Proceedings

Proceedings of
of the
the 3rd
3rd IFAC
IFAC Conference
Conference on
on
Advances
Advances in
Proceedings
in Proportional-Integral-Derivative
of the 3rd IFAC Conference on Control
Available
Proportional-Integral-Derivative online at www.sciencedirect.com
Proceedings of
Ghent, Belgium,
Belgium,
Advances the 3rd IFAC
May 9-11, Conference
9-11, 2018
2018 on Control
in Proportional-Integral-Derivative Control
Ghent,
Advances in May
Proportional-Integral-Derivative Control
Ghent, Belgium, May 9-11, 2018
Ghent, Belgium, May 9-11, 2018
ScienceDirect
IFAC PapersOnLine 51-4 (2018) 101–106

Adaptive
Adaptive PI
PI Controller
Controller for
for Slip
Slip controlled
controlled
Adaptive
Adaptive PI
PI Controller
Controller for
for Slip
Slip controlled
controlled
Belt
Belt Continuously
Continuously Variable
Variable Transmission
Transmission
Belt Continuously
Belt Continuously Variable Transmission
Variable Transmission
∗,∗∗∗ ∗,∗∗,∗∗∗
Florian Verbelen
Florian Verbelen ∗,∗∗∗ Michiel Michiel Haemers
Haemers ∗,∗∗,∗∗∗
∗,∗∗∗
FlorianDe
Jasper
Jasper DeVerbelen
Viaene
Viaene
∗,∗∗
∗,∗∗ Michiel
∗,∗∗∗Stijn
Stijn Haemers ∗,∗∗,∗∗∗
Derammelaere
Derammelaere
∗∗,∗∗∗
∗∗,∗∗∗
∗,∗∗,∗∗∗
Florian
Jasper Kurt DeVerbelen
Viaene
Stockman ∗,∗∗
∗,∗∗∗Michiel
Stijn
∗,∗∗∗ Peter Haemers
Derammelaere
Sergeant ∗,∗∗∗
∗,∗∗∗
∗∗,∗∗∗
Jasper Kurt De Stockman
Viaene ∗,∗∗ Peter
Stijn
∗,∗∗∗ Sergeant ∗,∗∗∗∗∗,∗∗∗
Derammelaere
Kurt Stockman ∗,∗∗∗ Peter Sergeant ∗,∗∗∗
∗ Kurt Stockman Peter Sergeant
∗ Department of Electrical Energy, Metals, Mechanical Constructions
∗ Department of Electrical Energy, Metals, Mechanical Constructions
∗ Department and of
and Electrical
Systems,
Systems, Energy,
Ghent
Ghent Metals,Belgium
University,
University, Mechanical
Belgium Constructions
(e-mail:
(e-mail:
Department and of Electrical
Systems, Energy,
Ghent Metals,Belgium
University,
florian.verbelen@ugent.be). Mechanical Constructions
(e-mail:
∗∗ and Systems, florian.verbelen@ugent.be).
Ghent University, Belgium (e-mail:
florian.verbelen@ugent.be).
∗∗ Department of Electromechanics, Op3Mech,
Op3Mech, University
University of of Antwerp,
∗∗ Department of Electromechanics,
florian.verbelen@ugent.be). Antwerp,
∗∗ Department of Electromechanics, Belgium
BelgiumOp3Mech, University of Antwerp,
Department
∗∗∗ of Electromechanics,
∗∗∗ Member of EEDT partner Belgium of Op3Mech, University of Antwerp,
∗∗∗ Member of EEDT partner Belgium of Flanders
Flanders Make,
Make, www.eedt.ugent.be
www.eedt.ugent.be
∗∗∗ Member of EEDT partner of Flanders Make, www.eedt.ugent.be
Member of EEDT partner of Flanders Make, www.eedt.ugent.be
Abstract: The
Abstract: The control
control of of slip
slip inin aa belt
belt Continuously
Continuously Variable
Variable Transmission
Transmission (CVT) (CVT) has has been
been
Abstract:
the
the subject
subject The The
of control
many
of many of
research
research slip in a
papers. belt Continuously
Optimal control Variable
of the belt Transmission
CVT is of (CVT)
major has
importance been
Abstract:
the
for thesubject
the efficiency control
of many of slippapers.
research
as demonstrated
demonstrated in in
a belt
papers.
Optimal
literature.
control of
Continuously
Optimal control of
The challenge
the beltTransmission
Variable
the belt
challenge
CVT is of major
CVT is ofthat
in optimizing
optimizing (CVT)
major
that
importance
has been
importance
efficiency is in
in
for
the efficiency
subject of as
many research in literature.
papers. Optimal The
control of the in
belt CVT is of major efficiency
importance is
for
the the
the the efficiency
reduction
reduction of as
the
of the demonstrated
necessary in
clamping
necessary clamping literature.
forces The
while
forces The challenge
the
whilechallenge in
stability
the stability optimizing
of the that
variable
of the variable efficiency
transmission
transmission is in
for
the efficiency
reduction
is maintained.
maintained. of as demonstrated
the
Although necessary
these problemsin literature.
clamping
problems forcesalready
have while the
already been in optimizing
stability
tackled ofto
tothe that extent,
variable efficiency
transmission is in
is
the reduction Although
of the these
necessary clamping have
forces while been
the tackled
stability of aa certain
the
certain
variable extent, mostly
mostly
transmission
is maintained.
fairly
fairly complex Although
complex controllersthese
controllers are problems The
are proposed.
proposed. havegoal
The already
goal been
of this
of this paper
papertackled totopropose
is to
is a certain
propose extent, mostly
aa straightforward
straightforward
is maintained.
fairly
though complex
effective Although
controllers
method these
toare
controlproblems
proposed.
slip. Thehave
The mainalready
goal of
idea been
this
of paper
the tackled
paper is to
isto ause
propose
to certain a extent,
straightforward
linearized mostly
equations
thoughcomplex
fairly effectivecontrollers
method toare control slip. The
proposed. The main
goal idea
of of the
this paper paper is is to
to use linearized
propose a equations
straightforward
though
of the
of the slip effective
slip dynamics method
dynamics to to
to update control
update the slip.
the controllerThe main
controller parameters idea
parameters inof the paper
in function
function of is to use linearized
of the operating
operating point. equations
point. This
This
though
of the slip
approach effective
dynamics
allows method
to reduce
reduce to the
to update control
the slip.nonlinear
The main
controller
highly idea ofto
parameters
system theafunction
in paper
first orderisoftothe
order theuse linearized
operating
transfer equations
point.
function This
which
approach
of allows
the slipcontrolled
dynamics to to update the highly
the nonlinear
controller system
parameters to a first
in afunction transfer
ofsimulations
the operating function
point. which
This
approach
is easily
is easily allows
controlled to with
reduce
with thecontroller.
aa PI
PI highly
controller. nonlinear
Results
Results system
based on
based to
on first order
extensive
extensive transfer
simulations function
show
show that
thatwhich
the
the
approach
is easily
controller allows
controlled
is robustto reduce
with
against a the
PI highly
controller.
torque nonlinear
Results
disturbances system
based
and speedto
on a first
extensive
ratio order transfer
simulations
variations. function
show thatwhich
the
controller
is is robust with
easily controlled against a PI torque disturbances
controller. Resultsandbasedspeed ratio variations.
on extensive simulations show that the
controller is robust against torque disturbances and speed ratio variations.
© 2018, IFAC
controller (International
is robust againstFederation of Automatic and
torque disturbances Control)
speed Hosting
ratioby Elsevier Ltd. All rights reserved.
variations.
Keywords: Adaptive
Keywords: Adaptive control,
control, PI PI controllers,
controllers, mechanical
mechanical systems,
systems, dynamic dynamic models models
Keywords: Adaptive control, PI controllers, mechanical systems, dynamic models
Keywords: Adaptive control, PI controllers, mechanical systems, dynamic models
1. INTRODUCTION
1. INTRODUCTION ation by
ation by preventing
preventing gross gross slipslip at at all
all times
times (Bonsen
(Bonsen et et al.
al.
1. INTRODUCTION ation
(2005b)).by preventing
This urge gross
for slip
stability at all
will times
increase (Bonsen
the et
losses al.
in
1. INTRODUCTION (2005b)).
ation This urge for
by preventing grossstability
slip atwill all increase
times (Bonsen the losseset in
al.
(2005b)).
the hydraulic This urge
circuit for stability
significantly. will increase
Another the
negative losses in
side-
Manufacturers of
Manufacturers of advanced
advanced drive drive trains
trains specialized
specialized in the the
in the the
effect
hydraulic
(2005b)).
hydraulic
of this
this
circuit
Thiscontrol
urge for
circuit
significantly.
stability
significantly.
strategy is will
Another
increase
Another
increased
negative
theand
negative
wear
side-
losses in
side-
thus
Manufacturers
automotive industry of advanced
industry encounter drive
encounter an trains
an ever specialized
ever growing
growing pressure in the
pressure effect effect
the of
hydraulic control
circuit strategy
significantly. is increased
Another wear
negativeand thus
side-
automotive
Manufacturers of advanced drive reduced of life
thisspancontrol
span of the strategy
the is increased wear and thus
transmission.
automotive
due to
due industry
to increasingly
increasingly encounter
stringent
stringent antrains
ever specialized
regulation
regulation growing
concerning
concerning
in fuel
pressurethe reduced
fuel
effect of life
reduced thisspan
life controlof
of the
transmission.
strategy is increased wear and thus
transmission.
automotive
due to
consumption industry
increasingly
and encounter
stringent
emissions. As an ever they
regulation
a result, growing
concerning
are pressure
forced fuel
to The
reducedefficiency
life spancan ofbe theincreased
transmission. by lowering
lowering the the clamping
clamping
consumption
due to increasinglyand emissions.
stringent Asregulation
a result, they are forcedfuel
concerning to The The
efficiency can be increased by
consumption
innovate and and
one emissions.
of those As a
innovationsresult, is they
the are forced
Continuously to forcesefficiency
forces but than
but thancan a be increased
a robust
robust slip controller
slip by lowering
controller is the clamping
is necessary.
necessary. Bonsen
Bonsen
innovate
consumption and and one emissions.
of those innovations
As a result, is they
the Continuously
are forced to The
forces efficiency
but than can
a be increased
robust slip by lowering
controller is theon
necessary. clamping
Bonsen
innovate and
Variable Transmissionone of
Transmission (CVT).those innovations
(CVT). In In aa CVT is
CVT thethe Continuously
the speed
speed ratioratio et et al. (2005b)
al. (2005b) elaborated
elaborated such a controller
suchcontroller
a controller based
based on Bonsen the slip
the slip
Variable
innovate and one of those innovations is the Continuously forces
et al. but than
(2005b) a robust
elaborated slip
such a controller is necessary.
based on the slip
Variable
can be Transmission
varied continuously (CVT). In
between a CVT
two the
finite speed
values. ratio dynamics of the CVT and the PID design methodology
can
can
be varied
Variable continuously
Transmission
be varied continuously (CVT). between
In a two
between CVTfinite
two finite speed ratio dynamics
the values.
values.
et al. (2005b)
dynamics
described
of the CVT and
elaborated
of Panagopoulos
by the CVT and
Panagopoulos such the
the
et aal. PID design
controller
PID
(2002).designbased
The
methodology
on theshow
methodology
results slip
Many
can types
be types
varied of of CVT are
continuously are between
discussed twoininfinite
literature,
values. e.g. described
dynamics
e.g. described by
of the CVT and et
the al. (2002).
PID The
design results
methodology show
Many CVT discussed literature, a substantial
substantial by Panagopoulos
increase in in thetheet efficiency
al. (2002).of
efficiency ofThetheresults
drive show
train
Many
toroidaltypes
toroidal CVT of(Carbone
CVT CVT are
(Carbone et discussed
et al. (2004)),inbelt
al. (2004)), literature,
belt CVT (Car-
CVT e.g. aadescribed
(Car- substantial
while slip by
is
increase
Panagopoulos
increase
still adequately in theetcontrolled.
al. (2002).ofThe
efficiency
the
the
drive
results
drive
train
show
train
Many ettypes
toroidal
bone et CVT
al. of(Carbone
(2007)), CVT Milner are discussed
etCVT
al. (2004)),
CVT (Akehurst inbelt
literature,
et CVT e.g. while
(Car-
al. (2007)),
(2007)), slip is still
a substantial adequately
increase in thecontrolled.
efficiency of the drive train
bone
toroidal al.
CVT (2007)),
(Carbone Milner et al. (Akehurst
(2004)), beltet al.
CVT (Car- while slip is still adequately controlled.
bone et
hydraulic al.CVT(2007)), Milner
(Kempermann CVT (Akehurst
(2007)), wheelet al.
type (2007)),
CVT The
while objective
slip is stillof this
adequately papercontrolled.
is to
to demonstrate that that thethe
hydraulic
bone et al.CVT(2007)), (Kempermann
Milner CVT (2007)),
(Akehurst wheelet type
al. CVT The objective of this paper is to demonstrate
(2007)), The objective of this paper is
demonstrate that the
hydraulic
(Chen et CVT
et al.
al. (2017)), (Kempermann
(2017)), .. .. .. However,
However, the(2007)),
the belt wheel
belt CVTCVT is type
is the CVT
the most highly
most highly non linear
non linear slip dynamics can be reduced to a first
(Chen
hydraulic CVT (Kempermann (2007)), wheel type CVT The
orderobjective of slip
this dynamics
paper is tocandemonstrate
be reduced to thata first
the
(Chen
commonly
commonly et al. (2017)),
used
used CVT,
CVT, . . .among
However,
among all,
all, the
in belt
in CVT is
automotive
automotive the most highly
applica-
applica- order
highly
non linear
system.
system.
non
slip
Therefore
Therefore
linear slip
dynamics
no complex
no
dynamics
complex can be
can
reduced
control
control
be reduced
design
designto
to
arules,
a
first
rules,
first
(Chen
commonly et al. (2017)),
used CVT,
tions (Srivastava
(Srivastava . . .
and Haque However,
among(2009)).
Haque the belt CVT is
all, in automotive applica- as
(2009)). the most order
as used system.
used in in Bonsen Therefore
Bonsen et et al. no
al. (2005b),complex
(2005b), are control
are necessary.
necessary. design rules,
tions
commonly used CVT, and among(2009)).
all, in automotive applica- as order
used system.
in Bonsen Therefore no complex
et al. (2005b), are control
necessary. design rules,
tions (Srivastava and Haque This
as usedpaperin is
Bonsenstructured
et al. as follows.
(2005b), are Section
necessary. 2 shortly dis-dis-
The
tions ratio variation
(Srivastava andabilities
Haque of the
(2009)). belt CVT
The ratio variation abilities of the belt CVT are used in a This are used in a paper is structured as follows. Section 2 shortly
The ratio
vehicle to variation
enable the abilities
Internal of the belt
Combustion CVT are
Engine used
(ICE) intoa This
cussespaper
cusses the is structured
the operating
operating as follows.
principle
principle of the
of Section
the belt
belt CVT.
CVT. 2 shortly
Thereafter
Thereafter dis-
vehicle to enable theabilities
InternalofCombustion Engine (ICE) to This paper is structured as follows. Section 2inshortly
The ratio
vehicle
operate to variation
inenable
its most
most theoptimal
optimal the beltpoint,
Combustion
Internaloperating
operating CVT are used
Engine
which (ICE) into
results a cusses
the slip
the slipthe operating
dynamics
dynamics areprinciple
are derived
derived of
and
andthelinearized
belt CVT.
linearized sectiondis-
inThereafter
section 3.
3.
operate
vehicle in
to its
enable the Internal Combustion point, which
Engine results
(ICE) to cusses
the slipthe operating
dynamics are principle
derived of
andthe belt CVT.
linearized in Thereafter
section 3.
operate
in lower in
lower fuel its most
fuel consumption optimal
consumption as operating
as demonstrated point,
demonstrated by which
by Carboneresults
Carbone Section Section 4 discusses the design and implementation of the
in
operate infuel
itsand most optimal operating the slip 44dynamics
discussesare thederived
design and and linearized
implementation in sectionof the
section 3.
in
et lower
et al.
al. (2001)
(2001) consumption
and Van
Van as
der Sluis
der Sluis al. point,
demonstrated
et al.
et (2006).which
(2006). by
Despite
Despite
results
Carbone the Section
the
controllers
controllers
Section 4
discusses
while the
while
discusses
the theresults
the
designare
results
design
andpresented
are
and
implementation
presented
implementation
in
in of the
section
of
5.
5.
the
in lower fuel consumption asa significant
demonstrated partby of Carbone the controllers
Finally, in while the
in Section
Section results are presented the in sectionare 5.
et
et
al.
al.
(2001)
potential
potential
(2001)
of the
of and
the
and
Van der Sluis
technology,
technology,
Van der
et al. (2006).
a significant
Sluis et al. part
(2006).
Despite
of
Despite
the fuel
the fuel
the Finally,
controllers while the 66 the
the
results
conclusions
conclusions
are presented
of the
of research
research
in sectionare
are
5.
potential
savings is isoflost
lost theduetechnology,
due to thethe low a significant
low efficiency part of the
theof variable Finally,
formulated.
the fuel formulated.
variable in Section 6 the conclusions of the research
savings
potential of the to
technology, a efficiency
significant of
part of the fuel Finally, in Section 6 the conclusions of the research are
savings is lost due to the low efficiency of the variable formulated.
transmission.
transmission.
savings is lost due to the low efficiency of the variable formulated. 2.
transmission. 2. OPERATING
OPERATING PRINCIPLE PRINCIPLE
The main
transmission. reason for the CVTs
The main reason for the CVTs low efficiency are the low efficiency are the 2. OPERATING PRINCIPLE
The
high main
clamping reason forcesfor which
the CVTs
which are used
usedlow totoefficiency
transfer are the The
torque. The belt
belt CVT CVT 2. consists
OPERATING of 22 pulleysPRINCIPLE
pulleys which are are composed
composed
high
The clamping
main reason forcesfor the are
CVTs low transfer
efficiency torque.
are the consists of which
high
These
These clamping
clamping
clampingforces forces
forces, which
forces,which are
delivered
delivered used
by to
by hydraulictransfer
hydraulic torque.
cylinders,
cylinders, The
of
of a belt
fixed
a belt
fixedCVT CVT
half consists
and
half consists a
and a movable of
movable 2 pulleys
half which
(sheave),
half (sheave), are composed
see Fig. 1.1.
high
These
are clamping
clamping
typically chosenforces,
30% too are
deliveredhighused
byto to transfer
hydraulic
ensure stable torque.
cylinders,
oper- The
of
By a fixed
adapting halfthe and
positiona ofof2the
movable pulleys
half
movable which
(sheave), aresee
sheaves, see
Fig.
composed
Fig.
the 1.
belt
are
Thesetypically
clamping chosen 30%delivered
forces, too high to hydraulic
ensure stable oper- By of aadapting theand position of the half movable sheaves, seetheFig.belt
are typically chosen 30% too highby to ensure stable cylinders,
oper- By fixed half
adapting the position a movableof the movable (sheave),
sheaves, the 1.
belt
are typically chosen 30% too high to ensure stable oper- By adapting
2405-8963 © 2018, IFAC (International Federation of Automatic Control) Hosting by Elsevier Ltd. All rights reserved.the position of the movable sheaves, the belt
Copyright
Copyright
Peer review© 2018
2018 IFAC
© under IFAC 101
101 Control.
responsibility of International Federation of Automatic
Copyright © 2018 IFAC
10.1016/j.ifacol.2018.06.045 101
Copyright © 2018 IFAC 101
IFAC PID 2018
102 Belgium, May 9-11, 2018
Ghent, Florian Verbelen et al. / IFAC PapersOnLine 51-4 (2018) 101–106

runs on another radius and as a result the speed ratio τs Which can be rewritten as:
is altered. Movement of the sheaves and thus control of ω˙s ωp −ω˙p ωs
τ̇t ωωps − ωp2 τt
the speed ratio is done via 2 hydraulic cylinders which ν̇ = (4)
τt2
deliver a force on the primary Fp and secondary Fs pulley.
Consequently the belt CVT is a Multiple Input Multiple The derivatives of the primary ω˙p and secondary speed ω˙s
Output (MIMO) system: the forces (input) will not only in eq. (4) can be replaced by respectively eq. (5) and eq.
be used to control the speed ratio (output) but also to (6).
control the slip (output) as mentioned in the introduction. TICE − Tp
ω˙p = (5)
Moreover the clamping force on the secondary pulley is Jp
used for slip control purposes while the ratio of both Ts − Tload
clamping forces is used to control the speed ratio as ω˙s = (6)
Js
demonstrated by Carbone et al. (2007).
Considering eq. (5) and (6), eq. (4) can be redrafted in
terms of torque and inertia:
3. LINEARIZED SLIP DYNAMICS
τ̇t ωs Ts − Tload (TICE − Tp ) (1 − ν)
ν̇ = 2 − + (7)
As demonstrated by Bonsen et al. (2005b), it is vital to τt ωp ω p J s τt ωp Jp
model the slip dynamics in order to design a proper slip By using eq. (1), the equation describing the slip dynamics
controller. The model is based on Fig. 1 in which TICE and can be finalized:  
Tload are respectively the engine torque and load torque, τ̇t ωs 1 2Fs µRs Tload
the inertia of both pulleys is denoted by Jp and Js and the ν̇ = 2 + − +
τt ωp ω Js cos βτt J τ
speed by ωp and ωs . Subscript p stands for the primary side  p s t (8)
while subscript s stands for the secondary side. 1−ν 2Fs µRs τt TICE
+ − +
ωp Jp cos β Jp
Besides engine TICE and load torque Tload , Tp and Ts are
The variation in slip is thus a function of 10 variables:
mentioned on Fig. 1. These variables represent the torque
produced via friction and are calculated as follows (Bonsen ν̇ = f (ν, τ̇t , τt , ωs , ωp , Fs , Rs , TICE , Tload , µ) (9)
et al. (2005b)): Of these variables, only the secondary clamping force
2Fs µ (ν) Rp,s Fs can be used to actively control the slip. All other
Tp,s = (1) variables are controlled at a higher level by for example
cos β
the energy management system which selects optimal
with β the half cone angle of the pulleys, µ the friction operating points (TICE , ωp , . . . ) or are induced by the
coefficient which depends on ν the slip and Rp,s the considered load profile (Tload , ωs , . . . ). Therefore, these
running radii of the belt. variables are considered as disturbances D.
Driver pulley
Notice that the primary clamping force Fp does not appear
Fp in the equation. However, this does not mean that Fp has
Jp Rp no impact on the slip. As shortly mentioned in section 2,
the speed ratio τt of the CVT depends on the ratio of
Tp TICE ωp the clamping forces. Consequently, an increase of Fp will
lead to a variation in speed ratio τt . As the speed ratio
ωs Tload Ts
has an impact on the slip dynamics (see eq. (8)), Fs will
Fs also have to change to maintain the desired slip value. The
Rs Js secondary clamping force has thus an indirect effect on the
slip.
2β In the following paragraphs, the slip dynamics are lin-
Driven pulley earized for control purposes. However, the model of the
CVT which is used to test the behavior of the elaborated
Fig. 1. Belt CVT controller, is not simplified to any extend. If eq. (8) is lin-
When the speed ratio is varied or there is any load earized, ignoring the variables considered as disturbances
disturbance or variation, slip variations will occur. Slip D, the following equation is found:
ν is defined by the ratio of the actual speed ratio τs and ∂f 
ν̇ = Dτ̇t + Dτt + Dωs + Dωp +  (Fs − Fs∗ ) + DRs
the theoretical or geometrical speed ratio τt , see eq. (2). ∂Fs ∗
In eq. (2), the actual speed ratio τs is calculated as the ∂f 
ratio of the secondary ωs and primary speed ωp while the + DTload + DTICE +  (ν − ν ∗ ) + Dµ
∂ν ∗
theoretical speed ratio τt is defined by the ratio of the (10)
primary Rp and secondary radius Rs .
−τs
ωs ∂f  ∂f 
ν= +1=
ωp
+1 (2) ν̇ = D +  (Fs − Fs∗ ) + NRs +  (ν − ν ∗ ) (11)
τt Rp ∂Fs ∗ ∂ν ∗
Rs
As it is the objective to control the variations in slip, the Which can be rewritten as:
derivative is considered: ∂f  ∗ ∂f  ∗ ∂f  ∂f 
ν̇ = D −  Fs −  ν +  Fs +  ν (12)
τ̇t τs − τ̇s τt ∂F ∗ ∂ν ∗ ∂F ∗ ∂ν
 ∗
ν̇ = (3)  s    s 
τt2 K L M

102
IFAC PID 2018

Ghent, Belgium, May 9-11, 2018 Florian Verbelen et al. / IFAC PapersOnLine 51-4 (2018) 101–106 103

In which the partial derivative of Fs equals: with the proportional term Kp and the integral term Ti .
∂f  2µRs (1 − ν) 2µRs τt The integral term should be equal to the time constant of
 =− − =L (13) the first order system depicted by eq. (16). If the system
∂Fs ∗ ωp Js cos βτt ωp Jp cos β
pole is canceled by the controller zero, the dynamical
And the partial derivative of ν: behavior can be uniquely defined by the control parameter
∂f  2Fs µRs τt TICE Kp . Ti is therefore equal to:
 = − =M (14)
∂ν ∗ ωp Jp cos β ωp Jp
−1
Ti = (18)
The next step is to take the Laplace of eq. (12): M
ν(s)s = K + LFs (s) + M ν(s) (15) Note that this can only be done if the pole is in the left
half plane (see Fig. 3 a)) which means that M needs to
Note that the disturbance term has been removed as it be negative. This will be the case if TICE > Ts τt which is
has no impact on the actual linearized transfer function. valid for almost all operating conditions. Only when the
The term K can also be removed as it is merely an offset ICE is turned off while the belt is still rotating this method
which can be ignored for control purposes. Therefore, the becomes erroneous. However this is an unlikely event as
following transfer function is found: the ICE is only turned off when the vehicle is close to a
−L standstill or is at standstill. In that case, slip control is no
ν(s) M
= −1 (16) longer necessary. To calculate the proportional term, the
Fs (s) M +s 1
closed loop system is taken into account:
This means that the complex nonlinear behavior of the slip
dynamics is now converted to a first order system which −L
M Kp
can be easily controlled. T FCL = (19)
Ti s + −L
M Kp
4. CONTROLLER DESIGN AND IMPLEMENTATION Which means that the closed loop pole defining the closed
loop dynamics can be found at:
4.1 Control architecture L
Kp
s= M (20)
The belt CVT is a MIMO system as highlighted in section Ti
2. To decouple the control loop of the ratio and the slip the By which the proportional gain of the controller can be
scheme presented in Fig. 2 is used. In this scheme, the ratio defined with one degree of freedom α:
controller yields the ratio of the forces on the primary and αTi −α
secondary pulley while the slip controller yields the force Kp = L = (21)
on the secondary pulley. M
L

ν∗ Fs∗ The value α defines the position of the closed loop pool
Slip
Controller as demonstrated in Fig. 3 (red line) and is related to the
- ability to handle disturbances as will be demonstrated in
ν
Fp∗ section 5. The black cross resembles the system pole while
τ∗ Fp∗
Ratio Fs∗ × the blue cross and circle resemble respectively the pole and
- Controller zero of the PI controller. Only case a) is discussed in this
τs paper because of the previously mentioned reason.

Fig. 2. Control architecture

4.2 Ratio controller

As control algorithm, a PI controller is chosen, which has

been tuned based on the work of Simons et al. (2008). A
differential action is not necessary as the shifting process
exhibits a sufficient amount of damping (Simons et al.
(2008)). The implementation of this PI controller with
fixed Kp and Ti is rather trivial and is therefore not further
discussed.
a) b)
Fig. 3. a) Root locus for the case in which TICE > Ts τt :
left half plane system pole. b) Root locus for the case
in which TICE < Ts τt : right half plane system pole.

4.3 Slip controller The implementation of the PI controller comes thus down
to the determination of parameters L and M . Parameter
Based on section 3 it is known that the slip dynamics L is calculated based on eq. (13) while parameter M can
can be described as a first order system. Therefore a PI be determined with eq. (15).
controller should result in acceptable slip behavior. The The many parameters in those equations define the oper-
transfer function of a PI controller
 canbe written as: ating point based on which the linearization is done. As all
Ti s + 1 inputs of eq. (13) and eq. (15) are constantly varying, the
T FP I = Kp (17)
Ti s equations need to be solved repeatedly during simulation

103
IFAC PID 2018
104 Belgium, May 9-11, 2018
Ghent, Florian Verbelen et al. / IFAC PapersOnLine 51-4 (2018) 101–106

to feed the PI controller with the optimal Kp and Ti at all for the specific load variation. Furthermore it is possible
times. However, the computational effort of solving these to conclude that there is no further improvement of the
static relations is negligible. response of the controller for α values larger than 250.
Therefore, the remaining simulations have been performed
Another complexity of the determination of L and M is the with α equal to 250.
number of parameters which need to be obtained to solve
the equations, see Fig. 4 a). This is of course not an issue
in simulation but when tested on a test bench this becomes 8
important. To counter this concern, some simplifications α = 20
are proposed to reduce the number of variables. 7 α = 100
α = 250
α α α = 500
C C 6
Js Js
C C

ν [%]
Jp Jp 5
C C
β β
C L C L
ν ν∗ 4
C
µ µ
C
τt τs∗ 3
M C M
Rs Rs
ωp ωp 2
2.45 2.5 2.55 2.6 2.65 2.7
TICE TICE Time [s]
Fs Fs∗
Fig. 5. Slip variation due to a stepwise load variation of
a) b) 25Nm at t=2.5s for varying values of α.

Fig. 4. a) Overview on the inputs to calculate L and M . b)

Overview on the strictly necessary inputs to calculate
L and M . The value C stands for a constant.
As constant slip is expected due to proper control, the
friction coefficient µ will be fairly stable. The first adap-
tation is thus to consider a fixed friction coefficient µ.
Furthermore, on the hypothesis of proper slip control,
the slip is presumed equal to the setpoint. Based on the
assumption of proper slip control and low values for slip it
is also possible to equalize the theoretical and the actual
speed ratio (see eq. (2)). Combined with an appropriate
ratio controller it is possible to obtain τt and the running
radii Rp,s directly from the speed ratio setpoint. The last
simplification is to use the output of the slip controller to
estimate the force on the secondary pulley Fs instead of
measuring the clamping force.
Due to these changes, only 2 measurements remain: speed
of the primary pulley ωp and the torque on the primary
pulley TICE . Of these variables speed is easily measured
and torque of the ICE could be estimated based on engine
characteristics (Zweiri et al. (2006)) to avoid an expensive
torque sensor.
5. RESULTS
5.1 Impact of the parameter α
As mentioned in section 4.3, the proportional gain Kp was
defined with one degree of freedom left. This allows the
control engineer to test the behavior of the controlled sys-
tem on torque disturbances. Fig. 5 shows the response for
varying α values. At t equal to 2.5s there is a stepwise load
variation of 25Nm which is in correspondence with the test
done by Bonsen et al. (2005a). In all cases, the controller
is able to stabilize the slip, however the larger the value
for α, the more immune the controlled system becomes
5.2 Varying load at constant speed ratio
In the previous subsection, the impact of one specific load
disturbance is highlighted to analyze the impact of the
factor α. In this subsection the results are discussed of
both positive and negative load disturbances of varying
magnitude. Fig. 6 shows that the magnitude of the load
variation has no significant impact on the settling time of
the controller. Only the overshoot increases for increasing
load. Fig. 6 also shows that not only the force on the
secondary pulley changes due to the load variations but
also the force on the primary pulley. The reason for this
effect is that the speed ratio of the CVT is defined by the
ratio of both forces on the pulleys as already mentioned in
subsection 4.2. As the secondary force needs to be changed
to maintain a constant value for the slip, the primary force
needs to change as well to maintain the desired, constant,
value for the speed ratio.
5.3 Varying speed ratio at constant load
According to eq. (10), the speed ratio could also be a
disturbance for the slip controller. However, Fig. 7 shows
that the proposed control architecture has no difficulties
with sustaining constant slip values while the speed ratio
is following a desired curve. Note that there are 2 peaks in
the profile of the secondary clamping force. By increasing
the force on the secondary clamping force the torque
Ts developed by that pulley increases. As a result the
secondary pulley is accelerated by which the imposed
speed ratio τs∗ value is maintained.
5.4 Varying speed ratio and load
Fig 6 and 7 clearly showed the interaction between the slip
and ratio controller but in those cases speed ratio or load

104
IFAC PID 2018

Ghent, Belgium, May 9-11, 2018 Florian Verbelen et al. / IFAC PapersOnLine 51-4 (2018) 101–106 105

contrast with the integral term Ti where a steep decrease is

4 noticeable after 3s. Furthermore Ti also varies during ratio
ν [%]

3 variation. This can also be clarified as Ti depends uniquely

on M (see eq. (18)) which is a function of, among others,
2 torque and speed ratio. Fig. 10 thus clearly demonstrates
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
the dependency of the controller settings on the operating
10 conditions.
Fp,s [kN]

5 Fp
Fs 3
0

ν [%]
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 2
1
0
Tload [Nm]

1 1.5 2 2.5 3 3.5 4 4.5 5 5.5 6

50
2

τs [%]
0 τs∗
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 1.5
Time [s] τs
1
1 1.5 2 2.5 3 3.5 4 4.5 5 5.5 6

Tload [Nm]
Fig. 6. Slip ν and clamping forces Fp,s for load torque Tload
variations in function of time at constant speed ratio 50
τs of 1.
0
1 1.5 2 2.5 3 3.5 4 4.5 5 5.5 6
5 Time [s]
ν [%]

0 Fig. 8. Slip ν, speed ratio τs and load torque Tload in

2 4 6 8 10 12 function of time.
10
Fp,s [kN]

Fp 22
5
Fs 20
0
2 4 6 8 10 12 18
16 Fp
2
τs [/]

14
Fp,s [kN]

1
12
2 4 6 8 10 12
Time [s] 10
Fs
8

Fig. 7. Slip ν, clamping forces Fp,s and speed ratio τs in 6

function of time for a constant load torque Tload of
30Nm.
torque were constant. Fig. 8 shows that the controllers
can also handle simultaneous variations of load torque
and speed ratio. As in Fig. 6, a small increase in slip is
noticeable due to the increased load torque. The impact
due to ratio variation is non-existent.
Fig. 9 shows the corresponding clamping forces which show
to be in coherence with the previously presented data.
Again due to ratio variation, a peak in the clamping force
profile of the primary pulley appears in order to increase
the torque on that pulley. After 3s the clamping forces
increase due to varying load torque similar to the results
shown in Fig. 6.
Besides the corresponding clamping forces, it is also useful
to plot Kp and Ti for this simulation. Fig. 10 shows that,
for this case, Kp starts varying when the speed ratio
changes. Moreover, the impact of the stepwise load torque
variation is not visible in the value for Kp . This makes
sense as Kp is only a function of α and L (see eq. (21))
which do not depend on the load torque Tload . This is in
4
2
1 1.5 2 2.5 3 3.5 4 4.5 5 5.5 6
Time [s]
Fig. 9. Corresponding clamping forces Fp,s for Fig. 8 in
function of time.
6. CONCLUSION
In this paper a method is elaborated to design a PI
controller for a slip controlled belt CVT. In contrast to
earlier literature, this paper proposes a straightforward
adaptive controller without losing robustness. To this end
the complex nonlinear slip dynamics were simplified to
a first order transfer function which enabled the use of
basic control design rules. The results which were obtained
based on a detailed model, show that the slip controller
can handle torque disturbances and is immune for speed
ratio variations. These results justify the simplifications
which were made in the controller design. Based on these
promising results, measurements on a test bench can be
done to see wheter these simplifications still hold.

105
IFAC PID 2018
106
Ghent, Belgium, May 9-11, 2018 Florian Verbelen et al. / IFAC PapersOnLine 51-4 (2018) 101–106

Panagopoulos, H., Åström, K.J., and Hagglund, T. (2002).

-0.5 Design of PID controllers based on constrained optimi-
sation. IEE Proceedings - Control Theory and Applica-
s ]
Kp [ kgm

-1
tions, 149(1), 32.
-1.5 Simons, S., Klaassen, T., Veenhuizen, P., and Carbone,
G. (2008). Shift dynamics modelling for optimisation of
-2 variator slip control in a pushbelt CVT. International
1 1.5 2 2.5 3 3.5 4 4.5 5 5.5 6 Journal of Vehicle Design, 48(1/2), 45.
Srivastava, N. and Haque, I. (2009). A review on belt and
0.15 chain continuously variable transmissions (CVT): Dy-
namics and control. Mechanism and Machine Theory,
0.1 44(1), 19–41.
Ti [s]

Van der Sluis, F., Van Dongen, T., Van Spijk, G.J., Van
0.05
der Velde, A., and Van Heeswijk, A. (2006). Fuel con-
0 sumption potential of the pushbelt CVT. In Proceedings
1 1.5 2 2.5 3 3.5 4 4.5 5 5.5 6 of FISITA 2006 World Automotive Congress, 1–12.
Time [s] Zweiri, Y.H., Whidborne, J.F., and Seneviratne, L.D.
(2006). Diesel Engine Indicated and Load Torque
Fig. 10. Corresponding controller parameters Kp and Ti Estimation Using Sliding Mode Observer. Journal of
for Fig. 8 in function of time. Automobile Engineering, 220(6), 95–103.

ACKNOWLEDGEMENTS

This research is carried out for the EMTechno project

(project ID: IWT150513) supported by VLAIO and Flan-
ders Make, the strategic research center for the manufac-
turing industry.

REFERENCES
Akehurst, S., Parker, D.A., and Schaaf, S. (2007). Dynamic
modeling of the Milner continuously variable transmis-
sion - The basic kinematics. Journal of Mechanical
Design, 129(11), 1170–1178.
Bonsen, B., Klaassen, T., Pulles, R., Simons, S., Steinbuch,
M., and Veenhuizen, P. (2005a). Performance optimi-
sation of the push-belt CVT by variator slip control.
International Journal of Vehicle Design, 39(3), 232. doi:
10.1504/IJVD.2005.008473.
Bonsen, B., Pulles, R.J., Simons, S.W.H., Steinbuch, M.,
and Veenhuizen, P.A. (2005b). Implementation of a slip
controlled CVT in a production vehicle. In Control
Applications, 2005. CCA 2005. Proceedings of 2005
IEEE Conference on, 1212–1217.
Carbone, G., Mangialardi, L., Bonsen, B., Tursi, C., and
Veenhuizen, P.A. (2007). CVT dynamics: Theory and
experiments. Mechanism and Machine Theory, 42(4),
409–428.
Carbone, G., Mangialardi, L., and Mantriota, G. (2001).
Fuel Consumption of a Mid Class Vehicle with Infinitely
Variable Transmission. In SAE Technical Papers, 2474–
2484.
Carbone, G., Mangialardi, L., and Mantriota, G. (2004). A
comparison of the performances of full and half toroidal
traction drives. Mechanism and Machine Theory, 39(9),
921–942.
Chen, X., Hang, P., Wang, W., and Li, Y. (2017). De-
sign and analysis of a novel wheel type continuously
variable transmission. Mechanism and Machine Theory,
107(4800), 13–26.
Kempermann, C. (2007). Hydrostatic variators for stepless
transmissions - Aspects of CVT technology from a
supplier’s viewpoint. In VDI-Berichte, 121–126.

106