Comparison of Tailpipe Gaseous Emissions for RDE and WLTC Using SI Passenger Cars

Abstract

The drive characteristics and gaseous emissions of legislated Real Driving Emissions (RDE) test data from 8 different spark ignition vehicles were compared to data from corresponding Worldwide harmonized Light vehicles Test Cycle (WLTC) tests. The effect of the official RDE exclusion of cold start and idling on the RDE test, and the effect of the use of the moving averaging window (MAW) analysis technique, were simultaneously investigated. Specific attention was paid to differences in drive characteristics of the three different driving modes and the effect this had on the distance-based CO2, CO and NOx emission factors for each. The average velocity of the RDE tests was marginally greater than the WLTC tests, while the average acceleration was smaller. The CO2 emission appeared on average 4% lower under the RDE tests compared to the WLTC tests, while the CO was 60% lower. The NOx values were 34% lower under the RDE testing, and appeared to be linked to the average acceleration. No link was seen for the maximum acceleration or deceleration, indicating that this is not a good indicator for test cycle emissions. The exclusion of cold start and idling decreased all RDE emissions. RPA (Relative Positive Acceleration) had little correlation with CO2, CO and NOx distance-based emissions, and was shown to be uncorrelated with any mass-rate emissions. The range of RPA values seen was much greater for RDE tests than WLTC tests, with individual RDE tests having variable values for each drive mode. The application of the MAW technique had minimal effect on the CO2 distance-based emission, but it appeared to decrease CO and NOx emissions by 12% and 21% on average respectively. The MAW also decreased the variation in emissions across different modes.

Full text

Abstract
The drive characteristics and gaseous emissions of legislated Real
Driving Emissions (RDE) test data from 8 dierent spark ignition
vehicles were compared to data from corresponding Worldwide
harmonized Light vehicles Test Cycle (WLTC) tests. The eect of the
ocial RDE exclusion of cold start and idling on the RDE test, and
the eect of the use of the moving averaging window (MAW)
analysis technique, were simultaneously investigated. Specic
attention was paid to dierences in drive characteristics of the three
dierent driving modes and the eect this had on the distance-based
CO2, CO and NOx emission factors for each. The average velocity of
the RDE tests was marginally greater than the WLTC tests, while the
average acceleration was smaller. The CO2 emission appeared on
average 4% lower under the RDE tests compared to the WLTC tests,
while the CO was 60% lower. The NOx values were 34% lower under
the RDE testing, and appeared to be linked to the average
acceleration. No link was seen for the maximum acceleration or
deceleration, indicating that this is not a good indicator for test cycle
emissions. The exclusion of cold start and idling decreased all RDE
emissions. RPA (Relative Positive Acceleration) had little correlation
with CO2, CO and NOx distance-based emissions, and was shown to
be uncorrelated with any mass-rate emissions. The range of RPA
values seen was much greater for RDE tests than WLTC tests, with
individual RDE tests having variable values for each drive mode. The
application of the MAW technique had minimal eect on the CO2
distance-based emission, but it appeared to decrease CO and NOx
emissions by 12% and 21% on average respectively. The MAW also
decreased the variation in emissions across dierent modes.
Introduction
There are three main legislated steps to the method to control
regulated pollutant emissions [1]. A type approval test ensures that
any new vehicle designs adhere to the aforementioned emission
standards. Conformity of production then requires that all cars be
manufactured to those same standards. Finally, in-service conformity
and durability requirements ensure that the vehicle maintains similar
emissions factors after being sold [2]. Since the 1990’s a set of
European Emission Standards for light duty and heavy duty gasoline,
diesel, LPG and CNG vehicles have been launched for European
Union (EU) type approval testing. From Euro 3 up until Euro 6,
Europe employed the New European Driving Cycle (NEDC) for the
certication of cars [3]. However the NEDC test has some downfalls,
which have been widely discussed in the literature [4, 5, 6]. The main
arguments are that the NEDC test procedures are outdated for current
vehicle technologies and unrepresentative of real-world driving, as
well as too lax, allowing car manufacturers to ‘play the system’ to
their advantage and give emissions values that can never be achieved
in the real world [7,8].
In response to these concerns, the World Forum for the
Harmonization of Vehicle Regulations (WP.29) of the United Nations
Economic Commission for Europe (UNECE) launched a program to
develop a new Worldwide harmonized Light vehicles Test Cycle
(WLTC) and procedure (WLTP) [8]. They aimed to develop a cycle
that represented average worldwide driving characteristics, and to
have it tested using a world-harmonized type approval testing
procedure. The WLTC is not a single cycle, but a set of dierent
cycles to be used on dierent vehicles with dierent characteristics.
WLTC cycle class 3 is used by the majority of European cars [8].
The WLTC has been formulated for the currently enforced Euro 6
legislation, and diers from the NEDC in various ways, including
more aggressive driving styles and a greater range of engine
operating points [9]. Although the WLTC test more accurately
replicates the type of driving behaviors seen in real world driving
than the NEDC, it still carries the same disadvantages of any standard
laboratory test cycle. One disadvantage is that these laboratory test
cycles cannot adequately cover the wide range of ambient and driving
A Comparison of Tailpipe Gaseous Emissions for RDE and
WLTC Using SI Passenger Cars
2017-01-2391
Published 10/08/2017
Daisy Thomas and Hu Li
University of Leeds
Xin Wang, Bin Song, Yunshan Ge, and Wenlin Yu
Beijing Institute of Technology
Karl Ropkins
University of Leeds
CITATION: Thomas, D., Li, H., Wang, X., Song, B. et al., "A Comparison of Tailpipe Gaseous Emissions for RDE and WLTC Using
SI Passenger Cars," SAE Technical Paper 2017-01-2391, 2017, doi:10.4271/2017-01-2391.
Copyright © 2017 SAE International
Downloaded from SAE International by Daisy Thomas, Thursday, October 26, 2017

conditions seen in real world vehicle use [1]. This means that while
vehicles may comply with emission limits in laboratory tests, they
could have substantially higher emissions on the road under
conditions outside those tested [10]. These standard cycles are also
very predictable, allowing car manufacturers the possibility of
‘cheating’ the tests, most notably in the VW scandal of 2016 [11].
It has therefore been deemed necessary for a complimentary test
procedure to be formulated alongside the WLTC to address the
above issues, called a real driving emissions (RDE) test. This will
limit the trend of overly narrow optimization of emissions control
technologies that is currently such a problem for climate and air
quality, as well as encouraging the adoption of novel emission
abatement technologies [1].
The individual results from an RDE test are not reproducible, and this
lack of repeatability creates uncertainties that have to be accounted
for when designing emissions limits [12]. For example, driver
behavior is problematic. While driver behavior is pre-determined
with random cycle testing, this is not the case for Portable Emissions
Measurement System (PEMS) on-road testing. Similarly, weather
conditions, particularly temperature, are not dened. It was therefore
necessary to dene appropriate boundary conditions to limit the
ability of the manufacturer to manipulate results [1], which will be
discussed in the next section.
Many papers have previously explored the regulated emissions
from real world driving compared to type approval testing.
However, most have been limited to comparisons between the old
NEDC test and non-legislated on-road driving emissions. Merkisz
et al. [13] compared road tests for a 2006 year gasoline vehicle
with NEDC tests. Each road test consisted of the same route, 76
km long, with emissions of carbon monoxide (CO), nitrous oxide
(NOx) and carbon dioxide (CO2) measured in real-time using a
Semtech DS analyzer. They found that the road tests gave lower
values of NOx and CO compared to the NEDC test, while they
gave higher values of CO2. This same trend was seen under pure
urban driving conditions while for pure extra-urban driving the CO
values under real driving were increased past the NEDC values,
and NOx increased but remained marginally below the NEDC
values. The trend for increasing on-road CO2 emission was also
witnessed by Weiss et al. [14], which studied the on-road emissions
of 12 light duty gasoline and diesel vehicles of Euro 3-5 emission
limits, and compared them to the NEDC results. This paper found
that NOx emissions were also higher for on-road testing than
NEDC tests, in contrast to Merkisz et al. [13]. The CO2 exceeded
the 130 g/km emission limits, while NOx remained below its Euro
limit. For CO this paper saw varying test results, but with a trend
toward increasing emission under real-world conditions than the
NEDC, so again contrasting with Merkisz et al. [13]. May et al.
[15] used a Semtech-D PEMS analyzer to measure the emissions of
a gasoline vehicle over pre-selected routes, and these were then
compared to NEDC and WLTC tests performed for that same
vehicle. Their results showed increasing emissions of CO2, CO and
NOx compared to both NEDC and WLTP tests, with the WLTP
results being higher than the NEDC results for all tests. Both CO2
and CO emissions were still within their legislated limits, while
NOx was on average 23% above the limit. These results are
generally in agreement with those of Weiss et al. [14]. Merkisz et
al. [16] conducted an extensive study, testing 150 dierent Euro4
and Euro 5 cars on RDE tests and comparing results to
corresponding NEDC and WLTC tests. A Semtech DS analyzer
was the PEMS used for this study, and the protocol being
considered for EU RDE legislation at the time was employed. The
protocol used does however vary from that in legislation today,
having dierent speed delineations between modes and the
inclusion of cold start and idling. This paper found that CO and
NOx emission on RDE tests were both around 80% of their
legislated limits, while WLTC gave values equal to just 31% and
60% of those respective limits. The NEDC test gave far lower NOx
emissions than the WLTC in this study, while the CO gave
approximately equal values. No work into CO2 was conducted.
Merkisz et al. [13] performed an analysis into the engine operating
parameters during their road tests, and related the dynamic behavior
to the emission of pollutants. They found that the highest emissions
of CO per unit time were for accelerations from -0.6m/s2 to 1.4 m/s2
for vehicle speeds from 2 to 24m/s. They found the CO emission
clearly increases for higher levels of acceleration. For NOx, speeds of
4 - 12 m/s with -0.6 - 1.8 m/s2 acceleration, and 10 - 26 m/s with -0.2
- 1 m/s2 acceleration produced the highest emission peaks. On
comparison of the distance-specic emissions during individual test
portions to equivalent NEDC portions, this study again concluded
that acceleration and cruise velocity for these sections are the most
inuential factors on toxic CO and NOx emissions. Few other papers
have gone into detail regarding the engine operating conditions
during testing and their correlation with emissions. Similarly, few
other studies have investigated how the drive properties aect the
production of gaseous pollutants.
In order to elucidate the link between the emission of CO2, CO and
NOx over the RDE compared to the WLTP cycles, further study is
clearly required. In particular, work relating the dynamic behavior
required of the newest version of the RDE test procedure to the
WLTP procedure would be useful to see how the changes in driving
styles aect the pollutant emissions. It is the aim of this paper to
compare the driving styles used in the latest RDE testing to those
legislated in the WLTC, and investigate how this appears to aect the
relative pollutant emissions.
EU RDE Legislation
An RDE element using PEMS, as discussed above, is now being
brought into EU legislation [17]. This consists of portions of
dierent driving styles, performed where possible in the order
presented below:
1. 34% urban operation, characterized by vehicle speeds up to
60 km/h
2. 33% rural operation, characterized by vehicle speeds between
60 and 90 km/h
3. 33% motorway operation, characterized by vehicle speeds over
90 km/h
There are also environmental conditions that must be adhered to in
order for the test to be valid:
1. The test must be performed under ambient conditions of 0°C ≤
T ≤ 30°C or ‘extended conditions’ of -7°C ≤ T ≤ 0°C or 30°C ≤
T ≤ 35°C.
Downloaded from SAE International by Daisy Thomas, Thursday, October 26, 2017

2. The test must be performed at a moderate altitude of less than or
equal to 700 meters above sea level, or an ‘extended altitude’ of
700 m ≤ altitude ≤ 1300 m.
The emissions values from a RDE test can be calculated using the
Moving Averaging Window (MAW) method outlined in the current
EU legislation [18]. The second-by-second emission rates in g/s
are averaged over moving averaging windows, the duration of
which is determined by a reference quantity of CO2. The principle
is that the mass emissions are not calculated for the complete data
set, but for sub-sets of the complete data set, the length of these
sub-sets being determined so as to match the CO2 emissions over
the WLTC. The window then moves forward in the same
increments as the measurement interval once this reference
quantity has been reached. The value of average CO2 emissions for
each window are recorded and compared to the vehicle CO2
emissions versus average speed measured at type approval on the
WLTC test, called the “vehicle CO2 characteristic curve”. The
windows are also categorized into the three speed classes (urban,
rural and motorway) dened above. The test is ‘complete’ when it
is comprised of at least 15% urban, rural and motorway windows,
out of the total number of windows. The test is ‘normal’ when at
least 50% of the windows are within the primary tolerance
(normally ±25%) of the characteristic curve values. The windows
are weighted according to their similarity to the reference CO2
curve, and then total emissions from the test, per km or per kWh,
along with the average concentrations, are calculated from the
normal windows.
The following data points are excluded from the calculation of the
CO2 mass:
1. Cold start (when engine coolant temperature has not yet reached
70 °C)
2. Durations with vehicle speeds under 1 km/h (idle)
3. Durations where the vehicle engine is switched o
Methodology
Testing Vehicles and Ambient Conditions
A range of gasoline powered spark ignition passenger vehicles were
selected, and used to perform WLTC and RDE testing cycles in
compliance with the latest EU emissions legislation. The vehicles
were all light duty vehicles of class M1, and were suciently varied
so as to contain a mixture of port fuel injection (PFI) and direct
injection (DI) fuel delivery systems, and a mixture of naturally
aspirated (NA) and turbocharged (T) engines. All vehicles are Euro
5 emission compliance, with the exception of a hybrid electric
vehicle, of Euro 2 emission compliance. All vehicles used the same
three way catalyst (TWC) emission reduction technique. Table A1
in the appendix summarizes some key characteristics of these
vehicles. The ambient conditions under which the RDE and WLTC
tests were performed and the drive properties of the tests are given
in table A2 in the appendix.
RDE Equipment and Testing Procedure
The RDE tests were performed using PEMS devices. Six of the tests
were performed using the Horiba OBS-ONE-GS PEMS equipment,
while two of the tests were performed using the AVL M.O.V.E Gas
PEMS iS equipment. Both systems measure the gaseous emissions of
CO, CO2, NO and NO2, the latter two of which can be used to
calculate NOx. Exhaust ow rates were also measured by these two
systems, to allow calculation of tailpipe mass emissions. These
PEMS devices allow gases regulated in current legislation, in
combination with a range of vehicle operating parameters, to be
measured in real time at a 10Hz frequency. The instruments used are
given in table 1.
Table 1. Instruments used by the Horiba and AVL PEMS devices.
For each test, the particular equipment was placed into the rear of the
vehicle, with the tailpipe attachment and heated sample probe
connected to the tailpipe as per the corresponding manufacturer
manual (gure 1). A global positioning system (GPS) antenna and an
ambient temperature and humidity sensor were placed on the roof of
the vehicle, while an on-board diagnostics (OBD) interface unit was
connected to the OBD port. These sensors were all connected to the
main gas sensor unit via universal serial bus (USB) cables. The
equipment was connected to batteries placed inside the car, so that
power was provided by a source external to the engine. The
equipment was then warmed up and calibrated according to the
manufacturer’s instructions before testing commenced.
Each test was performed according to the RDE test procedure
guidelines set out in Commission Regulation (EU) 2016/427 [18].
The tests were performed in two dierent Chinese cities; Beijing and
Xiamen. All tests began with an urban driving mode section,
followed by a rural driving mode section and then a motorway
driving mode section. These were apportioned to give percentage
ratios of approximately 34%, 33% and 33%, respectively.
Each vehicle test was commenced from a cold start, so in accordance
with the legislated RDE test procedure the following exclusions were
made to a set of the data: the warm-up period and any idling periods
where the speed was less than 1km/h were cut. In this paper, the term
“RDE exclusions” will refer to the exclusion of cold start and idling
periods of velocity less than 1km/h. The data with such exclusions is
called “MAW unapplied, with RDE exclusions”. A set of such data
then had the MAW method applied to it, and is called the “MAW
applied, with RDE exclusions”. For comparison, characteristics of the
tests are also presented both without the exclusions required by the
RDE legislation and without application of the MAW method. These
are termed “raw RDE”.
Downloaded from SAE International by Daisy Thomas, Thursday, October 26, 2017

Figure 1. Examples of a) the Horiba OBS ONE and b) the AVL M.O.V.E
PEMS devices in vehicles prior to RDE testing.
WLTC Equipment and Testing Procedure
The WLTC tests were conducted according to the standard test
procedure for WLTC testing (WLTP) as outlined in EU legislation
[19]. Cold start tests were performed for each car following the
legislated soak period. The WLTC cycle is split into four speed phase
sections: low, medium, high and extra high. However, the results in
this paper are presented as three dierent speed phases, with the
medium and high speed phases combined into one phase. This
increases the parity with the RDE results. The WLTC data presented
in this paper for each vehicle has been attained from the WLTC report
for that particular vehicle.
Data Analysis
The raw data for each RDE test consisted of a raw, pre-aligned data
le giving a range of operational parameters and pollutant emission
concentrations at 10 Hz frequency. These were averaged to give a 1
Hz frequency. Using the exhaust ow rate, the pollutant emission
concentrations were converted into mass emission rates. The velocity
distributions were then studied to discern which sections belonged to
the urban, rural and motorway driving modes. Figure 2 gives an
example velocity-time plot in green, with the mode divisions
indicated by red lines.
Figure 2. Example of the visual inspection and selection process to determine
drive mode divisions. The modes are separated by red lines, being classified
into urban, rural, motorway, urban, rural and urban sections chronologically.
In accordance with the ocial EU RDE legislation, the cold start and
idle data points were removed from a copy of the data to comprise
the “MAW unapplied with RDE exclusions” dataset. As instructed in
the legislation, cold start was delineated by the period of time before
the engine coolant temperature rst reached 70 °C, and idling was
counted as any time with velocity less than 1 km/h. The originals,
with cold start and zero velocity included, comprise the ‘raw RDE’
dataset. Quality Assurance on the data was performed and any
outliers were removed. The same processing was then performed on
both datasets for each vehicle, as described below.
Average distance-based mass emission factors for the dierent modes
of each dataset were calculated by summing the total mass emissions
for each section and dividing by the distance covered during that
section. Similarly, the average drive properties for each test were
calculated by averaging the particular property for each drive mode,
and then the total value for the whole cycle was an average of the
values for the three individual drive modes, taken proportional to the
distance covered in each mode.
Some of the properties were not in the raw data le, but were instead
calculated from other information. The stop time was calculated as a
sum of all the times when the velocity within the Raw RDE dataset
was less than 1km/h, while the number of stops was calculated as the
number of discontinuities in the ocial RDE data. Both of these
values were then divided by the distance covered to give distance-
based values.
The relative positive acceleration (RPA) was calculated using the
same method as outlined in May et al. [15], taken from methods
used to characterize vehicle trips in the development of the WLTC.
The calculation is described by equation 1, where ai is the
acceleration at time step i if ai is greater than 0 m/s2, vi is the
vehicle speed at time step i (m/s), Δt is the time increment and s is
the total trip distance (m).
(1)
An analysis was conducted into the eect of the use of the moving
averaging window (MAW) method on the emission of pollutant
gases. In this case the MAW technique was applied to the “MAW
unapplied with RDE exclusions” dataset manually (without the use of
the automated PEMS RDE software) to attain mass emissions for
each drive mode individually. The resulting total values were
compared to the RDE report output total values by PEMS to conrm
the manual technique was accurately applied for each RDE test. The
emissions factors of the “MAW applied with RDE exclusions”
datasets were then compared back to those of the “MAW unapplied
with RDE exclusions” datasets. The divisions between urban, rural
and motorway drive modes are dierent for the MAW applied data
than the “raw RDE” and “MAW unapplied with RDE exclusions”
results, due to the dierent methods used to select these drive modes.
The WLTP data consisted of the output report, which gave average
mass emission values (g) and rates in g/s and g/km for each drive
section. For the characterisation of the drive cycle, the legislated
velocity distribution with time was used, as given in the Annex to EC
715/2007 [20]. The drive characteristics were calculated as described
Downloaded from SAE International by Daisy Thomas, Thursday, October 26, 2017

above, with the exception of the stop number, for which the value
given in Tutuianu et al. [7] was used. There are four parts of the
WLTC cycle, so for the purposes of comparison with the RDE cycle,
the second and third parts were combined into one for data analysis.
These three sections were then assigned the same names as the RDE
cycle (urban, rural and motorway) for the data presentation.
Results and Discussion
Test Cycle Driving Characteristics Comparisons
As discussed in the methodology, the RDE testing took place in 2
dierent locations, and at dierent times. As a result, each test varies
from the others to various extents. Figure 3 gives an overview of the
speed characteristics of the dierent RDE tests 1-8, and of the WLTC
test, for comparison. One can see that the velocity patterns across
dierent RDE tests vary, but all are most dissimilar to the WLTC test,
which is far shorter.
As discussed in the methodology, a comparison of various drive
characteristics was performed, looking at the raw RDE data, the
“MAW unapplied with RDE exclusions” data, and the WLTC
properties. Average values were taken in each test mode (urban, rural
or motorway), and then the distance-based average of these
individual sections gave the average value. For some examples, the
sum of individual modes gave a total value instead.
Figure 4 displays the average velocities across dierent modes. One
can see that the average velocity is slightly higher under the “MAW
unapplied with RDE exclusions” than the WLTC for all drive modes,
being 62 km/h compared to 60km/h. The average velocity is
increased from 59km/h by the application of the RDE exclusions,
indicating that these exclusions are the main cause of the dierence.
Figure 5 shows the maximum velocities across the dierent modes. It
indicates that the maximum velocities are marginally lower for the
RDE cycles than the WLTC cycle in this case, except for the urban
section. The reason for this may lie in the visual selection of the drive
modes, described in the methodology.
Figure 6 displays the average acceleration over dierent cycles, and
results indicate that the RDE cycles have lower acceleration in
general, except for the rural drive mode. Average “MAW unapplied
with RDE exclusions” values were marginally lower than the WLTC.
Figure 7 shows the average magnitude of deceleration, with the WLTC
having the greater average deceleration for all modes. One can also
conclude that the exclusions employed by the RDE legislation
increase the average values of acceleration and deceleration, taking
the average acceleration from 0.33 m/s2 - 0.36 m/s2.
Figures 8 and 9 show that concerning the maximum rate of
acceleration and deceleration, the RDE tests have a 2.5 times greater
magnitude than the WLTC. These ndings are in contrast to the
average accelerations and decelerations seen, indicating that the
maximum values may not be a good indicator of the respective
general behaviors.
Figure 3. Velocity, acceleration and cumulative distance distributions of all
RDE tests and the WLTP test performed.
Downloaded from SAE International by Daisy Thomas, Thursday, October 26, 2017

Figure 4. Average velocities for the raw RDE, MAW unapplied with RDE
exclusions, and WLTC tests, divided into their relative drive modes.
Figure 5. Maximum velocities for the raw RDE, MAW unapplied with RDE
exclusions, and WLTC tests, divided into their relative drive modes. The total
value is an average of all modes.
Figure 6. Average accelerations for the raw RDE, MAW unapplied with RDE
exclusions, and WLTC tests, divided into their relative drive modes.
Figure 7. Average decelerations for the raw RDE, MAW unapplied with RDE
exclusions, and WLTC tests, divided into their relative drive modes.
Figure 8. Maximum accelerations for the raw RDE, MAW unapplied with
RDE exclusions, and WLTC, divided into their relative drive modes. The total
value is an average of all modes.
Figure 9. Maximum decelerations for the raw RDE, MAW unapplied with
RDE exclusions, and WLTC tests, divided into their relative drive modes. The
total value is an average of all modes.
Downloaded from SAE International by Daisy Thomas, Thursday, October 26, 2017

Whereas the average acceleration was smaller for the RDE cycles
than the WLTC cycle, the RPA shows more similar values. The RPA
in the rural section of the RDE is greater than the WLTC, while the
urban and motorway sections show the opposite trend to a small
degree, giving an average value only marginally greater for the RDE
than WLTC (0.15 m/s2 compared to 0.14 m/s2). This relationship is
displayed in gure 10. The average RPA values are clearly larger than
the average RPA values calculated for the on-road tests by May et al.
[15], who reported 0.117 m/s2.
Figure 10. Relative Positive Accelerations (RPA) for the raw RDE, MAW
unapplied with RDE exclusions, and WLTC tests, divided into their relative
drive modes. The total value is an average of all modes.
Figure 11. Average number of stops per kilometer for the raw RDE, MAW
unapplied with RDE exclusions, and WLTC tests, divided into their relative
drive modes. The total value is a total of all modes.
Figures 11 and 12 display the average number of stops per kilometer
and the average stop duration per kilometer, respectively. One can see
that there are certainly far fewer stops involved in the RDE tests than
the WLTC test, but that the average time stopped per kilometer
remains approximately equal. This indicates that the duration of
individual stops may be greater for the RDE test. Of course, with the
idle exclusions involved in the “MAW unapplied with RDE
exclusions” results, these values decrease to zero. It would be
interesting to investigate whether the longer stops involved in the
RDE testing may cause a noticeable cold start eect, and if so,
whether this lingers beyond the time the vehicle is travelling below
1km/h and so aects the legislated RDE emissions measurement.
Figure 12. Total duration of stops per kilometer for the raw RDE, MAW
unapplied with RDE exclusions, and WLTC tests, divided into their relative
drive modes. The total value is a total of all modes.
Emission Comparisons
Figure 13 shows that the CO2 emissions across all sections of the
RDE and WLTC cycles are above the EU passenger cars 2015 CO2
target (130 g/km). It should be noted that the target value of 130 g/km
is based on the NEDC cycle which is less aggressive than WLTP and
RDE, and is for new cars sold in 2015 in the EU. The RDE cycles
appear to produce slightly lower levels of CO2, being 4% lower than
the WLTP cycles. The RDE exclusions clearly appear to have
contributed to this reduction in the urban drive mode. Comparing the
patterns between RDE and WLTC results across the modes for gures
6 and 13, it appears that CO2 may be positively correlated with the
average acceleration, but the correlation is weak.
Figure 13. Average CO2 emission per kilometer for the raw RDE, MAW
unapplied with RDE exclusions, and WLTC tests, divided into their relative
drive modes. The total value is a distance-weighted average of all modes. A
horizontal line at 130 g/km delineates the 2015 CO2 limit.
Downloaded from SAE International by Daisy Thomas, Thursday, October 26, 2017

Figure 14 shows that the CO emissions of the RDE cycles appear to
be 60% lower than the WLTC, and well below the Euro 5 legislated
emission limit of 1 g/km. The rural section, however, gave larger CO
emissions for the RDE cycle, which would be interesting to
investigate in more detail in future work. Again, it is clear that the
ocial RDE exclusions of cold start and idling have an eect on the
level of emission in the urban section, decreasing the value by 4%.
One can see some similarities between the pattern of CO emissions
across modes with that of the average acceleration and RPA in gures
5 and 9, when comparing the RDE and WLTC trends. Increases in
RPA and average acceleration may contribute to CO emission.
Figure 14. Average CO emission per kilometer for the raw RDE, MAW
unapplied with RDE exclusions, and WLTC tests, divided into their relative
drive modes. The total value is a distance-weighted average of all modes. A
horizontal line at 1g/km delineates the Euro 5 legislated emissions limit.
Figure 15. Average NOx emission per kilometer for the raw RDE, MAW
unapplied with RDE exclusions, and WLTC tests, divided into their relative
drive modes. The total value is a distance-weighted average of all modes. A
horizontal line at 0.06g/km delineates the Euro 5 legislated emissions limit.
The RDE emissions of NOx were 34% lower than the WLTC
emissions, as shown in gure 15. The dierence is greatest for the
urban drive sections, and this is, of course, where the RDE exclusions
also appear to have had the greatest impact, reducing NOx emissions
by 41%. However, it is clear that this is not the sole cause of the
decrease in NOx emission. The rural section shows a dierent trend,
with NOx emissions having similar values. It appears that NOx
emission may be positively correlated with average acceleration
(gure 5), indicating that the higher engine speeds required for
acceleration may increase combustion temperatures, leading to
increased NOx emission.
Looking at the patterns in the emissions studied, it appears that none of
them are correlated in any way with the maximum acceleration and
deceleration. This would indicate that these parameters, though useful
in dening the type of cycle, are not indicative of the emissions
behavior of such cycles. The emissions results tend to agree with those
of Merkisz et al. [13] regarding the decrease in on-road emission
factors compared to those of chassis dynamometer tests, but are in
contrast to those of Merkisz et al. [16] and May et al. [15]. This is
particularly interesting given that the average RPA of the current work
was found to be greater than in the latter of these two studies. This may
be explained by large dierences in other characteristics between tests
and the greater number of cars used in the current study. The ndings
regarding the association of CO and NOx to average acceleration also
agree with the ndings of Merkisz et al. [13]. It should be noted,
however that none of these studies used the same RDE legislation as is
presented in the current work, and so cannot be fully compared.
A Closer Investigation of Emission Trends with RPA
RDE and WLTC Distance-Based Emissions
The CO2, CO and NOx grams per kilometer for each vehicle and each
drive mode (urban, rural and motorway) were plotted for the RDE
and WLTC data. The RDE data used abides by the ocial RDE
requirement of exclusion of cold start and idling.
Looking at gures 16, 17, 18, 19, 20 and 21, little correlation for
CO2, CO and NOx with RPA is observed. There is an increase in the
range of RPA values attained through RDE testing compared to
WLTC testing, and average RPA is highly variable across dierent
RDE tests. It is also interesting to see that many individual cars
exceeded the NOx limits for some drive modes of the WLTC test.
One can conclude from the results displayed below that RPA is an
unreliable indicator for distance-based CO2, CO and NOx emissions.
Figure 16. Average CO2 mass per km for different modes of the RDE cycle
(MAW unapplied, with RDE exclusions) plotted against RPA. A horizontal
line at 130 g/km delineates the 2015 CO2 limit.
Downloaded from SAE International by Daisy Thomas, Thursday, October 26, 2017

Figure 17. Average CO2 mass per km for different modes of the WLTC cycle
plotted against RPA. A horizontal line at 130 g/km delineates the 2015 CO2 limit.
Figure 18. Average CO mass per km for different modes of the RDE cycle
(MAW unapplied, with RDE exclusions) plotted against RPA. A horizontal
line at 1 g/km delineates the desired fleet-average limit.
Figure 19. Average CO mass per km for different modes of the WLTC cycle
plotted against RPA. A horizontal line at 1 g/km delineates the desired
fleet-average limit.
Figure 20. Average NOx mass per km for different modes of the RDE cycle
(MAW unapplied, with RDE exclusions) plotted against RPA. A horizontal
line at 0.06 g/km delineates the desired fleet-average limit. The values for
Vehicle 2 have been removed as were anomalous.
Figure 21. Average NOx mass per km for different modes of the WLTC cycle
plotted against RPA. A horizontal line at 0.06 g/km delineates the desired
fleet-average limit.
RDE Mass Emission Rate
The mass emission rate (g/s) for each vehicle and each drive mode
was plotted against RPA for the RDE tests as shown in gures 22, 23
and 24 for CO2, CO and NOx respectively. It appears that there is no
correlation between RPA and emission rates of CO2, CO and NOx. It
can therefore be inferred that RPA is not an indicator for mass
emission rates of CO2, CO or NOx.
A possible reason for lack of correlations between the RPA and
emissions (g/km or g/s) is that the emissions were the average values
for each mode (urban, rural and motorway) which mingled the
acceleration and deceleration movements. It would be worthwhile to
separate acceleration and deceleration events and this is planned for
the future work.
Downloaded from SAE International by Daisy Thomas, Thursday, October 26, 2017

Figure 22. Average CO2 mass per s for different modes of the RDE cycle
(MAW unapplied, with RDE exclusions) plotted against RPA.
Figure 23. Average CO mass per s for different modes of the RDE cycle
(MAW unapplied, with RDE exclusions) plotted against RPA.
Figure 24. Average NOx mass per s for different modes of the RDE cycle
(MAW unapplied, with RDE exclusions) plotted against RPA.
Effect of the MAW Technique on Emissions
The MAW applied RDE values of emissions across the dierent drive
modes were compared to the MAW unapplied RDE values. This
allows deeper analysis of how the employment of the MAW method
alters the perceived levels of emissions. The plot for CO2 in gure 25
indicates that the use of the MAW method has minimal eect on the
overall CO2 emissions, but does act to normalize the data by way of
minimizing dierences between modes. Again, all values are above
the desired eet-average values.
Figure 25. Average CO2 concentrations resulting from the MAW applied
data compared to that with the MAW unapplied (both with the RDE official
exclusions employed). A horizontal line at 130 g/km delineates the 2015
CO2 limit.
Figure 26. Average CO concentrations resulting from the MAW applied data
compared to that with the MAW unapplied (both with the RDE official
exclusions employed). A horizontal line at 1g/km delineates the Euro 5
legislated emissions limit.
Figure 26 shows that the MAW is reducing the average CO emissions
for the RDE test, acting primarily in the motorway and urban modes
to reduce the overall value by 12%.
Downloaded from SAE International by Daisy Thomas, Thursday, October 26, 2017

The NOx emissions, displayed in gure 27, show a stronger trend of
decreasing values under the use of MAW processing. This time most
change comes from the rural section, followed by the motorway
section, giving an overall reduction of 21%.
In all cases, the use of the MAW processing method appears to smooth
out dierences between driving modes, which is not surprising given
the nature of the technique. It would be interesting to further
investigate exactly why the use of the MAW technique decreases some
of the perceived CO and NOx emission levels, and why this is
accompanied by a normalization of those results between vehicles.
Figure 27. Average NOx concentrations resulting from the MAW applied data
compared to that with the MAW unapplied (both with the RDE official
exclusions employed). A horizontal line at 0.06g/km delineates the Euro 5
legislated emissions limit.
Conclusions
In summary, the legislated RDE test results from 8 dierent spark
ignition vehicles were compared to corresponding WLTC test results
for the same vehicles. The eect of the ocial RDE exclusion of cold
start and idling was simultaneously investigated. Specic attention
was paid to dierences in drive characteristics and the eect this
appeared to have on the CO2, CO and NOx emissions. The relationship
between RPA and the emissions of CO2, CO and NOx was investigated
in more depth. Finally, the MAW technique was applied to investigate
how this changes the perceived levels of CO2, CO and NOx emissions.
The investigation found that the average velocity is 4% greater under
the legislated RDE test than the WLTC test, with the exclusion of
cold start and idling appearing to be the main cause of this increase.
The maximum velocity, however, was 2.4% lower for the RDE test
than the WLTC test, with the exception of the urban drive mode, for
which there was an increase. The average acceleration and
deceleration were lower for the RDE than the WLTC, while the
maximum acceleration and deceleration were greater. The eect of
the RDE exclusions was to increase the average values seen by
around 10%, while it had negligible eect on the maximum values.
The values of RPA between the RDE and the WLTC tests showed
little variation on average, nor did the dierence between the “raw
RDE” and “MAW unapplied with RDE exclusions” values. There are
fewer stops per kilometer on average for the “raw RDE” than the
WLTC, but a longer stop duration per kilometer, indicating that the
average time per stop in the RDE is greater.
When studying the CO2 emissions pattern, it seems that the RDE
produces 4% lower values of CO2 per km than the WLTC, and that the
ocial RDE exclusions of cold start and idling decrease the emission of
CO2 in the urban drive mode by 8%. The RDE showed a large decrease
in distance-based CO emissions of 60% compared to the WLTC, with
the RDE exclusions leading to 18% of this decrease. The remainder is
due to the dierences in the driving characteristics of the RDE tests
compared to WLTC test. Some correlations between CO emissions and
average acceleration and RPA were seen from the results. NOx
emissions were 33% lower under RDE testing than WLTC testing, with
18% of this decrease again attributable to the ocial RDE exclusions.
NOx emissions also appeared to be related to the average acceleration.
No pollutants were correlated with the maximum acceleration and
deceleration, indicating that, although a useful cycle characteristic, these
variables are not suitable indicators for pollutant emission.
There were hardly any correlations observed between the RPA and
CO2, CO and NOx distance-based emission factors. This was the case
for both RDE and WLTC results. The range of RPA values was much
greater for the RDE than the WLTC test, and dierent RDE tests
showed a large variation in values. No correlation was seen between
the RPA and mass emission rates for any of the pollutants, indicating
that RPA is not correlated with emission rates. The use of the MAW
technique had minimal eect on the distance-based CO2 emissions,
but did appear to decrease the CO and NOx results by 12% and 21%
respectively. For all three pollutants the MAW acted to decrease the
variation between drive modes.
Recommendations
In future work it would be interesting to more closely study the
variation in emissions with drive characteristics by looking at each
vehicle test individually. It would also be desirable to investigate
whether other drive characteristics are more closely aecting the
pollutant emission, such as vehicle specic power (VSP). A more
thorough study into the eect the MAW technique has on the pollutant
emission values is also necessary, in order to see why the CO and NOx
decreased under its use. The cold start and idling exclusions for the
ocial RDE test is a topic of hot debate currently, and so further
investigation into the eects of these exclusions would be prudent.
References
1. Vlachos, T., Bonnel, P., Perujo, A., Weiss, M. et al., "In-Use
Emissions Testing with Portable Emissions Measurement
Systems (PEMS) in the Current and Future European Vehicle
Emissions Legislation: Overview, Underlying Principles and
Expected Benefits," SAE Int. J. Commer. Veh. 7(1):199-215,
2014, doi:10.4271/2014-01-1549.
2. AVL Emission Testing Systems, AVL Emission Testing
Handbook - 2016, 80, 2016.
3. Marotta, A., Pavlovic, J., Ciuffo, B., Serra, S., and Fontaras,
G., “Gaseous Emissions from Light-Duty Vehicles: Moving
from NEDC to the New WLTP Test Procedure,” Environ. Sci.
Technol. 49(14):8315-8322, 2015, doi:10.1021/acs.est.5b01364.
Downloaded from SAE International by Daisy Thomas, Thursday, October 26, 2017

4. Dings, J., “Mind the Gap! Why official car fuel economy figures
don’t match up to reality?,” 2013.
5. Kadijk, G., Verbeek, M., Smokers, R.T.M., Spreen, J., Patuleia,
A., Ras, M. van, Norris, J.O.W., Johnson, A., O’Brien,
S., Wrigley, S., Pagnac, J., Seban, M., and Buttigieg, D.,
“Supporting analysis regarding test procedure flexibilities and
technology deployment for review of the light duty vehicle fleet
CO2 regulations,” Brussels, 2012.
6. Mock, P., German, J., Bandivadekar, A., and Riemersma, I.,
“Discrepancies between type- approval and ‘real-world’ fuel-
consumption and CO 2 values: Assessment for 2001-2011
European passenger cars,” Berlin, 2012.
7. Tutuianu, M., Bonnel, P., Ciuffo, B., Haniu, T., Ichikawa, N.,
Marotta, A., Pavlovic, J., and Steven, H., “Development of
the World-wide harmonized Light duty Test Cycle (WLTC)
and a possible pathway for its introduction in the European
legislation,” Transp. Res. Part D Transp. Environ. 40:61-75,
2015, doi:http://dx.doi.org/10.1016/j.trd.2015.07.011.
8. Ciuffo, B., Marotta, A., Tutuianu, M., Anagnostopoulos,
K., Fontaras, G., Pavlovic, J., Serra, S., Tsiakmakis, S., and
Zacharof, N., “Development of the Worldwide Harmonized Test
Procedure for Light-Duty Vehicles: Pathway for Implementation
in European Union Legislation,” J. Transp. Res. Board
(2503):pp 110-118, 2015, doi:10.3141/2503-12.
9. Kageson, P., “Cycle-Beating and the EU Test Cycle for Cars,”
Eur. Fed. Transp. Environ. 98/3(November 1998):10, 1998.
10. Weiss, M., Bonnel, P., Hummel, R., Provenza, A., and Manfredi,
U., “On-road emissions of light-duty vehicles in Europe,”
Environ. Sci. Technol. 45(19):8575-8581, 2011, doi:10.1021/
es2008424.
11. Johnson, T., "Vehicular Emissions in Review," SAE Int. J.
Engines 9(2):1258-1275, 2016, doi:10.4271/2016-01-0919.
12. Els, P., “Random Cycle Test vs . Portable Emission
Measurement System,” 2013.
13. Merkisz, J., Fuc, P., Lijewski, P., and Bielaczyc, P., "The
Comparison of the Emissions from Light Duty Vehicle in On-
road and NEDC Tests," SAE Technical Paper 2010-01-1298,
2010, doi:10.4271/2010-01-1298.
14. Weiss, M., Bonnel, P., Hummel, R., and Manfredi, U.,
“Analyzing on-road emissions of light-duty vehicles with
Portable Emission Measurement Systems (PEMS),” ISBN
9789279190728, 2011, doi:10.2788/23820.
15. May, J., Bosteels, D., and Favre, C., "An Assessment of
Emissions from Light-Duty Vehicles using PEMS and Chassis
Dynamometer Testing," SAE Int. J. Engines 7(3):1326-1335,
2014, doi:10.4271/2014-01-1581.
16. Merkisz, J., Pielecha, J., Bielaczyc, P., and Woodburn, J.,
"Analysis of Emission Factors in RDE Tests As Well as in
NEDC and WLTC Chassis Dynamometer Tests," SAE Technical
Paper 2016-01-0980, 2016, doi:10.4271/2016-01-0980.
17. Department for Transport, “Vehicle Emissions Testing
Programme,” 2016.
18. European Parliament and Council of the European Union,
“Commission Regulation (EU) 2016/427 of 10 March 2016
amending Regulation (EC) No 692/2008 as regards emissions
from light passenger and commercial vehicles (Euro 6) (Text with
EEA relevance),” Off. J. Eur. Union 82(31/03/2016):1-98, 2016.
19. United Nations, “Addendum 15: Global technical regulation No
. 15 - Worldwide harmonized Light vehicles Test Procedure,”
Glob. Regist. ECE/TRANS/:1-234, 2014.
20. European Commission, “ANNEXES to the Commission
Regulation supplementing Regulation (EC) No 715/2007 of the
European Parliament and of the Council on type-approval of
motor vehicles with respect to emissions from light passenger
and commercial vehicles (Euro 5 and Euro 6) and,” 2007.
Contact Information
Daisy Thomas
University of Leeds
School of Chemical and Process Engineering
Energy Building
LS2 9JT
py11dbt@leeds.ac.uk
Dr. Hu Li
University of Leeds
School of Chemical and Process Engineering
Energy Building
LS2 9JT
fuehli@leeds.ac.uk
Telephone: +44 113 3437754
Acknowledgments
The author wishes to thank Beijing Institute of Technology for their
help in the collection of data used to produce this manuscript.
Definitions/Abbreviations
CO - Carbon monoxide
CO2 - Carbon dioxide
DI - Direct injection
EU - European Union
GPS - Global positioning system
MAW - Moving averaging window
NA - Naturally aspirated
NEDC - New European drive cycle
NOx - Nitrous oxides
OBD - On board diagnostics
PEMS - Portable emissions measurement system
Downloaded from SAE International by Daisy Thomas, Thursday, October 26, 2017

PFI - Port fuel injection
RDE - Real driving emissions
RPA - Relative positive acceleration
T - Turbocharged
TWC - Three way catalyst
USB - Universal serial bus
VSP - Vehicle specic power
WLTC - World harmonized light duty test cycle
WLTP - World harmonized light duty test procedure
SI - Spark ignition
Downloaded from SAE International by Daisy Thomas, Thursday, October 26, 2017

APPENDIX
Table A1. Characteristics of vehicles tested.
Table A2. Ambient conditions and trip properties. All values reported were measured during testing, except where otherwise indicated. The values over all WLTC tests
performed in a single test center were very similar to each other, so an average value for each city test center is reported for the WLTC tests.
Table A3. Trip drive characteristics and properties. All values reported were measured during testing and taken from the official RDE output report, except where
otherwise indicated. Because the WLTC test is a predefined cycle, the values for this are taken from literature.
The Engineering Meetings Board has approved this paper for publication. It has successfully completed SAE’s peer review process under the supervision of the session organizer. The process
requires a minimum of three (3) reviews by industry experts.
All rights reserved. No part of this publication may be reproduced, stored in a retrieval system, or transmitted, in any form or by any means, electronic, mechanical, photocopying, recording, or
otherwise, without the prior written permission of SAE International.
Positions and opinions advanced in this paper are those of the author(s) and not necessarily those of SAE International. The author is solely responsible for the content of the paper.
ISSN 0148-7191
http://papers.sae.org/2017-01-2391
Downloaded from SAE International by Daisy Thomas, Thursday, October 26, 2017