Available online at www.sciencedirect.

com

ScienceDirect
IFAC PapersOnLine 55-24 (2022) 103–110
Jerk
Jerk Minimization for a Novel Engine Crank Mechanism using LQG
Jerk Minimization
Minimization for
for a
a Novel Engine
Engine Crank
Novel Control Crank Mechanism
Mechanism using
using LQG
LQG
Jerk Minimization for a Novel Engine
Engine Crank Mechanism using LQG
Control
Jerk Minimization for a Novel Control Crank Mechanism using LQG
Neeraj Shidore*,Control
Shidore*,Control
Madhusudan Raghavan*
Neeraj
Neeraj Shidore*, Madhusudan
Madhusudan Raghavan*
Raghavan*
Neeraj Shidore*, Madhusudan Raghavan*
*General Motors
Neeraj ResearchMadhusudan
Shidore*, and Development, Raghavan* Warren, USA
*General Motors Research and Development, Warren,
Warren, USA
(Tel:Neeraj
*General
*General Motors
Motors Shidore*,
586-985-9487; Research
Research Madhusudan
e-mail: and Development, Raghavan*
and neeraj.shidore@gm.com).
Development, Warren, USA
USA
(Tel:
*General
(Tel: 586-985-9487;
(Tel: 586-985-9487;
586-985-9487;
Motors Research e-mail:
e-mail:
e-mail: neeraj.shidore@gm.com).
neeraj.shidore@gm.com).
and neeraj.shidore@gm.com).
Development, Warren, USA
*General Motors Research and Development, Warren, USA
Abstract: The paper presents (Tel:the
586-985-9487;
developmente-mail: of LQGneeraj.shidore@gm.com).
(Linear Quadratic Gaussian) Control to minimize
Abstract: The paper (Tel:the
presents 586-985-9487;
development e-mail:
of neeraj.shidore@gm.com).
Abstract:
Abstract:
the jerk , during The
The paperpaper presents
the presents the development
actuationtheofdevelopment
a novel engine of LQG
of LQG
LQG (Linear
crank(Linear
(Linear
mechanism
Quadratic
Quadratic
Quadratic Gaussian)
,for Gaussian)
Gaussian)
a hybrid electric
Control
Control to
to minimize
to
Control vehicle. minimize
minimize The
the
the jerk
jerk
Abstract: ,
, during
during the
the actuation
actuation of
of a
a novel
novel engine
engine crank
crank mechanism
mechanism ,for
,for a
a hybrid
hybrid electric
electric vehicle.
vehicle. The
The
operation of the novel mechanism is explained in detailed. Open loop simulation results minimize
the jerk , The
during paper the presents
actuation theofdevelopment
a novel engine of LQG
crank (Linear
mechanism Quadratic ,for Gaussian)
a hybrid Control
electric to
vehicle.showThe an
Abstract:
operation Thethe paper presents the development of LQG (Linear Quadratic Gaussian) Control to minimize
the jerk , of
operation
operation
undesirable of
during
of jerkthe
the novel
novel
the
novel
, which
mechanism
mechanism
actuation of a novel
mechanism
is perceptible
is
isthe
tois
explained
explained
engine
driver, crank
explained in
in detailed.
in
during detailed.
mechanism
detailed. Open
Open,for
Open
the actuation
loop
loop
loop simulation
simulation
a hybrid
of this simulation
mechanism.electricresults
results
Avehicle.
results show
showloop
show
closed
an
an
The
an
the jerk , during
undesirable the actuation of a novel engine crank mechanism ,for a hybrid electricA Avehicle. The
undesirable
operation
undesirable
LQG controller ofjerk
jerk
jerktheis,,, which
which
novel
which
then
is
is perceptible
is perceptible
mechanism
perceptible
formulated
to
toisthe
to
to minimize the
the driver,
driver,
explained
driver,
the jerk
during
during
in and
during the
the
detailed.
the actuation
actuation
Open
actuation
provide a smooth
of
of this
loop
of this
this mechanism.
mechanism.
simulation start. results
mechanism.
engine The A closed
showloop
closed
closed
formulation loop
an
loop
operation
LQG
LQG controller
controller
undesirable of theis
is
jerkis, which novel
then
then mechanism
formulated
formulated
is perceptibleto
to is
minimize
minimize explained
to the driver, the
the in
jerk
jerk
during detailed.
and
and provide
provide Open
a
a smooth
smoothloop simulation
engine
engine start.
start. results
The
The show
formulation
formulation an
LQG
of thecontroller
plant model, then formulated
observer designtoisminimize
explained. theIncremental
jerk and the actuation
provide
control a smooth
is usedof this mechanism.
engine
to start. The
approximate Ajerkclosed
fromloop
formulation the
undesirable
of
of the
the plant
plant jerk , which
model,
model, observer
observeris perceptible
design
design is
is to the driver,
explained.
explained. during thecontrol
Incremental
Incremental actuation
control is
is of this
used
used to
to mechanism.
approximate
approximate Ajerk
jerkclosed
from
from loop
the
the
LQG
of thecontroller
output plant
equation. model, isFinally,
then formulated
observer design
closed looptoisminimize
explained.
simulation the jerk and
Incremental
results show provide
acontrol
smooth a smooth
isengine
used to engine start.
whileThe
approximate
crank formulation
jerk
minimizing from the
LQG
output controller
equation. isFinally,
then formulated
closed loopto minimize the jerk and provide a smooth engine start. The formulation
output
of equation.
the plant
output
undesirable equation. jerk. Finally,
model, observer
Finally, closed loopissimulation
loop
design
closed explained.results
simulation
simulation results
Incremental
results show
show aaacontrol
show smooth
smoothisengine
smooth used tocrank
engine
engine crank
crank while
while minimizing
while
approximate jerk from the
minimizing
minimizing the
the
of the plant model,
undesirable
undesirable jerk.
jerk. observer design is explained. Incremental control is used to approximate jerk from the
output
undesirable
Keywords:
Copyright equation.
© jerk.
Planetary Finally, closed
Starter,
The Authors.
2022Finally, LQG loop
This simulation
control,
is an tip
open in results
response,
access show
article P2 a smooth
hybrid
under the .
CC engine
BY-NC-ND crank while
license minimizing the
output
Keywords: equation.Planetary closed
Starter, LQG loop simulation
control, tip in results
response, showP2 ahybrid
smooth. engine crank while minimizing the
undesirable
Keywords:
Keywords: jerk.
Planetary
Planetary
(https://creativecommons.org Starter, LQG control,
Starter, /licenses/by-nc-nd/4.0/) tip in response,
LQG control, tip in response, P2 hybrid . P2 hybrid .
undesirable jerk.
Keywords: Planetary Starter, LQG control, tip in response, P2 hybrid .
Keywords: Planetary Starter, LQG control, tip in response, P2 hybrid .
As shown in figure 1, two starting mechanisms are
1. INTRODUCTION As
As shownshown used in
in figure
figure 1,
1, the two
two starting
starting mechanisms
mechanisms are
are
1.
1. INTRODUCTION
INTRODUCTION As shown
traditionally in figure
to crank 1, two starting
engine, eithermechanisms
a pinion starter are
1. INTRODUCTION traditionally
traditionally used
used to crank
to crank
crank the
the engine,
engine, either
either a pinion starter
A hybrid architecture, where-in the electric motor generator As
[4], shown
traditionally
or the BAS in (Belt
used figure
to 1, the
Alternator two starting
engine,
Starter) either
Machine aa pinion
pinion
mechanisms
[5]. starter
are
starter
Each of
A hybrid architecture, 1. INTRODUCTION
where-in the electric motor generator As or
[4], shown
or the BAS in (Beltfigure
(Belt 1, twoStarter)
Alternator starting
Starter) mechanisms
Machine [5]. Each are
of
A
unithybrid
A (MGU)architecture,
hybrid architecture,
is connected where-in
1. INTRODUCTION
where-in
to the driveline the
the electric
electric motor
motorthe
in between engine [4],
generator
generator the
traditionally
[4],
theseortwo
traditionally
BAS
thedevices
BAS used (Belt
used offer
offer Alternator
to crank
Alternator
to crank
the engine,
a trade-off Starter)
the engine,
between Machine
either
Machine
either
several [5].
a pinion
[5].
a pinion
Each
Each
factors: of
starter
of
the
starter
unit
unit (MGU)
(MGU) is
is connected
connected to
to the
the driveline
driveline in
in between
between the
the engine
engine these
these two
two devices
devices offer a
a trade-off
trade-off between
between several
several factors:
factors: the
the
A
andhybrid
unit (MGU)
the architecture,
is connected
transmission where-in
(Figure to the is the
1)driveline electric
referredin to motor
between
as a ‘P2’.the engine [4],
generator tip-inortwo
these the BAS (Belt
devices
drivability offer Alternator
metric, fuelStarter)
a trade-off between
economy Machine
several [5]. Each
factors:
benefits, of
the
cost,
A
and hybrid
the architecture,
transmission where-in
(Figure 1) is the electric
referred to motor
as generator [4], tip-in or the BAS (Beltmetric,
drivability Alternator
metric, fuel Starter)
economy Machine [5]. Each
benefits, of
cost,
and
and the
unit transmission
(MGU)
the is connected
transmission (Figure
(Figure to the 1) is
is referred
1)driveline
referredin to as aaa ‘P2’.
as
between
to ‘P2’.
the engine tip-in
‘P2’. these
tip-in
packaging,twodrivability
devices
ease offer
drivability ofmetric,
control, fuel
a trade-off
fuel economy
between
economyand
robustness, benefits,
several factors:
benefits,
cold weathercost,
the
cost,
unit (MGU) is connected to the driveline in between the engine packaging, these two devices
packaging, ease
ease offer
of a trade-off
control, between
robustness, several
and coldfactors:
weather the
and the transmission (Figure 1) is referred to as a ‘P2’. tip-in
packaging,
performance. drivabilityThe of
ease of control,
metric,
control,
authors have robustness,
fuel economy
robustness,
proposed aand
and
novel cold
benefits,
cold weather
weather
mechanism cost,
and the transmission (Figure 1) is referred to as a ‘P2’. tip-in drivability
performance.
performance. The
The metric,have
authors
authors havefuelproposed
economy
proposed a
a novel
novelbenefits,
mechanism
mechanism cost,
packaging,
performance.
to crank a P2 hybrid ease of
The authors control,
powertrain, robustness,
have proposed
as an alternative and cold
a noveltomechanismweather
the above
packaging,
to
to crank a P2 ease
hybrid of control,
powertrain, robustness,
as an and
alternative cold
to theweather
above
twocrank
performance.
to crank aa P2
P2 hybrid
mechanisms, The authors
hybrid powertrain,
powertrain,
called have as
the proposedan
an alternative
asplanetary a novel
alternative starterto the
[6].above
tomechanism
the aboveBy
performance.
two
two mechanisms,
mechanisms, The authors called have
the proposed
planetary a novel
starter mechanism
[6]. By
to
two crank a P2the
mechanisms,
eliminating needcalled
hybrid calledfor a the
powertrain, the asplanetary
starter an
oralternative
planetary a BAS starter
starterto the
machine,[6].
[6].above By
By
this
to crank
eliminating
eliminating a P2 the
thehybrid
need
need powertrain,
for
for a
aa the as
starter
starter an
or
or alternative
a BAS to the
machine, above
this
two
mechanism mechanisms,
eliminating the needcalled
eliminates forthe starter to aacarry
needplanetary
or BAS
BAS machine,
starter
machine,
two [6]. this
electrical By
this
two mechanisms,
mechanism
mechanism eliminates
eliminates called the theneed planetary
to carry starter
two [6]. By
electrical
eliminating
mechanism
machines onboard the need
eliminates (thefor P2the
theaMGU need
starter
need beingto
or
to the acarry
BAS
carryprimarytwo electrical
machine,
two electrical
machine). this
eliminating
machines the need
onboard (the for
P2 aMGU
MGU starterbeingor the a BAS machine, this
machines
This paperonboard
mechanism
machines eliminates
onboard
discusses (the
(the P2
P2the
a controlsMGU need being
being
approachto the carry
the primary
to primary
primary
improve
machine).
two machine).
electrical
machine).
the tip-in
mechanism
This
This paper
paper eliminates
discusses
discusses the
aaa controls
controls need
approach
approachto carry
to
to two
improve
improve electrical
the
the tip-in
machines
This
metric paper
with onboard
discusses (the
the planetary P2 MGU
controlsstarter being
approach the to primary
by elimination improve thetip-in
ofmachine).
the tip-in
jerk
machines
metric
metric
This with
with
paper onboard
the
the
discusses (the
planetary
planetary a P2 MGU
controlsstarter
starter being
by
by
approach the primary
elimination
elimination
to improve of
ofmachine).
the
the
the jerk
jerk
tip-in
metric
during enginewith the planetary
crank. Other aspectsstarter by likeelimination
packaging, cost, of thewhich jerk
This
during
during paper
engine
engine discusses
crank.
crank. a
Other
Othercontrols
aspects approach
like to improve
packaging, the
cost, tip-in
which
metric
during
are equally withimportant
engine the planetary
crank. whenaspects
Other starter
makingby
aspects like
like packaging,
final choicecost,
theelimination
packaging, of thewhich
of
cost, jerk
which
start
metric
are
are equally
equallywithimportant
the planetary
important when
when starter
making
making bythe
the elimination
final
final choice
choice of
of
of the
the jerk
start
start
during
are engine
equally crank.
important Other
when
mechanism, are beyond the scope of the paper. aspects
making like
the packaging,
final choice cost,
of the which
start
during
mechanism,
mechanism, engine are
are crank.
beyond Other the aspects
scope oflike
the packaging,
paper. cost, which
Figure 1. P2 hybrid vehicle architecture.
Figure are
mechanism, are beyond
equally important beyond whenthe scope
scope of
the making ofthethe
thefinalpaper.
paper.choice of the start
Figure 1.
Figure 1. P2
1. P2 hybrid
P2 hybrid vehicle
hybrid vehicle architecture.
vehicle architecture.
architecture. are equally
1.1 Literature important
review when making the final choice of the start
mechanism,
1.1 Literature are beyond
review the scope of the paper.
A Figure
significant
1. P2 challenge
hybrid vehicle the P2 hybrid architecture , is the 1.1
for architecture. 1.1 Literature
mechanism,
Literature arereview
beyond the scope of the paper.
review
A
A significant
Figure
A significant challenge
1. P2 challenge
significant hybrid vehicle for
for the P2
architecture.
the P2 hybrid
hybrid architecture
architecture ,
, is
is the
the Wu
1.1 et et al discuss
Literature the different powertrain architectures in [1] .
review
need for quickchallenge
reconnection for the P2 hybrid
of the architecture
disconnected engine, to is the Wu Wu et al discuss the different powertrain architectures in [1]
need
need for
for
A significant quick
quick reconnection
reconnection
challenge for the of
of the
the disconnected
disconnected
P2 hybrid engine
engine
architecture to
to the
the 1.1
Wu
The et al
al
tradeoffdiscuss
Literature
discuss the
the different
review
between different
drivability powertrain
and fuelarchitectures
powertrain economy forin
architectures [1] ...
[1]
ina mild
need for quick
drivetrain reconnection
, during a ‘tip-in’ of change
or the disconnected
of mind engine
event. A, to
is
tipthe in The The tradeoff
tradeoff between
between drivability and fuel economy for a mild
A significant
drivetrain
drivetrain
need
or change ,,, of
for quick
drivetrain
challenge
during
during aaa ‘tip-in’
reconnection
during
mind
‘tip-in’
‘tip-in’
for the
event occurs
or
or P2 hybrid
change
of change
or the during
change of
of architecture
mind
of mind
disconnected
mind
vehicle
event.
event.
engine
event.braking
A
A, is
A to tip
tip the
tipthe in
in Wu
The
hybrid
in hybrid
or
et al
P2discuss
tradeoff between
drivetrain isdrivability
the different
drivability
discussed and
powertrain
and
in [2]. fuel
fuel Juacheconomy
architectures
economy
et al [3]for
forin aa mild
[1] .
mild
discuss
Wu et al
hybrid P2
P2 discuss
drivetrain
drivetrainthe different
is
is discussed
discussed powertrain
in
in [2].
[2]. architectures
Juach
Juach et
et al
al [3]
[3] in [1] .
discuss
discuss
need
or
or for
change
change quick of reconnection
mind event of
occurs the disconnected
during vehicle engine
braking to the
or The
hybrid
a tradeoff
model P2 between
drivetrain
for drivability is drivability
discussed
and control and
in fuel
[2].
for aJuacheconomy
hybrid et al for
[3]
vehicle. a mild
discuss
Hao
drivetrain
or change
coasting , of
of
(engine mind
during
mind isaoff event
event and occurs
‘tip-in’ or change
occurs
disconnected)during vehicle
of mind
during and isevent.
vehicle braking
braking
initiated A tip byor ina aaThe
or tradeoff
model for between
drivability drivability
and control and forfuela economy
hybrid for
vehicle. a mild
Hao
drivetrain
coasting , of
during
(engine is aoff
‘tip-in’
and or change
disconnected) of mind
and isevent.
initiated A tip by aetmodel
in hybrid al [4]P2for
model for drivability
drivetrain
drivability and
and control
is discussed control in for
[2].
for aaJuach
hybrid
hybrid vehicle.
aset an
al [3]
vehicle. Hao
discuss
Hao
coasting
or change
coasting
sudden (engine
(engine
accelerationmind is
is off event
off
demandand disconnected)
and occurs driver.and
by theduring
disconnected) is
is initiated
vehicle
and
In such braking
initiated
a scenario, byoraaa et
by hybrid
et al
al [4]
[4]
discuss
P2for drivetrain
discuss
discuss
a high-performance
is discussed
aaa high-performance
high-performance in for
[2].
starter
Juach
starter
starter as
aset an
alternative
al [3]
an discuss
alternative
alternative
or
sudden
suddenchange of
acceleration
accelerationmind event
demand
demand occurs
by
by the during
driver. vehicle
In such braking
a scenario, or a
et model
al
option [4]to the drivability
discussconcept and
high-performance
proposed controlin this a hybrid
starter
paper. as
Somevehicle.
an alternative
details Haoon
coasting
sudden
the engine (engine
acceleration
must be is off demand
cranked by the
and disconnected)
and the driver.
driver.and
connected In
to such
In is
such aa scenario,
theinitiated scenario,
driveline byina option aoption
modeltoto for
the
the drivability
concept
concept and control
proposed
proposed in
in for paper.
this
this a hybrid
paper. Some
Somevehicle.
details
details Haoon
on
coasting
the
the engine (engine
must is
be off
crankedand disconnected)
and connected andto is
the initiated
driveline by in a et
theal
option [4]to
planetarydiscuss
the concept
startera high-performance
canproposed
be foundin this
in starter
paper.
the as
patentSomean
by alternative
details
Raghavan on
the engine
sudden engine
shortest must
acceleration
must be
be cranked
possible time, and
demand
cranked and
to connected
by fulfill
the driver.
connected to
In
the driver the
to such driveline
a scenario,
the acceleration
driveline in
in the et
thealplanetary
[4] discuss
planetary starter
startera high-performance
can
can be
be found
found in
in starter
the
the as an
patent
patent by
by alternative
Raghavan
Raghavan
sudden
the
the acceleration
shortest
shortest possible
possible demand
time,
time, by fulfill
to
to the driver.
fulfill the
the In suchacceleration
driver
driver a scenario, and
acceleration option
the to the[6].
planetary
Shidore concept
starter canproposed
be foundin this
in paper.
the patentSome by details on
Raghavan
the
demand.engine
shortest
This mustis be
possible
even cranked
time,
more importantand connected
to fulfill the
when tothe
driverthemotor
driveline
accelerationpower in and option to the[6]. concept proposed in this paper. Some details on
the
demand.
demand.engineThis
This mustis
is be
even cranked
more and connected
important when tothe
themotor
driveline power in and the Shidore
Shidore
Shidore [6].
and planetary starter can be found in the patent by Raghavan
[6].
the
demand.shortest
is a small This
fractionis even
possible
even
of the more
time,
more engine important
to fulfill
powerthe
important when
anddriver
when the
the
cannot motor
acceleration
motor
support power
power the the planetary starter can be found in the patent by Raghavan
the
is shortest possible time, to fulfill the driver acceleration and Shidore [6].
is aaa small
is
demand.small
smallThis
acceleration
fraction
fraction
fractionis even
power
of
of
of the the
more
the engine
engine
engine
requirements
power
power
important
power and
and
on when
and cannot
its cannot
the motor
cannot
own.
support
support
support
The tip the
power the
the
in 1.2 and Shidore [6].
Paper overview
demand.
acceleration
acceleration This is even
power
power more
requirements
requirements important on
on when
its
its the motor
own.
own. The
The power
tip
tip in
in 1.2 Paper overview
is a small
acceleration fraction
maneuver is therefore of the
power requirements engine
a critical drive power and
on qualitycannot
its own. metric Thefortipa the
support in 1.2
P2 1.2 Paper
Paper overview
overview
is a small fraction
maneuver is thereforeof the engine power
aaa critical
critical drive and cannot
quality metric support
for the
maneuver
acceleration
maneuver
[1], [2]. This is
is therefore
power
tip-in requirements
therefore criticaldrive
related drive
drive quality
onquality
its own.
quality metric
metric
metric Thehas fortipaaatwo
for P2
in The
P2
P2 flow of
1.2 Paper
The flow
the paper is as follows: Section 2 gives an
overview
of the paper is as follows: Section 22 gives an
acceleration
[1],
[1], [2].
[2]. This
This power
tip-in
tip-in requirements
related
related drive
drive on its
quality
quality own.
metric
metric The has
has tip two
twoin The flow
1.2 Paper
The flow
overview of the
overview
of
of paper
paper is
the planetary
the as
as follows:
is starter follows: Section
Sectionwhich
architecture, 2 gives
gives is thean
an
maneuver
[1], [2]. is
This therefore
tip-in a critical
related drive
drive quality
quality
components to it: the time to acceleration (from the initial overview metric
metric for
has a P2
two overview of the planetary starter architecture, which is the
maneuver
components
components is therefore
to it: the a critical
time to drive quality
acceleration metric
(from thefor a
initial P2 The flow
overview
novel of
of
of
mechanism the
the
the planetary
paper
planetary
for is
engine starter
as
starter
cranking architecture,
follows: Section
architecture,
that will which
2
which
be gives is
is
controlled the
an
the
[1], [2]. Thisto
components
accelerator to
pedalit:
it: the
tip-in the time
time to
related
command), acceleration
todrive the (from
andquality
acceleration metricthe
(from
quality has
the ofinitial
two
initialthe novel The flow
novel of the for
mechanism paper
for is ascranking
engine follows:that
cranking thatSection
will be 2 controlled
gives an
controlled
[1], [2].
accelerator
accelerator This tip-in
pedal
pedal related
command),
command), drive and
and quality
the metric
quality hasof two
the overview
novel
by the mechanism
of
mechanism
LQG the
control. planetary
for engine
engine
This starter
cranking
section architecture,
also that will
will
provides be
which
be is
controlled
detail on the
components
acceleration, to
accelerator pedal
whichit: the time the
command),
includes and the
to acceleration
disturbance quality
(from
the during
qualitythe theofofinitial
engine the
the by overview
by the LQG of control.
the planetary This starter also
section architecture,
also provides which
detail is the
on
components
acceleration,
acceleration,
accelerator
acceleration,
cranking event
to
which
whichit: the
[3]. includes
pedal
which
time the
includes
command),
includes
to acceleration
the disturbance
the disturbance
and the during
disturbance
(from
during
qualitythe
during
the engine
the
theof
initial
engine
engine the vehicle by the
novel LQG
drive control.
themechanism
LQG control.
scenario for Thisengine
This
(changesection
cranking
section also that
of mind provides
provides
engine detail
willcrank)
be
detail on
on the
controlled
and the
accelerator pedal command), and the quality of the vehicle novel
vehicle mechanism
drive
drive scenario
scenario for engine
(change
(change cranking
of
of mind
mind that
engine
enginewill be
crank)
crank) controlled
and
and the
the
cranking
cranking
acceleration, event
event [3].
[3].
which by the
vehicle LQG
drive control.
propulsionscenario This
system (changesection
model of also
mind
that provides
engine
will be detail
crank)
used to on
and
accessthe
cranking event [3]. includes the disturbance during the engine vehicle by the LQG control.system
propulsion This section
system model also will
that provides
will be detail
used to on
accessthe
acceleration, which includes the disturbance during the engine vehicle vehicle
the propulsion
drive
propulsion
effectiveness scenario
of system
the (change
LQG model
model of
Control. that
mind
that engine
will
Open be
be
loop used
crank)
used to
to
results access
and
access
showthe
cranking event [3]. vehicle
the drive
effectiveness scenario
of the (change
LQG of
Control.mind Openengine loop crank)
results and showthe
cranking event [3]. the
the effectiveness
vehicle
needpropulsion
effectiveness
of closed of the
loopLQG
of system
the LQG Control.
model
Control.
control, thatOpen
during will
Open loop
thebe
loop results
used show
to access
results
actuation show
of the
vehicle
the
the need
need propulsion
of closed system
loop model during
control, that will
during thebeactuation
used to access
actuation of the
need of of closed
the effectiveness closed loop
of the
loopLQG control,
Control.
control, during Open the
theloop resultsof
actuation the
ofshow
the
the effectiveness of the LQG Control. Open loop results show
2405-8963 Copyright © 2022 The Authors. This is an open access article under the CC BY-NC-ND license. the need of closed loop control, during the actuation of the
Peer review under responsibility of International Federation of Automatic Control.
the need of closed loop control, during the actuation of the
10.1016/j.ifacol.2022.10.269
104 Neeraj Shidore et al. / IFAC PapersOnLine 55-24 (2022) 103–110

novel crank mechanism. Section 3 lays out the development of since the engine is disconnected from the drivetrain and is
the control algorithm in detail. Section 4 shows simulation assumed to be at zero speed. All speeds are shown on the
results that compare the results of engine crank with and primary Y axis. At about 43.03 seconds, there is a need to
without the LQG controller. Section 5 is the conclusion. crank the engine, due to a sudden tip in by the driver. As a
result, brake B is engaged (pink and pink Y axis). This causes
the carrier (engine speed) to increase from zero. As soon as the
2. PLANETARY STARTER
engine speed has reached firing speed , the brake is dis-
2.1 Architecture engaged, and the engine is able to increase its own speed using
spark and fuel (around 43.4 seconds). There is a need to bring
Figure 2 shows the architecture of the planetary starter. A
the engine up to firing speed before throttle and spark are
simple planetary gear is used along with a clutch and brake (or
actuated, and therefore , the brake actuation precedes throttle.
clutch) to realize the planetary starter.
The engine is connected to the carrier node, while the P2 MGU
is connected to the sun gear. The other end of the P2 MGU
shaft is connected to the torque converter pump. The Ring
Gear is free-wheeling but can be forced to be at zero speed by
using brake (B), which is between the ring gear and a non -
moving element like the transmission housing. The clutch (C
), when closed, forces the ring gear and the carrier to turn at
the same speed, thus forcing the planetary gear ratios to reduce
to unity. Schematically, the clutch C can be between any two
elements of the planetary. Table 1 shows the different modes
possible with the clutch (C) and the brake (B) being in the
closed or open position.
Figure 3. Engine crank actuation with planetary starter

The engine speed is then controlled to match the P2 MGU

speed (around 43.8 seconds). When the carrier (engine) and
the sun (P2 MGU) speeds are close, the clutch C is closed
(cyan, command 0 to 1), which causes all elements of the
planetary to spin at the same speed, i.e. the engine and the P2
motor are now at the same speed, as should be the case in a P2
where there is no gear ratio between the engine and the P2
motor. Once clutch C is closed, engine can provide traction
torque to support the acceleration demand.
2.2 P2 vehicle model for planetary drive quality evaluation
Figure 2. Planetary Starter Architecture To emulate the plant, the entire driveline was modelled in
AMESIM.To be able to view the details of the powertrain
Table 1. Planetary States of Operation model, the figure representing the powertrain model is
presented as a full page in Appendix A at the end of the paper.
Vehicle Brake (B) Clutch ( C )
This model also acts as the plant for the LQG [7] controller
State
described in the next section. The planetary gear is shown in
EV Driving Open Open blue. All significant elements of the drivetrain have been
modeled in red. Table 2 lists the parameter variable names,
Engine Closed Open and description. Nominal values for a class 2B truck have been
Crank chosen for the simulation but are not stated in the paper, for
brevity. In table 2, the ‘transformer’ is a notation commonly
Hybrid Open Closed used in bond graphs, to suggest the ratio of flow and effort
Driving elements on either bond of the transformer. In the current
context, these are used to express the gear ratios associated
with the planetary (TF1 and TF2 ) and 2nd gear (TF3)
During EV driving, both brake B and clutch C are open ,
respectively.
effectively disengaging the engine from the drive train. The
engine crank and the hybrid driving vehicle states are 2.3 Open loop simulation of the P2 drivetrain for a change of
explained in detail below. Figure 3 shows the operation of the mind engine crank
planetary starter during an engine crank event. At 43 seconds,
the vehicle is in EV mode , and the P2 motor is at around 1650 A vehicle drive scenario, which would provide the máximum
RPM (blue). This is also the speed of the sun gear. The ring jerk to the driveline during an engine crank event with the
gear is ‘free-wheeling’ and based on the planetary gear ratio, planetary starter was chosen. This was with the vehicle at low
is spinning at 500 RPM. The carrier (engine) speed is zero speeds (around 7 to 8 meters per second) , in 2nd gear and with
 Neeraj Shidore et al. / IFAC PapersOnLine 55-24 (2022) 103–110 105

the torque converter clutch closed. Figure 4 shows the vehicle
acceleration (red) along with the brake torque (blue), engine
throttle (black) and clutch command (cyan). The brake torque
and the throttle commands are in the form of impulse inputs
(very short time spans).The initial brake torque causes a
sudden negative vehicle acceleration (around 39.3 seconds),
which can be sensed by the driver and will negatively impact
the drive quality rating. The impulse nature of the torque
commands also causes a sustained undesirable oscillation on
the driveline. The initial jerk and sustained oscillations have a
negative impact on vehicle drive quality and need to be
reduced/eliminated.
state. Following are the six states of the system: engine speed,
motor speed, turbine speed, wheel speed, angle between motor
and turbine inertias , i.e. due to the spring with stiffness Kd,
angle between turbine and wheel inertias, due to spring with
stiffness K. All speeds are measured, while the angles are
estimated using a Kalman filter. There are two modes of
operation, when clutch C is open, the plant model has 6 states
(stated above). When clutch C is closed engine and motor
speeds are the same, so that the system collapses to a 5 th order
system. The control objective is to control the motor torque to
reject disturbances – throttle, brake, clutch, and vehicle load
torque.
Equation (1) below shows the plant model in standard state
space format .
̇ = 𝐴𝐴𝑂𝑂 𝑋𝑋(𝑡𝑡) + 𝐵𝐵𝑂𝑂 𝑈𝑈(𝑡𝑡) + 𝐵𝐵𝑂𝑂𝑂𝑂 𝑊𝑊(𝑡𝑡)
𝑋𝑋(𝑡𝑡) (1)
𝑦𝑦(𝑡𝑡) = 𝐶𝐶𝑜𝑜 𝑋𝑋(𝑡𝑡) + 𝑣𝑣(𝑡𝑡)

Figure 4. Actuation torques and resultant acceleration (open loop)
3. LQG CONTROL DEVELOPMENT
3.1 Plant Model Development
The dynamic equations for the plant (vehicle driveline with the
planetary gear) are derived using Bond Graphs [8]. The
variable names in the bond graph representation in the figure
5 below are the same as Table 2. The use of bond graphs is not
a necessary step in the state space formulation.
The subscript ‘o’ is for clutch C being open. 𝑋𝑋(𝑡𝑡)is the state
vector, 𝑣𝑣(𝑡𝑡) is the vector of sensor noise (modeled in the plant
as a zero-order hold and added gaussian noise), 𝑈𝑈(𝑡𝑡) is the
control input vector (motor torque) and 𝑊𝑊(𝑡𝑡) is a vector of
disturbance inputs (clutch torque, throttle, brake, vehicle load).
In continuous time domain, wheel acceleration (𝑥𝑥4̇ ) can be
considered an output, which is directly linked to vehicle
acceleration by a constant ratio. Using the bond graph, the
state, input, and output matrices in equation (1) are as below
in equations (2) through (5). When the clutch is closed, the
system collapses to a 5th order. Equation (6) below is the state
equation for the 5th order system. Other matrices, when clutch
is closed, have not been stated in the paper for brevity.
Table 2. Vehicle drivetrain Parameters for Plant Model
Variable Parameter Explanation Unit/Formula
Name
Ns # Teeth on Sun Gear

Nr # Teeth on Ring Gear

I_f Final Drive Gear Ratio

I_t Second Gear Ratio

K Axle ,Prop Shaft Stiffness Nm/deg

C Axle ,Prop Shaft Damping Nm/deg/s

Je Equivalent Engine Inertia kg m^2

rw Wheel Radius m
Figure 6. Bond Graph Representation of the Powertrain Model
Bv Vehicle viscous damping 0.5*rw^2
To have integral causality for element ‘Je’, i.e. engine speed, Nm/rad/s
the inertia of the brake element/ring gear has not been
considered. Inserting an elastic element between the brake and M Vehicle Mass kg
the ring gear would enable inclusion of brake inertia but would rho Air Density kg/m^3
have increased the order of the system. The inertia of the
carrier and sun gears have been neglected in this exercise, but Mu_roll Rolling Resistance Coeff
they can easily be combined with the engine and the motor
inertia respectively. Integral causality for ‘Je’ in the bond Cw Drag Coefficient
graph is important since it allows us to use engine speed as a
106 Neeraj Shidore et al. / IFAC PapersOnLine 55-24 (2022) 103–110

Av Frontal Area m^2 −𝑓𝑓𝑟𝑟− 𝐶𝐶𝐶𝐶 𝐶𝐶𝐶𝐶

0
−𝐾𝐾𝐾𝐾
0
𝐽𝐽𝐽𝐽𝐽𝐽𝐽𝐽 𝐽𝐽𝐽𝐽𝐽𝐽𝐽𝐽 𝐽𝐽𝐽𝐽𝐽𝐽𝐽𝐽
Jw Wheel Inertia kgm^2 𝐶𝐶𝐶𝐶 −𝐶𝐶𝐶𝐶 𝜇𝜇3 2 𝐶𝐶 𝜇𝜇3 𝐶𝐶 𝐾𝐾𝐾𝐾 −𝜇𝜇3 𝐾𝐾
−
𝐽𝐽𝐽𝐽 𝐽𝐽𝐽𝐽 𝐽𝐽𝐽𝐽 𝐽𝐽𝐽𝐽 𝐽𝐽𝐽𝐽 𝐽𝐽𝐽𝐽
𝐴𝐴𝑐𝑐 = 𝜇𝜇3 −𝐶𝐶 𝐵𝐵𝐵𝐵 𝐾𝐾 (6)
Jv Equivalent Vehicle Inertia 2*Jw+M*rw^2 0 𝐶𝐶 − 0
𝐽𝐽𝐽𝐽 𝐽𝐽𝐽𝐽 𝐽𝐽𝐽𝐽 𝐽𝐽𝐽𝐽
Kd TCC Damper stiffness Nm/deg 1 −1 0 0 0
[ 0 𝜇𝜇3 −1 0 0 ]
Cd TCC Damper Damping Nm/deg/sec

Jm Motor Inertia kgm^2 𝑊𝑊 = 𝑇𝑇𝐿𝐿 . 𝑋𝑋 (7)

Jp TC Pump Inertia kgm^2
𝐴𝐴𝑜𝑜,𝑠𝑠 = 𝑇𝑇𝐿𝐿 . 𝐴𝐴𝑜𝑜. 𝑇𝑇𝑅𝑅 (8)
Jtu TC Turbine Inertia kgm^2
𝐵𝐵𝑜𝑜,𝑠𝑠 = 𝑇𝑇𝐿𝐿 . 𝐵𝐵𝑜𝑜 (9)
Jtr Transmission Inertia (2nd gear, kgm^2
reflected to input) 𝐶𝐶𝑜𝑜,𝑠𝑠 = 𝐶𝐶𝑜𝑜 . 𝑇𝑇𝑅𝑅 (10)
Jpr Prop shaft inertia reflected to kgm^2
trans input in 2nd gear
To design the Kalman filter gains, matrices 𝐶𝐶𝑜𝑜𝑜𝑜 and 𝐶𝐶𝑐𝑐𝑐𝑐 are
µ1 Ratio for transformer TF1 (in -Nr/(Ns+Nr) defined as in (11) and (12). Matrix 𝐶𝐶𝑜𝑜𝑜𝑜 is the output matrix for
bond graph notation) the Kalman filter when the clutch C is open, while 𝐶𝐶𝑐𝑐𝑐𝑐 is the
µ2 Ratio for transformer TF2 (in (Ns+Nr)/Ns
output matrix for the Kalman filter when the clutch is closed.
bond graph notation)
1 0 0 0 0 0
𝐶𝐶𝑜𝑜𝑜𝑜 = [0 1 0 0 0 0] (11)
3.2 Scaling and Discretization 0 0 0 1 0 0
Scaling has been used to formulate a numerically well- 𝐶𝐶𝑐𝑐𝑐𝑐 = [
1 0 0 0 0
] (12)
conditioned problem, given the large difference in the values 0 0 1 0 0
of some states (speed versus angle) . MATLAB ‘Prescale' Above state space systems are then discertized (for the
command creates diagonal scaling matrices 𝑇𝑇𝐿𝐿 and 𝑇𝑇𝑅𝑅 . The feedback and Kalman filter gains).
transformation maps state vector 𝑋𝑋 to 𝑊𝑊 , and the state space
matrices are also suitably transformed as shown in equations 3.3 Controllability and Observability
(7) to (10). The suffix ‘s’ in the matrix names denotes a scaled For the state space system where the clutch is open, i.e. the
matrix. The same operation is performed on the 5th order
state space system (𝐴𝐴𝑜𝑜,𝑠𝑠 , 𝐵𝐵𝑜𝑜,𝑠𝑠 , 𝐶𝐶𝑜𝑜,𝑠𝑠 ) , or its discrete form, the
system. The state space systems (𝐴𝐴𝑜𝑜,𝑠𝑠 , 𝐵𝐵𝑜𝑜,𝑠𝑠 , 𝐶𝐶𝑜𝑜,𝑠𝑠 ) and
controllability matrix is not full rank. This is intuitively
(𝐴𝐴𝑐𝑐,𝑠𝑠 , 𝐵𝐵𝑐𝑐,𝑠𝑠 , 𝐶𝐶𝑐𝑐,𝑠𝑠 ), are used to design the state feedback gains obvious since, when the clutch C is open, engine speed state
(LQR). (𝑥𝑥1) cannot be controlled by the control input Tm, i.e. motor
torque. But, the observability matrix is still full rank. This is
−𝑓𝑓 because the engine speed state can still be deduced (it is zero,
0 0 0 0 0
𝐽𝐽𝐽𝐽 since the engine is disconnected from the drivetrain).
0
−𝐶𝐶𝐶𝐶 𝐶𝐶𝐶𝐶
0
−𝐾𝐾𝐾𝐾
0 Controllability is important during design of the state feedback
𝐽𝐽𝐽𝐽𝐽𝐽 𝐽𝐽𝐽𝐽𝐽𝐽 𝐽𝐽𝐽𝐽𝐽𝐽
𝐶𝐶𝐶𝐶 −𝐶𝐶𝐶𝐶 𝜇𝜇3 2 𝐶𝐶 𝜇𝜇3 𝐶𝐶 𝐾𝐾𝐾𝐾 −𝜇𝜇3 𝐾𝐾
gains, while observability is important for designing the
𝐴𝐴𝑜𝑜 = 0
𝐽𝐽𝐽𝐽 𝐽𝐽𝐽𝐽
−
𝐽𝐽𝐽𝐽 𝐽𝐽𝐽𝐽 𝐽𝐽𝐽𝐽 𝐽𝐽𝐽𝐽 (2) observer, i.e. Kalman filter. Therefore, when designing the
0 0 𝐶𝐶
𝜇𝜇3 −𝐶𝐶
−
𝐵𝐵𝐵𝐵
0
𝐾𝐾 LQR gains, a reduced order model of (𝐴𝐴𝑜𝑜,𝑠𝑠 , 𝐵𝐵𝑜𝑜,𝑠𝑠 , 𝐶𝐶𝑜𝑜,𝑠𝑠 ) is used.
𝐽𝐽𝐽𝐽 𝐽𝐽𝐽𝐽 𝐽𝐽𝐽𝐽 𝐽𝐽𝐽𝐽
For designing the observer, the full order system is maintained.
0 1 −1 0 0 0
[0 0 𝜇𝜇3 −1 0 0 ]
𝑇𝑇 3.4 Kalman filter/Observer Design
1
𝐵𝐵𝐵𝐵 = [0 𝐽𝐽𝐽𝐽𝐽𝐽
0 0 0 0] (3)
−1 1 As stated above, two of the states (the angles) must be
0 0
𝐽𝐽𝐽𝐽 𝜇𝜇1. 𝐽𝐽𝐽𝐽 estimated using a Kalman filter, since these are not
−1 −1
0 0 measurable. Steady state Kalman filter gains are calculated for
𝐽𝐽𝐽𝐽𝐽𝐽.𝜇𝜇1 .𝜇𝜇2 𝐽𝐽𝐽𝐽𝐽𝐽
𝐵𝐵𝐵𝐵𝐵𝐵 = 0 0 0 0 (4) both modes. Process covariance matrix (Equation 13) and
0
−1
0 0
sensor noise covariance matrix (Equation 14) are defined as
𝐽𝐽𝐽𝐽 follows:
0 0 0 0
[0 0 0 0 ] 𝑠𝑠𝑠𝑠11 0 0 0
𝐶𝐶𝜇𝜇3 −𝐶𝐶 𝐵𝐵𝐵𝐵 𝐾𝐾 0 𝑠𝑠𝑠𝑠22 0 0
𝐶𝐶𝐶𝐶 = [0 0 − 0 ] (5) 𝑆𝑆𝑆𝑆 = 𝜌𝜌 [ ] (13)
𝐽𝐽𝐽𝐽 𝐽𝐽𝐽𝐽 𝐽𝐽𝐽𝐽 𝐽𝐽𝐽𝐽 0 0 𝑠𝑠𝑠𝑠33 0
0 0 0 𝑠𝑠𝑠𝑠44
 Neeraj Shidore et al. / IFAC PapersOnLine 55-24 (2022) 103–110 107

𝑠𝑠𝑠𝑠11 0 0 scaled, discretized versión of the output matrix 𝐶𝐶𝐶𝐶 , when the
𝑆𝑆𝑆𝑆 = 𝜌𝜌𝑣𝑣 [ 0 𝑠𝑠𝑠𝑠22 0 ] (14) clutch is open. Change in wheel acceleration over a time step
0 0 𝑠𝑠𝑠𝑠33 is given by equation 17.
The noise values were based on engineering judgment of prior Δ𝑤𝑤4 [𝑘𝑘] = 𝐶𝐶𝑜𝑜,𝑠𝑠,𝑑𝑑 . Δ𝑊𝑊[𝑘𝑘] (17)
work. No specific measurements were undertaken for this
work. Where 𝜌𝜌 and 𝜌𝜌𝑣𝑣 tuning scalars. Matlab command Where W is the state vector after the scaling.
‘kalman’ is used to obtain Kalman filter gains for the observer
for the two modes. State estimation is implemented as a two- The jerk can therefore be approximated as
step process as shown with equations 15 and 16 below:
Step 1: Prediction: Δ𝑊𝑊[𝑘𝑘] Δ𝑊𝑊[𝑘𝑘]
𝐽𝐽[𝑘𝑘] = 𝐶𝐶𝑜𝑜,𝑠𝑠,𝑑𝑑 .
Δ𝑇𝑇
= 𝐶𝐶𝑜𝑜,𝑠𝑠,𝑑𝑑 .
Ts
(18)
𝑊𝑊 − [𝑘𝑘] = (𝛼𝛼. 𝐴𝐴𝑜𝑜,𝑠𝑠,𝑑𝑑 + (1 − 𝛼𝛼). 𝐴𝐴𝑐𝑐,𝑠𝑠,𝑑𝑑 ) ∗ 𝑊𝑊 + [𝑘𝑘 − 1]
+ (𝛼𝛼. 𝐵𝐵𝑜𝑜,𝑠𝑠,𝑑𝑑 + (1 − 𝛼𝛼). 𝐵𝐵𝑐𝑐,𝑠𝑠,𝑑𝑑 ). 𝑈𝑈[𝑘𝑘] (15) Let 𝑄𝑄𝑜𝑜,𝑠𝑠,𝑑𝑑 be the state weighting matrix in the cost function,
defined as
Where
𝐶𝐶𝑜𝑜,𝑠𝑠,𝑑𝑑 𝑇𝑇 𝐶𝐶𝑜𝑜,𝑠𝑠,𝑑𝑑
𝑊𝑊 − [𝑘𝑘] is the predicted state from the previous state 𝑊𝑊 + [𝑘𝑘 − 1] 𝑄𝑄𝑜𝑜,𝑠𝑠,𝑑𝑑 = 𝜚𝜚. ( ) ( ) (19)
𝑇𝑇𝑇𝑇 𝑇𝑇𝑇𝑇

𝑈𝑈[𝑘𝑘] is the input from the current time step
𝐴𝐴𝑜𝑜,𝑠𝑠,𝑑𝑑 is the state matrix for the open clutch case (suffix ‘o’)
which is scaled and discretized (suffices ‘s’ and ‘d’
respectively).
𝐵𝐵𝑜𝑜,𝑠𝑠,𝑑𝑑 is the input matrix for the clutch open case , similarly
scaled and discretized.
𝛼𝛼 is a variable to switch from open clutch case to close clutch
case, i.e. when clutch is open, 𝛼𝛼 = 1, when clutch is closed,
𝛼𝛼 = 0. Decision on the value of 𝛼𝛼 is based on the speed
difference between engine and motor speed.
Then, the cost function to minimize jerk can be written in
quadratic form as
1
𝔍𝔍 = ∑ Δ𝑊𝑊 𝑇𝑇 𝑄𝑄𝑜𝑜,𝑠𝑠,𝑑𝑑 Δ𝑊𝑊 + Δ𝑈𝑈 𝑇𝑇 𝑅𝑅Δ𝑈𝑈 (20)
2
Where 𝑅𝑅 is a suitable weightage on incremental input (Motor
Torque) Δ𝑈𝑈. 𝜚𝜚 is a tuning coefficient. Thus, if the LQR state
feedback control is implemented in the incremental form [9]
(i.e. Δ𝑊𝑊 and Δ𝑈𝑈) instead of 𝑊𝑊 𝑎𝑎𝑎𝑎𝑎𝑎 𝑈𝑈, then the Jerk can be
very easily incorporated in the cost function, and the feedback
control law is
Δ𝑈𝑈𝑓𝑓𝑓𝑓 = (𝛼𝛼. 𝐾𝐾𝑜𝑜,𝑠𝑠,𝑑𝑑 + (1 − 𝛼𝛼). 𝐾𝐾𝑐𝑐 ,𝑠𝑠 ,𝑑𝑑 ). Δ𝑊𝑊 (21)

Step 2: Correction: Where

The state vector prediction in Step 1 of the Kalman filter
(equation 15) is now corrected with the most recent
measurement (vector ‘meas’) in equation 16, and the Kalman
filter gains 𝐿𝐿𝑜𝑜 , 𝐿𝐿𝑐𝑐 . The vector ‘meas’, would be a column
vector with engine, motor and wheel speed measurement
samples.
𝑊𝑊 + [𝑘𝑘] = 𝑊𝑊 − [𝑘𝑘] + (𝛼𝛼. 𝐿𝐿𝑜𝑜 + (1 − 𝛼𝛼). 𝐿𝐿𝑐𝑐 ). (meas − (𝛼𝛼. 𝐶𝐶𝑜𝑜𝑜𝑜,𝑠𝑠,𝑑𝑑 +
(1 − 𝛼𝛼) . 𝐶𝐶𝑐𝑐𝑐𝑐,𝑠𝑠,𝑑𝑑 ). 𝑊𝑊 − [𝑘𝑘]) (16)
Where
𝐾𝐾𝑜𝑜,𝑠𝑠,𝑑𝑑 are the state feedback gains if clutch C is open, 𝛼𝛼 = 1
𝐾𝐾𝑐𝑐,𝑠𝑠,𝑑𝑑 are the state feedback gains if clutch C is closed, 𝛼𝛼 = 0.
The LQR gains are calculated using the Matlab function ‘dlqr’.
Feedforward control has been discussed in literature as an
effective method to reject disturbance [10] if it can be modeled
with some accuracy. Brake torque (Tb) was considered using
feedforward control. The incremental control law (22) is
modified to add the feedforward component

𝑊𝑊 + [𝑘𝑘] is the state after correction of 𝑊𝑊 − [𝑘𝑘] ̂

Δ𝑈𝑈 = −(𝛼𝛼. 𝐾𝐾𝑜𝑜,𝑠𝑠,𝑑𝑑 + (1 − 𝛼𝛼). 𝐾𝐾𝑐𝑐 ,𝑠𝑠 ,𝑑𝑑 ). Δ𝑊𝑊 − 𝐾𝐾𝑓𝑓𝑓𝑓 . Δ𝑇𝑇𝑇𝑇 (22)

𝐶𝐶𝑜𝑜𝑜𝑜,𝑠𝑠,𝑑𝑑 is the output matrix (Equation 11) scaled and Where

discretized, for the mode when clutch C is open.
𝐶𝐶𝑐𝑐𝑐𝑐,𝑠𝑠,𝑑𝑑 is the output matrix (Equation 12) scaled and 𝐾𝐾𝑓𝑓𝑓𝑓 is the feedforward gain, calculated in (23) below.
discretized when the clutch C is closed. State vector W can be −1
−1
converted back into the X domain by inverting the scaling 𝐾𝐾𝑓𝑓𝑓𝑓 = − (𝐶𝐶𝑜𝑜,𝑠𝑠,𝑑𝑑 . (𝐴𝐴 − 𝐵𝐵𝑜𝑜,𝑠𝑠,𝑑𝑑 ∗ 𝐾𝐾𝑜𝑜,𝑠𝑠,𝑑𝑑 ) . 𝐵𝐵𝑜𝑜,𝑠𝑠,𝑑𝑑 ) . 𝐶𝐶𝑜𝑜,𝑠𝑠,𝑑𝑑 . (𝐴𝐴 −
operation performed in equation 7. −1
𝐵𝐵𝑜𝑜,𝑠𝑠,𝑑𝑑 . 𝐾𝐾𝑜𝑜,𝑠𝑠,𝑑𝑑 ) . 𝐵𝐵𝑜𝑜𝑜𝑜,𝑠𝑠,𝑑𝑑 (23)
3.5 Cost Function, Incremental Control for LQR and
feedforward control The Kalman filter and the feedback, feedforward code was
embedded in Simulink as a m-function and cosimulated with
The objective of the state feedback control is to minimize the the AMESIM plant model.
jerk, i.e. the rate of change of acceleration, during engine crank
and its connection to the drive train. Wheel speed is the fourth
state of the system. Matrix 𝐶𝐶𝑜𝑜,𝑠𝑠,𝑑𝑑 in equation 17 represents the
108 Neeraj Shidore et al. / IFAC PapersOnLine 55-24 (2022) 103–110

4. SIMULATION RESULTS

The objective of closed loop control is to eliminate the initial

jerk and subsequent sustained oscillations due to actuation of
the brake, the throttle, and the clutch . In addition, since not
all states are measurable, a Kalman filter was developed to
estimate the states associated with the two springs in the
system. While elimination of jerk is the objective of the closed
loop control, acceleration is a physical measurement that is
most used to study measure and quantify disturbances and has
been plotted in the results section. Indeed, visual comparison
of two acceleration signals can provide a good qualitative idea
of the difference in jerk .
4.1 Kalman filter operation
Figure 7. Prediction action of the Kalman filter.
Figure 6 below shows the measured wheel speed (Red) and the
predicted wheel speed (blue) using the Kalman filter when the
clutch is open and closed. Filtering action of the Kalman filter
can be easily seen. To verify the predictive nature of the
Kalman filter, the wheel speed measurement (state 4) was
removed in figure 7. Figure 7 shows the wheel speed predicted
with and without the noisy measurement. The predicted value
without the measurement (green) corelates well with the value
with measurement (blue).

Figure 8. Vehicle acceleration with and without closed loop control

Figure 6. Filtering action of the Kalman filter for wheel speed

measurement.

4.2 Rejection of initial jerk during engine crank with closed

loop control
Figure 8 below shows the vehicle acceleration response with
and without the LQG control (green and red lines,
respectively). Comparing the vehicle acceleration with and
without the closed loop control (green and red respectively),
two observations can be made. The initial jerk (rapid
deceleration) between 39 and 39. 5 seconds on the time axis
has been elimited. It can be seen that the vehicle acceleration
with closed loop control (green) is no longer negative. Thus
the driver will not feel a sudden retardation of the vehicle as
the engine is cranked. The sudden retardation is caused by the
brake (B) acting on the free wheeling ring gear. The negative
brake torque is trasferred to the wheels, resulting in the
negative acceleration. Inspection of figure 9 suggests that the
Figure 9. Vehicle acceleration with and without feedback control
cancellation of the initial jerk can be attributed primarily to the
feedforward action of the control. Certainly, the motor torque
command in figure 10, between 39 and 39.5 seconds, shows
significant similarities to the brake torque command (in blue)
in figure 4.
4.3 Dampening of the vehicle oscillations (ringing effect)
As discussed in section 2.3, the impulse nature of the brake
torque, the throttle command, and the clutch, result in a
sustained oscillation along the driveline , as can be seen in the
red signal in figure 8. Close loop control eliminates this
negative acceleration (green signal in figure 8). Inspection of
 Neeraj Shidore et al. / IFAC PapersOnLine 55-24 (2022) 103–110
figure 9 suggests that the dampening of the oscillations can be
attributed to both the feed-forward and the feedback torques.
This makes intuitive sense, because elimination of the brake
torque impulse by the feedforward torque should reduce the
ringing, but not completely eliminate it. The contribution to
the ringing effect by the throttle and clutch is eliminated by the
feedback control. The sustained oscillations, with the open
loop actuation, would have manifested as vibrations felt by the
driver and would have been undesirable from a drive quality
perspective. Figure 10 shows the motor torque action towards
cancelling the sustained oscillations.
Figure 10. Motor Power for disturbance Rejection
4.4 Motor Utilization to reject disturbances
Figure 10 shows the motor power to reject the disturbance. The
peak motor power is about 2.5 kW. The engine crank scenario
and the vehicle state during this simulation is one of the
harshest from a drive quality perspective. Therefore, it can be
concluded that 2.5 kW of motor power should cover all the
engine crank disturbances ( at other gears and vehicle states).
In general, about 2.5 kW of motor power needs to be
‘reserved’ to be available to reject any disturbance during
engine crank, and cannot be used for traction, unless the engine
is already operational. This can have an adverse effect on the
hybrid vehicle fuel economy. One approach, to ensure that 2.5
kW of motor power is always ‘reserved’ to cancel engine crank
disturbance, is to command /force an engine start as the motor
power approaches this 2.5 kW limit. This anticipatory control
will eliminate any situation where there is insufficient power
reserve, and the engine crank quality suffers.
5. CONCLUSIONS
The simulation results show that the planetary starter can
effectively crank an engine for a flying start type of maneuver
and Closed loop LQG control has been effective in minimizing
the disturbance during the engine cranking and the subsequent
oscillations.
In addition, the following comments can be made about this
work:
1. The motor torque required to cancel the jerk is
heavily influenced by the planetary gear ratio. A
lower gear would decrease the brake torque
requirement, but would increase the jerk.
109
2. Certain simplifications are assumed in the model, i.e.
no gear lash during transition from negative to
positive torque. This would increase the complexity
of the model and the associated control. Similarly,
motor torque is assumed to be instantaneous, motor
electrical time constants, being significantly smaller
than the mechanical time constants , have been
neglected.
3. Hardware validation of the planetary starter and the
control mechanism are needed to validate the
simulation results.
REFERENCES
[1] Wu, G., Zhang, X., and Dong, Z (2015). Powertrain
Architectures of electrified vehicles: Review,
classification, and comparison, Journal of Franklin
Institute, Volume 352, Issue 2, February 2015, Pages
425-448
[2] Shidore, N., Bucknor, N., and Raghavan M. (2020). Fuel
Economy and Drivability Trade-Off for Mild Hybrid
Electric Vehicles, Advances in Industrial Machines and
Mechanisms, Select Proceedings of IPROMM 2020,
Pages 709-719.
[3] Christian Jauch, Santosh Tamilarasan, Giorgio Rizzoni
et al. (2018). Modeling for drivability and drivability
improving control of HEV, Control Engineering
Practice, Volume 70, January 2108, pages 50-62.
[4] Hao, L., Shidore, N., et al. (2019) .Brushless Fast Starter
for Automotive Start/Stop Application”, 2019 IEEE
Energy Conversion Congress and Exposition (ECCE),
29 Sep – 3rd Oct 2019, DOI: 10. 1109/ECCE
2019.8912617.
[5] Fulks,G., Roth, G., Fedawa.A. (2012). High Performance
Stop-Start System with 14 Volt Belt Alternator Starter.
SAE Int.J.Engines 5(3):2012, doi:10.4271/2012-01-
1041.
[6] Raghavan, M., and Shidore, N. (2020) Hybrid
Transmission with gear-based starter and method of
starting. US Patent # US10940748B2.
[7] Burl, J. (1998) Linear Optimal Control, H2 and H-infinity
methods, Addison Wesley Publications, California.
[8] Karnopp, D., Margolis, D., Rosenberg, R. (1990) System
Dynamics: A Unified Approach, 2nd edition, Wiley
Interscience Publication, New York.
[9] Peterka, V.(1991) Adaptation of LWG control design to
engineering needs, Advanced Methods in Adaptive
Control for Industrial Applications, Lecture notes in
control and information Sciences,Vol 158, Springer ,
Berlin.
[10] Sievers, L., Andreas H., (1989) Comparison of Two LQG
Based Methods for Disturbance Rejection. Proceedings
of the 28th conference on Decision and Control, Tampa,
Florida.
110 Neeraj Shidore et al. / IFAC PapersOnLine 55-24 (2022) 103–110

APPENDIX A: AMESIM PLANT MODEL OF THE P2 HYBRID POWERTRAIN