j o u r n a l o f t r a f fi c a n d t r a n s p o r t a t i o n e n g i n e e r i n g ( e n g l i s h e d i t i o n ) 2 0 1 8 ; 5 ( 5 ) : 3 8 6 e4 0 5

Available online at www.sciencedirect.com

ScienceDirect

journal homepage: www.elsevier.com/locate/jtte

Original Research Paper

Dynamic parameters of vehicles under heterogeneous

traffic stream with non-lane discipline: An
experimental study

Geetimukta Mahapatra*, Akhilesh Kumar Maurya

Department of Civil Engineering, Indian Institute of Technology Guwahati, Guwahati 781039, India

highlights

The probability distribution of longitudinal A/D and the lateral acceleration are analyzed.

The relationship with operating speed of vehicles are studied on roads with the different number of lanes for different vehicle types.

A two-term exponential and linear relationship with operating speed are observed for lateral and longitudinal A/D respectively.

These A/D values and lateral acceleration can be used in the operating speed prediction models and also establish the bounding values
in the case of simulation models.

article info abstract

Article history: On the heterogeneous and no lane disciplined traffic, the abreast maneuver of vehicles
Received 26 July 2017 depends upon the driver behavior, vehicle type and most importantly the traffic parameters
Received in revised form such as vehicle speed and acceleration. Hence, the drivers have two basic tasks while
20 January 2018 driving, first is to control the vehicle's position along longitudinal direction of motion and
Accepted 24 January 2018 second is to control the vehicle's position along lateral direction i.e. the width of the
Available online 29 July 2018 roadway. The relation between the dynamic parameters (speed and lateral/longitudinal
acceleration), can very well represent the driving behavior of vehicles in mixed with weak or
Keywords: no lane disciplined traffic. Hence, the study of longitudinal and lateral control of vehicles in a
Non-lane discipline traffic comprehensive manner is necessary. In the present paper, driving behavior of vehicles are
Mixed traffic studied by analyzing longitudinal and lateral acceleration/deceleration (A/D) with operating
Driving behavior speed of vehicles on different straight roadway sections. Data are collected from five major
Dynamic parameters cities of India using GPS based instrument (Video-VBOX) mounted on five different type of
Operating speed vehicles. The probability distribution of longitudinal A/D and the lateral acceleration are
g-g diagram analysed, and their relationship with operating speed of vehicles are studied on roads with a
different number of lanes for different vehicle types. A two-term exponential and linear
relationship with operating speed are observed for lateral and longitudinal A/D respectively.
© 2018 Periodical Offices of Chang'an University. Publishing services by Elsevier B.V. on
behalf of Owner. This is an open access article under the CC BY-NC-ND license (http://
creativecommons.org/licenses/by-nc-nd/4.0/).

* Corresponding author. Tel.: þ91 361 2582426; fax: þ91 361 2582440.
E-mail addresses: geetimuktamahapatra@gmail.com (G. Mahapatra), maurya@iitg.ernet.in (A.K. Maurya).
Peer review under responsibility of Periodical Offices of Chang'an University.
https://doi.org/10.1016/j.jtte.2018.01.003
2095-7564/© 2018 Periodical Offices of Chang'an University. Publishing services by Elsevier B.V. on behalf of Owner. This is an open
access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
 J. Traffic Transp. Eng. (Engl. Ed.) 2018; 5 (5): 386e405
1. Introduction
The task of driving can be defined as tracking a path or a lane
along a given roadway. Vehicles' stability control is one of the
major issues while driving and it affects the vehicle's attitude
and motion. The vehicle's stability mainly depends on how
accurately the driver interprets the vehicle motion to provide
added directional control. The directional control implies the
lateral and longitudinal control of the vehicles. The driver
typically expresses the directional control through the steering
angle. In mixed traffic, due to the variability in vehicle char-
acteristics, fast moving, as well as slow moving vehicles pre-
sent together in the same traffic scenario. The vehicles travel
abreast due to variation in their sizes and non-standard widths.
Due to the heterogeneous and non-lane discipline traffic, ve-
hicles occupy any lateral position throughout the roadway
width. Hence, in mixed and non-lane discipline traffic stream,
the drivers have two basic tasks, first is to control the vehicle's
position along the longitudinal direction of motion and second
is to control the vehicle's position along the lateral direction,
i.e., the width of the roadway. Hence, both the lateral and
longitudinal parameters are interdependent and go on simul-
taneously known as two-dimensional (2D) driver behavior.
Therefore, the study of longitudinal and lateral control of ve-
hicles in a comprehensive manner is necessary to understand
and analyze the driving behavior in the mixed traffic stream.
Speed and acceleration are two basic dynamic parameters
which define the movement of vehicles in any traffic stream. On a
given roadway section, the relation between the dynamic pa-
rameters (speed and lateral/longitudinal acceleration), can very
well represent the driving behavior of vehicles in mixed and non-
lane discipline traffic. Lateral acceleration is a useful parameter
for representing the realistic lateral behavior of vehicles while
generating simulation models for such traffic conditions. Also,
it can be converted into the lateral force coefficient of a highway,
which is a key factor for lateral stability and driving safety.
The main objective of this study is to analyze the longitu-
dinal and lateral driving behavior of different type vehicles on
straight roads during normal driving conditions using their
dynamic parameter like operating speed, lateral acceleration
and longitudinal acceleration/deceleration (A/D). Therefore,
in the present paper, the driving behavior of vehicles are
studied by studying the longitudinal and lateral accelerations
with respect to operating speed of vehicles on different
straight roadway facilities of India. Though the super-eleva-
tion rate and lateral stability are the important characteristics
in case of curved road sections, on Indian roads, the abreast
driving of the vehicles on straight road sections also leads to
the serpentine motion of vehicles resulting in higher lateral
and longitudinal acceleration/deceleration or braking of
different vehicles while passing or filtering through the traffic.
Hence studying both the longitudinal and lateral behavior
became an critical issue in the present scenario.
2. Background
The longitudinal interaction of the vehicles in traffic stream
has been studied for quite a long time. However, limited
387
studies had been found on lateral acceleration by observing
the vehicle trajectory in the non-lane discipline mixed traffic
conditions. This section is divided into two subsections, i)
studies on the longitudinal acceleration of vehicle and ii)
studies on the lateral acceleration of vehicles.
2.1. Studies on the longitudinal acceleration
Various speed profile models and A/D profile models are
developed for different applications, in the past studies. The
acceleration and deceleration (A/D) rates are both assumed to
be equal to 0.8 m/s2 by Lamm et al. (1988) or 0.85 m/s2 by
Echaveguren and Basualto (2003) and Richl and Sayed (2011).
In some studies, they are assumed to be within certain
range, based on the current operating speed which usually
yields a smaller acceleration at a higher speed (Liu et al.,
2010). Brooks (2012) reported that vehicle's acceleration
decreases with the increase in their speed. Researchers have
also used experimental data to establish the relationship
between acceleration rate and speed. However, these models
do not differentiate between A/D (Camacho-Torregrosa
et al., 2013; Fitzpatrick et al., 2000; Montella et al., 2014).
Majority of the studies were conducted the study on
acceleration behavior on signalized intersections (Akcelik
and Beseley, 2001; Akcelik and Biggs, 1987; Bham and
Benekohal, 2002; RaiChowdhury and Rao, 1989; Wang et al.,
2004). Limited work is done in the past on deceleration
modelling in comparison to acceleration modelling (Akcelik
and Biggs, 1987; Wang et al., 2005).
Limited studies have been conducted on A/D behavior in
mixed and weak lane discipline traffic like India (Bokare and
Maurya, 2016; Dey and Biswas, 2011; Maurya and Bokare, 2012;
RaiChowdhury and Rao, 1989). Arasan and Koshy (2005)
concluded that rate of acceleration decreases with increase
in speed. Lower rates of acceleration are reported for heavy
vehicles like buses, trucks and light commercial vehicles as
compared to cars, motorized three-wheelers, and two-
wheelers. The maximum and average acceleration at
different speeds for different vehicle types, are measured by
Mehar et al. (2013), Maurya and Bokare (2012) and Bokare
and Maurya (2016) observed that deceleration behavior is
different for different vehicle types. However, the study was
conducted at the signalized intersection at a short stretch
under controlled condition. Because of weak lane discipline
heterogeneous traffic, it became inconsistent and difficult to
analyze the data at intersections.
2.2. Studies on lateral acceleration
Though the longitudinal interaction of the vehicles in traffic
stream has been studied for quite a long time, however, the
lateral behavior of the vehicles (one of the most important
criteria for the weak lane discipline mixed traffic conditions)
has not been explored much. The driver controls his speed or
trajectory to keep his lateral acceleration in a comfortable
range and to ensure a safety margin (Felip and Navin, 1998;
Reymand et al., 2001). Ritchie et al. (1968) conducted a study to
find the relation between longitudinal speed and lateral
acceleration on curves during normal driving and result
showed that lateral acceleration was inversely proportional
388 J. Traffic Transp. Eng. (Engl. Ed.) 2018; 5 (5): 386e405
to speed at curves (Reymand et al., 2001). It was predicted that
the extreme lateral acceleration values decrease quadratically
with speed in curves. Again Odhams and Cole (2004) observed
a drop-off in lateral acceleration with an increase in the speed.
Biral et al. (2005) found that longitudinal acceleration remains
constant at lower to medium speed and linearly decreases
with the increase in speed at medium to high speed.
Similarly, the lateral acceleration increases from low to
medium speeds and then decreases linearly at medium to
high-speed conditions. It has also been observed from the
literature that lateral acceleration up to 4 m/s2 is easy to
deal with by the majority of the average drivers and 4e6 m/s2
is challenging, whereas 6e8 m/s2 cannot be dealt by the
drivers (Jimenez et al., 2008). Xu et al. (2015) stated that lateral
acceleration negatively related to driving speed. Eboli et al.
(2016) proposed a relationship between lateral and
longitudinal accelerations with speeds.
It is observed that most of the studies were conducted to
study the maximum acceleration capabilities of vehicles
(during an emergency or at intersections). The adoption of a
constant value of acceleration by the authors does not repre-
sent the acceleration rates comprehensively. In case of non-
lane disciplined mixed traffic, the presence of many types of
vehicles with different dynamic and vehicular capabilities
leads to the frequent acceleration and deceleration or braking
of the vehicles. Also, insufficient studies have reported the
deceleration behavior of vehicles in such traffic scenario. It
has also been observed that lateral acceleration behavior of
vehicles on the straight mid-block section for the normal
driving condition has not been evaluated in the past.
Kanagaraj et al. (2015) studied the traffic characteristics of
mixed traffic and stated that lateral movement of the
vehicles is one of the important characteristics of the
vehicles in the mixed and non-lane based traffic stream.
Hence, the driving behavior models for mixed traffic stream
should account for these characteristics of different
vehicles. Therefore, the present paper analyses all three
dynamic parameters of different vehicles (longitudinal A/D
and lateral acceleration, operating speed) on the straight
mid-block section for the normal driving condition in mixed
and weak lane discipline traffic on Indian roads. The overall
driving behavior is observed by plotting the g-g diagram,
considering both the lateral and longitudinal A/D.
Fig. 1 e VBOX. (a) Details of video VBOX data recording unit. (b) Installed in a car.
3. Study methodology
The present study explores the operating speed and longitu-
dinal/lateral acceleration of different vehicles (i.e., sports
utility vehicles (SUV), sedan and hatchback (H-back) cars,
motorized three wheelers and two wheelers) on roads in
different major cities of India. The road trajectory includes
straight roads, curves, and roads with different number of
lanes (four lane divided roads (2-lane each side) and six lane
divided roads (3-lane each side)). Video VBOX along with (GPS
with 10 Hz data logging frequency) a display unit (for real-time
monitoring) are installed in the subject vehicles (hereafter
referred as an instrumented vehicle) which was used for data
collection (Fig. 1).
The accuracy of the VBOX data logger is tasted by the
manufacturer of the VBOX automotive using Doppler Shift
method in the GPS carrier signal. The accuracy of velocity data
logging is 0.1 km/h. Many researchers analysed the accuracy
of the VBOX data logger and proved that such data recording
facility has an improved accuracy level (Adnan et al., 2013;
Durrant and Hill, 2005; Katzourakis et al., 2012; Zhang et al.,
2010).
Field data was collected by running different instrumented
vehicles over different routes in the five major cities of India
(Fig. 2) in sunny weather during moderate to high flow
conditions. The blue lines in the maps indicate the marked
road section. The road surface was in good condition, and
proper visible lane markings were present on all the selected
routes.
Five type of instrumented vehicles like motorized two-
wheeler (2W), motorized three-wheeler (3W), SUV, sedan and
H-back cars (three categories of passenger cars) (Car and Bike
Team, 2016) are used for data collection. In the present study,
the variability in the driver and vehicle are considered
simultaneously by varying the 8e10 different drivers for
each vehicle type from different ages, i.e., 20e50 years old.
Each instrumented vehicles were run on roads of five
different cities of India namely, Delhi (40 km), Kolkata
(40 km), Bengaluru (70 km), Mumbai (40 km) and Pune
(65 km). The driver was instructed to travel at an average
speed, interacting with the surrounding vehicles. Also, a list
had been made which includes the number of vehicles
overtaking our test vehicle and the number of vehicles
 J. Traffic Transp. Eng. (Engl. Ed.) 2018; 5 (5): 386e405 389

Fig. 2 e The road trajectory of data collection in five major cities of India. (a) Delhi. (b) Kolkata. (c) Bengaluru. (d) Pune.
(e) Mumbai.

overtaken by our test vehicle from the cameras fitted on the
test vehicle to ensure that the vehicle was travelling
approximately with an average stream speed. The detailed
procedure for data collection and its extraction are provided
in Mahapatra and Maurya (2013, 2015) and Mahapatra et al.
(2016).
The lateral and longitudinal acceleration data along with
longitudinal speed are extracted from the recorded VBOX data
using VBOX-Tools software. Data from all five different cities
for all different types of instrumented vehicle are extracted.
The longitudinal acceleration is divided into two parts, i.e.
acceleration and deceleration data. Statistical t-tests are
conducted between the A/D behavior as well as the lateral
acceleration of different vehicles for each city to know the
significant mean difference between the 2-lane and 3-lane
roads by assuming equal variance with a hypothetical mean
difference of zero. The details of t-test results are presented
below in Table 1.
From the above table, it is clear that there is no signifi-
cant difference between the longitudinal A/D of for 2-lane
and 3-lane roads of each city for all five types of vehicles (t-
statistics > t-critical). However, a significantly different data
is observed for the lateral acceleration between the 2-lane
and 3-lane roads, indicating the road width has a significant
impact on the lateral maneuver of vehicles. Hence, the
behavior of longitudinal A/D is studied by combining the
data of 2-lane and 3-lane roads, whereas the lateral accel-
eration behavior is studied for different types of vehicles on
the straight mid-block sections for different number of
lanes separately. To know the variation in driving behavior
390 J. Traffic Transp. Eng. (Engl. Ed.) 2018; 5 (5): 386e405

Table 1 e Details of t-test for all different types of vehicles of each city.
City Vehicle type Acceleration Deceleration Lateral acceleration
t-stat t-critical t-stat t-critical t-stat t-critical
Delhi 3W 0.90 1.74 0.57 1.74 1.75 1.73
H-back 1.11 1.72 0.09 1.72 1.79 1.72
Sedan 0.70 1.74 0.43 1.71 1.74 1.71
SUV 2-lane data not available
Kolkata 3W 0.87 1.78 0.89 1.78 2.13 1.78
H-back 0.14 1.70 0.15 1.70 1.90 1.81
Sedan 0.80 1.70 0.37 1.70 1.88 1.70
SUV 0.60 1.71 0.62 1.71 3.00 1.73
Bengaluru 3W 0.59 1.73 0.22 1.73 2.43 1.74
H-back 0.93 1.78 0.25 1.79 1.82 1.73
Sedan 0.00 1.73 0.24 1.73 2.93 1.71
SUV 1.29 1.74 0.46 1.74 2.44 1.79
Mumbai 3W 1.20 1.73 0.31 1.73 1.76 1.73
H-back Data not available
Sedan 1.08 1.70 0.28 1.70 1.90 1.70
SUV 0.57 1.71 0.28 1.70 1.71 1.70
Pune 3W 1.26 1.74 0.54 1.74 1.76 1.74
H-back 0.12 1.72 0.22 1.71 1.77 1.71
Sedan 0.84 1.71 0.73 1.71 1.84 1.71
SUV 0.38 1.71 0.62 1.71 3.01 1.71

of drivers in different cities, a comparative study of A/D
behavior and lateral acceleration of five types of vehicles in
different cities is done. Analysis of variance (ANOVA) test is
conducted for each vehicle type data collected from
different cities to know the significant differences between
them. It is found that there is very less pair of cities where
the difference between means is observed for A/D and
lateral acceleration of different vehicles at 5% significance
level. So, for further analysis, the A/D and lateral accelera-
tion data of each vehicle type from different cities are
combined to see the general behavior of different vehicles at
no-lane disciplined mixed traffic throughout the country.
The A/D and lateral acceleration behavior of different ve-
hicles are studied during normal driving conditions.
The analysis and results of this study are presented in
three sections i) study of longitudinal A/D ii) study of lateral
acceleration and iii) relationship between lateral and lon-
gitudinal A/D. Each section includes the distribution of the
extracted data and the establishment of a relationship be-
tween A/D and lateral acceleration data with operating
speed.
4. Study of longitudinal acceleration/
deceleration (A/D)
The detailed analysis of longitudinal A/D is conducted in two
different subsections, i) the percentile distribution and the
probability distribution of A/D for all five types of vehicles and
ii) the effect of operating speed on A/D of different vehicles.
4.1. Distribution of longitudinal A/D
The percentile distribution of A/D is plotted to see the detailed
behavior of a different types of vehicles on the mixed non-
lane based traffic. The following figure depicts the distribution
of combined A/D data for different vehicle types.
Fig. 3 shows the cumulative frequency curves of A/D for
each type of vehicle. The 85th and 95th percentile values are
also indicated in Fig. 3 as well as presented in Table 2. It is
observed that the 85th and 95th percentile values of
deceleration are greater than acceleration for all vehicles.
The deceleration characteristics and the braking
performance depends on the braking coefficient provided to
the road surfaces (f) and the pavement conditions, the tire
condition and type of braking system. On the other hand,
the acceleration capability of a vehicle is dependent on its
mass, the resistance to motion and power of the vehicle
(power to weight ratio). Both the acceleration and
deceleration are two different phenomena of a vehicle while
driving. In the micro simulation frameworks, the
acceleration and deceleration of vehicles are given as input
parameters separately to describe the vehicle's driving
characteristics. The differences in acceleration rates for
different vehicles are quite significant in mixed traffic
stream (because of heterogeneity in vehicle types). Hence, it
is important to analyze the A/D behavior separately for each
type of vehicles to study the microscopic behavior of traffic
stream.
It has been observed that the 85th percentile acceleration
values are different for different vehicle types. The values are
higher in case of motorized 2W and 3W; then it reduced for
cars (H-back, sedan, and SUV cars). This is due to the differ-
ence in engine torque, the power to weight ration and drive
mechanism of different vehicles. The following table lists the
85th, 95th percentile A/D rates extracted from cumulative
percentile curves.
In the past, though many studies have been conducted
on observing the A/D range of vehicles, most of them
referred to the maximum acceleration capabilities applied
by the drivers. Drivers rarely use their maximum
 J. Traffic Transp. Eng. (Engl. Ed.) 2018; 5 (5): 386e405 391

Fig. 3 e Percentile distribution A/D rates for all five different vehicle types. (a) 2W. (b) 3W. (c) H-back. (d) Sedan. (e) SUV.

capabilities of vehicles acceleration/braking except for any A/D (Echaveguren and Basualto, 2003). This is due to the
emergency situations. Those maximum capability values presence of vehicles with different acceleration or
are not useful in design except for fixing bounding values for braking capabilities in the mixed traffic stream. These
safe driving (Fildes et al., 1991; Liu et al., 2003; Long, 2000; characteristic percentile values can be used in the
Zhu et al., 2013). Hence, in the present paper, the A/D values operating speed prediction models and also used to
are obtained from the normal driving conditions. It can be establish the bounding values in case of simulation models.
observed that the 85th and 95th percentile deceleration The cumulative percentile curve of A/D increases and then
values of all five types of instrumented vehicles are higher maintains a constant value after attaining the maximum
than 1 m/s2. However, the 85th percentile acceleration value. The breakpoint is the point from where the slope of A/D
values lie below 1 m/s2 except motorized 2W and 3W trend changes. It has been observed that the breakpoints of
which is higher than the adopted value of 0.85 m/s2 for both the cumulative percentile curves do not appear exactly at 85th
percentile value in all cases. The point lies somewhere be-
tween 85th and 95th percentile. In most of the design practice,
the value of 85th percentile is taken for design purposes.
Table 2 e Observed 85th and 95th percentile A/D rates for
Hence, a fresh evaluation of the concept of the 85th percentile
different vehicles at different roads (2-lane/3-lane).
for design purposes may be needed. This result is similar to
Acceleration (m/s2) Deceleration (m/s2) the findings by Xu et al. (2017).
Average 85th 95th Average 85th 95th To describe the detailed statistical properties of A/D, the
2W 0.78 1.45 2.33 0.83 1.58 2.43 probability distribution function is analyzed for all the five
3W 0.57 1.07 1.65 0.61 1.18 1.82 types of vehicles. In an attempt to find an appropriate dis-
H-back 0.54 1.04 1.54 0.56 1.04 1.66 tribution for representing longitudinal acceleration data for
Sedan 0.53 0.95 1.45 0.56 1.06 1.77 all vehicle types, several (z49) single distribution models
SUV 0.52 0.94 1.38 0.55 1.03 1.63
were implemented in Easyfit software which ranks the fitted
392 J. Traffic Transp. Eng. (Engl. Ed.) 2018; 5 (5): 386e405

Table 3 e Comparison of performance measures for various probability models fitted for A/D of different vehicles.
Vehicle type Acceleration Deceleration
Sample size Distribution LL AIC Sample size Distribution LL AIC
2W 13,828 Lognormal 17,022 34,048 13,285 Lognormal 11,026 22,056
Gamma 10,729 21,462 Gamma 10,982 21,968
Burr 29,175 58,356 Burr 26,887 53,780
Weibull 10,634 21,272 Weibull 16,007 32,018
3W 32,671 Lognormal 7838.9 15,681.8 30,688 Lognormal 8289 16,582
Gamma 7829.9 15,663.8 Gamma 19,760 39,524
Burr 38,082 76,170 Burr 34,662 69,330
Weibull 22,132 44,268 Weibull 8204 16,412
H-back 78,778 Lognormal 21,874 43,752 74,954 Lognormal 65,484 130,972
Gamma 21,696 43,396 Gamma 21,853 43,710
Burr 120,670 241,346 Burr 120,220 240,446
Weibull 68,381 136,766 Weibull 21,365 42,734
Sedan 79,471 Lognormal 51,888 103,780 73,781 Lognormal 89,697 179,400
Gamma 14,373 28,750 Gamma 16,120 32,244
Burr 96,228 192,462 Burr 15,054 30,112
Weibull 13,708 27,420 Weibull 45,226 90,456
SUV 56,720 Lognormal 51,039 102,082 53,790 Lognormal 44,886 89,776
Gamma 12,411 24,826 Gamma 14,411 28,826
Burr 90,009 180,024 Burr 83,148 166,302
Weibull 12,366 24,736 Weibull 13,915 27,834

Fig. 4 e Best fitted probability distribution plot for longitudinal acceleration (x) of all five types of vehicles. (a) 2W. (b) 3W.
(c) H-back. (d) Sedan. (e) SUV.
 J. Traffic Transp. Eng. (Engl. Ed.) 2018; 5 (5): 386e405
distributions based on goodness-of-fit tests and accordingly
four statistical models (Lognormal, Gamma, Burr and Wei-
bull) are fitted well to our data at 5% significance level. To
know the best-fitted distribution amongst all four, log-like-
lihood (LL) value and akaike's information criterion (AIC)
values are estimated for all these probability distribution
functions (PDF). The formulations are presented below.
LLFðxÞ ¼ lnðPDF of xÞ (1)
AIC ¼ 2N  2LLmax (2)
where N is the number of fitted model parameters, LLmax is the
maximum LL.
The probability density functions of A/D separately for
different vehicle types at 5% significance level and the best-
fitted distributions are presented in Table 3. Though different
distributions are fitted well to the A/D data, Weibull and
Gamma distribution are best fitted to the acceleration data
of SUV, sedan and H-back cars, motorized 3W and 2W with
different shape and scale parameter. Similarly, the
deceleration data of SUV, H-back cars and motorized 3W
and 2W follows Weibull and Gamma distribution except for
the sedan cars. The deceleration data of sedan car follows
Fig. 5 e Best fitted probability distribution plot for longitudinal deceleration (x) of all five types of vehicles. (a) 2W. (b) 3W.
(c) H-back. (d) Sedan. (e) SUV.
393
Burr distribution as the best-fitted distribution at 5%
significance level. The best-fitted distribution for A/D
behavior of all five type of vehicles on different roads (2-
lane/3-lane) is depicted in the following Figs. 4 and 5
respectively.
The list of different parameter values of all the best-fitted
distributions shown in the above figure is presented below in
Table 4.
From Table 4, it is observed that the shape parameters of
Weibull distributions fit to different vehicles A/D have nearly
same value, indicating different distributions with the same
moment skewness. In case of Gamma distribution, the value
a >1 indicates that the Gamma distributions assume a
mounded (uni-modal), but the skewed shape. The scale
parameter (b) can stretch or compress the range of both
Gamma and Weibull distributions. The probability
distribution (PDF) of the longitudinal A/D of different
vehicles indicates that the PDF of all five types of vehicles
follows similar distribution (Weibull/Gamma) in all cases
except few (deceleration of sedan; Burr), at 5% significance
level. However, the shape and scale parameters of the fitted
distributions indicate different shapes and stretches of the
distributions for different vehicles. It is clearly visible from
Figs. 4 and 5 that drivers mostly apply the acceleration and
394 J. Traffic Transp. Eng. (Engl. Ed.) 2018; 5 (5): 386e405

Table 4 e Parametric values of all the best-fitted distributions to different vehicles on different roads.
Acceleration Deceleration
SUV Weibull a¼ 1.3487, b ¼ 0.63165 Weibull a ¼ 1.2674, b ¼ 0.6520
Sedan Weibull a¼ 1.2650, b ¼ 0.55672 Burr k ¼ 1.8655, a ¼ 1.6176, b ¼ 0.73534
H-back Gamma a¼ 1.5887, b ¼ 0.37620 Weibull a ¼ 1.2598, b ¼ 0.66852
3W Gamma a¼ 1.4763, b ¼ 0.41987 Weibull a ¼ 1.2513, b ¼ 0.71535
2W Weibull a¼ 1.2180, b ¼ 0.88973 Gamma a ¼ 1.3429, b ¼ 0.65155

Note: a ¼ shape parameter, b ¼ scale parameter for Gamma distribution and Weibull distribution; k, a ¼ shape parameters, b ¼ scale parameter
for Burr distribution.

deceleration in the range of 0.1e0.4 m/s2 for all vehicle types
during normal driving conditions on straight roads. In the
present study, the variability in the driver and vehicle are
considered simultaneously by varying the 8e10 different
drivers for each vehicle type from different ages. Hence,
these probability distribution models describe the of
different driver-vehicle behavior simultaneously as a single
entity for A/D of different types of vehicles.
However, a detailed understanding of the driver-vehicle
interaction with the traffic stream can be achieved by
considering the variation in A/D behavior with regard to traffic
stream parameters such as flow, density, and speed, etc. In
our present study, it is not possible to measure the flow and
density of the whole trajectory as the trajectory length is very
high (including different cities). Though the flow and density
have a direct impact on the traffic speed, speed is considered
as a stream parameter in this study. The variability of longi-
tudinal A/D concerning the speed at the normal driving con-
dition on straight road sections is briefly studied in the next
section.
4.2. Effect of driving speed on longitudinal acceleration/
deceleration (A/D)
To see the A/D behavior of such traffic stream, the data
collected from different major cities of India are combined.

Fig. 6 e Fitted regression models on the scattered acceleration data of all five types of vehicles. (a) 2W. (b) 3W. (c) H-back.
(d) Sedan. (e) SUV.
 J. Traffic Transp. Eng. (Engl. Ed.) 2018; 5 (5): 386e405 395

Fig. 7 e Fitted regression models on the scattered deceleration data of all five types of vehicles. (a) 2W. (b) 3W. (c) H-back.
(d) Sedan. (e) SUV.

The data is clustered with different speed groups of 5 km/h.
The 85th and 95th percentile A/D are calculated for different
speed groups. The variation of 85th and 95th percentile A/D is
observed with regard to the different operating speeds of
vehicles.
The extreme outliers are removed from the extracted data.
The box-whisker plot of A/D vs. speed is obtained for A/D by
gathering the data of five different cities (refer Figs. 6 and 7).
The 85th and 95th percentile values of A/D are indicated by
the blue and green scatter points on each box plot. The

Table 5 e Regression model coefficients and its statistical evaluation for 85th and 95th percentile deceleration and
acceleration data.
85th percentile 95th percentile
2
m n R RMSE m n R2 RMSE
Deceleration 2W 0.01030 1.837 0.8568 0.0673 0.0135 2.678 0.7403 0.1373
3W 0.01440 1.498 0.7960 0.1144 0.0226 2.304 0.8106 0.1717
H-back 0.00948 1.283 0.7674 0.1195 0.0150 2.021 0.8355 0.1523
Sedan 0.01354 1.497 0.8871 0.1208 0.0220 2.384 0.9465 0.1365
SUV 0.00752 1.269 0.8843 0.0780 0.0136 2.045 0.9289 0.1082
Acceleration 2W 0.01800 1.911 0.9057 0.0943 0.0170 2.746 0.8254 0.1265
3W 0.01290 1.378 0.9627 0.0398 0.0208 2.143 0.8220 0.1520
H-back 0.01080 1.332 0.8781 0.0928 0.0159 1.976 0.8425 0.1580
Sedan 0.01160 1.326 0.9414 0.0753 0.0180 1.963 0.9145 0.1430
SUV 0.00870 1.246 0.9721 0.0406 0.0136 1.836 0.8826 0.1362
396 J. Traffic Transp. Eng. (Engl. Ed.) 2018; 5 (5): 386e405
Fig. 8 e Percentile distribution of lateral acceleration for 2-lane and 3- lane roads. (a) 2W. (b) 3W. (c) H-back. (d) Sedan.
(e) SUV.
regression models are fitted to the 85th and 95th percentile A/
D values of each vehicle type.
It can be observed that both the 85th and 95th percentile A/
D curve shows a linearly decreasing trend for all vehicle types
at 5% significance level. Observations depict that the
maximum vehicle acceleration occurs at lower speeds when
the vehicle starting from rest, the ability to accelerate de-
creases with increase in vehicular speeds. The linearly
decreasing trend is modelled, and statistical tests are con-
ducted to model the 85th and 95th percentile trend of A/D of
all five types of vehicles. The general linear trend of A/D is
given in Eq. (3).
A or D ¼ mv þ n (3)
where A or D is acceleration or deceleration, v is vehicular
speed, m and n are regression coefficients.
Table 5 describes the details of all regression model
coefficients and the statistical details (i.e., the R2 value and
root mean squared error values) of the models describing
the A/D behavior respectively. The adjusted R2 value and the
RMSE value for each fitted model is calculated. It can be
observed that the linear model presents the observed data
with more than 75% adjusted R2 value in all cases of A/D.
The lower values of RMSE indicate a better fit of the linear
models at 5% significance level.
5. Study of lateral acceleration
The lateral acceleration is one of the important parameter,
which describes the vehicles lateral behavior on the road. In
case of mixed and non-lane discipline traffic, the lateral ma-
neuver of vehicles is quite prominent. In the present study,
the vehicle type wise data from different cities with no sig-
nificant difference are combined as discussed in section 3. It is
observed that the lateral acceleration data obtained from 2-
lane and 3-lane roads are significantly different from each
other for different vehicle types. Hence, the lateral
acceleration is analyzed for different vehicle types for 2-lane
and 3-lane roads separately.
5.1. Distribution of lateral acceleration
The probability distribution of lateral acceleration on different
roads, i.e., roads with different number of lanes for all five
types of vehicles and the effect of operating speed on lateral
 J. Traffic Transp. Eng. (Engl. Ed.) 2018; 5 (5): 386e405 397

It can be seen that motorized 2W and 3W are having a

Table 6 e 85th percentile lateral acceleration values for
higher average lateral acceleration compared to all type of
different vehicle types on 2-lane and 3-lane roads.
cars. The lane wise variation of lateral acceleration distribu-
Vehicle type Number of lanes Lateral acceleration
tion is not showing any pattern. In some cases, the 85th
(m/s2)
percentile value of lateral acceleration on 3-lane road is lower
Average 85th 95th than that of 2-lane roads whereas in some cases it is vice
2W 2-lane 0.83 1.78 2.09 versa. Hence, the average value of lateral acceleration cannot
3-lane Data not available explain the overall behavior of lateral maneuverability of ve-
3W 2-lane 0.55 1.33 2.54 hicles in case of non-lane based heterogeneous traffic stream.
3-lane 0.48 0.90 1.63
The lateral maneuverability of vehicles not only depends on
H-back 2-lane 0.40 0.94 1.66
3-lane 0.34 1.05 2.22
the road width or number of lane but also other traffic pa-
Sedan 2-lane 0.33 0.95 1.76 rameters like the surrounding traffic condition. In this study,
3-lane 0.39 0.79 1.55 we have considered the probe vehicle speed itself as a traffic
SUV 2-lane 0.38 0.96 1.93 stream parameter which reflects the average speed of the
3-lane 0.38 1.05 1.98 traffic stream.
According to past literature, the threshold value of lateral
acceleration for comfort is 1.8 m/s2 with medium comfort and
discomfort levels of 3.6 m/s2 and 5 m/s2 (Cheng, 2007; Dixon
acceleration are analysed. The 85th and 95th percentiles
and Rohani, 2008; Kilinc and Baybura, 2012; Voigt et al., 2003).
values of the lateral acceleration were obtained, and re-
In the present study, the 4-lane and 6-lane divided roads have
lationships are established with operating speed to see the
medium comfort level, and this comfort level accounted for
extreme trend of lateral acceleration of different vehicles.
more than 85% of the data. The result is similar to the values
Fig. 8 shows the percentile distribution of the lateral
obtained by Xu et al. (2015). The lateral acceleration was all
acceleration for combined data (five different cities data) for
smaller than 3 m/s2, and most were smaller than the
five types of instrumented vehicles for 2-lane and 3-lane
threshold value of comfort for normal driving conditions on
roads separately. The 85th percentile of lateral acceleration
straight mid-block sections.
is lesser for 3-lane roads from 2-lane roads. The 85th
To describe the detailed statistical properties of lateral
percentile lateral acceleration of 2W for 2-lane roads is
acceleration, the lane wise probability distribution function
highest, and then it decreases gradually for 3W, H-back,
for both types of roads (i.e., 2-lane and 3-lane) are analysed for
sedan, and SUV. The difference in values of 85th percentile
all the five types of vehicles. In an attempt to find an appro-
lateral acceleration between 2-lane and 3-lane increases as
priate distribution for representing lateral acceleration data
the size of the vehicle decreases. This is due to the higher
(3-lane and 2-lane roads) for all vehicle types, several single
maneuverability capability of 2W and 3W than the cars. The
distribution models were implemented in Easyfit software
85th percentile values of lateral acceleration for different
which ranks the fitted distributions based on goodness-of-fit
vehicle types on different roads are presented in Table 6.
tests and accordingly four statistical models (Normal,

Table 7 e Comparison of performance measures for various probability models fitted for lateral accelerations of different
vehicles on two-lane and three-lane roads.
Vehicle type 2-lane roads 3-lane roads
Sample size Distribution LL AIC Sample size Distribution LL AIC
2W 30,268 Normal 31,847 63,698
Exponential 48,860 97,722 Significant data not available
Gamma 21,321 42,646
Burr 72,011 144,028
3W 36,155 Normal 75,520 151,044 35,623 Normal 19,619 39,242
Exponential 134,500 269,002 Exponential 51,462 102,926
Gamma 37,455 74,914 Gamma 8546 17,097
Burr 242,580 485,166 Burr 86,440 172,886
H-Back 117,792 Normal 18,481 36,966 38,014 Normal 8437 16,879
Exponential 81,448 162,898 Exponential 70,403 140,808
Gamma 4168 8340 Gamma 1145 2286
Burr 138,210 276,426 Burr 115,000 230,006
Sedan 93,805 Normal 24,338 48,680 62,818 Normal 18,792 37,588
Exponential 137,930 275,862 Exponential 111,150 222,302
Gamma 793.7 1591.4 Gamma 784.2 1572.5
Burr 230,370 460,746 Burr 181,770 363,546
SUV 79,337 Normal 25,836 51,694.2 42,350 Normal 12,997 25,998
Exponential 125,220 250,451 Exponential 69,659 139,320
Gamma 5144 10,310 Gamma 1746 3497
Burr 204,720 409,473 Burr 114,560 229,126
398 J. Traffic Transp. Eng. (Engl. Ed.) 2018; 5 (5): 386e405

Fig. 9 e Probability distribution of all five types of vehicles on 2-lane and 3-lane roads separately. (a) 2W 2-lane. (b) 3W 2-
lane. (c) 3W 3-lane. (d) H-back 2-lane. (e) H-back 3-lane. (f) Sedan 2-lane. (g) Sedan 3-lane. (h) SUV 2-lane. (i) SUV 3-lane.

Exponential, Gamma and Burr) are fitted well to our data at 5%
significance level. To know the best-fitted distribution for all
five vehicle types, the log-likelihood (LL) value and akaike's
information criterion (AIC) values are estimated as described
in section 4.1. The probability density functions of lateral
acceleration for different vehicle types at 5% significance
level and the best-fitted distributions are presented below in
Table 7.
It can be observed that Normal, Exponential, Gamma and
Burr distributions are the distributions fitted well to all vehi-
cles at 5% significance level. Gamma distribution is resulting
in the maximum LL value and minimum AIC/BIC values for all
 J. Traffic Transp. Eng. (Engl. Ed.) 2018; 5 (5): 386e405
the vehicle types in all road conditions. The best-fitted dis-
tributions (Gamma PDF) for different vehicles at different
roads are shown in Fig. 9 below.
The parameters for all the best-fitted distribution are pre-
sented below in Table 8.
From Table 8 it is observed that the shape and scale
parameters of Gamma distribution fitted to different
vehicles have nearly same values of a and b, indicating a
similar stretch of the Gamma distributions fitted with same
moment skewness. The value a >1 indicates that the
Gamma distributions assume a mounded (unimodal), but
skewed shape. The similar values of shape parameters of all
the different population groups indicate they are the
different distributions with the same moment skewness.
Hence, the distributions fitted to the lateral acceleration of
different vehicles on both 2-lane and 3-lane roads describe
both the driver-vehicle behavior combined as a single entity.
However, a detailed understanding of the lateral maneuver-
ability of different vehicles on the non-lane based mixed
traffic stream can be achieved by studying the variation of the
lateral acceleration with the other stream parameters like
speed, density, etc., on roads with different width. So, in the
next section, the variability of lateral acceleration with respect
to operating speed at the different width of the road, i.e.
number of lanes is studied.
5.2. Effect of driving speed on lateral acceleration
In the present section the behavior of lateral acceleration with
regard to speed for straight roads in mixed non-lane discipline
traffic on 2-lane and 3-lane one way roads is studied. The re-
lationships are established for the 85th percentile lateral ac-
celeration data for different operating speed. The 85th
percentile value of lateral acceleration data for different speed
ranges (say every 5 km/h group size) is modelled. These 85th
percentile lateral acceleration values are plotted over the box
plots for each speed group (indicated by the blue scatter points
in Fig. 10). It can be observed that the lateral acceleration
follows a non-linear relationship with operating speed of the
vehicles. The two-term exponential models (Eq. (4)) fitted
well for the 85th percentile lateral acceleration values in all
the cases at 5% significance level.
y ¼ aebx þ cedx (4)
where y is lateral acceleration of different vehicles in m/s2, x is
operating speed of different vehicles in km/h.
Table 8 e Detail parameters of different PDFs of lateral
acceleration of Gamma distribution.
Vehicle type 2-lane 3-lane
a b a b
SUV 1.7990 0.24088 1.7991 0.23495
Sedan 1.7725 0.22774 1.8483 0.22075
H-back 1.6715 0.42285 1.9199 0.19462
3W 1.4164 0.41443 1.6352 0.31834
2W 1.3864 0.52477 Data not available
Note: a ¼ shape parameter, b ¼ scale parameter for Gamma
distribution.
399
In case of a two-term exponential function, there are two
decay modes. The two-term exponential model fitted the data
well as compared to the single term exponential function. The
details of the fitted models are presented below in Table 9.
It is observed that lateral acceleration increases with the
increase of speed at the lower speed ranges, i.e., less than
20 km/h. Then it obtains the maximum value of lateral ac-
celeration within the speed range of 15e25 km/h speed, for all
vehicle types in all cities. Afterwards, the lateral acceleration
decreases with the increase in speed. This is due to the higher
weaving action of vehicles at a lower speed depending on the
surrounding traffic. The vehicle feels safe to veer frequently at
a lower speed, whereas at higher speed vehicle feels unsafe to
veer laterally. It is observed that at higher speed the lateral
acceleration follows an inverse relationship, i.e., with the in-
crease in speed, the magnitude of lateral veer reduces. The
threshold value for changing the lateral behavior lies around
20 km/h speed range for all types of vehicles. It is also
observed that motorized 2W obtain a higher lateral accelera-
tion comparing to all vehicle types due to its high maneu-
verability. The lateral acceleration decreases as the size of
vehicles increases (or maneuverability decreases). The 95% of
lateral acceleration data lies within ± 2 m/s2, and 85% of data
lies below ± 1 m/s2 for all vehicle types.
The adjusted R2 value and the RMSE (root mean squared
error) value for each fitted model are also presented in Table 9.
It can be observed that the model presents the observed data
with more than 80% adjusted the R2 value in all cases. The
RMSE value indicates that the absolute fit of the model to
the data or the closeness of the observed data to the
predicted model. Here, the lower values of RMSE indicates a
better fit of the models at 5% significance level.
6. Relationship between lateral and
longitudinal acceleration
As observed from the literature (Biral et al., 2005; Biral and Lot,
2009; Hisaoka et al., 1999; Vaiana et al., 2014) that, the g-g di-
agram is the most efficient way to represent the driving
behavior by considering both the dynamic parameters (lateral
and longitudinal acceleration). Here we have plotted the g-g
diagram for each vehicle type with combined data of all cities.
The lateral and longitudinal acceleration data for each type of
vehicles are grouped in six equal intervals of speed every
20 km/h ranging from 0 to 100 km/h. The diagrams show the
lateral accelerations (ay) and longitudinal A/D (ax) normalized
with respect to gravity (g) in both horizontal and vertical axis
respectively for each speed group (Figs. 11e15). ax/g is longi-
tudinal acceleration/gravity (g) and ay/g is lateral acceleration/
gravity (g).
It can be observed that the g-g diagrams of all five type of
vehicles produce interaction envelopes which are a small
subset of the adherence circle or the friction circle of the tire.
These envelopes represent “driver capability envelope” Biral
et al. (2005). It can be observed that the interaction envelops
of the vehicles decreases as the size of vehicle increases. It is
observed that the g-g diagrams change their size with speed.
At lower speed (0e20 km/h) the g-g diagram covers a lesser
area compared to the speed group of 20e40 km/h. In the
400 J. Traffic Transp. Eng. (Engl. Ed.) 2018; 5 (5): 386e405

Fig. 10 e Fitted regression models on the lateral acceleration scattered data of all five types of vehicles on 2-lane and 3-lane
roads (significant data was not available for 2W 3-lane). (a) 2W 2-lane. (b) 3W 2-lane. (c) 3W 3-lane. (d) H-back 2-lane.
(e) H-back 3-lane. (f) Sedan 2-lane. (g) Sedan 3-lane. (h) SUV 2-lane. (i) SUV 3-lane.

case of lower to medium speed, the size of the g-g diagram speed. In all speed groups, the shape of the g-g diagram is
enlarges in all vehicle classes except motorised two- approximately round. The shape and size of each g-g
wheelers and three-wheeler. For all three type of cars, the diagram changes with the driving capability of a different
size of g-g diagram shrinks with the further increase in vehicle on plain terrain.
 J. Traffic Transp. Eng. (Engl. Ed.) 2018; 5 (5): 386e405 401

Table 9 e Details of all the fitted regression models.

a b c d R2 RMSE
2W 2-lane 2.5370 0.1219 2.3450 0.0301 0.8084 0.1539
3-lane Data not available
3W 2-lane 1.3550 0.0248 1.425 0.1127 0.8251 0.08833
3-lane 1.0300 0.0183 1.0940 0.1582 0.8641 0.1259
H-back 2-lane 0.8551 0.0183 0.9633 0.1618 0.8874 0.0540
3-lane 1.0700 0.0791 1.0580 0.0181 0.8292 0.0554
Sedan 2-lane 0.8291 0.0166 0.9158 0.1813 0.8407 0.0837
3-lane 0.8597 0.0092 0.8973 0.1653 0.8107 0.0673
SUV 2-lane 0.7202 0.0138 0.8216 0.2039 0.8054 0.0707
3-lane 2.4070 0.0784 2.3950 0.0344 0.8334 0.0812

Fig. 11 e Normalized longitudinal and lateral accelerations (g-g plot) for 2W. (a) 0e20 km/h (b) 20e40 km/h (c) 40e60 km/h.

Fig. 12 e Normalized longitudinal and lateral accelerations (g-g plot) for 3W. (a) 0e20 km/h (b) 20e40 km/h (c) 40e60 km/h.
402 J. Traffic Transp. Eng. (Engl. Ed.) 2018; 5 (5): 386e405

Fig. 13 e Normalized longitudinal and lateral accelerations (g-g plot) for H-back. (a) 0e20 km/h (b) 20e40 km/h (c) 40e60 km/h
(d) 60e80 km/h.

To see the interaction between the lateral and longitudinal
acceleration more clearly the relationship between the two
dynamic parameters are established for all five vehicle types.
The average lateral maneuver for different longitudinal ac-
celeration values of different vehicle types is presented in the
following Fig. 16.
It can be observed that the average lateral maneuver of all
types of vehicles linearly increases with the increase in lon-
gitudinal acceleration of the vehicles. In case of mixed traffic
condition, the vehicle tries to overtake and veer frequently.
Hence it accelerates and veers simultaneously to pass through
(filter) the traffic. Hence as the veering maneuver increases
the acceleration also increases while overtaking. From Fig. 16,
it is clear that the lateral and longitudinal maneuver
(accelerating) of the vehicles or the veering of motorized
two-wheeler is higher than the cars (as observed in Figs.
11e15).
7. Conclusions
In the current study, different types of vehicles (three types of
cars, motorized 3W and 2W) equipped with Video VBOX (GPS
based instrument) are driven in different traffic streams

Fig. 14 e Normalized longitudinal and lateral accelerations (g-g plot) for Sedan. (a) 0e20 km/h (b) 20e40 km/h (c) 40e60 km/h
(d) 60e80 km/h.
 J. Traffic Transp. Eng. (Engl. Ed.) 2018; 5 (5): 386e405
Fig. 15 e Normalized longitudinal and lateral accelerations (g-g plot) for SUV. (a) 0e20 km/h (b) 20e40 km/h (c) 40e60 km/h
(d) 60e80 km/h.
(varying from low flow to high) of different roads (4-lane to 6-
lane divided roads) located in five different cities of India.
Trajectories of several such vehicles (obtained from Video
VBOX at 10 Hz) are collected by driving them more than
250 km each and analyzed for their speed, longitudinal and
lateral acceleration and deceleration (A/D) values. Observed
A/D values in the study represent the A/D limits of different
vehicles under normal driving condition. Some of important
observations of this study are listed below.
(1) Cumulative percentage curve of longitudinal and lateral
A/D represents that A/D trends do not change their
slope significantly (also referred as a breakpoint) at
exactly 85th percentile value. The breakpoint lies in
between the 85th and 95th percentile. Hence the use of
85th percentile value philosophy for design purpose
needs a relook.
Fig. 16 e Relation between the average lateral acceleration
with the average longitudinal acceleration for all five types
of vehicles.
403
(2) The A/D limits (at 85%) are found comparable for each
vehicle type. However at 95th percentile level A/D show
significance difference (i.e., for SUV cars acceleration
and deceleration values are 1.38 m/s2 and 1.63 m/s2
respectively). The difference in lateral acceleration (i.e.,
0.95 m/s2 for cars to 1.78 m/s2 for motorized 2W at 85th
percentile level) due to the heterogeneity in vehicle
types (change in the dynamic and static capability of
different vehicles).
(3) The probability distribution (PDF) of the longitudinal A/
D of all types of vehicles follows Weibull and Gamma
distributions at 5% significance level except decelera-
tion of Sedan cars. However, PDF for lateral acceleration
exhibit Gamma distribution at 5% significance level for
all vehicle types.
(4) The 85th and 95th percentile A/D curves with the
operating speed, depict a linearly decreasing trend for
all five types of vehicles. The maximum acceleration
occurs at lower speeds (vehicle starting from rest), the
ability to accelerate decreases with increase in vehic-
ular speeds. However, the lateral acceleration in-
creases with the increase in speed initially and attains
the maximum value at around 20 km/h speed for all
type of vehicles. Lateral acceleration decreases with
further increase in speed. The 85th/95th percentile
lateral acceleration model concerning the operating
speed of vehicles follows a two-term exponential
relationship at 5% significance level for all vehicle
types.
(5) Using both the longitudinal A/D and the lateral accel-
eration data, the g-g diagram produces an interaction
envelope known as “driver capability envelope” for all
types of vehicles. It can be observed that the size of the
interaction envelope for motorized 2W is higher than
motorized 3W and cars. This indicates that the size of
404 J. Traffic Transp. Eng. (Engl. Ed.) 2018; 5 (5): 386e405
the interaction envelope increases with the increase in
vehicle maneuverability.
(6) The lateral acceleration increases linearly with the in-
crease in the longitudinal acceleration for all vehicles.
This indicates that for all vehicles, higher maneuver-
ability exists at lower speeds as longitudinal accelera-
tion are higher at a lower speed.
The relation between the dynamic parameters (speed and
lateral/longitudinal acceleration), can very well represent the
driving behavior of vehicles in mixed traffic with non-lane
discipline. These A/D values and lateral acceleration can be
used in the operating speed prediction models and also
establish the bounding values in case of simulation models.
The probability distribution models and the A/D behavior with
operating speed of different vehicles can also be used in
simulation models for the realistic representation of such
traffic stream. The model for lateral acceleration can be used
for evaluation of vehicles lateral weaving in simulated tra-
jectories for such traffic. Also, the lateral acceleration is an
indicative of the lateral force coefficient of the highway, which
is a key factor for lateral stability and driving safety. The
lateral acceleration (a measure of lateral comfort) is directly
related to the horizontal alignment and can be used to
determine the applicable range of super-elevation ratios. The
present study is limited to straight mid-block sections of 4-
lane and 6-lane divided roads in plain terrain. This study of
the lateral and longitudinal behavior of vehicles can be
extended to the roads with other geometry conditions like
horizontal and vertical curved roads, intersections, etc.
Further such behavioral study for heavy vehicles can also be
conducted.
Conflicts of interest
The authors do not have any conflict of interest with other
entities or researchers.
Funding
This research did not receive any specific grant from funding
agencies in the public, commercial, or not-for-profit sectors.
references
Adnan, M.A., Zaiddunin, N.I., Sulaiman, N., et al., 2013. Vehicle
speed measuring techniques using various speed detection
instrumentation. IEEE Business Engineering and Industrial
Application Colloquium 978 (1), 668e672.
Akcelik, R., Beseley, M., 2001. Acceleration and deceleration
models. In: 23rd Conference of Australian Institute of
Transport Research (CAITR 2001), Melbourne, 2001.
Akcelik, R., Biggs, D., 1987. Acceleration profile models for
vehicles in road traffic. Transportation Science 21 (1), 36e54.
Arasan, V.T., Koshy, R.Z., 2005. Methodology for modeling highly
heterogeneous traffic flow. Journal of Transportation
Engineering 131 (7), 544e551.
Bham, G., Benekohal, R., 2002. Development evaluation and
comparison of acceleration models. In: 81st Annual
Meeting of the Transportation Research Board, Washington
DC, 2002.
Biral, F., Lot, R., 2009. An interactive model of g-g diagrams of
racing motorcycle. In: 3rd (ICMEM), International Conference
on Mechanical Engineering and Mechanics, Beijing, 2009.
Biral, F., Lio, M.D., Bertolazzi, E., 2005. Combining safety margin
and user preferences into driving criterian for optimal
control-based computation of reference maneuvers for an
adas of the next generation. In: Intelligent Vehicle
Symposium, Proceedings of IEEE, Las Vegas, 2005.
Bokare, P.S., Maurya, A.K., 2016. Study of acceleration behaviour
of motorized three wheeler in India. Transportation
Research Procedia 17, 244e252.
Brooks, R.M., 2012. Acceleration characteristics of vehicles in
rural Pennsylvania. International Journal of Research and
Reviews in Applied Sciences 12 (3), 449e453.
Camacho-Torregrosa, F.J., Perez-Zuriaga, A.M., Compy-
Ungria, J.M., et al., 2013. New geometric design consistency
model based on operating speed profiles for road safety
evaluation. Accident Analysis and Prevention 61 (8), 33e42.
Car and Bike Team, 2016. Varand Bike: an NDTV Venture.
Available at: auto.ndtv.com/news/types-of-cars-1450327
(Accessed 9 May 2017).
Cheng, W.J., 2007. Study on the Evaluation Method of Highway
Alignment Comfortableness. Hebei University of Technology,
Tianjin.
Dey, P.P., Biswas, P., 2011. Acceleration of queue leaders at
signalized intersections. Indian Highways 39 (12), 49e54.
Dixon, K.K., Rohani, J., 2008. Methodologies for Estimating
Advisory Curve Speeds on Oregon Highways. FHWA-OR-RD-
08-04. FHWA, Washington DC.
Durrant, A.J., Hill, M.J., 2005. Technical Assessment of the DL2s
GPS Data Qulity (Including Comparative Data from VBOX3
and an Optical Speed Sensor. Race Technology Limited,
Strelley.
Eboli, L., Mazzulla, G., Pungillo, G., 2016. Combining speed and
acceleration to define car users' safe or unsafe driving
behaviour. Transportation Research Part C: Emerging
Technologies 68, 113e125.
Echaveguren, T., Basualto, J., 2003. Consistency analysis of
acceleration rate on the horizontal alignment with a
configuration of simple elements. In: Congreso Chileno de
Ingenieria de Transporte, Universidad de Concepcion.
Santiago, 2003.
Felip, E., Navin, F., 1998. Automobiles on horizontal curves:
experiments and observations. Transportation Research
Record 1628, 50e56.
Fildes, B.N., Rumbold, G., Leening, A., 1991. Speed Behaviour and
Drivers' Attitudes to Speeding, Wellington Road, Clayton.
Australia. Monash University Accident Research Centre,
Victoria.
Fitzpatrick, K., et al., 2000. Speed Prediction for Two Lane Rural
Highways. Federal Highway Administration, Washington DC.
Hisaoka, Y., Yamamoto, M., Okada, A., 1999. Closed-loop analysis
of vehicle behaviour during braking in a turn. JSAE Review 20
(4), 537e542.
Jimenez, F., Aparicio, F., Paez, J., 2008. Evaluation of in-vehicle
dynamic speed assistance in Spain: algorithm and driver
behaviour. IET Intelligent Transport Systems 2 (2), 132e142.
Kanagaraj, V., Asaithambi, G., Toledo, T., et al., 2015. Trajectory
data and flow characteristics of mixed traffic. Transportation
Research Record 2491, 1e11.
Katzourakis, D.I., Velenis, E., Abbink, D., et al., 2012. Race-car
instrumentation for driving behaviour studies. IEEE
Transactions on Instrumentation and Measurement 61 (2),
462e474.
 J. Traffic Transp. Eng. (Engl. Ed.) 2018; 5 (5): 386e405
Kilinc, A.S., Baybura, T., 2012. Determination of Minimum
Horizontal Curve Radius Used in the Design of
Transportation Structures, Depending on the Limit Value of
Comfort Criterion Lateral Jerk. TS06G- Engineering Surveying
Machine Control and Guidance, Rome.
Lamm, R., Choueiri, E.M., Hayward, J.C., 1988. Tangent as an
independent design element. Transportation Research
Record 1195, 13e20.
Liu, R., Bonsall, P., Young, W., 2003. Modelling safety-related
driving behaviour impact of parameter values. In: European
Transport Conference, Strasbourg, 2003.
Liu, J.P., Lin, S., Gao, J.S., 2010. Analysis and optimization for
operating speed model on straight section on highway.
China Journal of Highway and Transport 23 (S), 83e88.
Long, G., 2000. Acceleration characteristics of starting vehicles.
Transportation Research Record 1737, 58e70.
Mahapatra, G., Maurya, A.K., 2013. Study of vehicles lateral
movement in non-lane disciplined traffic stream on a
straight road. Procedia-Social and Behavioural Sciences 104
(2), 352e359.
Mahapatra, G., Maurya, A.K., 2015. Study on lateral placement
and speed of vehicles under mixed traffic condition. In:
Eastern Asia Society for Transportation Studies, Cebu, 2015.
Mahapatra, G., Maurya, A.K., Minhans, A., 2016. Lateral
movement of vehicles under weak lane discipline
heterogeneous traffic. Jurnal Teknologi 78 (4), 27e35.
Maurya, A.K., Bokare, P.S., 2012. Study of deceleration behaviour
of different vehicle types. International Journal for Traffic
and Transportation Engineering 2 (3), 253e270.
Mehar, A., Chandra, S., Velmurugan, S., 2013. Speed and
acceleration characteristics of different types of vehicles on
multi-lane highways. European TransportyTransport
Europei 55 (1), 1e12.
Montella, A., Pariota, L., Galants, F., et al., 2014. Prediction of drivers'
speed behaviour on rural motorways based on an instrumented
vehicle study. Transport Research Record 2434, 52e62.
Odhams, A.M., Cole, D.J., 2004. Models of Driver Speed Choice in
Curves. Cambridge University Engineering Department,
Cambridge.
RaiChowdhury, M., Rao, S., 1989. Acceleration characteristics of
vehicles at signalized intersection. Journal of the Institution
of Engineers (India) 70, 35e38.
Reymand, G., Kemeny, A., Droulez, J., et al., 2001. Role of lateral
acceleration in curve driving: driver model and experiments
on real vehicle and driving simulator. Human Factors 43 (3),
483e495.
Richl, L., Sayed, T., 2011. Effect of speed prediction models and
perceived radious on design consistency. Canadian Journal
of Civil Engineering 32 (2), 388e399.
Ritchie, M.L., McCoy, W.K., Welde, W.L., 1968. A study of the
relation between forward velocity and lateral acceleration in
curves during normal driving. Human Factors 10 (3), 255e258.
Vaiana, R., Iuele, T., Astarita, V., et al., 2014. Driving behaviour and
traffic safety: an acceleration-based safety evaluation procedure
for smartphones. Modern Applied Science 8 (1), 88e96.
405
Voigt, A.P., Fenno, D.W., Borchardt, D.W., 2003. Evaluation of
Vehicle Speeds on Freeway Connector Ramps in Houston,
Texas. FHWA/TX-03-4318-1. Texas Transportation Institute,
College Station.
Wang, J., Dixon, K., Li, H., et al., 2004. Normal acceleration
behaviour of passenger vehicles starting from rest at all-way
stop-controlled intersections. Transportation Research
Record 1883, 158e166.
Wang, J., Dixon, K., Li, H., et al., 2005. Normal deceleration
behaviour of passenger vehicles at stop sign controlled
intersections evaluated with in-vehicle global positioning
system data. Transportation Research Record 1937, 120e127.
Xu, J., Yang, K., Shao, Y., et al., 2015. An experimental study on
lateral acceleration of cars in different environments in
Sichuan, southwest China. Discrete Dynamics in Nature and
Society 2015, 494130.
Xu, J., Lin, W., Wang, X., et al., 2017. Acceleration and deceleration
calibration of operating speed prediction models for two-lane
mountain highways. Journal of Transportation Engineering,
Part A: Systems 143 (7), 04017024.
Zhang, X., Peng, J., Xia, P., 2010. Design of roadway test system for
motor vehicle brake performance and its evaluation methods.
In: International Conference on Computer, Mechatronics,
Control and Electronic Engineering, Changchun, 2010.
Zhu, X., Hu, X., Chiu, Y., 2013. Design of driving behaviour pattern
measurement using smartphone global positioning system
data. International Journal of Transportation Science and
Technology 2 (4), 269e288.
Geetimukta Mahapatra is currently a
research scholar in the Department of Civil
Engineering at Indian Institute of Technol-
ogy Guwahati. Her research interests include
driver behavior study in the mixed traffic
stream, traffic data collection and analysis
and road safety analysis.
Akhilesh Kumar Maurya is currently an
associate professor in Department of Civil
Engineering at Indian Institute of Technol-
ogy Guwahati. His research interests are
traffic data collection and analysis, traffic
flow theory and modeling, driver behavior
modeling, road safety and accident analysis,
vehicular emission modelling, etc.