1078122, 7:09AM —_—_—Maceling large vehicles operating speed characteristics on freeway alignment based on aggregated GPS data - Hou - 2020 -IET
Advertiseme
IET Intelligent Transport Systems / Volume 14, Issue 8 / p. 857-865
Special Section: Sustainably Intelligent Mobility (SIM) @lFree Access
Modelling large vehicles operating speed characteristics on
freeway alignment based on aggregated GPS data
Wei Hou, Nengchao Lyui&, Zhaoxin Liu, Xu Wang
First published: 30 June 2020
https://doi.org/10.1049/iet-its.2019.0563
Abstract
Operating speed is a basic measure for studying the safety of freeways.
However, whether the designed speed matches the actual operating
speed of large vehicles is worth studying. As such, this study included a
large number of aggregated vehicle global positioning system (GPS)
datasets used to analyse the actual speed of large vehicles under
different alignment conditions on the freeway; and supplement the
theoretical operating speed model. Using the traditional operating speed
prediction model, the theoretical speed of each section was calculated.
Two months of vehicle GPS data from freeways was extracted from a
vehicle monitoring platform, the actual operating speed of each section
was obtained. The similarities and differences between the theoretical
operating speed and the actual vehicle speed on the horizontal and
vertical curves were analysed and compared. The results indicate that the
hitpsetresearch.onlinelibrary.wiey.comdo!10.104@\iet-ts.2019.0569 1311078122, 7:09AM —_-Maceling large vehicles operating speed characteristics on freeway alignment based on aggregated GPS data - Hou - 2020 -IET
actual speed of large vehicles was obviously higher than the theoretical
speed operating downhill. When the vehicle was operating uphill, the
vehicle's actual speed was significantly lower than that of the theoretical
speed. The actual vehicle operating speed data can be used as a new
method for evaluating the safety of large vehicles on the freeway.
1 Introduction
Effective road management requires systematic evaluations of traffic safety.
In response, researchers have developed scientific frameworks that
accurately evaluate road traffic safety and formulate strategic safety
measures to support management and reduce crash rates. Among these
frameworks, there are methods based on operating speed, which include
the theoretical operating speed model [1], spatial sight of the speed model
[2], vehicle dynamics simulation of the operating speed model [3] and more.
However, most of the above-mentioned methods are based on the
theoretical model. At present, the road safety assessment methods that are
commonly used in China are mainly recommended by the Ministry of
Transportation: ‘Guidelines for Safety Audit of Highway’ and ‘Specifications
for Highway Safety Audit’ JTG BO5-2015 [4, 5]. Internationally, the model
most wildly used is the interactive highway safety design model (IHSDM),
which was developed in the United States [6]. After testing, it has been
found that results from the two methods are satisfactory and widely used.
1.1 Evaluation of road safety based on operating speed
Significant research has been conducted on road traffic safety. Analysis
methods include the probability-based statistics method, intensity analysis
method, comprehensive evaluation method, and more. These are based on
mathematical models from statistics. The intensity analysis method is more
reliable and scientifically sound than the probability mathematical statistics
method because it comprehensively analyses the factors affecting road
traffic safety. In the process of using the comprehensive evaluation method,
hitpsetresearch.onlinelibrary.wiey.comdo!10.104@\iet-ts.2019.0569 21311078122, 7:09AM —_-Maceling large vehicles operating speed characteristics on freeway alignment based on aggregated GPS data - Hou - 2020 -IET
there is a hierarchical analysis method [7], gray theory [8, 9], fuzzy synthesis
[10, 11], and a method of merging multiple theories [12] with better results.
However, the comprehensive evaluation method is more random and
ambiguous. This is due to the deviation in the numerical classification of the
indicators caused by differences in the knowledge and experience of
scientists in the evaluation of road traffic safety. There may be a significant
difference in scientific opinion regarding numerical classifications.
Additionally, the indicated value is interfered with by man-made or other
external measurement issues, resulting in the situation that the value in the
critical area meets the requirement of being divided into two different levels
[13]. The relatively new methods of analysis and evaluation include principal
component analysis [14] and neural networks [15]. The concept of the
operating speed clearly stipulates that an abrupt line-shaped change of road
position is required to check the operating speed to ensure that the vehicle
speed satisfies the coordination requirement [16].
The ‘Guidelines for Safety Audit of Highway’ stipulate that road alignment
indicators should be analysed based on vehicle speed to evaluate road
safety [4]. Studies have shown that the greater the difference in operating
speed between adjacent road units, the less coordination of the road
section and the poorer the safety of the vehicle when it is operating [17-19].
After its own investigation, the Australian Road Transport Bureau also
reiterates that when the operating speed increases, the risk factor of traffic
accidents increases [20]. At the application level, speed characteristic
analysis methods have been applied in safety evaluation. Yu et al. [21]
proposed a unit-based map matching technology to correlate vehicle global
positioning system (GPS) trajectories with road geometry. In this way, a
network-wide velocity matrix was generated, which can be used to identify
traffic situation; Chen et al. [22] researched on the longitudinal movement of
vehicles, using GPS data collected from taxis. The graphical method can
describe the characteristics of the driver's longitudinal acceleration
hitpsetresearch.onlinelibrary.wiey.comdo!10.104@\iet-ts.2019.0569 3311078122,7:09 AM —_—Maceling large vehicles operating speed characteristics on freeway alignment based on aggregated GPS data - Hou - 2020 -IET
behaviour and distinguish the differences between drivers. This method can
be used for driving safety analysis.
For a proposed new highway, the model presented herein can be used to
predict the vehicle speed, and for active roads, the model presented herein
can obtain the speed directly from the roadway. In fact, there exists a close
relationship between the operating speed and the linear continuity of the
road. Leisch and Leisch [23] proposed the use of operating speed to
evaluate the consistency of the longitudinal profile of the road and the
geometry of the plane. Lamm and Choueiri [24] believe that road safety
assessment needs to be considered in terms of speed coordination, and the
evaluation index is consistent with the specifications. However, the amount
of data used in these studies is relatively small.
1.2 Operating speed distribution and factors analysis
Speed distribution characteristics are significant to speed limit
management, traffic flow theory, and road network planning and design [25,
26]. The actual operating speed is affected in many ways. Highway capacity
manual found that the speed of vehicles on urban roads is affected by
factors such as traffic control, vehicle-to-vehicle interaction, and road
environment. The operating speed reflects the adaptability of the vehicle to
the actual road. The speed distribution can be described by three
observations [27]: the shape of the distribution, the degree of concentration
of the distribution, and the degree of dispersion of the distribution. Based
on this research, Helbing [28] established a relationship model between
average vehicle speed, vehicle speed variance, and density. Shankar and
Mannering [25] established that the discreteness of vehicle speed calculated
by a discrete vehicle speed model can be affected by adjacent lane speeds
and exogenously related to associated environmental, traffic flow, and
temporal factors [29]. Hong and Oguchi [30] and Shu et a/. [31] found that
road characteristics, including alignment, number of lanes, slope, and
hitpsetresearch.onlinelibrary.wiey.comdo!10.104@\iet-ts.2019.0569 4311078122, 7:09AM —_-Maceling large vehicles operating speed characteristics on freeway alignment based on aggregated GPS data - Hou - 2020 -IET
length, significantly impact operating speed. For free-flow speed, Zhu et al.
[82] found that the influencing factors mainly include: (i) traffic conditions,
such as traffic management, flow direction distribution, and vehicle
composition; (ii) road conditions, such as the nature of the road, lane width,
and surface roughness; and (iii) environmental conditions, such as
inclement weather. During a storm, e.g. rainfall can impact vehicle speed in
many ways, including wet slippage reducing surface friction and increasing
accident probability or heavy downpour narrowing the scope of vision and
reducing visibility.
Generally, these inclement weather conditions cause drivers to change their
behaviour and reduce their speed, which adversely affects traffic flow
capacity and travel efficiency [33, 34]. In particular, the operating speed of
large vehicles is greatly affected by the longitudinal curve. In this study, road
conditions are analysed.
Wang et al. [35] introduced a new method to estimate the speed change.
This method uses the high-frequency GPS data of a taxi to integrate the
spatio-temporal speed fluctuations of a single vehicle and the speed
difference between the vehicles. By analysing the data collected from a total
of 234 one-way sections, the research results show that the average speed
increased by 1%, the total collision increased by 0.7%, and greater speed
changes are also related to increasing the collision frequency. Dai et al. [36]
proposed a method to analyse the speed characteristics of different road
units by using actual GPS data of operational vehicles. This study analysed
the similarities and differences between the theoretical driving speed and
actual vehicle speed on the typical curve.
However, in previous research and applications, the data came mainly from
a typical section [37] and bore some deficiencies in both spatial and
temporal resolutions. When the vehicle speed analysis is performed based
on the predicted operating speed derived from the prescribed speed limit,
hitpsetresearch.onlinelibrary.wiey.comdo!10.104@\iet-ts.2019.0569 511078122,7:09 AM —_—Maceling large vehicles operating speed characteristics on freeway alignment based on aggregated GPS data - Hou - 2020 -IET
there is a gap between the actual speed and the predicted speed because
the vehicle is operating on a complex road environment. Therefore, the
results calculated using the operating speed prediction model do not fully
reflect the changes in the speed of the vehicle during operation. Especially in
the actual operation of large-scale vehicles, due to the difference in load and
dynamics, the operating speed may be significantly different from the
theoretical limit. Therefore, using actual operating data to develop a
prediction model for the operating speed of large vehicles is of practical
significance to road traffic safety.
1.3 Research objective
Owing to the difference in the dynamic performance, braking performance,
and loading characteristics of large vehicles, operating speed is greatly
influenced by the alignment of the road. To improve the accuracy of the
speed forecasting model for traffic safety evaluations and to help operating
speed evaluations render results that better resemble real-world roadways,
this study analyses the speed characteristics of various large vehicles using
GPS data and proposes a new method for road safety evaluation based on
the actual operating speed of vehicles rather than the theoretical speed. The
similarities and differences between the theoretical operating speed Vgs and
the actual vehicle speed Vg,. of large vehicles on horizontal and vertical
curves were analysed, and the causes of these differences are discussed
herein. Additionally, the influence of the radius of the horizontal curve and
the gradient of the longitudinal slope on the actual speed of a large vehicle
was also analysed, and the variation of the actual speed of different road
types was obtained. The new method presented in this study corrects and
supplements the prediction model of the operating speed of large vehicles
in the Guidelines for the Safety Evaluation of Highway Projects and also
provides a new method for evaluating the safety of freeways using
operating speed characteristics.
hitpsetresearch.onlinelibrary.wiey.comdo!10.104@\iet-ts.2019.0569 611078122, 7:09AM —_-Maceling large vehicles operating speed characteristics on freeway alignment based on aggregated GPS data - Hou - 2020 -IET
2 Method and data preparation
In both road design and safety evaluations, the vehicle speed is used as the
most basic measure of traffic safety. Therefore, the closest and most direct
parameter to road safety is speed. However, for any given road, there are at
least two speed considerations: (i) the theoretical driving speed calculated
(or predicted) according to the road alignment; and (ii) the operating speed
actually conducted and observed on the roadway. The vehicle speed studied
also includes two definitions: (i) the theoretical operating speed Vgs
calculated by the model; and (ii) the actual operating speed V,,, observed by
collecting a large number of GPS datasets.
2.1 Segmentation and speed prediction model
(i) Road segmentation : According to the IHSDM model, the operating speed Ves refers to
the safe speed that can be maintained by mid-level drivers according to actual road,
traffic, and weather conditions. Owing to their stature and lightweight, small cars can
maintain higher speeds on highways. Large cars, due to their bigger size and heavier
weight, have inferior performance to small cars in terms of power and handling. The
overall control-ability of large vehicles is poor when the vehicle turns and goes downhill
Therefore, to ensure safe driving, large vehicles operate at a lower speed
In this study, a 90 km stretch of mountain freeway with alignment variations was selected
as the test roadway, on which both the horizontal curve radius and longitudinal gradient
change over a large range. According to the various radii and slopes, the freeway was
divided into a straight line, horizontal curve, longitudinal slope, and curved slope section.
The segmentation method and criteria are shown in Table 1. The speed was calculated
based on the longitudinal driving direction of the vehicle. Uphill was considered positive,
while downhill was considered negative. In accordance with the above principle, the
freeway was divided into 141 sections.
Since in this study, the actual operating speed of the vehicle was obtained based on GPS,
and in the tunnel section, the GPS signal strength was poor, data became unavailable.
Therefore, in the theoretical speed calculation portion, the operating speed in the tunnel
section was not captured nor studied. In addition, the vehicle speed characteristics were
more complex due to interference in the vicinity of the interchange. Therefore, the impact
of the interchange was disregarded in this study. Removing the tunnel and interchange
sections, 92 valid sections were obtained for analysis.
hitpsetresearch.onlinelibrary.wiey.comdo!10.104@\iet-ts.2019.0569 71311078122,7:09 AM —_—Maceling large vehicles operating speed characteristics on freeway alignment based on aggregated GPS data - Hou - 2020 -IET
Firstly, the operating speed calculation method was used to determine the initial
operating speed Vo from the first analysis unit of the road section, and then the operating
speed Vgs was calculated at the end of the first analysis unit according to the driving
direction and the corresponding operating speed prediction model of the analysis unit.
This was used as the initial operating speed of the second analysis unit, and then the
operating speed of the end of the unit was calculated by substituting the calculation
formula corresponding to the second analysis unit. This method was successively
implemented until the last analysis unit was calculated. The subject of this study was a
large vehicle, and according to the related specification, the initial operating speed Vo was
75 km/h.
(ii) Typical road speed prediction model : In accordance with the ‘Specifications for Highway
Safety Audit’ [5], the operating speed operation model of a typical road type was
established, The road on the straight section was calculated using the method described
in ‘Specifications for Highway Safety Audit’ [5].
Table 1. Road segment division criteria
Vehicle type Vertical Horizontal
section
Horizontal curve radius Horizontal curve
> 1000 m radius $ 1000 m
passenger vehicles or slope<3% tength > 200 m, straight length = horizontal curve
large vehicles 200 m, short straight
slope23% longitudinal slope curved-slope
On the horizontal curve segment, because the segmentation was divided
based on the track, it was partitioned into four sections, and the calculation
of the operating speed at the midpoint of the curve and the exit of the curve
was calculated and predicted according to the formula in ‘Specifications for
Highway Safety Audit’ [5]. The model of the operating speed at the end point
on the longitudinal slope section was selected from the ‘Specifications for
Highway Safety Audit’ [5].
hitpsetresearch.onlinelibrary.wiey.comdo!10.104@\iet-ts.2019.0569 at1078122,7:09 AM —_—Maceling large vehicles operating speed characteristics on freeway alignment based on aggregated GPS data - Hou - 2020 -IET
Based on the highway alignment data provided by the highway survey and
design company, a full-length chart and a vertical table with a total length of
90.869 km were obtained. After dividing the entire road by the road unit
division method, there were 92 valid road sections, which included 66
straight sections, nine curve sections, 11 vertical sections, and six curved
slope sections. According to the above-mentioned operating speed
calculation model, the operating speed of the entire highway was calculated
and is listed in Table 2.
Table 2. Partial results of road segmentation and predicted driving speed
Number Start End Type | Length, | Radius, | Slope, Ves,
stakes | stakes m m % km/h
15 uKio+ UkiT+ straight 795 1500 -0851 75
835,000 630,000
16 uKit + uKi2+ straight 8s 1400 203 reat
630.000 214827
7 uKi2+ uKi2+ curved 281 1000 2.08 sas
214827 495.928 slope
18 uKi2+ uKi2+ curve 231 ° 09 mas
495929 m71076
19 uKi2+ uKI3+ straight 688 ° 09 75
m976 465,000
20 uK13+ uki4s straight 720 2500 193 7s
465,000 185,000
2.2 GPS data pre-processing and actual speed calculation
hitpsetresearch.onlinelibrary.wiey.comdo!10.104@\iet-ts.2019.0569 9311078122, 7:09AM —_-Maceling large vehicles operating speed characteristics on freeway alignment based on aggregated GPS data - Hou - 2020 -IET
(i) Operating GPS data collection method : The GPS data obtained in this study came from
the desensitisation data of an operating vehicle monitoring company, which is a domestic
company that specialises in large car networking. It has more than 10,000 freight vehicles
that operate throughout the country.
The data from this study was obtained through the vehicle-mounted truck information
service platform device. The data collected by the operating vehicle monitoring platform is
stored in a local SQL server database. This study required the screening and extraction of
data distributed only on the 90 km target highway. Owing to the vastness of China's
region and a large amount of data, it was first required to write a filter statement on the
vehicle data in the SQL server database to extract vehicle data points from the target area.
(ii) Data screening : Owing to a large amount of original data, the average daily database
data was in the range of 15-20G. Therefore, a total of 55 days (over two months) of valid
data was selected for the study. It was determined to be the best approach for this study
to write code directly in the SQL server database for data processing operations to.
improve efficiency. To facilitate data processing and analysis, the valid data obtained by
filtering was saved into. XLS format.
(iii) Data pre-processing : To correlate the GPS point data with the actual highway segment,
it was necessary to combine the vehicle data with the corresponding spatial position to
obtain the vehicle speed distribution at different locations on the map. In this study, the
map-matching method was used to match GPS data with the road sections. The target
road was broken into 142 road sections, and then the SQL server data filter was imported
into ARCGIS. More than 400,000 valid data points were matched. These points contained
key information, such as dimensions, longitude, time, heading angle, and speed. These
points served as basic data for the calculation of road speed. The high-quality data
obtained for this study were screened and mapped using ARCGIS and are shown in Fig. 1
Fig. 1
Open in figure viewer | &PowerPoint
hitpsetresearch.onlinelibrary.wiey.comdo!10.104@\iet-ts.2019.0569 10311078/22,7.09AM Modeling large vehicles operating speed characteristics on freeway alignment based on aggregated GPS data -Hou 2020 -IET
GPS distribution before and after preprocessing
3 Results and discussion
3.1 Overall speed profiles
According to the unit division method of the operating speed calculation,
the road was divided into 92 valid sections. Each road type was numbered:
straight section: 1; curved section: 2; longitudinal section: 3; and curved
slope section: 4, as shown in Fig. 2 (93 to 99 are tunnel sections, shown only
in the figure but not considered in the research). The vehicle speed analysed
in this study was divided into two types: the theoretical operating speed Vgs
calculated by the model and the actual operating speed Vp; collected
through operating vehicle GPS, as shown in Fig. 3.
Fig. 2
Open in figure viewer | ¥PowerPoint
Change of actual speed and operating speed with radius in a straight section
Fig.3
Open in figure viewer | %PowerPoint
bnps:Metresearch onlinelbrary.wilay.comidol/10.104aietts.2019.0563 811078122, 7:09AM —_-Maceling large vehicles operating speed characteristics on freeway alignment based on aggregated GPS data - Hou - 2020 -IET
Section map of the average of the theoretical speed and actual speed of the left and right lines
According to IHSDM, the theoretical speed is the speed value predicted by
the operating speed prediction model. For each piece of data that has been
divided, the entering vehicle speed, the middle vehicle speed, and the
exiting vehicle speed on the road section, both calculate the average
represents the theoretical vehicle speed Vgs on the road of this section, as
shown in Fig. 3. The speed curve of the target road shows that the
theoretical operating speed of large vehicles was generally between 65 and
75 km/h, most of which was distributed at 75 km/h and the theoretical
speed of large vehicles on the left and right lines was generally consistent.
The actual operating speed refers to the speed at which the GPS data-
collecting device was automatically positioned and measured under the
actual road conditions. According to the speed curve of the target road, the
actual operating speed of large vehicles was generally between 40 and
90 km/h, most of which was distributed at 75 km/h, and the distribution
range of the actual operating speed of large vehicles on the left and right
lines was generally the same.
Comparing the two actual speed curves and theoretical operating speed
curves in Fig. 3, it can be observed that when a large vehicle ran on this
freeway, the relationship between the operating speed Vgs and the actual
speed was calculated according to the difference of the road alignment, and
the law of change was also different. In addition, it can be observed that the
operating speeds Vgs on the left and right lines all fluctuated around
75 km/h; however, during the actual operation of the vehicle, the vehicle
speed was not completely maintained according to the theoretically
calculated results; its fluctuation range was larger than that of Vgs. As can be
observed from Fig. 3, the roads corresponding to the points where the
vehicle speed changed more (mainly as a decline) were either the road types
numbered 3 and 4 or the road sections near these two road types. It can be
hitpsetresearch.onlinelibrary.wiey.comdo!10.104@\iet-ts.2019.0569 1131078122, 7:09AM —_-Maceling large vehicles operating speed characteristics on freeway alignment based on aggregated GPS data - Hou - 2020 -IET
observed that in the vicinity of most number 3 road sections (longitudinal
slope sections), the actual speeds of the left and right lines all show opposite
trends, such as in sections 64 and 105. The main reason for this being that
when driving on a longitudinal slope section, if the left line is downhill, the
speed will increase as a whole, then the right line will be uphill, and the
overall speed will be lower. For sections with the same speed change trend,
the operating vehicle will decelerate for safety reasons. Therefore, when the
vehicle enters, exits, and is driving on a longitudinal slope section and
curved slope section, the speed will change significantly. The common
feature of these two road types is that the longitudinal slope is large. From
this, it can be inferred that longitudinal slopes have a greater impact on
large vehicles operating on the highway.
It can be observed that the speed changes on particular road sections are
relatively large, and the speed of some road sections is significantly
reduced. The speed of some road sections may be slightly higher than the
operating speed Vgs. Therefore, when designing a road, in many cases, the
route is designed according to a method in which a horizontal curve
includes a vertical curve [33].
3.2 Effect of road alignment on actual operating speed
To study the influence of road alignment on the actual operating speed of large vehicles, this
study extracted two key factors: the horizontal curve radius and vertical slope. Four types of
road units and their influence laws are also discussed herein. When studying the influence of
the radius of a horizontal curve on the speed of a large vehicle, it is necessary to control the
gradient of the longitudinal slope. On the target highway, the longitudinal slope of the whole
line was analysed. The section with a gradient of more than 3% belongs to the longitudinal
slope section and curved slope section.
(i) Straight line section: Horizontal curves were divided into straight segments and
horizontal segments. According to previous studies, a plane line with a radius of 1500 m
or more was classified as a straight line. According to a previous theoretical model study,
the radius of the curve at this time has little effect on the vehicle speed. For a section with
a longitudinal gradient of 1% or less, to study the variation law of the speed of a large
hitpsetresearch.onlinelibrary.wiey.comdo!10.104@\iet-ts.2019.0569 19311078122, 7:09AM —_—_—Maceling large vehicles operating speed characteristics on freeway alignment based on aggregated GPS data - Hou - 2020 -IET
vehicle with a straight section radius, after integrating the data of the left and right lines,
each section was arranged in the order of small to large radius to a total of 60 data units,
as shown in Fig. 2.
It can be observed that the theoretical operating speed Vs in the straight section was
generally maintained at 75 km/h. With an increase of radius, the actual speed of the
vehicle fluctuated around 75 km/h, showing no obvious change. In the range of 1500 m to
infinity, the average vehicle speed was 70.8 km/h and the standard deviation was
9.1 km/h. This shows that when a large vehicle is operating on a straight section, the
radius of the horizontal curve does not have a significant effect on the vehicle speed
To analyse the difference between the theoretical speed and the actual speed, IBM SPSS.
statistical analysis package was used to further analyse the statistics. A single sample T-
test was performed on the actual speed, The analysis result was P <0.05 indicating that
there was a significant difference in this data group. In the paired-sample T-test results
for the theoretical speed and the actual speed, P <0.05 indicates that there was also a
significant difference between the two groups of data. The theoretical speed and actual
speed were compared to make the boxplot shown in Fig. 4.
As can be observed in the figure, although the mean values of the two sets of data were
not significantly different, the standard deviation of the theoretical speed (1.34) and the
standard deviation of the actual speed (9.07) were significantly different. This shows that
when a large vehicle is driving on a straight section, although the radius has no significant
effect on the operating speed, in actual operation, the vehicle will still be affected by many
factors in some cases, leading to a new theory that states that operating speed and model
prediction speed are inconsistent.
(ii) Horizontal curve section: According to previous studies, a planar linear shape with a
radius of <1500 m can be classified as a horizontal curve [38] and other studies have
indicated that the radius of the curve at this time influences the vehicle speed (J. In
particular, if the radius of the curve is small, it may cause a rollover accident [39]. For the
section where the longitudinal slope was <1%, to study the variation rule of the speed of a
large vehicle with a radius on the curved section in the horizontal curve alignment, 14
data units on the left and right were arranged in ascending order of radius, as shown in
Fig. 5. Itis observed that the operating speed Ves of the large vehicle on the curve
segment was maintained at 67-75 km/h, which is in consideration of the influence of the
horizontal curve radius. There was only a speed drop of <10 km/h on the small radius
curve segment. The actual speed Vgps can be observed at 53-84 km/h. Overall, as the
radius decreased, the speed decreased. This suggests that on the horizontal curve
segment, the theoretical speed and the actual speed are similar.
To analyse the difference between the theoretical speed and the actual speed, a single-
sample T-test was performed. The analysis result was P <0.05, indicating that there was a
significant difference in this data group. The paired sample T-test results for the
hitpsetresearch.onlinelibrary.wiey.comdo!10.104@\iet-ts.2019.0569 vais1078122, 7:09AM —_-Maceling large vehicles operating speed characteristics on freeway alignment based on aggregated GPS data - Hou - 2020 -IET
theoretical speed and the actual speed P > 0.05, indicate that there was no significant
difference between the two groups of data. A box-plot was made based on the theoretical
speed and actual speed, as shown in Fig. 6
As can be observed in the figure, although the average values were not significantly
different, the standard deviation of the theoretical speed (2.82) was significantly different
from the standard deviation of the actual speed (9.48). Preliminary analysis shows that
although there was a significant difference in the actual speed data set, there was no
significant difference with the theoretical speed as a whole, which also proves the
accuracy of the proposed large speed forecasting model. Of course, given the limited
number of speed units in the curve segment, the amount of data can be increased in the
future and further analysed.
(iil) Uphill vertical curve : When studying the effect of vertical slope on the speed of large
vehicles, the radius of the horizontal curve must be controlled to eliminate the effect of
the radius being too small on the vehicle speed. Therefore, a section with a radius
>1500 m was selected for this study. For an uphill section with a radius >1500 m, the
horizontal axis was arranged from small to large according to the size of the vertical slope.
The results are shown in Fig. 7.
In terms of theoretical speed, with an increase of slope, the theoretical operating speed
Vas of large-scale vehicles remained 75 km/h in a large range. Ata vertical slope of 3%, a
significant speed drop occurred. In the section with a gradient >3%, the vehicle's operating
speed on the vertical section was maintained at 73 km/h. In terms of actual speed, when
the uphill slope was <1%, the operating speed Ves of the large vehicle was maintained at
75 km/. When the slope uphill gradient was >1%, the actual vehicle speed was <75 km/h.
In the range of 1-3%, the actual operating speed average was 65 km/h, which was
obviously lower than the operating speed. In the 34% range, the average operating
speed was 56 km/h, and the lower speed was 45.7 km/h.
This is due to the greater mass of large vehicles and the weight of cargo. When the vehicle
was in the uphill section, the vehicle possessed sufficient traction, which resulted in low
speed. Particularly in a mountainous area, a lower operating speed resulted in a larger
speed differential with other vehicles, which may cause crashes [40]. When designing
roads, particular attention should be paid to the danger brought by heavy vehicles such
that the effect of road conditions on long and large vertical slopes is mitigated.
To analyse the difference between the theoretical speed and the actual speed, a single
sample T -test was performed on the actual speed. The analysis result was P <0.05,
indicating that there was a significant difference in this data group. In the paired-sample T
-test results for the theoretical speed and the actual speed, P <0.05 indicates that there
was also a significant difference between the two groups of data. A box-plot developed
based on the theoretical speed and actual speed is shown in Fig. 8. As can be observed in
the figure, although the mean values were not significantly different, the standard
hitpsetresearch.onlinelibrary.wiey.comdo!10.104@\iet-ts.2019.0569 16311078122, 7:09AM —_-Maceling large vehicles operating speed characteristics on freeway alignment based on aggregated GPS data - Hou - 2020 -IET
deviation of the theoretical speed (1.37) and the standard deviation of the actual speed
(9.97) were significantly different. This shows that the operating speed of large vehicles on
the uphill section was inconsistent with the theoretical speed of the model prediction,
which also reflects the deviation of the previous large-vehicle speed forecasting model,
which needs improvement.
To quantitatively deduce the relationship between the actual speed of large vehicles and
the gradient of the uphill slope, a linear fitting of the actual speed of the uphill section and
the radius of the curve was made, as shown in Fig. 9. The relationship between actual
operating speed and slope on the uphill section can be described by Y= ~ 4,6728X +
72.986, where Yis the operating speed (km/h), X's the slope (%) on uphill. The goodness
of fit was R® = 0.241, and the degree of fit was satisfactory. At the same time, an analysis
of the correlation between the actual speed of the uphill section and the radius of the
smooth curve was performed: P = -0.491. The fitting result shows that there is a
significant correlation, the operating speed of large vehicles decreases with an increase of
slope when driving on an uphill section. This result provides a reference for the prediction
model of large vehicle speed on an uphill-slope.
The deductions from the comparisons and analyses above indicate that the difference
between the theoretical speed and the actual speed on the uphill slope is relatively large;
especially when the slope is large, the speed of the large vehicle drops greatly. This may
be due to the fact that the current vehicle power is different from the theory assumption,
and the impact of overload on the operating speed cannot be ignored. Therefore, when
designing the road, these factors that cause the speed drop should be fully considered,
and the operating speed model should be revised. in actual traffic management, severe
overweight behaviours should also be paid attention to avoid safety risks caused by
excessively low vehicle speeds.
(iv) Downhill vertical curve : Similar processing as above, through the actual operating data,
it was possible to analyse the impact of the downhill slope on the speed of large vehicles.
For a downhill section with a radius >1500 m, the horizontal axis was arranged from small
to large in accordance with the size of the longitudinal slope, the results of which are
shown in Fig, 10. it can be observed that on the downhill section when the longitudinal
slope was <1.5%, the actual operating speed Vgs of the large vehicle was maintained at
75 km/h. As the gradient of the downgrade section increased, the actual operating speed
gradually deviated from 75 km/h. It can be clearly observed that the actual speed was
higher than the theoretical speed. When the downhill slope was close to 3%, the actual
measured speed was significantly higher than the theoretically calculated operating speed
Ves. The longitudinal gradient was 3.9%, and the average speed was 88.87 km/h. The
downhill section of the graded slope shows a systematic high speed. This is due to the
prevalence of overloading problems in some areas. When the vehicle descended, the
speed increased as the slope increased. The higher speed caused by the vertical gradient
hitpsetresearch.onlinelibrary.wiey.comdo!10.104@\iet-ts.2019.0569 1631‘182, 7.09 AM Modeling age vehicles operating speed characteristics on freeway algnment based on aggregated GPS data -Hou-2020-IET
significantly impacts traffic accidents (41). For road design, such conditions should be
avoided, especially on slopes with large slope values and on which the braking
requirement for large vehicles is high. These requirements also reflect a relatively large
potential safety hazard. At this time, large vehicle drivers should be reminded to
decelerate before entering the vertical slope.
To analyse the difference between the theoretical speed and the actual speed, a single sample
T test was performed on the actual speed. The analysis result was P <0.05, indicating that
there was a significant difference in this data group. In the paired-sample T test results for the
theoretical speed and the actual speed, P <0.05 indicates that there was also a significant
difference between the two groups of data. A box-plot was developed based on theoretical
speed and actual speed and is shown in Fig. 11
wi al a cay
Spe
Fig. 10
Open in figure viewer | ¥PowerPoint
Changes of actual speed and operating speed with slope on downhill
Open in figure viewer | PowerPoint
Fitting of actual speed and slop on uphill section
bnps:Metresearch onlinelbrary.wilay.comidol/10.104aietts.2019.0563 v3107822, 7.09 AM
‘erst on reeway alignment based on aggregated GPS data How - 2020 -IET
‘Speedin,
‘Theoretical speed ‘Actual average speed
Fig. 8
Open in figure viewer | PowerPoint
Theoretical and actual speed of the uphill
Fig. 7
Open in figure viewer | #PowerPoint
Changes of actual speed and operating speed with slope on uphill
Intps:etresearch.onineibrary.wiley.comidol/10.1049ietts.2019.0563 1831107822, 7.09 AM
‘erst on reeway alignment based on aggregated GPS data How - 2020 -IET
Theoreical speed ‘Accualaverage speed
Fig. 6
Open in figure viewer | #PowerPoint
Theoretical and actual speed on the curve
Fig. 5
Open in figure viewer | #PowerPoint
Change of the actual speed and the operating speed with a radius of the curved section
Intps:etresearch.onineibrary.wiley.comidol/10.1049ietts.2019.0563 19311018/22,7.09 AM Modeling large vehicles operating speed characteristics on freeway alignment based on aggregated GPS data -Hou 2020 -IET
=a
Sa
Sy
“
“Theoreteal speed
ig. 4
Theoretical and actual speed of straight section
Intps:etresearch.onineibrary.wiley.comidol/10.1049ietts.2019.0563
Open in figure viewer
¥ PowerPoint1078122, 7:09AM —_-Maceling large vehicles operating speed characteristics on freeway alignment based on aggregated GPS data - Hou - 2020 -IET
2
g.| =
2 “>
3
Theoretical speed Actual average speed
Fig. 11
Open in figure viewer | PowerPoint
Theoretical and actual speed of the downhill
It can be observed that although the average values were not significantly
different, the standard deviation of the theoretical speed (1.39) and the
standard deviation of the actual speed (7.97) were significantly different.
This shows that the operating speed of large vehicles on a downhill section
is inconsistent with the theoretical speed of the model prediction, which
also reflects the flaws in the previous large vehicle speed forecasting
models, which require improvement.
To quantitatively analyse the relationship between the downhill slope and
the actual speed, a linear fitting of the actual speed of the downhill section
and the radius of the curve was made, as shown in Fig. 12. The relationship
between actual operating speed and slope on downhill section can be
described by Y = — 3.2419X + 70.901, where Yis the operating speed (km/h),
hitpsetresearch.onlinelibrary.wiey.comdo!10.104@\iet-ts.2019.0569 21311078122, 7:09AM —_-Maceling large vehicles operating speed characteristics on freeway alignment based on aggregated GPS data - Hou - 2020 -IET
X is the slop (%) on the downhill. The goodness of fit in the results was R? =
0.1746, and the degree of fit was satisfactory. At the same time, a
correlation between the actual speed of the downhill segment and the
radius of the curve was made. The analysis result shows that P = -0.418,
indicating a significant correlation—that is to say, the speed of a large
vehicle increased with the increase of the slope when driving on a downhill
section, which provides a reference for the prediction model of the speed of
large vehicles on the down-slope.
Fig. 12
Open in figure viewer | %PowerPoint
Fitting of actual speed and slop in downhill section
To better investigate the variation of the speed of large vehicles on uphill
and downhill slopes, the uphill speed was compared with the downhill
speed. It can be observed that in the downhill section, the theoretical
operating speed Vgs and the actual operating speed Vg,. differed by
~2 km/h. Although there was a gap, the gap was not great. On the uphill
section, the theoretical operating speed Vgs and the actual operating speed
Vaps differed by about 10 km/h. Therefore, on the uphill section, the speed
model proposed herein is of urgent need.
The deductions from the comparisons and analyses above indicate that the
difference between the theoretical speed and the actual speed on the
hitpsetresearch.onlinelibrary.wiey.comdo!10.104@\iet-ts.2019.0569 221811078122, 7:09AM —_—_—Maceling large vehicles operating speed characteristics on freeway alignment based on aggregated GPS data - Hou - 2020 -IET
downhill section is relatively large, especially when the slope is relatively
large, the actual speed is at a high value. The reason for this phenomenon
may be due to the fact that on the downhill section, the driver's speed
expectation is relatively high, coupled with possible overload and braking
performance decreasing, resulting in higher vehicle speeds. When designing
the road, these factors that cause high vehicle speeds should be fully
considered, and the operating speed model should be revised. In addition,
in actual traffic management, the possible risks brought by higher vehicle
speeds should also be considered, and necessary speed control facilities
should be set up to avoid operational risks caused by uncontrolled vehicle
speeds.
4 Conclusion and prospect
In this study, a comparative analysis of highway speeds based on the operating speed model of
highway safety was conducted, A large number of actual vehicles operating speed datasets
were used to match the specific road segments through map-matching methods to describe
the speed profile of the freeway. Based on four types of road sections, the influence of highway
alignment on the actual speed of large vehicles was explored. The main findings are as follows:
(i) The actual operating speed reflects the actual variation of the vehicle speed when a
large vehicle is operating on the freeway. By comparing with the theoretical operating
speed, it can be observed that the result is reliable and can be used to analyse the
characteristics of the speed change of large vehicles.
(ii) On all types of sections, the speed of the vehicle is affected by the gradient of the
longitudinal slope more than the radius of the horizontal curve. The difference of the
uphill section Vas —Vgps is significantly greater than the value of the downhill section,
indicating the universality of over-speeding in road transport, which has a greater impact
‘on the prediction of operating speed.
(iii) It is reasonable to use the actual speed in the analysis of the characteristics of the
speed of large vehicles operating on the road. Using the actual vehicle GPS monitoring
data to statistically analyse the speed changes on different types of road sections, anew
method has been developed to evaluate the safety of vehicles operating on the freeway
using speed characteristics.
(iv) This study can also provide references for road designers and managers. When the
road design encounters an inevitable longitudinal road section, the driver should be told
to drive as carefully as possible in advance. Especially for the uphill section, there is a
hitpsetresearch.onlinelibrary.wiey.comdo!10.104@\iet-ts.2019.0569 231311078122, 7:09AM —_-Maceling large vehicles operating speed characteristics on freeway alignment based on aggregated GPS data - Hou - 2020 -IET
significant difference in the accelerating capacity of large vehicles and small vehicles. It
may also be a good suggestion that large vehicles and small vehicles be driven in different
lanes to reduce rear-end collisions. In the management of existing roads, the method of
this study can be used to find hidden road sections that are liable to over speeding; and
corresponding deceleration facilities, such as speed limit signs and deceleration markings,
can be added to suitable locations to improve traffic safety.
Although the study compares the actual speed with operating speed characteristics, it provides
new ideas for the operation of road traffic safety evaluation, and the analysis has achieved the
above results, However, the following aspects involved in this study still need further research:
(i) Although this study uses as much data as possible to analyse the impact of various road
variables on vehicle speed, it still does not cover all road types. Therefore, more types of
roads should be analysed.
(ii) Although this study excluded as much as possible the impact of extraneous factors,
such as tunnels, intercommunication etc., there are still other potential influencers, such
as night-time, adverse weather, and construction, which affect the speed on road
sections. An understanding of how to effectively exclude these factors can improve the
reliability of the results.
(iii) At present, the data from operating large vehicles is relatively easy to obtain, and the
traffic data of a large number of small cars on the freeway is still relatively difficult to
obtain. Therefore, obtaining the data of small cars and performing the same analysis has
stronger application significance for improving the speed model on the freeway.
5 Acknowledgments
This study was supported by the National Nature Science Foundation of
China: modelling driving cognition behaviour and driving performance
improvement methodology considering workload environment (51678460);
the Qilu Transportation Development Group Co, Ltd (2016B20); the Natural
Science Foundation of China (51775396); and Wuhan Science and
Technology Bureau Enterprise Technology Innovation Project
(2018010402011175).
hitpsetresearch.onlinelibrary.wiey.comdo!10.104@\iet-ts.2019.0569 ars1078122, 7:09AM —_-Maceling large vehicles operating speed characteristics on freeway alignment based on aggregated GPS data - Hou - 2020 -IET
1 Suwarto F, Basuki K.H.: ‘The application of traffic conflict technique as a road
safety evaluation method: a case study of Hasselt intersection’, Appl. Mech.
Mater., 2016, 845, pp. 394-403
Crossref | Google Scholar
2 Alfonsi R. Persia L. Antonino T. et al.: ‘Advancements in road safety
management analysis’, Transp. Res. Procedia, 2016, 14, pp. 2064~ 2073
Crossref | Google Scholar
3 Wen CP. Tian Q.Q.: ‘Assessment of urban road traffic safety based on Bayes
discriminant analysis method’, Adv. Mater. Res., 2013, 639-640, (1), pp. 544-547
Crossref | Google Scholar
4 JTG/T B05-2004.: ' Guidelines for safety audit of highway’ ( China
Communication Press, Beijing, 2004)
Google Scholar
5 JTG BOS-2015: ' Specifications for highway safety audit’ ( China Communication
Press, Beijing, 2016)
Google Scholar
6 XiJ.F. Gao ZH. QiB. et al.: Evaluation of traffic safety of curved section with
gray theory based on the driving track’, Adv. Mater. Res., 2012, 588-589, pp.
2117-2120
Crossref | Google Scholar
7 Wen-Ting H.U. Tan Z.F. Yang].: ‘Road safety comprehensive evaluation based
on analytic hierarchy process (AHP)’, Transp. Stand., 2010, 31, (2), pp. 94- 102
Google Scholar
8 Niu H.Y.: ‘Study on safety evaluation of urban road traffic based on gray
theory’, China Safety Sci. J, 2005, 15, (9), pp. 92-97
Google Scholar
hitpsetresearch.onlinelibrary.wiey.comdo!10.104@\iet-ts.2019.0569 25311078122,7:09 AM —_—Maceling large vehicles operating speed characteristics on freeway alignment based on aggregated GPS data - Hou - 2020 -IET
9 Yue X. Ding Y.: ‘Evaluation of urban road traffic safety based on the grey
cluster and AHP’, Forest Eng., 2007, 23, (6), 64~ 68 (in Chinese)
Google Scholar
10. Yan S. Yu J.: ‘Variable fuzzy assessment method and its application in
comprehensive evaluation of road safety’, J. Safety Environ., 2010, 10, (6), pp.
2991-2996
Google Scholar
11 Wang). Lei M. Zhou)J.: ‘Application of fuzzy comprehensive evaluation
model in study of road traffic safety’, Urban Roads Bridges Flood Control, 2010,
(11), pp. 97-99 (in Chinese)
Google Scholar
12. Sun H.X.: ‘An evaluation model to the road traffic safety based on grey
relational grade of fuzzy hierarchy’, Traffic Transp., 2010, (2), pp. 136-139 (in
Chinese)
Google Scholar
13. LiX. Zhao N. Zheng X.: ‘Road traffic safety evaluation based on cloud mode’,
J. Beijing Univ. Technol., 2015, 41, (8), pp. 1219- 1224 (in Chinese)
Google Scholar
14 Xue D.W. Jun-Ling J.|. Bai Z.: ‘Study on traffic safety evaluation for freeway
based on principal component analysis’, j. Heilongjiang Inst. Technol., 2014, 28,
(2), pp. 46 49 (in Chinese)
Google Scholar
15. Li G.:‘ Road safety evaluation based on artificial neural network’, 2015,
Google Scholar
16 Yuan Q. Li. Xing S. et al.: ‘Correlation between residual speedometer
needle reading and impact speed of vehicles in traffic accidents’, int. J. Autom.
hitpsetresearch.onlinelibrary.wiey.comdo!10.104@\iet-ts.2019.0569 261311078122, 7:09AM —_-Maceling large vehicles operating speed characteristics on freeway alignment based on aggregated GPS data - Hou - 2020 -IET
Technol., 2015, 16, (6), pp. 1057— 1063
Crossref | Web of Science® | Google Scholar
17 XuJ. Lin W. Wang X. et al.: ‘Acceleration and deceleration calibration of
operating speed prediction models for two-lane mountain highways’, J. Transp.
Eng. A, Syst., 2017, 143, (7), p. 04017024
Crossref | Web of Science® | Google Scholar
18 LiXH. Zhu S.L, Sun Y.B.: ‘The research on the alignment continuity of the
prairie highway based on possible speed’, Adv. Mater. Res., 2011, 255-260, pp.
4123-4127
Crossref | Google Scholar
19 Zhou Z.G. Wang Y.P. Long KJ.: ‘Highway traffic safety evaluation method
based on operating speed’, Traffic Sci. Eng., 2009, 25, (1), pp. 91-95
Google Scholar
20 Curnow G.:' Australian transport safety bureau heavy truck crash
databases what do the statistics tell us?’. National Heavy Vehicle Safety Seminar,
Melbourne Victoria Australia, 2002
Google Scholar
21 YuJ. Stettler M. Angeloudis P. et al.: ‘Urban network-wide traffic speed
estimation with massive ride-sourcing GPS traces’, Transp. Res. C, 2020, 112, pp.
136-152
Crossref | Web of Science® | Google Scholar
22. Chen C. Zhao X. Zhang Y. et al.: ‘A graphical modeling method for individual
driving behavior and its application in driving safety analysis using GPS data’,
Transp. Res. F, Psychol. Behav., 2019, 63, pp. 118-134
Crossref | Web of Science® | Google Scholar
23 Leisch J.£. Leisch J.P.: ‘New concepts in design-speed application’, Transp.
Res. Rec., 1977, 631, pp. 4-14
hitpsetresearch.onlinelibrary.wiey.comdo!10.104@\iet-ts.2019.0569 271311018122,7.09 AM Modeling ange vehicles operating speed characeiscs on reeway alignment based on agargoied GPS data - Hou -2020-IET
Google Scholar
24 Lamm R. Choueiri E.M.: ‘Recommendations for evaluating horizontal design
consistency based on investigations in the state of New York’, 1987
Google Scholar
25. Shankar V. Mannering F.: ‘Modeling the endogeneity of lane-mean speeds
and lane-speed deviations: a structural equations approach’, Transp. Res. A,
Policy Pract., 1998, 32, (5), pp. 311-322
Crossref | Web of Science® | Google Scholar
26 Pei Y.L. Cheng G.Z.: ‘Research on the relationship between discrete
character of speed and traffic accident and speed management of freeway’,
Zhongguo Gonglu Xuebao/China J. Highway Transp., 2004, 17, (1), pp. 74-78
Google Scholar
27 JiaJ.:‘ Statistics’ ( Tsinghua University Press, Beijing, 2006), pp. 56— 60
Google Scholar
28 Helbing D.: ‘Derivation and empirical validation of a refined traffic flow
model’, Physica A, 1998, 233, (1-2), pp. 253- 282
Crossref | Google Scholar
29 Cheng G.Z. Pei Y.L.: ‘Analysis and limit of Ves of highway’, Highway Eng.,
2008, 33, (1), pp. 62- 66
Google Scholar
30 Hong SJ). Oguchi T.: ‘Evaluation of highway geometric design and analysis of
actual operating speed’, j. Eastern Asia Soc. Transp. Studies, 2005, 6, pp. 1048—
1061
Google Scholar
hitpsetresearch.onlinelibrary.wiey.comdo!10.104@\iet-ts.2019.0569 281311078122, 7:09AM —_—_—Maceling large vehicles operating speed characteristics on freeway alignment based on aggregated GPS data - Hou - 2020 -IET
31 Shu H.B. Shao Y.M. Xu). ‘Effects of geornetric features of highway
horizontal alignment on steering behavior of passenger car’, J. Vibroeng., 2016,
18, (6), pp. 4086~ 4104
Crossref | Web of Science® | Google Scholar
32 Zhu S. Wang H. Wang W..: ‘The quantitative analysis influence factor of free
flow speed of neural network method’, Highway Traffic Technol., 2001, 18, (3), pp.
60-63
Google Scholar
33 Chen F. Chen S.R.: Reliability-based assessment of vehicle safety under
adverse driving conditions’, Transp. Res. C, Emerg. Technol., 2011, 19, (1), pp.
507-518
Crossref | Web of Science® | Google Scholar
34 Zhang L. Feng X. Chen D. et al.: ‘Designing a hazardous materials
transportation network by a bi-level programming based on toll policies’,
Physica A, 2019, 534, p. 122324
Crossref | CAS | Web of Science® | Google Scholar
35. Wang X. Zhou Q. Quddus M. et al. ‘Speed, speed variation and crash
relationships for urban arterials’, Accident Anal. Prev., 2018, 113, pp. 236-243
Crossref | PubMed | Web of Science® | Google Scholar
36 Dai. LyuN. Hu Y.:' Truck speed characteristics analysis of typical highway
segments based on GPS data’. The 4th Int. Conf. on Transportation Information
and Safety, Banff, Canada, 2017, pp. 535-540
Crossref | Google Scholar
37 MaX. Chen F. Chen S.: ‘Modeling crash rates for a mountainous highway
using refined-scale panel data’, Transp. Res. Rec., 2015, 2515, (1), pp. 10-16
Crossref | Google Scholar
hitpsetresearch.onlinelibrary.wiey.comdo!10.104@\iet-ts.2019.0569 291311078122, 7:09AM —_-Maceling large vehicles operating speed characteristics on freeway alignment based on aggregated GPS data - Hou - 2020 -IET
38 Xie L. WuC. LyUN. et al.: Effects of freeway alignment on driving
performance and workload based on simulated driving’. Proc. CICTP2017,
Shanghai, China, 2017
Google Scholar
39 Sun C. Wu GZ. Chu D.F. et al.: ‘Safety evaluation of driving on curves based
on analysis of vehicle lateral stability’, . Transp. Inf. Safety, 2014, 32, (6), pp. 95—
100
Google Scholar
40 Pei xX. Wong S.C. Sze N.N.: The roles of exposure and speed in road safety
analysis’, Accident Anal. Prev., 2012, 48, pp. 464-471
Crossref | PubMed | Web of Science® | Google Scholar
41. Chen F. Chen S.R. Ma X.X.: ‘Crash frequency analysis using hurdle models
with random effects considering short-term panel data’, int. J. Environ. Res.
Public Health, 2016, 13, (11), p. 1043
Crossref | CAS | Web of Science® | Google Scholar
Download PDF
-=
The Institution of
mmm Bt Engineering and Technology
ABOUT THE IET
IET PRIVACY STATEMENT
CONTACT IET
Copyright (2022) The Institution of Engineering and Technology. The Institution of Engineering and
Technology is registered as a Charity in England & Wales (no 211014) and Scotland (no SC038698)
hitpsetresearch.onlinelibrary.wiey.comdo!10.104@\iet-ts.2019.0569 30311078/22,7.09.AM Mailing large vehicles operating speed characteristics on freeway alignment based on aggregated GPS data -Hou - 2020 -IET
About Wiley Online Library
Privacy Policy
Terms of Use
About Cookies
Manage Cookies
Accessibility
Wiley Research DE&l Statement and Publishing Policies
Help & Support
Contact Us
Training and Support
DMCA & Reporting Piracy
Opportunities
Subscription Agents
Advertisers & Corporate Partners
Connect with Wiley
The Wiley Network
Wiley Press Room
999-2022 John Wiley & Sons, Inc. All rights reserved
Intps:etresearch.onineibrary.wiley.comidol/10.1049ietts.2019.0563 3131