39 |
|
Dalin-Minsyong |
250-257k |
3 |
1 |
4 |
|
Total |
1043 |
1227 |
2270 |
Data Sources: Central Region Office, National Freeway Bureau, MOTC,
Taiwan, 2016 [18]
4.1.2. Data Analysis
The analysis
proceeds through the following steps: First, after excluding ineffective data,
the total number of accidents is reduced to 1,254. Second, types of accidents:
The original data categorizes accidents into rear-end accidents and others.
Third, identification of induced backup: The original data indicates whether
backup occurred or not. Finally, counting driveways occupied: The original data
classifies accidents based on the number of lanes occupied (one lane, two
lanes, and three or more lanes).
Fig. 6. Flowchart of variable screening
Preliminary
analysis results, outlined in Tab. 5, reveal that among the effective
accidents, 1,050 are rear-end accidents (approximately 80 percent), and 1,048
accidents induce backup in vehicle flow. Further analysis, as shown in Fig. 7
and Fig. 8, is conducted separately for situations with or without backup in
the vehicle flow. It is evident that, regardless of whether backup is induced,
the majority of rear-end collisions typically involve the occupation of only
one lane. This pattern is attributed to the nature of rear-end accidents, where
the following vehicle often fails to pay attention to the leading vehicle, and
instances of deviation from the driveway are relatively uncommon.
Tab. 5
Variable analysis for
rear-end accidents
|
Not Have Queuing Line |
|
|
1048
|
2
|
Data Sources: Central Region Office, National Freeway Bureau, MOTC,
Taiwan, 2016 [18]
As rear-end accidents accounted for
nearly 80 percent of the accidents, this study classifies special accident
types such as sideswipe collisions against the side rail and rollover as other
accidents with the analysis steps remaining the same as aforementioned. As
shown by Tab. 6, as in rear-end accidents, after the accident happens, queening
line in the vehicle flow may be induced for further analysis. From Fig. 9 and
Fig. 10, we can see that most accidents would only occupy one lane, but three
or more lanes may be occupied under special circumstances owing to specific
types of accidents.
Fig. 7. Statistics of
rear-end accidents where queuing is induced
Fig. 8. Statistics of
rear-end accidents where queuing is not induced
Tab. 6
Variable analysis for other
accidents
|
Have
Queuing Line |
Not
Have Queuing Line |
|
201 |
3 |
Data Sources: Central Region Office, National Freeway Bureau, MOTC,
Taiwan, 2016 [18]
4.2. Accident
Prediction
The objective of the intelligent
traffic information platform on the Freeway is to enhance the efficiency and
safety of the Freeway, as depicted in Fig. 11. In this segment, the author aims
to develop a predictive model for accidents on the Freeway by integrating
traffic information such as accident records, current vehicle flows, and
dynamic data on future vehicle fleet composition. Apart from providing
real-time alerts for potential accidents, the platform has the potential to
mitigate the risk of accidents for drivers. Furthermore, the information
platform can be customized for emergency rescue efforts, thereby minimizing the
time required to address accidents and subsequently diminishing the impact of
accidents on the Freeway's overall efficiency.
Fig. 9. Statistics of other
accidents where queening is induced
Fig. 10. Statistics of
other accidents where queening is not induced
Fig. 11. Enhancing Freeway
efficiency and safety with intelligent traffic information platform
This research focuses on the section
of the Freeway between two interchange roads, using Empirical Bayes Estimation
to calculate the total numbers and expected values of all accidents (Fig. 12),
rear-end accidents (Fig. 13) and sideswipe accidents (Fig. 14) on individual
sections of Freeway No.1. These findings will serve as a reference for
allocating emergency vehicles.
Fig. 12. Total number and
expected value of all accidents on each section of Freeway No.1
To deliver immediate accident alerts
to highway drivers, this research presents an accident prediction model
developed from the accident warning system. While conventional studies commonly
depend on dynamic traffic data, the distinct data essential for accident
warnings cannot be substituted with OBU information. Consequently, this
investigation is compelled to employ expressway vehicle flow data to initially
evaluate the probability of accidents under various traffic conditions on each
expressway segment. Subsequently, utilizing OBU data, the study identifies
distinct causes of accidents within varied traffic flows, validating accident
alerts tailored for different accident types.
Given the variation in accident
warnings for different accident types, this section will initially focus on
seeking warnings for rear-end accidents, as outlined in Tab. 7. The data
necessary for analyzing rear-end accident warnings involve the speeds,
longitudinal accelerations, and decelerations of both the leading and following
vehicles. However, this data might not be obtainable through the existing OBU.
While the OBU can gather information about the driver's position, vehicle
speed, and even lateral and longitudinal accelerations and decelerations, it
lacks the capability to collect data about the leading and following vehicles'
positions, speeds, and related accelerations and decelerations. To confirm the
accident warning, data from the driver's OBU alone is insufficient; information
from surrounding vehicles is also essential. This necessity prompts the use of
alternative methods to refine the accident warning system on the Freeway.
Fig. 13. Total number and
expected value of rear-end accidents on
each section of Freeway No.1
Tab. 7
The rear-end accident
warnings
|
|
|
|
||||
|
|
|
|
|
|
|
|
|
|
P |
C |
C |
P |
C |
P |
|
|
P |
P |
P |
I |
P |
I |
|
|
P |
C |
C |
I |
C |
I |
Where:
P = Conflicts are likely to happen (Possible);
C = Conflicts happen (Conflict occur);
I = Conflicts are unlikely to happen (Impossible).
Fig. 14. Total number and
expected value of sideswipe accidents on
each section of Freeway No.1
4.3. Dispatch
of the Rescue Vehicle
Dispatching
a vehicle should be planned for system optimization after acquiring prediction
data. The relationship between the required input data and the available output
data is illustrated in Fig. 15. Input data encompasses existing data and
prediction data. Existing data may consist of four items: (1) The location of
each fleet branch; (2) The number of available vehicles in each fleet branch;
(3) The relative distance between each fleet branch and the standby location.
Notably, as prediction data cannot precisely determine the exact accident
location, it can only predict the probability of accident occurrence in each section
of the Freeway. In practical accident handling, vehicles may stand by at the
entrance of interchange roads in the upstream section of the Freeway or passing
bays within the Freeway section. Therefore, this study only estimates the
relative distance based on the standby location. Additionally, (4) the relative
time required can be calculated based on the relative distance and the vehicle
speed.
Through
the prediction results, this study can organize data into three items for
further use by Planning Support Tools: (1) Estimating the time when an accident
may occur; (2) Deriving available standby locations within the section of the Freeway;
(3) Estimating or dispatching the number of vehicles required for accident
handling based on the probability of accident occurrence.
The
output data may include the four items: (1) The expected time of arrival
indicates the time required for each fleet branch to arrive at the standby
location; (2) The number of attending vehicles indicates the actual number of
vehicles from each fleet branch attending the accident. The number of attending
vehicles shall at least satisfy the need for vehicle evacuation from the
accident; (3) The locations of departure indicate the locations of fleet
branches where the attending vehicles come from; (4) Service routes shall
indicate the routes from the fleet branches to the standby locations; (5) The
support conditions indicate the conditions and numbers of available vehicles
between the fleet branches.
Fig. 15. The relationship
between the input data and the output data
Freeway
No. 1, from 165k to 207k (Taichung system-Puyan system), is the focus of
analysis in this study. Along this stretch, there are eight interchange roads
and two fleet branches associated with the National Freeway Police Stations.
These are the Tainan fleet branch, responsible for National Freeway No. 1 from
Sanyi interchange road to Nantun interchange road, and the Yuanlin fleet
branch, overseeing National Freeway No. 1 from Nantun interchange road to Hsilo
interchange road. Consequently, the author will provide further details on the
background of the data.
Using
mathematical planning models and network flows, this study will construct an
optimized model for the deployment and dispatching of vehicles. The model will
take into account various constraints, including the number of available
vehicles from each fleet branch and the distance between the fleet branch and
each section of the Freeway. The study will employ network effectively flows to
identify vehicle movements in the spatial-temporal network, aiding in model
development.
The
flows of vehicles in the spatial-temporal network primarily serve to identify
vehicle movements at a specific time and location, as illustrated in Fig. 16. A
network layer represents the service route of a vehicle. If the numbers of
available vehicles from two fleet branches are m and n, respectively, there
will be a total of m+n network layers. The lateral axis denotes the
distribution of spaces where vehicles might stop, including each fleet branch
and each section of the Freeway, while the longitudinal axis represents a
continuous time schedule.
Fig. 16. The network
diagram of vehicle evacuation due to an accident
The
spaces between the nodes along the longitudinal axis indicate time intervals,
such as 1 hour, 4 hours, 6 hours, or 12 hours. This study will use a 1-hour
time interval for further planning. Planners can design an appropriate time
interval to achieve accurate prediction results based on their future needs
when using the model. The length of the network indicates the planning period's
duration, the nodes in the network represent the spatial/temporal nodes of a
vehicle at a specific time, and the nodal line indicates the vehicle's
activities between two spatial/temporal nodes. The vehicle flow along the nodal
line illustrates the vehicle's movement in its activities. The nodal line can
be further divided into the vehicle stagnant nodal line and the vehicle moving
nodal line, while the nodes can be further divided into the supply node, the
demand node of the fleet branch, and the demand node of a section of the
Freeway. Details of the two types of nodal lines and the three types of nodes
are outlined as follows:
(1) The stagnant nodal line: The stagnant nodal
line is the line extended downward from each spatial/temporal node, indicating
that a vehicle remains stationary at a specific space point during a particular
period. The cost of the stagnant nodal line is intentionally set to 0. It's
worth mentioning that if the vehicle flow starts from the supply node,
representing the fleet branch, and ends at the demand node of the same fleet
branch along the stagnant nodal line, it means the vehicle from the fleet
branch is never put to use.
(2) The moving nodal line: The moving nodal line
connects different spatial nodes, indicating the spatial-temporal flow of the
vehicle. The cost of the moving nodal line is the cost of flow, typically set
as the driving distance. Notably, the nodal line representing vehicle flow from
the fleet branch into a section of the Freeway network moves at the same point
in time, as the time interval required to travel from the fleet branch to the
section of the Freeway is greater than the driving time. However, when a vehicle
moves between sections on the Freeway or returns from a section to the fleet
branch, it is considered as providing services at the next point in time.
Therefore, the nodal line will be connected to the next point.
(3) Supply node: The first spatial/temporal node
at each fleet branch position serves as the supply node of the network, with
the supply amount set at 1.
(4) Nodes on the section of the Freeway:
Different colors indicate different probabilities of accident occurrence.
(5) Demand node of the fleet branch: The final
spatial/temporal node at each fleet branch position represents the demand node
of the network, with the supply amount set at 1.
5. CONCLUSIONS
Regarding
accident occurrence, Data Mining and preliminary statistical analysis have been
completed. There are currently 6000 vehicles equipped with the OBU, with
roughly 70 percent of them having been on the Freeway. Subsequently,
approximately 100 vehicles that have traveled on National Freeway No. 1 (from
Taichung system to Puyan system) were used for data analysis, incorporating
information from their OBU, eTag data, and VD data, to understand changes in
vehicle flow after an accident occurs.
This
study primarily utilizes the expected number of accidents on Freeway No. 1 as a
reference for emergency vehicle allocation. Initially, vehicle flow data (from
VD) and accident data are combined to analyze the accident risk (such as
rear-end accidents) associated with different states of vehicle flows.
Subsequently, data from the OBU is employed to understand the driving behaviors
exhibited by drivers in different states of vehicle flows and identify
forewarnings for rear-end accidents. The ultimate goal is to establish a
real-time Freeway accident risk prediction model, with warnings issued through
a smart traffic information platform.
Concerning
the dispatch of rescue vehicles, the preliminary construction of the
spatial-temporal network based on vehicle flows has been completed. The next
step involves establishing an optimized mathematical model for vehicle
evacuation from accidents based on the spatial-temporal network of vehicle flows.
Acknowledgment
The author would like to give many
thanks and acknowledge the support from Innovation Center for Intelligent
Transportation and Logistics, Feng Chia University, Taichung for providing the
data sets.
References
1.
“Taiwan RFID-based ETC Total Solution”. Available
at: https://www.roc-taiwan.org/public/USlax_en_events/5860173671.pdf.
2.
Xie T., Y. Liu. 2022. “Heterogeneous
traffic information provision on road networks with competitive or cooperative
information providers”. Transportation Research Part C: Emerging
Technologies 142: 103762. ISSN: 0968-090X.
3.
Chen C.H. 2021. “Temporal-Spatial
Feature Extraction Based on Convolutional Neural Networks for Travel Time
Prediction”. arXiv preprint arXiv:2111.00149.
4.
Casteigts A., A. Nayak, I.
Stojmenovic. 2011. “Communication protocols for vehicular
ad hoc networks”. Wireless
Communications and Mobile Computing 11(5): 567-582. ISSN: 1530-8677.
5.
Rolison J.J., S. Regev, S. Moutari,
A. Feeney. 2018. “What are the factors that contribute
to road accidents? An assessment of law enforcement views, ordinary
drivers’ opinions, and road accident records”. Accident Analysis & Prevention 115: 11-24. ISSN: 0001-4575.
6.
Khorasani-Zavareh D., R.
Mohammadi, H.R. Khankeh, L. Laflamme, A. Bikmoradi, B.J. Haglund. 2009.
“The requirements and challenges in preventing of road traffic injury in
Iran. A
qualitative study”. BMC public health 9(1): 1-9. ISSN: 1471-2458.
7.
Aloul F., I. Zualkernan, R. Abu-Salma,
H. Al-Ali, M. Al-Merri. 2015. “iBump: Smartphone application to
detect car accidents”. Computers & Electrical Engineering 43:
66-75. ISSN: 0045-7906.
8.
Sacchi E., T. Sayed. 2016. “Conflict-Based
Safety Performance Functions for Predicting Traffic Collisions by Type”. Transportation
Research Record: Journal of the Transportation Research Board 2583: 50-55.
ISSN: 0361-1981.
9.
Guo M., X. Zhao, Y. Yao, P. Yan,
Y. Su, C. Su, D. Wu. 2021. “A study of freeway crash risk prediction
and interpretation based on risky driving behavior and traffic flow
data”. Accident Analysis & Prevention 160: 106328.
10. Hu Y., Y. Yang, J. Liu, M. Bai. 2023. “Estimating freeway
accident-prone sections: research on vehicle dynamic simulation and accident
prediction model”. Proceedings of the Institution of Civil
Engineers-Transport: 1-43.
Thomas Telford Ltd.
11. Villiers
C., L.D. Nguyen, J. Zalewski. 2019. “Evaluation of traffic
management strategies for special events using probe data”. Transportation
research interdisciplinary perspectives 2: 100052.
12. Karrouchi
M., I. Nasri, M. Rhiat, I. Atmane, K. Hirech, A. Messaoudi, M. Melhaoui,
K. Kassmi. 2023. “Driving behavior assessment: A practical
study and technique for detecting a driver's condition and driving style. Transportation
Engineering 14: 100217.
13. Gross F.,
B.N. Persaud, C. Lyon. 2010. “A guide to developing quality crash
modification factors (No. FHWA-SA-10-032)”. United States. Federal
Highway Administration. Office of Safety.
14. Office of the Assistant Secretary for Research and
Technology (OST-R). “Integrated Vehicle-Based Safety Systems
(IVBSS)”. Available at: http://www.its.dot.gov/ivbss/.
15. Dingus T.A., S.G. Klauer, V.L. Neale, A. Petersen, S.E. Lee, J.
Sudweeks, M.A. Perez, J. Hankey, D. Ramsey, S. Gupta, C. Bucher, Z.R.
Doerzaph, J. Jermeland, R.R. Knipling. 2006. “The
100-Car Naturalistic Driving Study: Phase II – Results of the 100-Car
Field Experiment (Report No. DOT HS 810 593)”. Washington, DC: National
Highway Traffic Safety Administration, USDOT.
16. Buzzacchi
L., T.M. Valletti. 2005. “Strategic price discrimination in
compulsory insurance markets”. The Geneva Risk and Insurance Review 30:
71-97.
17. Paefgen J., T. Staake, E. Fleisch. 2014. “Multivariate exposure modeling of accident risk: Insights from
Pay-as-you-drive insurance data”. Accident Analysis and Prevention
61: 27-40.
18. Ministry of Transportation and Communications (MOTC). 2016. “Research of Information Service Sharing in Traffic
Field Application - Discussion on Related Subject of Traffic Information
Service”. Institute of Transportation.
Received 02.12.2023; accepted in
revised form 29.01.2024
Scientific Journal of Silesian University of Technology. Series
Transport is licensed under a Creative Commons Attribution 4.0
International License
[1] Faculty of Civil Engineering, University of Transport and
Communications, Hanoi, Vietnam. Email: gianglk@utc.edu.vn. ORCID:
https://orcid.org/0000-0002-6295-8578