Paper Code: F02V067

DRIVEABILITY ASSESSMENT OF A CVT-POWERTRAIN WITH

MECHANICAL TORQUE ASSIST
Serrarens, Alex1
Veldpaus, Frans2
1
Drive Train Innovations, The Netherlands, 2 Technische Universiteit Eindhoven, The Netherlands
KEYWORDS
automotive control, continuously variable transmission, driveability, ywheel, fuel economy
ABSTRACT
Due to rotating inertias within the engine and transmission, the response of a vehicle during large
and/or fast engine speed shifts may appear reluctant or even counteractive. Reminiscent of com-
parable behaviour seen in aircraft jet-propulsion, this phenomenon is also referred to as jet-start.
To overcome this behaviour, a CVT powertrain is augmented with a powersplitting planetary gear
stage and compact steel ywheel. The new transmissioncoded Zero Inertia (ZI) powertrain
seamlessly combines two contradictive features: the driveability in terms of the pedal-to-wheel
response is greatly improved and a large leap towards optimal fuel economy can be made. This is
achieved by cruising the vehicle at extremely low engine speeds as of the large ratio-coverage of
the CVT. The ywheel acts as a peak shaver during engine speed shifts: it delivers power during
(semi-) pedal kick down and absorbs kinetic energy of the engine at pedal back-out.
This paper presents results from simulations of, and eld experiments with a test vehicle incorpo-
rating the ZI powertrain. The evaluation concentrates especially on longitudinal driveability and is
compared with that of basic CVT and 4-gear automatic powertrains. Results on the substantially
improved fuel economy are summarized at the end of the paper. The driveability has an accu-
rately controllable character, i.e., the accuracy with which desired drive shaft torque excursions
can be realized by the powertrain. The subjective driveability depends highly on how drive pedal
deections are translated into desired drive shaft torque excursions. Evaluation by experts and
non-experts in general determines the quality of this translation. A non-rigorous attempt to for-
malize the translation and use it in a powertrain control strategy is made. Preliminary conclusions
on possible acceptance by public are drawn. From this, directions for ongoing research are given.

INTRODUCTION

Air Drag: 2.6%

Accessories: 2.2%
Rolling Resistance 4.2%
Driveline
Losses: 5.6%
Inertia
Braking: 5.8%

Standby &
Idle: 17.2%

Heat Losses: 62.4%

Figure 1: Energy distribution from fuel to wheels for city driving

Powertrains for passenger cars have been undergoing substantial improvements throughout the last
few decades. Successful eorts have been undertaken to improve the fuel economy by adapting

1
the engine-, aerodynamical, [Seiert and Walzer, 1989] and tire design, [Junio et al., 1999]. The
wide-spread application of electronic engine control systems, see [Hirschlieb et al., 1999], by far
furnished a tremendous improvement of both fuel economy and emissions. In Figure 1, from [DOE
and EPA, 2001], the distribution of the potential energy in petrol fuel amongst all heat generating
processes involved in vehicle propulsion is presented in a pie-chart.
Apart from somehow reducing the combustion heat losses and the fuel combusted during idling,
reduction of vehicle mass (rolling resistance, inertia braking) seems to have the highest potential
but is hardly realizable in practice if at all. This is caused by especially the increased safety and
comfort standards. The desire of society for a healthy and clean environment, preferably for a
low price, encouraged car manufacturers to launch new and innovative solutions for better fuel
economy and lower emissions. Mostly these innovations are forced though also nancially supported
by governments.
Since about ten years it is fully understood that improved fuel economy can also be realized by the
transmission in the powertrain. Not only higher transmission eciency but more importantly the
way the engine and transmission jointly cooperate to increase the engines combustion eciency is
the key to a substantial fuel economy improvement. The capabilities in this area can be intensied
by utilizing a continuously variable transmission (CVT). A CVT can transmit the engine torque
and change its speed continuously. Besides this advantage above other transmission (manuals,
stepped automatic) it possesses a relatively high overdrive ratio with low penalties on volume.
For example, the CVTs overdrive can lower the engine speed to about 1500 rpm at 80 km/h,
whereas standard transmissions manipulate at least 1900 to 2200 rpm at this vehicle speed. The
low engine speeds can be shown to gain about 10 to 15% fuel economy advantage. However, a
disadvantage of the high overdrive is the lack of vehicle responsiveness (driveability) when pushing
the accelerator pedal. Mostly because of safety reasons this kind of operation is socially unaccepted
and solutions are mandatory to break the paradox between driveability and fuel economy.
Zero Inertia Powertrain

Figure 2: Zero Inertia Powertrain

In the framework of the EcoDrive project at the Technische Universiteit Eindhoven a technical
innovation is elaborated that indeed banishes the mentioned paradox. The innovation originates
from the idea to exploit the inertia of a ywheel benecially for realizing an initial and persistent

2
response at request. Connecting a planetary gear stage with the ywheel in parallel to the CVT, see
Figure 2, it is possible to decelerate the ywheel while accelerating the engine with the CVT (for
example to facilitate a take-over manoeuvre). The torque stemming from the ywheel is directly
transmitted to the wheels until the moment where the engine can deliver the requested power
sustainably on its own. A unique state of the powertrain can be identied where the total torque
stemming from the engine sided inertias are compensated exactly by that of the ywheel. This
(transient) state is termed zero inertia hence the denomination Zero Inertia Powertrain.
A proper coordination of engine torque and shift speed of the CVT ratio during and after the
engine transitions is extremely important. By carefully choosing the setpoints for the engines
electronic air throttle and for the hydraulically controlled CVT, the driveability and fuel economy
can be optimized at all times. This procedure is termed coordinated powertrain control and in
[Serrarens, 2001] such controllers are developed for both CVT and ZI powertrains.
ENGINE, DRIVEABILITY AND FUEL ECONOMY
Driveability is for the larger part determined by the instantly available power after pressing the
accelerator pedal. The so-called power reserve is the product of the actual engine speed and the
torque reserve. The latter is dened as the dierence between the actual engine torque Te and
maximum engine torque TWOT , see Figure 3. At the expense of fuel economy a much larger power
reserve results when the engine is operated at a higher engine speed e and a lower engine torque
Te . In Figure 3 this can be easily observed by comparing the torque reserve of the economy line
(or E-line) and the driveability line available at, say, 25 [kW], i.e., at operating points A and
B, respectively.
This observation is more exemplied in Figure 4, where the trade-o betweem fuel consumption
and power reserve is visualized. In this gure the surplus of specic fuel consumption with respect
to that on the E-line is plotted against the additional power reserve that would emerge when
leaving the E-line for less fuel economic operating points with the same stationary power (here 5,
10, 20, 25 and 30 [kW]).

Wide Open Throttle (WOT) torque torque reserve

150

75 [kW]
250
engine torque [Nm]

Ao
0

100
26

driveability line
270

B 280
50 [kW]
economy line o
290
50 300
320

350 BSFC [g/kWh] 25 [kW]

400
500
700
5 [kW]
0
100 200 300 400 500 600
engine speed [rad/s]

Figure 3: Economy and driveability operating lines in the engine map

Clearly, the driveabilitytacitly dened as the power reserve is improved tremendously when
leaving the E-line for an operating line below it, e.g., the driveability line. For instance, at a
stationary engine power of 25 [kW] the driveability line degrades the fuel eciency by about 12%,
whereas the power reserve is increased by about 4.5 times.
In the previous reasoning it was assumed that the engine torque can be changed instantaneously
and without penalty. Changing the engine speed could (eventually) increase the output power as
well, though the dynamic response of the vehicleand thus its driveabilityis penalized impor-
tantly when the engine acceleration is too large.

3
 40

35 5 [kW]

30 20 [kW]
% more BSFC 10 [kW]
25 25 [kW]
driveability line
20 30 [kW]

15

10
stationary power level
5

0
1 2 3 4 5 6 7 8
multiples of the Eline power reserve

Figure 4: Trade-o between fuel economy and driveability

The physical background for this can be found in the primary-sided or, for short, primary inertia,
i.e., the total inertia of the components directly connected to and including the engine. Although
the primary inertia is relatively small compared to the vehicle inertia, for good driveability it is of
great relevance that the engine can be swiftly accelerated by shifting down the transmission, i.e.,
lowering the transmission ratio (dened as wheel sided speed divided by the engine sided speed).
The engine power reaches its maximum at a fairly high engine speed (approximately 5500 [RPM]).
If the transient to this maximum power is required to be fast, the acceleration of the primary inertia
(shortly) requires a power level similar to the power needed for the actual propulsion of the vehicle.
When in such a case the engine was operating at low speed and high torque, i.e., with a small
power reserve, fullling an acceptable power transient is deemed impossible. Instead, to facilitate
the acceleration of the primary inertia power ows from the vehicle to the engine, consequently
shortly decelerating the vehicle instead of accelerating it.
When shifting down the transmission more slowly, the increase of wheel power might show a latency
as the available power reserve is just enough to accelerate the primary inertia and cannot be used
yet for a prompt increase of wheel power. This degrades the sensitivity of the accelerator pedal as
the requested higher power is delivered after the primary inertia has accelerated. This phenomenon
is referred to as jet start (also known as shift-shock, rubber belt eect, torque hole, worn clutch
eect). Jet start behaviour is known from aircraft propulsion, where initially all combustion power
must be used to accelerate the turbines before the increased engine power can be used to propel
the aircraft. For pedal back-out the phenomenon also occurs but then reversely, as the instant
decrease of power is counteracted by the torque generating deceleration of the primary inertia.
Jet start can be minimized by persistently operating the engine at high speeds at the expense of
fuel economy, for example using the driveability line in Figure 3 and 4. Even then, the transmission
ratio shift speed cannot be arbitrarily high, so an advanced control strategy is required to yield
an acceptable compromise between fuel economy and driveability. The advantage of using controls
(software) to counteract the driveability problem is that such a solution can be exible and cheap,
as opposed to rigid and generally more expensive hardware solutions. On the other hand, pure
hardware solutions will be more robust and often inherently stable, whereas software solutions
cannot extend physical system limitations. The additional costs for modied hardware (and to a
lesser extent, software) largely depend on production numbers.
An example of a pure hardware solution is lowering the primary (engine) inertia while simultane-
ously increasing the number of cylinders in order to compensate for the lack in torsional damping,
as suggested, e.g., in [Guo et al., 1988]. Such a solution is quite expensive and moreover rather

4
ineective, since the remaining inertia can still adversely inuence driveability. If the primary
inertia were zero, the engine output torque would be independent of the engine acceleration. The
Zero Inertia solution will be shown to partly realize this.
If the compromise between fuel economy and driveability is considered unacceptable, E-line track-
ing demands an additional assisting power source. The additional power source has to ll up
the power gap between the actual power level and a (higher) requested power level. This should
be done as long as the engine itself has not reached the requested power level yet, i.e., during
acceleration of the primary inertia through altering the transmission gear ratio.

50

Pcombustion
40
P
Pe,total desired
30
27
23
20
+ =

P 10
assist

0
Pinertia
7
t
10

Figure 5: Realization of a power transient with power assist; power levels in [kW]

Figure 5 shows an illustrative example of the power assist mechanism. In the example, it is assumed
that the driveability level reached with the driveability line is acceptable in practice. In other words,
the additional power source should at least raise the limited power reserve emerging from E-line
tracking up to the power reserve reached when tracking the driveability line, see Figure 4. The
left plot in Figure 5 depicts the engine combustion power Pcombustion , the power from the primary
inertia Pinertia , and their combined total Pe,total = Pcombustion + Pinertia . The right plot shows the
wheel power Pdesired as it is assumed to be desired. To arrive at the desired power, an additional
power, depicted in the middle as Passist , is needed. In this example, a stepwise power transient from
20 [kW] (roughly corresponding to a stationary vehicle speed of 110 [km/h]) to 50 [kW] is desired.
When starting from the E-line, the engine power reserve is only 10 [kW], hence an additional
20 [kW] is needed to resemble the power reserve available at the driveability line. However, for
a prolonged increase of wheel power, the engine must be accelerated to arrive at a speed where
it can maintain the desired power level of 50 [kW]. Operating on the driveability line, this step
can be performed without increasing the engine speed. This can be seen in Figure 4, where the
intersection of the driveability line with the 20 [kW] stationary power level reveals about four
times the power reservethat is 40 [kW]from the E-line. Using the E-line, in this example the
engine speeds up from 230 [rad/s] to 330 [rad/s] (corresponding to 20 and 50 [kW] on the E-line,
respectively) in a time interval t = 0.8 [s]. Hence, an additional power Pinertia between 7 and
10 [kW] is required to perform the acceleration of the engine inertia. As a result, the necessary
power level Passist of the assist source amounts up to 27 [kW], such that Pe,total + Passist = Pdesired .
conclusion
In this section it was shown that the way the stationary engine operating points are chosen largely
determine the driveability and fuel economy. The driveability was assumed to be objectivated by
the engine power reserve, which is dened for any chosen (stationary) operating line. In practice,
the calibration of the driveability line may lead to acceptable compromises between the drive-
ability and fuel economy. On the other hand, when applying some sort of power assist system,
the sizing and controls of this system and the transmission are crucial in order to resemble the
driveability level encountered with the optimally tuned driveability line. For that matter it is im-
portant to objectivate (longitudinal) driveability in a more general way. This is the subject of the
next section.

5
OBJECTIVATION OF LONGITUDINAL DRIVEABILITY
Most propulsion principles compromise between driveability and fuel economy for reasons discussed
throughout this paper. Minimal fuel consumption is often less stressed than acceptable driveability.
However, driveability is rather subjective and is rated through panels of test drivers carrying out
driving test manoeuvres and elucidating their experiences via forms and/or interviews. Therefore
endorsing an acceptable driveability is a costly and time-consuming undertaking, which is often
not reproducible (see [List et al., 1998]). This renders a complete reiteration for every released
powertrain variant inevitable. If, on the other hand, driveability could be measured and rated
more objectively, incorporating it somehow in the design of powertrain components and control
strategies could take place much earlier and more decisively in the design process, leading to
apparent benets.
terminology of longitudinal driveability
A unique description of vehicle driveability does not exist and grasping it into a single denition
would be a nuisance. First of all there is a large dierence between lateral and longitudinal driver-
vehicle interaction. In literature this is often dierentiated in the terms handling and driveability
respectively. However, according to the denition given in [Bergman, 1973] handling refers to a
more general context, viz.
An interaction between driver, vehicle and environment, which takes place
during transportation of people or goods from place to place.
This denition may also include the longitudinal behaviour of the vehicle. Subjective assessment
of the longitudinal behaviour is more often linked to driveability. Denitions of driveability are
also encountered in literature:
Driveability describes how dependably and smoothly a cars powertrain oper-
ates under all kinds of weather and operating conditions. Driveability does not
include ride and handling quality, braking performance, or abnormal combus-
tion phenomena such as knock, [Everett, 1971].
and
Good vehicle driveability is characterized by the driver having ease of control
of the vehicle and confidence in both predictable and desirable system responses
to the drivers demands. It is very much dominated by the performance of the
powertrain and vehicle in transient conditions, [Brace et al., 1999].
In the rst denition of driveability handling is explicitly excluded, basically violating the de-
nition of [Bergman, 1973]. This denition seems to stress comfort more than the ease of vehicle
manipulation. The second denition, on the other hand, focuses more on the interaction between
driver and vehicle and is considered more tailored to the application discussed in this paper.
Driveability has emphatically nothing to do with performance of the vehicle which is related to
the elapsed time for normed vehicle speed changes, e.g., 0 to 100 km/h or 80 to 120 km/h, [van
Druten et al., 2001]. On the other hand, a shorter elapsed time implies higher installed power
(relative to the vehicle mass), which generally has the potential to improve driveability as well.
characterizing longitudinal driveability
Up to now, no standardized ways are found to characterize driveability objectively. The subjective
driveability assessmentswhich experienced some standardization [CEC, 1983]rate driveability
aspects through judging the vehicle acceleration feel or (dis)comfort. Hence, vehicle acceleration
or wheel torque may be chosen as variables that reect the level of driveability, [Pesgens, 2001].
According to the denition by [Brace et al., 1999] driveability is determined predominantly by the
behaviour of the powertrain in transient situations. As driveability is a matter of human-vehicle
interaction, vehicle acceleration in general (thus also stemming from powertrain oscillations) might
be the most suited variable to objectively quantify longitudinal driveability. Strongly related to the
vehicle acceleration, the drive torque at the wheels is another employable candidate for driveability
characterization as well as for control purposes.
The key problem is to infer a desired value for either of the two candidate variables from the
information that is given by the driver pressing the drive pedal. However, the problem is more

6
tricky than this as for standard powertrains with moderate engines the desired and realizable
trajectories are closely intertwined. Clearly, the driveability problem, formerly denoted as the
jet start phenomenon, is the result of the engine not being able to fulll desired output power
requests in all transient circumstances. Especially for smaller engines this problem is intensied.
Driveability, expressed in either of the two variables, could be impaired if the control of the
engine transients is not examined carefully. However, it is generally harder to formulate what
behaviour the driver still accepts than actually realizing it. Using denitions of behaviour that
is not accepted or is judged badly, it is attempted next to qualitatively formulate acceptable
longitudinal driveability.
qualitative driveability objective
Accepting the exclusion of specic engine running problems, driveability will be formulated using
the following terms, [Everett, 1971], [Graham et al., 1988]:
Hesitation: a temporary lack (time delay) of initial response in the acceleration;
Sag: a short, sharp reduction in the acceleration, a loss of power;
Stumble: same as sag but the acceleration is actually negative for a short time;
Shue: fore- and aft longitudinal oscillations in vehicle acceleration;
Stretchiness: lack of acceleration performance during light to moderate accelerations.

2 acceleration [m/s2]

1.5 acceleration [m/s2]

1.5
pedal deflection
1 pedal deflection
1

0.5
0.5

0
0

sag (and shuffle)

hesitation
0.5 0.5
0 0.5 1 1.5 2 0 0.5 1 1.5 2 2.5 3 3.5 4
time [s] time [s]

(a) Hesitation (b) Sag

3 3

2
acceleration [m/s ] 2.5
2.5
shuffle

2 2
pedal deflection
1.5 1.5
acceleration

1 1
pedal deflection

0.5 0.5

0 Stumble 0

0.5 0.5
0 0.5 1 1.5 2 2.5 3 3.5 4 0 2 4 6 8 10
time [s] time [s]

(c) Stumble (d) Shue

Figure 6: Measurements of driveability phenomena induced with the CVT test vehicle

In Figure 6 measurements conducted with the CVT test vehicle (discussed further on) show four
of these driveability phenomena. Shue is most related to comfort, which can be seen as a subset
of driveability. All the mentioned aspects relate to situations where changes in the drive pedal
position are imposed by the driver.

7
Shue is caused by (induced) drive shaft oscillations in transients. Hesitation, stumble, sag and
stretchiness originate from accelerating the primary inertias dominating the wheel torque transient
as explained in the previous section.
Driveability is judged poor when the above aspects are manifest for the driver through accelerations
forced upon him or her via the seat- back and surface. Hence, driveability could be judged good
if the human body accelerations are:
1 imposed directly after altering the pedal position, both for pedal tip-in and for back-out. The
response time should typically be within the 0.2 [sec] time delay humans take to consciously
percept and judge actions, and
2 persistent, low frequent or described by functions showing one extremum i.e., similar as in
Figure 7.

acceleration, wheel torque

pedal deflection

time [s]

Figure 7: Possible course of desired vehicle acceleration after a positive drive pedal deection.
Acceleration and wheel torque descend again due to limited requested or installed power at an
increasing vehicle speed, thus road load.

For stepped transmissions this kind of acceleration patterns is absolutely unfeasible when the
vehicle acceleration is accompanied with gear shifts, see Figure 8. Even for powertrains with a
CVT the conditions posed above can not be met without compromising time optimality. If the
pedal deection is quite moderate, the ideal acceleration might be induced by raising the engine
torque only, basically common practice when driving a vehicle with manual transmission. On the
other hand, driveability is often judged less important if the driver demands for performance, for
instance imposed by a complete pedal kick down (demand for instant availability of maximum
engine power). Hence, the driveability denition might be tightened to all vehicle instationary
situations excluding kick downs. In practice also distinguished attention is given to kick downs,
in fact focusing on performance mostly. In this paper, it is assumed that driveability is important
up to step-wise changes in pedal deections in succession to the actual deection which
arguably are at most 50% of the total deection range. Above that, performance becomes impor-
tant implying that in such cases smooth acceleration patterns such as in Figure 7 may be dropped
as a control objective. Arguably driveability is still assumed to be important when pressing the
pedal from 49% to 99% of the total pedal travel, that is almost a kick down.
Further on in this paper, the qualitative driveability objective dened in this section is used to
judge the driveability of a 4AT, CVT and ZI powertrain.
quantitative driveability
To actively control the driveability, the pedal deection needs to be translated somehow into a
quantity related to driveability. In the previous paragraph, vehicle acceleration or wheel torque
were mentioned as variables related to driveability. In fact, this translation sets a reference input
for the powertrain control that attempts to make the closed loop system track this reference input.

8
 2

1.5 4th to 3rd gear

acceleration [m/s ]
2

0.5

2nd to 3rd gear

0
3rd to 2nd gear

0.5
1 2 3 4 5 6 7 8 9 10
time [s]

Figure 8: Vehicle acceleration response after an imposed kick down from 80 km/h in the case of a
powertrain with a 4 gear automatic transmission (4AT)

For several reasons, the vehicle acceleration av might not be the ideal choice for the translation.
This acceleration is the result of the drive torque Td in the drive shafts and the actual road load
Trv actig on the vehicle, i.e.,
Td Trv
av = , (1)
Rw mv
where Rw is the dynamic wheel radius and mv is the surplus of the vehicle mass and the projected
wheel inertia. Hence, translating the pedal deection into a desired vehicle acceleration always
requires the road load torque Trv to be known (measured or reconstructed), [Serrarens et al.,
2002]. The main problem, however, is that needs to be associated with stationary vehicle speeds
as well, since in many situations a change in means that the driver wants to drive at a higher
velocity rather than prescribe the acceleration itself. Furthermore, the driver generally does not
expect the powertrain to compensate for changing road loads. Clearly, if a certain change in
pedal deection at the actual velocity would correspond to a desired vehicle acceleration, then the
controller would try to achieve this acceleration (within the physical boundaries of the powertrain
of course) regardless ofpossibly rapidchanges in the road load. On the other hand, drivers are
often aware of sudden changes in road load, for example at the foot of a hill. If in such cases the
driver pushes his pedal to maintain speed, the vehicle will start accelerating instead, which may
lead to unexpected behaviour.
A normal driver is well able to control a desired vehicle acceleration through changing Td and to
compensate reasonably automatically for changes in Trv . This argues for a translation of in terms
of the wheel torque Td . Doing so, the driver can easily manipulate the relative position, speed and
acceleration of the vehicle with respect to other vehicles in trac, which gives him or her a feeling
of safety and reliability. Furthermore, if the response of the vehicle to pedal deections is direct
(without, or with a very short time delay) the drivers workload expressed in terms of the pedal
deections is low, possibly accompanied with a low level of annoyance. Some aspects of the above
reasoning are also seen in [Schmid et al., 1995] and [Wicke et al., 2000].
The translation of the pedal deection [0 , 1] into some desired wheel torque Td,d is relevant for
constituting a coordinated powertrain control system. The powertrain may comprise an internal
combustion engine with an electronic throttle valve and a transmission with automated gear
shifting (CVT, AT, AMT). This is important since the freedom in choosing the setpoints for the
throttle valve and transmission gear ratio can be freely chosen such that the desired wheel torque
is (approximately) realized under the constraint of (near) optimal fuel consumption. In Figure 9
a conceivable translation of into Td,d is depicted.

9
 1800

1600

1400

desired wheel torque Td,d [Nm]

1200

1000 Td,d (w, =1)

=0.5
800

600

400

200

200

400 Td,d (w, =0)

0 20 40 60 80 100 120 140 160

wheel speed [rad/s]
w

Figure 9: Drive pedal translation: desired wheel torque Td,d as a function of pedal position and
actual wheel speed w

The translation as in Figure 9 is not trivial, but two boundaries are fairly straightforward. At the
rst boundary occuring for maximum pedal deection, = 1, the desired wheel torque Td,d should
resemble the maximal possible torque, i.e.,
 

T w


WOT rd rud

for w < w
ud
rd rud
Td,d(w , = 1) = (2)



P

e,max for w w
ud
,
w
where
w
ud
= rud rd e,max , (3)
rud is the lowest possibleunderdrivetransmission ratio and rd is the nal drive ratio. All trans-
mission ratio in this paper are dened as the wheel sided speeds over the engine sided speeds.
Furthermore e,max is the engine speed at which the engine output power is maximal, that is
Pe,max , i.e.,
Pe,max = TWOT (e,max) e,max . (4)
Clearly, the maximally desired wheel torque Td,d (w , = 1) in equation (2) is governed by the
maximal engine torque TWOT and the maximal engine power Pe,max . As long as the wheel speed
w is below w ud
, dened in equation (3), the maximal engine power Pe,max can not be reached
as the engine speed e is kinematically restricted by the underdrive transmission ratio rud . The
maximally possible wheel torque is then realized by completely opening the throttle valve and
keeping the transmission ratio in rud .
At the second boudary occuring for complete pedal back-out, the throttle valve should be closed
completely. In general the driver then wants to (moderately) brake the vehicle using the engine
drag torque TDRAG (not shown in Figure 3). Furthermore it is assumed that this braking occurs
at the engine speed required to maintain the actual wheel speed w using a fuel optimal strategy.
Thus, the desired drag torque Td,d (w , = 0) at the wheels is dened assuming that the pedal

10
is released only when the vehicle speed is stationary. Now, using the maximum and minimum for
Td,d , the translation of pedal deection into desired wheel torque is chosen as


() Td,d (w , = 1) for >

Td,d(w , ) = (5)


T ( , = 0) for ,
d,d w

where () is some proportionally (possibly linearly) increasing function of the pedal deection
. The threshold value to detect pedal back-out or tip-in, is generally very small and is set by
practical constraints at = 0.05.
In the foregoing it was assumed that the wheels have enough traction to move the vehicle with the
exerted wheel torque. This will not always be true. Especially at wet and slippery surfaces wheel
skid frequently occurs. The driver then looses the interaction between drive pedal deections and
response of the vehicle. Trimming of the control inputs for the engine throttle and the transmission
shift speed is essential in order to limit or nullify wheel skid. This area of powertrain control, is
referred to as traction control, [Park and Kim, 1999], and lies outside the tenor of this work.
Nevertheless it is noted that, if the wheel torque is carefully manipulated, traction control has a
large potential to be seamlessly integrated in the coordinated powertrain control.
COORDINATED POWERTRAIN CONTROL
Electronic powertrain control has become very attractive ever since the dramatic reduction of
costs and the increase of capabilities of automotive sensors, real time digital controllers and actu-
ators. Back around 1990 it wascontrary to Japannot widely spread in Europe and the USA,
[Schwab, 1990]. Since then this situation has changed dramatically, and not only for powertrain
control, [Harms, 2001]. Control engineers, researchers and scientists have been eager in formulat-
ing sophisticated control strategies for various purposes related to the operation of the existing
large diversity of powertrains, e.g., with manual transmission, automated manual transmission,
automatic transmission, continuously variable transmission, and hybrid propulsion. Most of the
proposed strategies focus either on minimization of fuel consumption, maximizing performance,
optimizing comfort, driveability and safety or combinations thereof (see [Narumi et al., 1990]),
and commercially available passenger cars in fact constitute a compromise amongst all of them.
Concerning electronic powertrain control, for clarity a threesome of denitions is made:
The electronic powertrain control system is dened as both the control system hardware
(sensors, digital controllers, actuators) and the control strategies captured in the software;
The software constituting control and/or steering strategies for engine and transmission is
termed coordinated powertrain control;
The software constituting control of local components in the powertrain, e.g., actuators, is
referred to as component control
Coordinated powertrain control for common vehicle propulsion systems refers to scheduling the
setpoints for the engine torque and the transmission ratio. Besides the intrinsic mechanical and
engine properties of any powertrain, coordinated powertrain control is a key factor in achieving the
intended fuel savings. The engine and transmission need to function in concert in order to obtain
the maximally possible fuel economy. In this paper the fuel saving principle earlier dened as E-
line tracking is pursued and the main control problem there is to schedule the engine operating
points dynamically such that a desired response of the vehicle (driveability) and best fuel economy
emerges. This research question is drawn up for both a conventional CVT powertrain as well as the
ZI powertrain. For comparision a similar strategy is applied for a 4 gear automatic transmission
(4AT).
The idea of coordinated powertrain control is not new. Especially for CVT powertrains interesting
control solutions are found in literature, see [Guzzella and Schmid, 1995], [Kolmanovsky et al.,
1999], [Mayer and Schroder, 1998], [Sackmann and Krebs, 1999], [Schmid et al., 1995] and [Shafai
and Geering, 1996]. An overview of methods applied in practice can be found in [Liu and Paden,
1997] and [Pner, 1999]. Most methods focus both on fuel economy and driveability but control
solutions for the latter are often exotic and strand into moderately tunable closed loop structures.

11
throttle valve and CVT ratio setpoints
In this paragraph, the coordinated powertrain control laws implemented in the test vehicle (see next
section) are discussed. The discussed closed loop structure is envisaged in Figure 10. The control

Powertrain
Control

..
rcvt,d w d
vv
.
rcvt , rcvt Vehicle Response Te
CVT Ratio Engine Torque

Pulley Pressures Air Intake (Fuel, Spark)

Hydraulic

Throttle
Control
Valve Stem Positions
Control

Air Throttle Angle

Solenoid Currents DC Motor Current

Control Voltages Control Voltage

Figure 10: Layers in electronic powertrain control

laws are applied for both the basic CVT and the ZI powertrain, but are based on the dynamics
seen in the ZI powertrain. To derive the desired setpoints for the engine throttle valve and the
CVT ratio, two input variables are distiguished, being the engine torque Te and the acceleration
of the primary inertia p . The CVT ratio rcvt is not used directly as a control variable. Instead
the desired acceleration p,d of the primary inertia is determined bearing in mind that the net
transmission output torque must resemble the desired drive shaft torque Td,d as close as possible,
preferably taking excursions (qualitatively) similar as plotted in Figure 7. Integration of p,d leads
to a value for the desired primary speed p,d and by relating it to the actual wheel speed, nally
a setpoint for the CVT ratio rcvt,d emerges. This setpoint is attempted to be realized as good as
possible by the hydraulic CVT pulley clamping controller, see [Vroemen, 2001].
The desired engine torque Te,d is determined by continuously mapping the desired wheel power
Td,d w hyperbolas onto the engine map and nding intersecting engine torques at the E-line (see
Figure 3 and 9). Clearly this assumes a transmission eciency of 100%, or alternatively it assumes
that the driver will compensate for limiting transmission eciencies by further increasing the drive
pedal angle to obtain a slightly higher setpoint Td,d .
Through altering the throttle valve angle using the measured engine speed e the desired engine
torque Te,d can be manipulated with fair accuracy. Hence a setpoint d for the throttle valve angle
being a function of e and Te,d is required. This setpoint is directed to the throttle valve controller
manipulating the setpoint d into an actually realized throttle opening .
control law for d
The engine torque setpoint Te,d and the measured engine speed e are substituted into the inverse
of the throttle map Te (, e ), see Figure 11.
This leads to the setpoint d for the throttle valve:
d = Te1 (Te,d, e ) , 0% d 100% (6)
 
Td,d w
Te,d = min TEline (Td,d w ), Td,d rd rod , (7)
e,min
In equation (7) the setpoint for the engine torque on the E-line is corrected whenever the overdrive

12
 160
=100 %
140
120

e
stationary engine torque T 100 TEline

80
60
40 constant throttle opening
20
0 =0 %
20
40
0 100 200 300 400 500 600 700
engine speed [rad/s]
e

Figure 11: Stationary engine torque T e as a function of stationary throttle and speed e

ratio rod or the minimally operable engine speed e,min are reached. In those cases E-line tracking
has to be dropped in order to manipulate the desired wheel torque.
control law for rcvt,d
The control law for rcvt,d is equal for both the CVT and the ZI powertrain and will be derived
in this paragraph. The CVT ratio control law is based on a extremely simplied powertrain
model. The control model in fact is the equivalent primary inertia J1 loaded with the torques
u1 and Td rd rcvt , see Figure 12. The reason for using just this simple powertrain model is that it
nonetheless embodies all the dynamics causing the driveability problem pointed out in this paper.
Applying the dynamic equation of the model in Figure 12 and the objective to realize Td = Td,d
J1*

Te Td rd rcvt

Figure 12: Simplified powertrain control model

as close as possible, the setpoint for the CVT ratio rcvt,d is given by:
w
rcvt,d = , (8)
rd p,d
Td,d w
Te,d
p,d
p,d = (9)
J1 (rcvt,d )
subject to
rud rcvt,d rod and e,min p,d e,max (10)

13
Laws (6) and (8) obviously simplify the coordinated control at the cost of accuracy. Through
discarding all (equivalent) inertias downstream from the primary inertia in the control model,
they are not compensated for by the control laws (6) and/or (8). The consequence of this will be
shown by simulations in the next section.
1

0.8

ZI
0.6 J1 J CVT
equivalent inertias [kgm ]
2

1
0.4 J*
1
0.2

0.2

0.4 r ud r r zi r od
gn

0.6

0.8
0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2 2.2
CVT ratio r []
cvt

Figure 13: Equivalent primary inertias: J1CVT (constant) is the total of the primary inertias seen in the
basic CVT powertrain, J1ZI (function of the CVT ratio) is the total of the equivalent primary inertias
observed in the ZI powertrain and J1 is an adapted version of J1ZI .

Observing the stability of the shift law (9) points out that stabilization of the engine operating
points is only possible if J1 ,i.e., the equivalent primary inertiais negative! Naturally, physical
inertias can not be negative. On the other hand due to the specic gyristor dynamics that emerge
from the flywheel and power splitting planetary gearset in parallel to the CVT 1 the equivalent
primary inertia is indeed negative for the larger part of the CVT ratio range. In Figure 13 the
equivalent inertias seen in the CVT powertrain and a ZI powertrain are drawn. Considering the ZI
gyristor dynamics two characteristic CVT ratios can be identied. First, the geared neutral CVT
ratio rgn ; this is the ratio where the ywheel exactly remains at zero speed and thus the equivalent
inertias of the CVT and ZI powertrain are exactly equal: J1ZI = J1CVT . Secondly, the zero inertia
CVT ratio rzi ; this is the ratio where the equivalent primary inertia J1ZI in the ZI powertrain is
exactly zero. As said the control law (9) leads to stable solutions for J1 < 0, which is the reason
that J1ZI is redened (in the control law only) into

J1 = min J1ZI , s s < 0, (11)

where s is some negative constant tuning value. In the present case s = 0.1 is chosen.
For control purpose, this adaptation of the physical variables is sucient to obtain satisfactory
closed loop results. The shift laws (8), (9) are used for both the CVT and ZI powertrain, even
though J1 as in equation (11) has no strict meaning for the CVT powertrain, e.g., compare J1
and J1CVT in Figure 13. However, for mutually comparing the dynamic dierences between the
CVT and ZI powertrains, it is fair to implement the same shifting law for both powertrains.
ASSESSING THE DRIVEABILITY OF THE 4AT, CVT and ZI POWERTRAINS
In this section the 4AT, CVT and ZI powertrains are compared on their driveability performance.
First the structure of a dynamic simulation model is discussed. With this powertrain model simula-
tions are performed illustrating the characteristic dierences between the mentioned powertrains.
Then the test vehicle is introduced with which driveability (but also fuel economy) experiments
are conducted. Some of these experiments are also discussed in this section.

1 in [Serrarens, 2001] this combination is termed Inertial Variable Shunt

14
powertrain model
In Figure 14 the dynamic model of the CVT powertrain is schematically envisaged. The ZI power-
train is modelled basically with the same structure as this CVT powertrain model. As said for the
Lockup Clutch
Mechanical Final
Engine Torque CVT Losses Gear
Engine l Converter Driveshafts
Drag Vehicle Mass & Body
Je Jt DNR J1 CVT
Wheels
e i t p 11
00
b
00
11
00
11
00
11 Jw
Tires
Jv
00
11
00rcvt
11 d w v o
e t p 11
00
00
11 J2
Jl 00
11
00
11 rd
00
11
00
11 s w v
Pump
s

Figure 14: Generic CVT powertrain model

ZI powertrain the parameters J1 and J2 , i.e., the equivalent primary and secondary pulley iner-
tias, depend on the CVT ratio rcvt . The experimentally obtained torque losses due to mechanical
friction in the variator and the hydraulic pump losses due to the pressurizing of the CVT pulleys
are also incorporated in the model. Apart from the model in Figure 14 also simple models for the
mean value engine torque and the CVT variator dynamics are introduced. Moreover also a variant
of the model is constructed describing the powertrain dynamics in case of a 4AT transmission, see
also [Cho and Hedrick, 1989]. The hydraulic control of the clutches in the 4AT is implemented
rather straightforwardly. Better solutions can be found in [Minowa et al., 1994], [Minowa et al.,
1996], [Petterson and Hedstrom, 1997] and [Yang et al., 2001]. The 4AT basically replaces the
CVT in Figure 14.
The models are implemented in Simulink as well as the coordinated controller discussed in the
previous section. The model is used for analysis of the CVT and ZI powertrain and for validation
of the controllers prior to implementation in the test vehicle. The coordinated controller for the
4AT is dierent from the one discussed in the previous section. For the 4AT the engine torque
compensates for dierences in the desired and realized wheel torque instead of the CVT ratio in
the CVT and ZI powertrains. The 4AT shifts to a higher or lower gear whenever the shift law (8)
approaches a transmission ratio that corresponds with a particular xed gear ratio in the 4AT.
Hence using the shift law (8) also for the 4AT, the desired primary acceleration p,d according
to (9) must also be computed. The desired engine torque Te,d used in (9) equals that described
by (7) but the desired engine torque used for the throttle setpoint as in (6) now equals
Te,d = Td,d rd ri , (12)
where ri is the actually selected gear ratio in the 4AT, thus either r1 , r2 , r3 , or r4 . Clearly,
the desired engine torque now compensates for the desired wheel torque Td,d given the actually
selected xed transmission ratio.
simulation results
With the setpoint control laws (8), (6) and (12) (for the 4AT) simulations are performed. The
simulation shows the results of a subsequent drive pedal tip-in and back-out, see Figure 15.
In this gure a multitude of observations can be made. Comparing the CVT with the ZI wheel
power transient for tip-in shows that the ZI persistently (with only a slight hesitation) increases
the power to the nal level. The CVT powertrain increases the wheel slightly after which the power
hesitates during about 0.7 [sec]. In that period the engine power is used to furnish the acceleration
of the primary inertias. The tip-in transient of the 4AT is hardly comparable with that of the
CVT and ZI. Shortly after a prompt increase of wheel power by completely opening the electronic
throttle, a shift from 4th to 3th gear occurs. This shift is accompanied with a steep decrease of
power during the shift and an even more steep increase of power at termination of the shift. The
observations for pedal back-out are basically the same as pedal tip-in only then in the opposite
sense.

15
 wheel power Tdw [kW]
50
T
d,d w
40 2
4AT
30 ZI CVT

20 3
10 1
0
pedal deflection [%]

45
backout
35 tipin
25
15
5
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
time [s]

Figure 15: Desired and realized wheel power transients upon pedal a subsequent pedal tip-in and back-out
manoeuvre simulated for the 4AT, CVT and the ZI powertrain

To illustrate the transients more in depth in Figure 16 the transients are plotted in the engine
map (plot (a)) in terms of the vehicle acceleration (plot(b)), the vehicle speed (plot(c)) and the
transmission gear ratio (plot(d)). The power levels indicated in Figure 15 by 1, 2 and 3 are also
indicated in Figure 16(a). Even though these power levels are carefully maniulated by the engine
and CVT, they are not actually realized (after the transient) at the wheels, see Figure 15. This is
caused by the limited eciency of the transmissions. On top of this the ZI powertrain accelerates
the ywheel along with the increasing vehicle speed after the engine speed transient. The 4AT,
on the other hand, needs to accelerate the engine inertia along with the increasing vehicle speed
as it is directly linked to the wheel speed by the xed third gear. In Figure 16(a) one can see
for the 4AT that along the second power iso-curve (39 [kW]) the engine torque is decreased with
increasing engine speed in order to keep the engine output power at the 39 [kW]. This obviously
brings the engine into worse fuel economic operating regions. These dierences clearly illustrate
one of the benets of the CVT in that it is able to operate the engine in a xed (fuel economic)
operating point even though the vehicle is accelerating (or decelerating).
The acceleration of the vehicle in Figure 16 shows about the same characteristics as the wheel
power transients. The decrease in acceleration after pedal tip-in is caused by the increasing vehicle
road load Trv in equation (1) along with the increasing vehicle speed, see Figure 16(c)
This vehicle speed shows latencies for the CVT at both pedal tip-in and back-out, which illustrates
one of the drawbacks of the CVT compared to the ZI and 4AT powertrains. Clearly the CVT is
not able to increase or decrease the wheel power level fast enough to be able to properly control
the vehicle speed without overshoots.
In Figure 16(d) the shifting patterns of the CVT, ZI and the 4AT are shown. At t = 0 [sec] the
gear ratio is in overdrive rod for the CVT and CVT, whereas for the 4AT it is in fourth gear.
Clearly the 4th gear has a much smaller gear ratio than rod . This also explains the higher engine
speed of the 4AT in the engine map before tip-in and after the back-out, see plot (a). Plot (d)
also illustrates that the decision to shift from 4th to 3th gear and back coincides with the time
point where the continuous gear ratio of the CVT approaches the respective xed gears. Clearly
this emerges since the shifting law (8) is also used for the 4AT.
From a driveability point of view it can be concluded that the ZI powertrain performs best in
realizing acceleration patterns such as described by the theoretical curve in Figure 7. As expected
the 4AT may be judged is relatively uncomfortable, i.e., it performs with periods of sag and shue,
see also Figures 6(b) and -(d). The CVT powertrain is rather unpredictable since it performs with
signicant hesitations, see also Figure 6(a).

16
 160 1.75
TWOT
140 1.5 4AT
2
120 1.25
engine torque [Nm]

acceleration [m/s ]
T

2
Eline
100 1
3
80 0.75 ZI
CVT, ZI 39 [kW]
CVT
60 3 0.5
1 17 [kW] 4AT
40 0.25
1
20 5.5 [kW] 0

0 0.25
0 100 200 300 400 500 600 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
engine speed [rad/s] time [s]

(a) Engine map (b) Vehicle acceleration

100
2.2
95
CVT
90 2
transmission ratio []
vehicle speed [km/h]

85 1.8
4AT
80
1.6
ZI
75
1.4
70 4th gear
1.2 ZI
65 CVT

60 1
3rd gear
4AT
55 0.8
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
time [s] time [s]

(c) Vehicle speed (d) Transmission ratio

Figure 16: Simulation results for pedal tip in and back with CVT, ZI and AT4 powertrains

In this paper nothing is said about interior sound generated by the respective powertrains at pedal
tip-in and back-out. The interior sound in the case of the CVT and ZI powertrains is the same, but
the sound generated by the 4AT is very dierent. For the 4AT, especially during the acceleration
of the vehicle the engine sound correlates highly with the increasing vehicle speed. As said the
CVT quickly brings the engine speed to a xed value and keeps it their during the entire vehicle
acceleration interval. In practice this is often remaked as a disadvantage of the CVT, since the
engine sound does not correlate with the increasing vehicle speed.
test vehicle
In Figure 17 the test vehicle (a), the ZI powertrain with engine and transmission (b) and the
control hardware (c) are illustrated.
Through experiments, the qualitative driveability, see Figure 7 and the fuel economy of the 4AT,
CVT and ZI powertrain are investigated. For the experiments, the test vehicle is equipped with a
basic CVT transmission (Van Doornes P920 transmission) extended with a planetary gearstage.
Later on the ywheel is installed to obtain the ZI powertrain. The test vehicle is a Volkswagen
Bora (Jetta in The Americas) with a 1.6  gasoline engine and originally equipped with a 4-gear
automatic transmission (4AT). Modications within the engine compartment were required to
install the developed prototype ZI transmission, which is about 70 mm longer than a commercially
available CVT transmission. More details about the mechanical design of this prototype are given
in [van Druten, 2001].

17
 (a) Test vehicle on the test track

(b) ZI powertrain, engine is right (c) Control hardware

Figure 17: VW Bora test vehicle

Software is developed for the control of the hydraulic actuation of the CVT (see [Vroemen, 2001]),
for the control of the air throttle (see [Serrarens, 2001]) and to generate setpoints for the CVT and
the throttle valve. These setpoints constitute the realization of the fuel economy and driveability
objectives as derived in this paper.
experimental results
Two types of experiments are shown in this section. First the responses of pedal jogging, i.e.,
pushing the pedal in and out in a (near) periodical manner are given. Secondly, the responses of a
full kickdown for the CVT, ZI and 4AT powertrains are shown. As explained earlier in this paper,
kickdowns are not a part of the driveability objective but rather show the maximal performance
of the powertrain. However, it is decided to investigate up to what extent the powertrains at issue
are able to fulll the driveability objective in a performance test.
-Pedal jogging
With the test vehicle discussed in the previous paragraph two pedal jogging experiments are
performed. In the rst experiment, the ywheel was still dismounted from the CVT powertrain,
whereas in the second experiment the ywheel was installed adapting the CVT- into a ZI power-
train. For all types of automated powertrains, this type of driveability experiments are generally
the toughest to handle properly. Especially in the case of an automated stepped transmission,
e.g., a 4AT, the decision to whether or not changing the transmission gear is very crucial in the
driveability feel of the vehicle in this kind of experiments. Although the CVT can change its
transmission ratio continuouslythus, cancelling the desicion problemthe responsiveness may
nonetheless be penalized importantly due to the continuously accelerating and decelerating pri-

18
mary inertias. At this level the ZI powertrain can be expected to show signicant driveability
improvements.
In Figures 18(a) and 18(b) the (scaled) torque responses upon a jogging pedal are shown. Besides
the engine speed in both graphs, Figure 18(b) also shows the absolute ywheel speed f
1 1
T
d

[], 103 T [Nm], 103 , 103 [rad/s]

0.9
T
Td [Nm], 10 e [rad/s]

0.8 0.8

f
0.7
3

e
0.6 0.6

0.5 f

0.4 0.4
3

e
[], 10

0.3

d
0.2 0.2

0.1 e

0 0
150 155 160 165 170 270 275 280 285 290
time [s] time [s]

(a) pedal jogging with the CVT powertrain (b) pedal jogging with the ZI powertrain

Figure 18: Pedal deflection , wheel torque Td , engine speed e and flywheel speed f for pedal jogging
experiment

The experiments for both powertrains were conducted with about the same jogging frequency,
but the induced pedal deections larger than zero have shorter time intervals in case of the CVT
powertrain. This can obviously be attributed to impreciseness of the driver. It could, on the
other hand, also be related to the specic behaviour of either of the powertrains that reects
dierently upon closed loop actions of the driver through sound, vibration, acceleration feel, etc.
In a qualitative sense the experiments are comparable though. Pedal jogging results in up and down
shifting of the CVT ratio to increase and decrease the engine speed according to equations (8)
and (9).
Regarding the experimental results for pedal jogging, the following observations are made:
accompanied by slight inverse response the torque responses on positive pedal deections of
the CVT powertrain seems to have the same time delay though is far less persistent than
the response of the ZI powertrain;
contrary to the CVT powertrain, the wheel torque for the ZI powertrain decreases steadily
after a pedal back-out;
the CVT powertrain even shows a relatively high increase of wheel torque after the pedal
has backed out already. Again, this is caused by the torque stemming from the decelerating
primary inertias. In the ZI powertrain, on the other hand, this torque is used to accelerate
the ywheel;
the average vehicle speed increases upon this sequence of pedal deections and this will also
increase the average (absolute) ywheel speed at approximately constant average engine
speed. This can indeed be observed in Figure 18(b).
These observations lead to the following conclusions:
the driveability penalty on pedal back-out is remarkably high for the CVT powertrain. This
is predicted, although to a lesser extent in the simulations of Figure 16(b);
the ZI powertrain fullls the driveability objective equally well for both pedal tip-in and
back-out.
the response of the ZI powertrain shows a more or less repeatable response, whereas for the
CVT powertrain this is less predictable. Most likely this is caused by a combination of lack
of responsiveness and a multitude of drive shaft oscillations induced after every pedal tip-in
and back-out. The ZI powertrain has the tendency to phase out such oscillations more than
the CVT powertrain, see [Serrarens et al., 2001].

19
-Kick-down

1200
moment of kickdown
1000

wheel torque T [Nm]

d
800
ZI

600

400

200 4AT

0
CVT

200
0 0.5 1 1.5 2 2.5 3 3.5
time [s]

Figure 19: Kickdowns from 80 [km/h] for 4AT, CVT and ZI powertrains

The response after a full pedal kick down is said to be related to the performance instead of
driveability. However, it is interesting to investigate to what extent the driveability objective is
met for such a performance demand. Therefore, kick down experiments are undertaken for the
vehicle with 4AT and after replacing it by the ZI transmission also with the CVT and ZI mode.
The wheel torque responses are shown in Figure 19.
All powertrains started at their respective overdrive ratios which is rod = 2.15 for CVT and ZI
and r4 = 1.25 (4th gear) for the 4AT. For the present case, the kick down starts with e = 270
[rad/s] for the 4AT and at e = 157 [rad/s] for the other two powertrains. This dierence of initial
operating point, results in a power reserve for the 4AT of 27 [kW], whereas for the CVT and ZI
this is about 9 [kW], at least if the 80 [km/h] is driven stationary.
The initial values of the wheel torques, however, dier somewhat. This is reected in slightly
dierent engine torques for CVT and ZI although the engine speeds are equal. In practice it is
dicult to have equal initial torques because they have to be controlled by the driver who is
focused on approximately maintaining a stationary vehicle speed instead of torque.
The transients of the three powertrains dier remarkably. The 4AT initially has a more prompt
response due to the much higher power reserve. However, when the 4AT starts to shift from the
fourth gear towards the third gear at t = 1.2 [sec], the wheel torque shows a severe transient that
lasts until the shift is nished at t = 1.8 [sec]. Shortly thereafter, at t = 2.1 [sec], the automatic
gearbox shifts further down to the second gear ratio. Again accompanied by an even more severe
transient, the wheel torque nally becomes (quasi-) stationary at t = 2.8 [sec], i.e., more than two
seconds after the pedal was fully pressed down.
The CVT transient shows the expected jet start behaviour (initial lack of wheel torque response),
but nevertheless reaches the nal torque level about 0.8 [sec] earlier than the 4AT. Finally, the ZI
powertrain shows a prompt and persistent increase of the wheel torque for the entire observation
interval. There is however, a slightly longer time interval between the pedal kick-down and the
response than for the 4AT. Morover, the initial response of the 4AT is steeper. For the ZI pow-
ertrain, the increase is inicted merely by the torque emerging from the (decelerating) ywheel,
i.e., by down shifting the CVT. The dynamic response of the hydraulically controlled CVT is
signicantly slower than the engine torque response and explains that the ZI powertrain initially
has a less steep response then the 4AT.
In conclusion, the ZI powertraincontrary to the CVT and 4AT powertrainsshows a response
that fullls the qualitative driveability objective fairly close even though a performance manoeuvre
is demanded.

20
CONCLUSIONS AND OUTLOOK
The nal section of this paper is preserved to present a brief overview of the results and conclu-
sions in this paper. Moreover it presents some directions for ongoing research on the Zero Inertia
Powertrain.
conclusions
In this paper a new CVT based powertrain is introduced. This powertrain coded Zero Inertia
Powertrain is able to combine optimal fuel economy with superior driveability. General ideas
on driveability for automated transmissions pointed out that the translation of the drive pedal
deection into a desired wheel torque trajectory is preferred. This trajectory should be promptly
increased or decreasing upon pedal tip-ins or back-outs, respectively. Furthermore, the wheel
torque trajectory should be smooth with preferably one extremum at the time instant where the
wheel power becomes constant.
These general ideas are incorporated in a coordinated powertrain controller. This controller emits
setpoints for the CVT and the engine electronic throttle valve such that the driveability is actively
controlled in transients and the fuel economy is preserved in quasi-stationary situations. Using
a powertrain model simulations are performed with the CVT and the ZI powertrain, moreover
they are compared with the simulated response of a 4 gear automatic transmission. From these
simulations it is concluded that the ZI powertrain is almost capable of realizing the preferred
driveability trajectory for both pedal tip-in and back-out. The CVT torque trajectory typicaly
hesitates and thus lacks the requested promptness. The 4AT, on the other hand, is rather prompt
but shows several uncomfortable extrema due to dynamics introduced by the gear shifting.
Two types of experiments are conducted with the powertrains. The rst one is pedal-jogging,
that is a sequence of pedal tip-ins and back-outs. For pedal jogging, the responsiveness of the
ZI powertrain is superior to that of the CVT powertrain. Slightly hampered by the actuator
bandwidth the wheel torque response of the ZI powertrain is consistent with the pedal tip-ins
and back-outs. This is not true for the CVT powertrain, the responsiveness of which is also not
repeatable. For the CVT powertrain pedal back-outs seem to be penalized more than tip-ins.
The second experiment being a kickdown manoeuvre showed that the ZI powertrain is able to
establish an immediate and persistent wheel power response. Kick-downs are often used to measure
acceleration time intervals of the vehicle, e.g., from 50 to 100 [km/h], and are therefore more related
to performance than comfort or driveability. The shape of the trajectory towards this acceleration
is often less important then, but as said, the ZI powertrain is still able to show a prompt, smooth
and persistently increasing torque trajectory.
In this paper fuel economy experiments are not discussed explicitly. Even though such experiments
are conducted pointing out that the ZI powertrain has fuel consumption benet with respect to
the 4AT of 10% on the NEDC cycle. For constant speed driving the fuel savings are about 20%,
see [Serrarens, 2001].
outlook
In this paper driveability was seen as a rather open-loop issue. In practice, however the driveability
is judged by a driver closing the control loop between vehicle and pedal. The driver is generally
exposed to trac situations where smooth control of the vehicle speed and relative position is
imprortant for the perception of the vehicles driveability. At this moment experiments concerning
this aspect of driveability are not performed with any of the powertrains yet. On the other hand
it may be expected that the improvement of the open-loop driveability due to the ZI principle
is instrumental to optimize the driveability in a closed loop sense. Directions in this area can be
found in [Passmore et al., 2001]. One of the conclusions in there was that the sensitivity of the
drive pedal to vehicle accelerations should be lower for smooth (part load) accelerations and ease
of longitudinal vehicle control than that required for best perceived vehicle performance.
The test vehicle is driven by CVT experts as well as non-experts. Both categories pointed out that
the ZI powertrain as applied did not improve the perception of engine sound in correlation with
the perceived vehicle acceleration. For semi-kick downs such as shown in Figures 15 and 16, all
three powertrains can be tuned to realize the same vehicle accelerations (apart from the transients
towards this acceleration). On the other hand, the CVT and ZI realize this acceleration keeping
the engine speed at a xed point. In case of the 4AT, however, the engine speed increases along
with the increasing wheel speed, which is generally leading to a better perceived driveability. Of
course, the CVT can be controlled as if it were a stepped automatic transmission, but fuel economy
would decrease then. It is interesting, however, to nd a new optimum of perceived driveability and

21
fuel economy explicitly accounting for the described cognitive phenomenon. Stressing especially
on driveability conceptually sound ideas for this optimization are presented in [Nasdal and Link,
2002].
To facilitate this investigation it is considered useful to install a tip-shift system in the test vehicle.
This brings up another aspect of driveability, namely that the perceived performance of the vehicle
is generally also correlated with the number of actions a driver has to perform to make his requests
explicit. The prompt response during shifting towards another xed transmission ratio in the case
of the ZI powertrain will probably be perceived fairly dierent for either automated shifting or
tip-shifting.
ACKNOWLEDGEMENT
This work forms part of the EcoDrive project which was nalized at the end of 2001. EcoDrive was
subsidized by the Dutch government through EET (Economy, Ecology and Technology). Project
partners were the Technische Universiteit in Eindhoven, Van Doornes Transmissie in Tilburg ,
and TNO-Automotive in Delft, The Netherlands.

References
Bergman, W. Measurement and subjective evaluation of vehicle handling. SAE Technical Paper
Series, no. 730492, 1973.
Brace, C. J., Vaughan, N. D., Burrows, C. R., and Deacon, M. Integrated control strategies for a
direct-injection diesel engine and CVT. In: Proc. EAEC European Automotive Congress, pp.
249260. Barcelona, Spain, 1999.
CEC. CEC Cold weather driveability test procedure. Coordinating European Council, 1983.
Cho, D. and Hedrick, J. K. Automotive powertrain modeling for control. Trans. of ASME,
Journal of Dynamics Systems, Measurement, and Control, vol. 111, pp. 568576, 1989.
DOE and EPA. Energy technology and fuel economy.
http://www.fueleconomy.gov/feg/atv.shtml, U. S. Departement of Energy, U. S. Environ-
mental Protection Agency, 2001.
van Druten, R. M. Transmission design of the zero inertia powertrain. Ph.D. thesis, Technische
Universiteit Eindhoven, 2001.
van Druten, R. M., Serrarens, A. F. A., Vroemen, B. G., van den Tillaart, E., and de Haas, J.
Mild hybrids with CVT: comparison of electrical and mechanical torque assist. VDI Berichte,
vol. 1610, pp. 331345, 2001.
Everett, R. L. Measuring vehicle driveability. SAE Technical Paper Series, no. 710137, 1971.
Graham, J. P., Leard, A. T., and Valade, C. T. Coordinating research council high and interme-
diate temperature driveability programm. SAE Technical Paper Series, no. 881671, 1988.
Guo, Z. Y., Yang, X. C., Yang, D., and Frank, A. A. On obtaining the best fuel economy and
performance for vehicles with engine-CVT transmissions. SAE Technical Paper Series, no.
881735, 1988.
Guzzella, L. and Schmid, A. M. Feedback linearization of spark-ignition engines with continuously
variable transmissions. IEEE Transactions on Control Systems Technology, vol. 3, no. 1, pp.
5460, 1995.
Harms, K. Perspectives in automotive control. In: Proc. 3rd IFAC workshop Advances in
Automotive Control, pp. 112. Karlsruhe, Germany, 2001.
Hirschlieb, G. C., Schiller, G., and Stottller, S. Engine control. In: Ronald K. Jurgen (editor),
Automotive Electronics Handbook, chap. 12. McGraw-Hill, 1999.
Junio, M., Roesgen, A., and Corvasce, F. Rolling resistance of tires. VDI Berichte, vol. 1505,
pp. 255298, 1999.
Kolmanovsky, I., Sun, J., and Wang, L. Coordinated control of lean burn gasoline engines with
continuously variable transmissions. In: Proc. of the American Control Conference, pp. 2673
2677. San Diego, CA, 1999.

22
List, O. H., , and Schoeggl, P. Objective evaluation of vehicle driveability. SAE Technical Paper
Series, no. 980204, 1998.
Liu, S. and Paden, B. A Survey of todays CVT controls. In: Proc. 36th IEEE Conf. Decision
& Control, pp. 47384743. San Diego, CA, USA, 1997.
Mayer, T. and Schroder, D. Nonlinear adaptive control of a CVT based parallel hybrid passenger
car. In: Proc. 2nd IFAC Wokshop Advances in Autmotive Control, pp. 115121. 1998.
Minowa, T., Kimura, H., Ishii, J., Morinaga, S., Shiraishi, T., and Ozaki, N. Smooth Gear Shift
Control System Using Estimated Torque. SAE Technical Paper Series, no. 941013, 1994.
Minowa, T., Kurata, K., Kuroiwa, H., Ibamoto, M., and Shida, M. Smooth Torque Control
System Using Dierential Value of Shaft Speed. SAE Technical Paper Series, no. 960431,
1996.
Narumi, N., Suzuki, H., and Sakakiyama, R. Trends of powertrain control. SAE Technical Paper
Series, no. 901154, 1990.
Nasdal, R. and Link, M. GALOP-IAVs Universal Speed Ratio Selection Strategy for ATs, CVTs
and Hybrid Drivetrains. SAE Technical Paper Series, no. 2002-01-1256, 2002.
Park, J. H. and Kim, C. Y. Wheel slip control in traction control system for vehicle stability.
Vehicle System Dynamics, vol. 31, no. 1, pp. 263278, 1999.
Passmore, M. A., Patel, A., and Lorentzen, R. The inuence of engine demand map design on
vehicle perceived performance. Int. J. Vehicle Design, vol. 26, no. 5, pp. 509522, 2001.
Pesgens, M. F. M. Driveability issues of the zero-inertia powertrain. Tech. Rep. DCT 2001.23,
Faculteit Werktuigbouwkunde, Technische Universiteit Eindhoven, 2001.
Petterson, M. and Hedstrom, L-G. Transmission-Torque Control for Gear Shifting with Engine
Control. SAE Technical Paper Series, no. 970864, 1997.
Pner, R. Steuerstrategien von CVT-getrieben: Ubersicht und Ausblick. VDI Berichte,
vol. 1467, pp. 313331, 1999.
Sackmann, M. and Krebs, V. Nonlinear control of a continuously variable transmission using
hyperstability theory. In: Proc. of the European Control Conference. Karlsruhe, Germany,
1999.
Schmid, A., Dietrich, P., Ginsburg, S., and Geering, H. P. Controlling a CVT-equipped hybrid
car. SAE Technical Paper Series, no. 950492, 1995.
Schwab, M. Electronically-controlled transmission systemscurrent position and future develop-
ments. SAE Technical Paper Series, no. 901158, 1990.
Seiert, U. and Walzer, P. Automobiltechnik der zukunft. VDI Verlag, 1989.
Serrarens, A., Shen, S., and Veldpaus, F. E. Control of a Flywheel Assisted Driveline with CVT.
submitted to ASME Jrnl. of Dynamic Systems, Measurement and Control , 2002.
Serrarens, A., Vroemen, B., and Veldpaus, F. A new CVT powertrain without jet start behaviour:
analysis of design, dynamics and control. In: P. Lugner and K. Hedrick (editors), Selected
papers from the 20th International Congress of Theoretical and Applied Mechanics, pp. 181
205. Zwets & Zeitlinger, 2001.
Serrarens, A. F. A. Coordinated Control of the Zero Inertia Powertrain. Ph.D. thesis, Technische
Universiteit Eindhoven, 2001.
Shafai, E. and Geering, H. P. Control issues in a fuel-optimal hybrid car. In: Proc. IFAC 13th
Triennial World Congress, pp. 231236. 1996.
Vroemen, B. Component control for the zero inertia powertrain. Ph.D. thesis, Technische
Universiteit Eindhoven, 2001.
Wicke, V., Brace, C. J., and Vaughan, N. D. The potential for simulation of driveability of CVT
vehicles. SAE Technical Paper Series, no. 2000-01-0830, 2000.
Yang, K.-J., Hong, K.-S., and Cho, D.-I. A Robust Control for Engine and Transmission Systems:
Enhancement of Shift Quality. JSME Int. Jrnl. Series C , vol. 44, no. 3, pp. 697707, 2001.

23