Available online at www.sciencedirect.

com

ScienceDirect
IFAC PapersOnLine 52-28 (2019) 106–113
LPV-MPC
LPV-MPC Control
Control for for Autonomous
Autonomous
LPV-MPC Control
Vehiclesfor Autonomous
LPV-MPC Control
Vehiclesfor Autonomous
Vehicles
Vehicles 1,2
Eugenio Alcalá1,2
1,2 , Vicenç Puig1,2 , Joseba Quevedo2
2
Eugenio
Eugenio Alcalá
Eugenio Alcalá1,21,2 , Vicenç Puig1,2
1,2 ,, Vicenç
Vicenç Puig
Puig1,2 , Joseba
Joseba Quevedo
1,2 ,, Joseba Quevedo2222
Eugenio Alcalá
Alcalá1,2 Quevedo
1,2 , Vicenç Puig1,2 , Joseba Quevedo2
1,2
1 Eugenio
Institut Alcalái Informàtica
de Robòtica , Vicenç Puig , Joseba
Industrial, Quevedo
CSIC-UPC, Llorens i
1
1 Institut
Institut
Institut de
de
de Robòtica
Robòtica iii Informàtica
Artigas
Robòtica Informàtica
Informàtica Industrial,
Industrial,
4-6, 08028 Barcelona,
Industrial, CSIC-UPC,
CSIC-UPC,
Spain
CSIC-UPC, Llorens iii
Llorens
CSIC-UPC, Llorens
1
1
1
Institut de Robòtica
Artigas
Artigas i Informàtica
4-6,
4-6, 08028
08028 Industrial, Spain
Barcelona,
Barcelona, Spain Llorens i
1
Institut de Robòtica
Artigas
Artigas i Informàtica
4-6, 08028
4-6, Industrial, Spain
08028 Barcelona,
Barcelona, CSIC-UPC, Llorens i
Spain
2 Artigas
Research Center 4-6, 08028 Barcelona,
for Supervision, Safety andSpain
Automatic Control,
2 Research Center for Supervision, Safety and Automatic Control,
2
2
2 Research
UPC, Center
Rambla for
Sant Supervision,
Nebridi,
Supervision, Safety
s/n, 08022
Safety and
and Automatic
Terrassa,
2 Research Center for Supervision, Safety and Automatic
Research Center for Automatic Control,
Spain
Control,
Control,
2 UPC,
Research
UPC, Rambla
UPC, Center
Rambla Sant
Ramblafor
Sant
Sant Nebridi,
Supervision,
Nebridi, s/n,
Nebridi, s/n,
s/n, 08022
08022
Safety
08022 andTerrassa,
Terrassa,
Automatic
Terrassa, Spain
Spain
Control,
Spain
UPC, Rambla Sant Nebridi, s/n, 08022 Terrassa, Spain
UPC, Rambla Sant Nebridi, s/n, 08022 Terrassa, Spain
Abstract In this work, a novel approach is presented to solve the trajectory tracking problem
Abstract
Abstract
for autonomous In
In this
this work,
work,
vehicles. aa
a novel
novel approach
approach is
is presented
presented to
to solve
solve the
the trajectory
trajectory tracking
tracking problem
problem
Abstract
Abstract
for autonomous
In
In this
this work, This
work,
vehicles. a novel
This
method
novel approach
approach
method
is based
is
is
based
on the use
is presented
presented
on the usetoof
to a cascade
solve
solve
of a
the
cascade
control where
the trajectory
trajectory
control
the problem
tracking
tracking
where the
external
problem
external
for autonomous
loop autonomous
Abstract
for solves In vehicles.
thethis
position
work,
vehicles. This
control
a
This method
novel usingis
approach
method isabased
based
novel
is onLinear
the use
presented
on the useto ofsolve
of a cascade
Parameter
a cascade
the control- where
Varying
trajectory
control where
Model the
tracking external
Predictive
the problem
external
for
loop autonomous
solves
loopautonomous
solves the vehicles.
position
the vehicles.
position This
control
controlmethodusing
using is abased
novel
abased on the
Linear
novelonLinear
Linearuse of a
Parameter
Parameter cascade control
Varying
Varying - where
Model
- where
Modelthethe external
Predictive
Predictive
Control
for
loop (LPV-MPC)
solves the position approach
This
control and
methodusing the
is a internal
novel loop
the useisofina charge
Parameter cascade of the dynamic
control
Varying - Model control
external
Predictive of
loop
Control
Control solves the
(LPV-MPC)
(LPV-MPC) position approach
approachcontrol using
and
and the a novel
thea internal
internal
internal Linear
loop Parameter
loop Parameter
istechnique
is in charge
in Varying
chargeVarying
charge of the
of -
the dynamic
the Model
dynamic
dynamic Predictive
control
control of
of
the
loop vehicle
Control solves using
the
(LPV-MPC) a LPV
position approach - Linear
control Quadratic
using
and the novel Regulator
Linear
loop is in designed
of - via
ModelLinear Matrix
Predictive
control of
Control
the
the vehicle
vehicle(LPV-MPC)
using
using a
a LPV
LPVapproach -
- Linear
Linear and the internal
Quadratic
Quadratic loop istechnique
Regulator
Regulator in chargedesigned
technique of the dynamic
designed via
via control
Linear
Linear Matrix
Matrix of
Inequalities
Control
the
the vehicle
vehicle (LPV-LMI-LQR).
(LPV-MPC)
using
using aa LPV
LPVapproach -- Linear
LinearBoth
and techniques
the internal
Quadratic
Quadratic use
Regulator
Regulatoran LPV
loop istechnique
technique representation
in chargedesignedof the dynamic
designed of
via
via the
Linear
Linearkinematic
control
Matrix
Matrix of
Inequalities
Inequalities
and
the dynamic
vehicle
Inequalities (LPV-LMI-LQR).
(LPV-LMI-LQR).
models
using a LPV
(LPV-LMI-LQR).of the Both
Both
vehicle.
- LinearBoth techniques
techniques
The
Quadratic use
use an
mainRegulator
techniques an
contribution
use LPV
an technique
LPV representation
LPVofrepresentation
representation
the designed
LPV-MPC of
via
of the
Linear
the kinematic
oftechnique
the kinematic
kinematic
is its
Matrix
Inequalities
and dynamic
and dynamic
dynamic (LPV-LMI-LQR).
models
models of
of the the
the very Both
vehicle.
vehicle. techniques
The
The main use
main obtained an
contribution
contributionLPV of representation
the LPV-MPC
ofrepresentation
the LPV-MPC of the kinematic
technique
technique is its
is its
its
ability
and to calculate
Inequalities models solutions
(LPV-LMI-LQR).of close
Both
vehicle. to those
techniques
The main use an LPV
contributionby the of non-linear version butkinematic
oftechnique
the reducing
and
ability
abilitydynamic
to
to models
calculate
calculate of the very
solutions
solutions vehicle.
very close
close The to
to main obtained
those
those contribution
obtained by
by of the
the
the the LPV-MPC
LPV-MPC
non-linear
non-linear version
version technique
but
but
is
is its
reducing
reducing
significantly
and
abilitydynamic
to the
calculate computational
models of
solutions the very cost
vehicle.
close and
The to allowing
main
those the real-time
contribution
obtained by of
the operation.
the LPV-MPC
non-linear To demonstrate
version technique
but is
reducingthe
its
ability to calculate
significantly
significantly the
the solutions verycost
computational
computational close
cost and
and to allowing
those obtained
allowing the
the by the non-linear
real-time
real-time operation.
operation. version
To
To but reducing
demonstrate
demonstrate the
the
potential
ability to of
significantly
significantly the
calculate
the
the LPV-MPC,
solutions we
computational
computational verypropose
close
cost
cost and
and toa allowing
comparison
those
allowing the
the between
obtained by thethe
real-time
real-time non-linear
non-linear
operation.
operation. To
To MPC
version formulation
but reducing
demonstrate
demonstrate the
the
potential
potential of
(NL-MPC)
significantly
potential of the
of and
the
the LPV-MPC,
LPV-MPC,
the
the LPV-MPC,LPV-MPC
computational we
we propose
cost andaaaa allowing
we propose
propose
approach. comparison
comparison
comparison between
between
the between
real-timethe the
the non-linear
non-linear
operation.
non-linear MPC
MPC
ToMPC formulation
formulation
demonstrate
formulation the
potential
(NL-MPC)
(NL-MPC) of the
and
and LPV-MPC,
the LPV-MPC
the LPV-MPC
LPV-MPC we propose
approach.
approach. comparison between the non-linear MPC formulation
potential
(NL-MPC)
(NL-MPC) of the
and LPV-MPC,
the we propose
approach. a comparison between the non-linear MPC formulation
© 2019, IFAC and
(NL-MPC) and
the LPV-MPC
(International
theParameter
LPV-MPCFederationapproach.
of Automatic Control) Hosting by Elsevier Ltd. All rights reserved.
approach.
Keywords: Linear Varying Systems and Applications , MPC, Autonomous vehicles,
Keywords:
Keywords:
Self Linear
Linear
drivingLinear
Keywords: Parameter
Parameter Varying
cars Parameter Varying Systems and
Varying Systems
Systems and Applications
and Applications ,,, MPC,
Applications MPC, Autonomous
MPC, Autonomous vehicles,
Autonomous vehicles,
vehicles,
Keywords:
Self
Self driving Linear
drivingLinearcars Parameter Varying Systems and Applications , MPC, Autonomous vehicles,
cars Parameter Varying Systems and Applications , MPC, Autonomous
Keywords:
Self
Self driving
driving cars cars vehicles,
Self driving cars
1. INTRODUCTION position, velocities, etc). Then, the trajectory planning
1. INTRODUCTION
INTRODUCTION position,
1.
1.
1. INTRODUCTION
INTRODUCTION position,isvelocities,
position,
module
position, responsibleetc).
velocities,
velocities,
velocities, for Then,
etc).
etc).
etc). Then,
generating
Then,
Then,
the
the trajectory
the
the trajectory
the route using
trajectory
trajectory
planning
planning
planning
planning the
In the last recent1. years, INTRODUCTION
we have experienced a great position, module
module is
modulevehicle
actual is responsible
isvelocities,
responsible
position
responsible for
and
etc).
for generating
for Then,
generating
the desired
generatingthe the
the route
route
one.route using
using
This trajectory
trajectory
the planning
using the
the
the
In the last recent years, we have experienced a great module
actual
actual is
vehicle
vehicleresponsible
position
position for
and
and generating
the desired
the desired
desired the
one. route
This
one.route
Thisand using
trajectory
trajectory the
In
In the
advance
the last
in the
last recent
recent years,
technological
years, we
we have
career
have experienced
towards
experienced aa great
autonomous great is composed
module
actual is
vehicle ofposition
global and
responsible positions,
for generating
the orientations
the
one. This usingvehicle
trajectory the
In the
advance last
in recent
the years,
technological we have
career experienced
towards a
autonomous great actual
is vehicle
composed ofposition
global and the
positions, desired one.
orientations This trajectory
and vehicle
advance
driving.
In the
advance in
last
in the
Today, technological
we
recent
the can
years,see how
technological we career
research
have
career towards centers
experienced
towards autonomous
and a
autonomous com- is
great is
is
composed
velocities.
actual vehicle
composed
composed
of
Finally,
of global
position
global
of global
positions,
theand automatic
the
positions,
positions,
orientations
desiredcontrol
one.
orientations
orientations This and
generates vehicle
trajectory
and the
vehicle
and vehicle
advance Today,
in the we technological career towards autonomous
driving.
driving.
advanceinToday,
panies
driving.
driving. inthe
Today,
Today, we can
can
theautomotive see
see
technological
we
we can
can see
see
how
how
sectors
how
how
research
research
career aretowards
research
research
centers
centers
accelerating
centers
centers
and
and
and
and and
autonomous com-
com-
com-
com-in- velocities.
velocities.
control
is composed
velocities.
velocities.
Finally,
Finally,
actions
Finally,
Finally, globalthe
the
of (acceleration, automatic
automatic
thepositions,
the steering
automatic
automatic
control
control
orientations
control
control
generates
generates
and braking)and vehicle
generates
generates
the
for the
the
panies
panies
vesting in
in the
the
large automotive
automotive
amounts sectors
of sectors
money. are
are accelerating
In accelerating
addition, and
and in-
in- control
actuators actions
control actions
actions
using (acceleration,
(acceleration,
the the
computed steering
steering
sequence and braking)
andofbraking)
braking)
references for the
for and
the
driving.
panies
panies in
in Today,
the
the we can see
automotive
automotive how research
sectors
sectors are
are centers ifand
accelerating
accelerating
we
and
and
add
com-in-
in-
velocities.
control
control Finally,
actions (acceleration,
(acceleration, automatic
steering
steering control
and
and generates
braking) for
for the
the
vesting large amounts of money. In addition, if we
the add actuators using the computed sequence of references and
vesting
to
panies
vesting
vesting
large
this inprogress amounts
the automotive
large
large amounts
amounts
of
the advances money.
of sectors
of money.
money.
In
in are
In
In
addition,
legislation
accelerating
addition,
addition,and
andif
if
if we
we
and
we in- actuators
add
add
add
the position
control
actuators
actuators
using
actions
using
the
the computed
of (acceleration,
using the vehicle.
computedsteering
computed
thevehicle.
sequence
sequence
sequence andof
of references
ofbraking)
references
references for and
the
and
and
to
to this
this
creasing progress
progress
acceptance the
the advances
advances
ofadvances
the in
in
user,inwe legislation
legislation
converge and
onif thethe
the in-
in-
fact the
the position
position
actuators
the position of
of
using
of the
the vehicle.
thevehicle.
the computed sequence of references and
vehicle.
vesting
to this large
progress
to this progress amounts
the
the of of
advancesmoney. inwe In addition,
legislation
legislation and
and we
the add
thefactin-
in- the position of the
creasing
creasing acceptance
acceptance of the
theit user,
user, inwe converge
converge on
on the
that
to driving,
this
creasing
creasing
that
progressas we
acceptance
acceptance
driving, as we
theknow
ofadvances
of
know
the
theit today,
user,
user,
today,
we
we has
legislation
has
converge
converge
days on the
days counted.
and
on
counted.
the
the fact
theThe in- the
fact
fact
The
position of the vehicle.
The automatic control is the last piece in the sequence
that
that driving,
numerous
creasing
driving, as
as we
advantages
acceptance we know
of
knowthat
theit
it today,
the
user,
today, we has
autonomous
has days
converge
days counted.
vehicle
on
counted.the The
offers
fact
The The
that driving,
numerous
numerous as we know
advantages
advantages that
that it the
today,
the has days counted.
autonomous
autonomous vehicle
vehicle The of
offers
offers
The
The
Thetheautomatic
automatic
autonomous
automatic
automatic
control
control
vehicle
control
control
is
is the
is
is and last
the
the
the last
last
last
piece
the in
piece
one of
piece
piece in
most
in
in
the
the
the
sequence
sequence
theimportant
sequence
sequence
with
that respect
driving,
numerous to
as traditional
we
advantages knowthat it vehicles
today,
the hasare days
autonomous obvious. However,
counted.
vehicle The
offers of
of the
the autonomous
autonomous vehicle
vehicle and
and one
one of
of the
the most
most important
important
numerous
with
with respect
respect advantages
to
to that
traditional
traditional the autonomous
vehicles
vehicles are
are vehicle
obvious.
obvious. However,
However,offers tasks
The
of
of the
the since
automatic it
autonomous
autonomous is in charge
control
vehicle
vehicle is of
the
and
and guaranteeing
last
one
one piece
of
of the
the in its
most
mostthemotion.
sequence
important
important It
the most
numerous
with attractive
respect advantages
to is that
the great
traditional reduction
the autonomous
vehicles are of vehicle
obvious.accidents
However, on tasks
offers tasks since
since it
it is
is inin charge
charge and of
of guaranteeing
guaranteeing its
its motion.
motion. It
It
with
the
the respect
most to
attractivetraditional
is the vehicles
great are
reduction obvious.
of However,
accidents on is
of also
the
tasks the
since topic
autonomous
it is addressed
in vehicle
charge in
of this
one paper.
of the
guaranteeing From
most
its a model-
important
motion. It
withmost
the
the roads,
most attractive
respect
most which
attractive is
will
to traditional
attractive is
is the
lead
the
the great
to
great
great reduction
a huge
vehicles arereduction
reduction
reduction of
obvious.
of
of accidents
in
accidents
accidents deaths
However, on
on
on tasks
is
is also
also
based
since
the
the
control
it is in
topic
topicpoint
charge of
addressed
addressed
of view, in
in guaranteeing
this
this
the paper.
paper.
control
its motion.
From
From
problem a
a model-
model-
may
It
be
the
the roads,
roads,
on roads which
which
worldwide. will
will lead
lead to
to a huge
aa huge reduction
reduction in
in deaths
deaths tasks
is
is also
also since
the
the it
topic
topicis in charge
addressed
addressed of
in
in guaranteeing
this
this paper.
paper. its
From
From motion.
a
a model-
model- It
the
the most
roads,
roads, attractive
which
which is
will
will the
lead
lead great
to
to a reduction
huge
huge of
reduction
reduction accidents
in
in deaths
deathson based
based
mainly control
control
defined point
point
by of
of
two view,
view, the
the control
control
characteristics: problem
problem
the type may
may
of be
be
con-
on
on roads
roads worldwide. is also
based
based the
control
control topicpoint
pointaddressed
of
of view,
view, in this
the
the paper.
control
control From
problem
problem a model-
may
may be
be
the
on roads,worldwide.
on roads which will lead to a huge reduction in deaths mainly
worldwide. defined by two characteristics:
roads worldwide. mainly
trol
based
mainly
mainly defined
(lateral,
control
defined
defined by
point
by two
longitudinal
by twoof view,
two or the control the
characteristics:
integrated)
characteristics:
characteristics: andtype
the type
problem
the
the the may
type
type
of
of
type
of con-
con-
of con-
con-of
be
on roads worldwide. trol
model (lateral,
trol (lateral,
(lateral,
considered longitudinal
longitudinal
fortwo or integrated)
or integrated)
its design integrated)
(kinematic, and
and
linearthe type
thedynamic,
type of
of
In order to achieve complete autonomous driving, a series mainly trol
trol defined
(lateral, by
longitudinal
longitudinal characteristics:
or
or integrated) the
and
and type
the of
typecon-
type of
thedynamic, of
In
In order
order
of modules to
to achieve
achieve
are needed complete
complete
working autonomous
autonomous
in a sequential driving,
driving, and a series
aa series
orga- trol model
model considered
considered
simplified
model (lateral,non-linear
considered for
for
longitudinal
for its
its design
design
dynamics
its or
design (kinematic,
(kinematic,
or non-linear
integrated)
(kinematic, linear
linear
and
linear dynamic,
dynamics).
the type
dynamic, of
In
In order
order to
to achieve
achieve complete
complete autonomous
autonomous driving,
driving, a series
series model considered for its design (kinematic, linear dynamic,
of simplified non-linear dynamics or non-linear dynamics).
In modules
of
of
of modules
nized manner.
order
modules
modules
are
are
to achieve
are
are
needed
needed
First, working
working
the
complete
needed
needed vehicle
working
working
in
in
autonomous
in
in
aaa sequential
sensing
a sequential
network
driving,
sequential
sequential
and
and
and
and
orga-
orga-
a(GPS,
series
orga-
orga-
simplified
Jiang
model and non-linear
Astolfifor
considered
simplified
simplified non-linear
non-linear dynamics
(2018)
its and Yang
design
dynamics
dynamics or
or non-linear
et al. linear
(kinematic,
or non-linear
non-linear(2017)dynamics).
address
dynamic,
dynamics).
dynamics).
nized
nized
IMU,
of manner.
manner.
modules
nized
nized manner.
manner. are First,
encoders, First,
cameras,
needed
First,
First,
the
the vehicle
vehicle
LIDAR,
working
the
the vehicle
vehicle insensing
sensing
a etc)
sensing
sensing
network
network
collects
sequential
network
network andall (GPS,
(GPS,the Jiang
orga-
(GPS,
(GPS,
Jiang
the and
problem
simplified
Jiang
Jiang and
and
Astolfi
and non-linear
Astolfi
Astolfi
Astolfi
(2018)
(2018)
of lateral dynamics
(2018)
(2018)
and
and Yang
control
and
and
Yang
using et al.
al. (2017)
etnon-linear
or non-linear
Yang
Yang et
et al.
al.
(2017)
(2017)
(2017)
address
address
feedback
dynamics).
address
address
IMU,
IMU,
vehicle encoders,
encoders, cameras,
cameras,
and environment LIDAR,
LIDAR,
information etc)
etc) andcollects
collects
is treated all
all the
the the problem
the problem
problem
control
to Jiang
the and of
techniques.
Astolfi
of lateral
of lateral
lateral control
control
Optimal-based
(2018) and
control using
Yang
using etnon-linear
usingtechniques
non-linear
al. (2017)
non-linear feedback
feedback
like LQR
address
feedback
nized
IMU,
IMU, manner.
encoders,
encoders, First,
cameras,
cameras,the vehicle
LIDAR,
LIDAR, sensing
etc)
etc) network
collects
collects (GPS,
all
all the
the the problem of lateral control using non-linear feedback
vehicle
vehicle
extract and
and environment
environment
measurements information
information
of interest (vehicle and
and is
is
and treated
treated
obstacles to
to control
control
for lateral
the techniques.
techniques.
problem
control control
of
techniques. Optimal-based
Optimal-based
problem
lateral control
Optimal-based techniques
techniques
is formulated
using in Boyali
non-linear
techniques like
like LQR
LQR
et
feedback
like LQR al.
IMU,
vehicle encoders,
vehicle andand cameras,
environment
environment LIDAR,
information
information etc) andcollects
and andis treated
is treated all the
to
to control techniques. Optimal-based techniques like etLQR
extract
extract measurements
measurements of
of interest
interest (vehicle
(vehicle and obstacles
obstacles for
for lateral
lateral
(2018).
control
for control
control
Regarding
techniques.
lateral control problem
problem
the is
is formulated
formulated
longitudinal
Optimal-based
problem is in
in
control,
techniques
formulated in Boyali
Boyali
we can
like et
Boyali et al.
al.
find
LQR
et al.
vehicle
extract
extract
 and environment
measurements
measurements of
of information
interest
interest (vehicle
(vehicle and is
and
and treated
obstacles
obstacles to for lateral control problem is formulated in Boyali al.
This work has been partially funded by the Spanish Government (2018).
(2018).
LQR
for Regarding
Regarding
strategy
lateral
(2018). control
Regarding in the
the
Naeem
problem
the longitudinal
longitudinal
and
is
longitudinal Mahmoodcontrol,
control,
formulated
control, we
we
(2016);
in Boyali
we can
can
can find
find
Junaid
et al.
find
extract
 This
 This
and FEDER measurements
work through
work has been
has beenthe of
partially interest
projects
partially funded
CICYT
funded by(vehicle
by DEOCS
the Spanish
the and
Spanish
and obstacles
Government
SCAV
Government (refs. (2018). Regarding the longitudinal control, we can find


and
This work has been partially funded by the Spanish Government
This work
FEDER has been
through partially
the projects funded
CICYT by the
DEOCS Spanish
and Government
SCAV (refs.
LQR
LQR
et
LQR al.
(2018). strategy
strategy
(2005)
Regarding
strategy in
in
and
in Naeem
Naeem
the
Naeem
H in and
and
Naeem
longitudinal
and Mahmood
Mahmood
and
Mahmood (2016);
(2016);
Mahmood
control, we
(2016); canJunaid
Junaid
(2016).
find
Junaid
MINECO
and

and FEDER
FEDER
This workDPI2016-76493,
through
through
has beenthethe DPI2017-88403-R).
projects
projects
partially CICYT
CICYT
funded DEOCS
by DEOCS Thisand
the Spanish andwork
SCAV
SCAV has(refs.
Government also
(refs. LQR
et
et al.
al. strategy
(2005)
(2005) in Naeem
and H
2
in and Mahmood
Naeem and (2016); (2016).
Mahmood Junaid
(2016).
and FEDER
MINECO
been through
fundedthe
DPI2016-76493,
partially projects CICYT
DPI2017-88403-R). DEOCS Thisand deSCAV
work has(refs.
also However,
LQR
et al. theseand
strategy
(2005) andcontrol
in H 2 in
Naeem
H 2 in Naeem
strategies
and Mahmood
Naeem and
solve
and Mahmood
simplified
(2016);versions
Mahmood Junaid
(2016).
MINECO
MINECO
and FEDER
MINECO
DPI2016-76493,
DPI2016-76493,
through
DPI2016-76493, thebyprojects
AGAUR CICYTof Generalitat
DPI2017-88403-R).
DPI2017-88403-R).
DPI2017-88403-R). DEOCS This
Thisand
This
work
work Catalunya
SCAV
work
has also
has(refs.
has also
also et al.
However, (2005)
these andcontrol
H 2 in
2 Naeem
strategies and
solve Mahmood
simplified (2016).
versions
been
been partially
through
been the
partially
partially
funded
Advanced by
funded Control
funded by
AGAUR
AGAUR Systemsof
of Generalitat
by DPI2017-88403-R).
AGAUR of (SAC)This
Generalitat
Generalitat group de Catalunya
degrant
de (2017
Catalunya
Catalunya
However,
of
et the
al. real
However,
However, these
(2005) problem,
these
these andcontrol
control
control
2
Hi.e.
2 instrategies
theNaeem
strategies
strategies solve
integrated
solve
solve simplified
and control.
Mahmood
simplified
simplified Thisversions
work
(2016).
versions
versions
MINECO
been DPI2016-76493,
partially funded by AGAUR of Spanish
Generalitat work
degrant has
Catalunya also of the real problem, i.e. the integrated control. This work
through
SGR
through
through
been
the
482),
the Advanced
and by
Advanced
the Advanced
partiallyAdvanced
Control
AGAUR
Control
Control
funded Control
by AGAUR
Systems
and the
Systems
Systems
(SAC)
(SAC)
(SAC) group
of Generalitat
group
Research
group grant
groupdegrant
grant
(2017
Agency
(2017
(2017
Catalunya
of the
addresses
However,
of the real
real problem,
both
these
problem, the i.e. the integrated
longitudinal-lateral
controli.e. strategies
the solve
integrated control.
integrated
simplified
control. This
This work
control
versions
work
through
SGR the
482), and by AGAUR Systems
and the (SAC)
Spanish Research (2017
Agency of the
addresses
addressesreal problem,
both
both the i.e. the integrated
longitudinal-lateral
the longitudinal-lateral
longitudinal-lateral control.
integrated
integrated This work
control
control
through
SGR 482),
SGR
through the
482),
the Maria
and
and by
byde AGAUR
Advanced Maetzu
AGAUR
Control Seal
and
and ofthe
Excellence
the
Systems Spanish
Spanish
(SAC) togroup
IRI (MDM-2016-
Research
Research
grant Agency
Agency
(2017 problem
of the
addressesrealfor
bothautonomous
problem, the i.e. vehicles.
the integrated control.
integrated This work
control
SGR
through
0656). 482),
the
The and
Maria byde AGAUR
Maetzu and
Seal of the Spanish
Excellence to Research
IRI Agency
(MDM-2016- addresses
problem
problem both
for
for the
autonomous
autonomous longitudinal-lateral
vehicles.
vehicles. integrated control
through
through
SGR
through theauthor
the
482),
the
Maria
Maria
and by
Maria
is supported
de
de
de
Maetzu
Maetzu Seal
AGAUR
Maetzu
by of
Seal
and
Seal
athe
of
of
FI AGAUR
Excellence
Spanishto
Excellence
Excellence togrant
to
IRI (ref 2017
IRI (MDM-2016-
(MDM-2016-
Research
IRI Agency
(MDM-2016-
FI addresses
problem both
for the
autonomous longitudinal-lateral
vehicles. integrated control
0656).
0656). The
B00433).The author
author is
is supported
supported by a
a FI
by of FI AGAUR
AGAUR grant grant (ref
(ref 2017
2017 FIFI problem for autonomous vehicles.
0656).
through
0656).
B00433).
The
theauthor
The Maria de
author
is supported
Maetzu Seal
is supported
by a FI AGAUR
by a Excellence
FI AGAURtogrant
grant (ref
IRI (MDM-2016-2017
(ref 2017 FI
FI problem for autonomous vehicles.
B00433).
B00433).
0656). The author is supported by a FI AGAUR grant (ref 2017 FI
B00433).
B00433).
2405-8963 © 2019, IFAC (International Federation of Automatic Control) Hosting by Elsevier Ltd. All rights reserved.
Peer review under responsibility of International Federation of Automatic Control.
10.1016/j.ifacol.2019.12.356
 Eugenio Alcalá et al. / IFAC PapersOnLine 52-28 (2019) 106–113
Control strategies based on Linear Parameter Varying
(LPV) models aim to solve NL control problems using a
pseudo-linear reformulation which incorporates the origi-
nal non-linearities within new parameters. These parame-
ters depend on some system states and inputs which are
called scheduling variables. Some recent books, Tanaka
and Wang (2004), Gáspár et al. (2016), Rotondo (2017),
Ostertag (2011) and Duan and Yu (2013), presented the
study of the modeling and design of LPV under the for-
mulation based on LMI. Several design approaches can be
used such as pole positioning, H∞ , H2 and H2 -LQR. These
techniques have proven to be widely accepted in the field
of robotics, for example Rotondo (2017) and Blažič (2017).
Model Predictive Control (MPC) is another technique that
has proven to be one of the most interesting methods
in this field in recent years. This strategy allows to find
the optimal control action through the resolution of a
constrained optimization problem in which a mathemat-
ical model of the real system is evaluated in a future
horizon. Recent articles such as Rawlings and Risbeck
(2017), Corriou (2018) and Mayne (2014) present the latest
advances in MPC control outside the automotive field. In
the field of autonomous vehicles, we can find all kinds of
formulations for the MPC. From NL-MPC applications in
Ercan et al. (2017), where the lateral control problem is
solved, to MPC lateral control using a linearized model of
the vehicle in Xu et al. (2017). Working with non-linear
models usually gives the best results. However, when work-
ing with systems with fast dynamics this technique may
result non-viable since its excessive computational time.
This is the reason why recent exploration of other ways
opens the door to ideas such as Linear Parameter Varying
- Model Predictive Control (LPV-MPC). Cisneros et al.
(2016) and Besselmann and Morari (2009) present the
MPC strategy using LPV models. The advanatge of LPV
approach is that the non-linear model can be expressed
as a combination of linear models with parameter varying
with some scheduling variables without using linearization
(Sename et al., 2013).
The contribution of this paper focuses on the use of LPV
models for the automatic control strategy design for an
autonomous vehicle considering both longitudinal and lat-
eral dynamics. An MPC approach is proposed based on
the LPV kinematic formulation of the vehicle that leads
to a quadratic optimal problem. In addition, introducing
the terminal set concept, we are able to guarantee stability.
The paper is structured as follows: Section 2 gives an
overview of the work and describes the different types
of modelling used for control purposes. In Section 3, the
kinematic and dynamic control designs are developed.
Section 4 shows the simulation results and Section 5
presents the conclusions of the work.
2. OVERVIEW OF THE PROPOSED SOLUTION
In this work, we consider the problem of urban au-
tonomous guidance. To solve it, two important tasks have
to be carried out: the trajectory planning and the auto-
matic control.
107
On the one hand, the trajectory planning to be followed
by the vehicle has to fulfill particular specifications such
as continuous and differentiable velocity profiles. Thus,
this module is in charge of providing discrete and smooth
references to the automatic control stage. On the other
hand, the automatic control is in charge of following the
planned references, thus, moving the vehicle between two
ground coordinates as well as generating smooth control
actions for achieving a comfortable journey. In Fig. 1, we
show the planning-control diagram proposed in this work.
Observe that two control levels have been designed as a
cascade scheme, one for the position control and a faster
and inner one to control the dynamic behaviour of the
vehicle, i.e. linear and angular velocities.
The level of difficulty of a vehicle guidance control problem
comes often determined by two aspects: the type of control
(lateral, longitudinal or mixed) and the complexity of the
model to be controlled (kinematic, linear dynamic, non-
linear simplified dynamic or non-linear dynamic). In this
work we address one of the most complex configurations,
to solve the mixed non-linear dynamic problem. The
following subsection covers the formulation of the different
models used for solving the control problem.
2.1 LPV Control Oriented Models
Unlike controlling common mobile robots which operate
at a low interval of velocities, urban cars work in a higher
velocities and accelerations range. This fact makes indis-
pensable to study control techniques based on elaborated
dynamic models, articularly with the aim of being safer
and smoother in the control performance. In this work, two
model-based techniques cover the kinematic and dynamic
control at different layers and, hence, at different sampling
times. For that reason, the use of mathematical kinematic
and dynamic vehicle models are necessary. The kinematic
model is based on the mass-point assumption while for the
dynamic one the bicycle model has been considered. We
refer to the Appendix A for the complete model equations
used in this paper. In the following subsections, we present
the LPV formulation of both, the kinematic and dynamic
models.
Kinematic LPV model. The state, control and reference
vectors, respectively, are denoted as
 
xe    
v
x c = y e , uc = x , r c = d
v cos θe
, (1a)
ω ωd
θe
where xe , ye and θe are the position and orientation errors,
respectively. The inputs vx and ω are the longitudinal
and angular velocities, respectively. vd and ωd are the
longitudinal and angular reference velocities, respectively.
Then, defining the vector of scheduling variables as ρ(k) :=
[ω(k), vd (k), θe (k)] which are bounded in ω ∈ [−1.42, 1.42]
s , vd ∈ [0.1, 20] s and θe ∈ [−0.05, 0.05] rad, the non-
rad m
linear kinematic model (see Alcala et al. (2018b)) is trans-
formed into the Linear Parameter Varying representation
as follows
xc (k + 1) = Ac (ρ(k))xc (k) + Bc uc (k) − Bc rc (k), (1b)
where
108 Eugenio Alcalá et al. / IFAC PapersOnLine 52-28 (2019) 106–113

Figure 1. Autonomous guidance scheme composed by the trajectory planning stage and both control layers: kinematic
and dynamic.

Dynamic LPV model. The dynamic LPV model consid-

ered in this work is a transformation of the non-linear one
presented in Alcala et al. (2018a).
Then, the state and control vectors are denoted as

   
vx
δ
xd = vy , ud = , (4a)
a
ω
where vy , a and δ are the lateral velocity, rear wheel
longitudinal acceleration and steering angle, respectively
(see Fig. 2). The LPV model can be expressed as
xd (k + 1) = Ad (ϑ(k))xd (k) + Bd ud (k) , (4b)
Figure 2. Bicycle model used for control purposes.
with the time-varying scheduling vector as ϑ(k) :=
  [δ(k), vx (k), vy (k)] being δ ∈ [−0.25, 0.25] rad, vx ∈
1 ωTc 0
[0.1, 20] ms and vy ∈ [−1, 1] s and
m
Ac (ρ(k)) =  −ωTc 1 vd sinθeθe Tc  (1c)
0 0 1  
  1 + A11 Td A12 Td A13 Td
−1 0 Ad (ϑ(k)) = 0 1 + A22 Td A23 Td (4c)
Bc = 0 0 Tc , (1d) 0 A32 Td 1 + A33 Td
0 −1
with Tc being the outer loop sampling time. 1
Cd ρAr vx2 + µmg Cf sin δ
A11 = − 2 , A12 = (4d)
From this formulation, a polytopic representation for the mvx mvx
control design is obtained as Cf lf sin δ Cr + Cf cos δ
A13 = + vy , A22 = − (4e)
mvx mvx
Cf lf cos δ − Cr lr
(4f)
nc
2
 A23 = − − vx
xc (k + 1) = µi (ρ(k))Aci xc (k) + Bc uc (k) − Bc rc (k) , mvx
i=1 Cf lf cos δ − Cr lr Cf lf2 cos δ + Cr lr2
(2) A32 = − , A33 = −
Ivx Ivx
being nc the number of scheduling variables and Ai each (4g)
one of the polytopic vertex systems obtained as a combi-
nation of the extreme values of the scheduling variables.  
0 1
The expression µi (ρ(k)) is known as the membership Bd = B21 0 Td (4h)
function and is given by B31 0

Cf Cf l f
nc
 B21 = , B31 = , (4i)
µi (ρ(k)) = ξij (η0j , η1j ) , nc
i = {1, ..., 2 } (3a) m I
j=1 with Td being the sample time used in the dynamic control
loop, i.e. the inner control loop. m and I represent the
ρj − ρj (k) vehicle mass and inertia, respectively. lf and lr are the
η0j = distances from the center of gravity to the front and rear
ρj − ρj (3b)
wheel axes, respectively. Variables Cf and Cr represent
η1j = 1 − η0j , j = {1, ..., nc } , the tire stifness coefficient for the front and rear wheels. ρ,
Ar , µ and Cd are the density of the air, the front sectional
where ξij (η0j , η1j ) corresponds to any of the weighting area of the vehicle, the friction coefficient and the drag
function that depend on each rule i. coefficient, respectively.
 Eugenio Alcalá et al. / IFAC PapersOnLine 52-28 (2019) 106–113 109

As in the case of kinematic model, we look for a polytopic At each time k the values of xk and uk−1 are known and
dynamic formulation like the following the following optimization problem can be solved
2n d
 
xd (k + 1) = µi ϑ(k) Adi xd (k) + Bdi ud (k) , (5)
i=1 minimize Jk (∆Uk , Xk )
being nd the number of dynamic scheduling variables and ∆Uk
Adi represent each one of the polytopic vertex dynamic subject to
systems obtained as a combination of the extreme values 2

rc

of the dynamic scheduling variables. The membership xk+i+1 = µj (ρk+i )Aj xk+i
function µi (ϑ(k)) is the same than the one presented in j=1
(3) but using the dynamic scheduling vector ϑ(k).
+ Buk+i − Brk+i
3. CONTROL DESIGN uk+i = uk+i−1 + ∆uk+i , ∀i = 0, ..., N − 1
∆Uk ∈ ∆Π
In this section, we present the control scheme proposed for Uk ∈ Π
this work as well as its design. The control strategy of the
xk+N ∈ χ ,
vehicle has been divided into two nested layers, see Fig
1. The outermost layer controls the vehicle’s kinematics, (7)
i.e. position and orientation of the car, and works at a where
frequency of 10 Hz. On the other hand, the internal loop
   
controls the dynamic behavior of the vehicle, i.e. its speeds, ∆uk uk
at a frequency of 200 Hz. Next, both control loops are  ∆uk+1   uk+1 
described separately. ∆Uk = 
 .
..
 ∈ R m , Uk = 
  ..  ∈ Rm ,

.
3.1 Kinematic LPV-MPC Design ∆uk+N −1 uk+N −1
(8)
At this point, we present the formulation of the LPV-MPC
strategy, which focuses on solving position and orientation being m the number of inputs of the kinematic system. Π
control of the vehicle. and ∆Π are the constraint sets for the inputs and their
derivatives, respectively.
This strategy is based on the resolution of a linear
quadratic optimization problem by using the non-linear The set χ represents the terminal state set. Then, by
kinematic error model in its LPV polytopic representation introducing this constraint in the optimization problem,
(2). However, there exist the problem associated with the we force the states to converge into a stable region and
lack of knowledge of the matrix of scheduling variables then, to ensure the MPC stability. The computation of
through the entire prediction horizon. In Cisneros et al. this terminal safety set is carried out by solving two LMI-
(2016), the use of the optimized state sequence which is based problems.
obtained after each optimization is proposed. First, the controller for each polytopic system (Ai ) is found
In this work, the scheduling variables are states of the by solving the following LQR-LMI
system whose desired values are known since the trajectory
planner generates them. That is why we propose the use of  
such references as known scheduling variables for the entire Y (Ai Y + BWi )T Y WiT
optimization horizon being then the scheduling sequence  Ai Y + BWi Y 0 0 
 ≥0
Γ := [ρ(k), ..., ρ(k + N )]. In this way, we can compute the  Y 0 −1
QT S 0  (9)
evolution of the model more accurately and in anticipation. Wi 0 0 RT−1S
In addition, since the basic MPC formulation cannot ∀i = 1, ..., 2rc ,
guarantee the overall stability of the system, we propose
the addition of a terminal constraint and a terminal cost
to the optimization problem. with Y = Y T > 0, QT S = QTT S ≥ 0 and RT S = RTT S > 0.
This problem returns the matrices Y and Wi . Then, the
To formulate the problem, the polytopic LPV system resulting controllers are obtained by Ki = Wi Y −1 . Note
presented in (2) has been considered. In order to avoid that the terminal set matrix P in (6) is found to be equal
a difficult reading, the sub-index c is omitted in the rest to Y −1 .
of the subsection. Then, the focus is on a MPC scheme
where the cost function is defined as This LQR design is a particular formulation for the one
presented in Theorem 25 of Tanaka and Wang (2004). The
constant nature of kinematic input matrix Bc in (1) allows
N
 −1
  the use of this simplified LMI version.
Jk = xTk+i Qxk+i + ∆uk+i R∆uk+i + xTk+N P xk+N ,
i=0 The second problem consists on finding the largest termi-
(6) nal region χ. To do so, we solve the following constrained
optimization problem using the previously obtained con-
where Q = QT ≥ 0, R = RT > 0 and P = P T > trollers Ki
0 represent the states, input and terminal set tuning
matrices of apropriate dimensions, respectively.
110 Eugenio Alcalá et al. / IFAC PapersOnLine 52-28 (2019) 106–113

maximize Jk (Z) 15 (Intell core i7-8550U CPU @ 1.80GHzx8). The LPV-

Z
MPC, dynamic LPV-LQR and vehicle model parameters
subject to are listed in Tables 1, 2 and 4, respectively.
 
−Z Z(Ai + BKi )T
<0 Table 2. Dynamic LPV-LQR design parame-
(Ai + BKi )Z −Z
ters.
Ki ZKiT − u2 < 0 ∀i = 1, ..., 2rc .
Parameter Value
(10)
Q 0.9*diag(0.66 0.01 0.33)
The resulting variable is Z. Hence, we compute the largest R 0.1*diag(0.5 0.5)
terminal region as χ = {x|xT Sx ≤ 1}, with S = Z −1 . Note Td 0.005 s
that this problem is totally constrained by the maximum
values of the control actions. Solving the problem (10) to determine the largest terminal
set, we obtain matrix S as
 
3.2 Dynamic LPV-LQR Design 0.465 0 0
S= 0 23.813 76.596 . (12)
To design the dynamic controller we use the polytopic 0 76.596 257.251
system (5). Then, we solve offline the optimal LMI problem
(9) for computing the polytope vertex controllers Ki . The tests have been carried out in the circuit of Fig.
Finally, the dynamic controller gain is computed online 3 where the aim is to simulate a road driving at a
following variable speed. For assessment purposes, a perturbation is
2rd introduced in the coefficient of friction, varying it sharply

K(ϑ(k)) = µi (ϑ(k))Ki , (11) from 1 to 0.5 at t = 110 s and from 0.5 to 1 at t = 120
s. This scenario intends to show a real case in which the
i=1
vehicle passes through dry and wet asphalt surface while
where µi (ϑ(k)) represents the weighting function pre- turning a curve.
sented in (3) by using the dynamic scheduling vector
Fig. 4 shows both, the linear and angular speed profiles
defined in (4). The offline computation of polytopic con-
provided by the trajectory planning and the respective
trollers allows this control strategy to work at the desired
vehicle responses for both compared approaches. In Fig.
frequency of 200 Hz.
5, we illustrate the complete set of errors, i.e. xe , ye , θe , ve
and ωe . In both, Fig. 4 and 5, it is seen the close behaviour
4. SIMULATION RESULT between LPV-based and NL-based approaches. However,
the non-linear MPC is able to better handle the external
Table 1. Kinematic LPV-MPC design param- disturbances.
eters.
The respective control actions applied to the simulation
Parameter Value Parameter Value
vehicle are shown in Fig 6. Note that the LPV-MPC
Q 0.9*diag(0.33 0.33 0.33) u [1.4 20] response is as good as the NL-MPC one until the arrival
R 0.1*diag(0.8 0.2) u [-1.4 0.1] of the disturbance in the longitudinal axis. The steering
Tc 0.1 s ∆u [0.3 2]
behaviour is practically the same throughout the test.
N 20 ∆u [-0.3 -2]
RT S diag(1 3) QT S diag(1 1 3) An important aspect of control strategies based on opti-
mization is the computational time spent at each optimiza-
In this section, we validate the performance of the pro- tion procedure. In Fig. 7, we show the elapsed time at each
posed control scheme through simulation in MATLAB. kinematic MPC optimization for both compared methods.
The considered vehicle for running simulations is an elec- Note the computational time improvement achieved when
tric RWD one whose dynamics are presented in Appendix using the LPV-MPC strategy. Finally, a quantitative com-
B. To show the promising results of the LPV-MPC, we parison is made using the root mean squared error (RMSE)
perform a comparison against the non-linear MPC ap- as performance measurement as shown in Table 3.
proach (NL-MPC in resulting figures). The LPV-MPC
uses planning data to instanciate the state space matrices Table 3. Comparison using the root mean
at every control time step within the MPC prediction. squared error measure (RMSE).

Then, the optimal control problem (7) is solved at a Approach x y θ v ω

frequency of 10 Hz using the solver GUROBI (Gurobi LP V -M P C 0.589 0.238 0.016 0.302 0.014
(2014)) through YALMIP framework (Lofberg (2004)). N L-M P C 0.528 0.225 0.015 0.268 0.012
This solves the position control problem in a outer loop.
Note that the non-linear MPC problem has been computed These results conclude a similar performance of the LPV-
using IPOPT solver and considering the same adjustment MPC approach in comparison to the NL-MPC version.
and prediction horizon (see Table 1). In the inner loop, the
dynamic state feedback control problem (Section 3.2) is 5. CONCLUSION
solved at a rate of 200 Hz to follow the velocities provided
by the kinematic control.
In this work, a cascade control scheme (kinematic and
To verify the real-time feasibility of the presented strate- dynamic) was presented to solve the problem of integrated
gies, we perform the simulations on a DELL Inspiron control (lateral and longitudinal) of autonomous vehicles.
 Eugenio Alcalá et al. / IFAC PapersOnLine 52-28 (2019) 106–113 111

420 360

350
400
340

380 330

320
360
310
Y [m]

340 300

903 904 905 906 907

320

300

280

260 Reference LPV-MPC Road limits NL-MPC

550 600 650 700 750 800 850 900

X [m]

Figure 3. Simulation circuit used for testing the proposed control technique (LPV-MPC) and its non-linear version
(NL-MPC).

planner references. It was demonstrated that the LPV-

55
MPC technique works very well compared to the non-
Reference LPV-MPC NL-MPC linear control problem but in a much faster way (more
than 50 times faster).
Linear velocity [Km/h]

50

45
Table 4. Dynamic model parameters.
40

35 Parameter Value Parameter Value

lf 0.758 m Ar 1.91 m2
30 kg
lr 1.036 m ρ 1.184 m 3
25 m 683 kg Cd 0.36
70 80 90 100 110 120 130
I 560.94 kg m2 µ 1
N N
Cf 24000 rad Cr 21000 rad
15 d 2680 c 1.6
b 6.1
Angular velocity [deg/s]

10

A. VEHICLE MODEL FOR CONTROL

5

The non-linear equations employed for control purposes

0
are presented as
-5
70 80 90 100 110 120 130 ẋe = ωye + vd cos θe − vx
Time [s]
ẏe = −ωxe + vd sin θe
θ̇e = ωd − ω
Figure 4. Reference and response velocities for both com-
pared approaches. FyF sin δ Fdf
v̇x = a − − + ωvy
m m
FyF cos δ FyR
The novel kinematic control was designed using the MPC v̇y = + − ωvx
technique with the prediction model expresed in the LPV m m
formulation (LPV-MPC) without using any linearisation. FyF lf cos δ − FyR lr (13)
ω̇ =
On the other hand, the dynamic control was approached  I 
using the Linear Quadratic Regulator (LQR) strategy, vy lf ω
FyF = Cf δ − −
with a LPV modeling and using a LMI formulation of the vx vx
problem (LPV-LMI-LQR).  
vy lr ω
FyR = Cr − +
A comparison was made between two methods of solving vx vx
the control problem: using the non-linear MPC formula- 1
tion (NL-MPC) and using updated LPV-MPC with the Fdf = Fdrag + Ff riction = Cd ρAr vx2 + µmg
2
112 Eugenio Alcalá et al. / IFAC PapersOnLine 52-28 (2019) 106–113

Longitudinal error [m]

0.2
0
-0.2
-0.4
70 80 90 100 110 120 130
Lateral error [m]
0.2
0
-0.2
70 80 90 100 110 120 130
Orientation error [deg]
1
0
-1
70 80 90 100 110 120 130
Linear velocity error [km/h]
2
0
-2
70 80 90 100 110 120 130
Angular velocity error [deg/s]
2
1
0
-1
70 80 90 100 110 120 130
Time [s]
LPV-MPC NL-MPC

Figure 5. Time evolution of the tracking errors for each compared kinematic control strategy.

NL-MPC
12 2.2
LPV-MPC NL-MPC
Elapsed time / it. [s]

2
acceleration [m/s²]

10
Rear wheel

1.8
8
1.6

6 1.4

1.2
4 70 80 90 100 110 120 130
70 80 90 100 110 120 130

LPV-MPC
0.07
3
Elapsed time / it. [s]

0.06
Steering angle [deg]

2
0.05
1
0.04
0
0.03
-1
0.02
70 80 90 100 110 120 130
-2
70 80 90 100 110 120 130 Time [s]
Time [s]

Figure 7. Computational effort comparison when solving

Figure 6. Resulting control actions for both compared the kinematic MPC. Top: Elapsed time per iteration
methods. Top: traction wheel acceleration. Bottom: of non-linear MPC strategy. Bottom: Elapsed time per
steering angle of front wheels. iteration of the proposed LPV-MPC approach.
 Eugenio Alcalá et al. / IFAC PapersOnLine 52-28 (2019) 106–113
Refer to the Appendix A in Alcala et al. (2018b) for
the complete development of the kinematic error model.
Vehicle parameters are defined in Table 4.
B. VEHICLE MODEL FOR SIMULATION
For simulation purposes we use a higher fidelity vehicle
model. Unlike the model used for control design, this
considers a more precise tire model, i.e. the Pacejka "Magic
Formula" tire model where the parameters b, c and d
define the shape of the semi-empirical curve. Also, a more
accurate computation of the tire slip angles is given.
ẋ = vx cos θ − vy sin θ
ẏ = vx sin θ + vy cos θ
θ̇ = ω
FyF sin δ Fdf
v̇x = a − − + ωvy
m m
FyF cos δ FyR
v̇y = + − ωvx
m m
FyF lf cos δ − FyR lr
ω̇ =
I (14)
FyF = d sin (c tan−1 (bαf ))
FyR = d sin (c tan−1 (bαr ))
 
−1 vy lf ω
αf = δ − tan −
v vx
 x 
vy lr ω
αr = − tan−1 + Springer Science & Business Media.
vx vx
1 Rawlings, J.B. and Risbeck, M.J. (2017). Model predictive
Fdf = Fdrag + Ff riction = Cd ρAr vx2 + µmg
2 Automatica, 78, 258–265.
All parameters are properly defined in Table 4.
REFERENCES
Alcala, E., Puig, V., Quevedo, J., and Escobet, T. (2018a).
Gain scheduling lpv control for autonomous vehicles
including friction force estimation and compensation
mechanism. IET Control Theory & Applications.
Alcala, E., Puig, V., Quevedo, J., Escobet, T., and Coma-
solivas, R. (2018b). Autonomous Vehicle Control Using
a Kinematic Lyapunov-based Technique with LQR-LMI
Tuning. Control Engineering Practice, 73, 1–12.
Besselmann, T. and Morari, M. (2009). Autonomous
vehicle steering using explicit lpv-mpc. In Control
Conference (ECC), 2009 European, 2628–2633. IEEE.
Blažič, S. (2017). Two approaches for nonlinear control
of wheeled mobile robots. In Control & Automation
(ICCA), 2017 13th IEEE International Conference on,
946–951. IEEE.
Boyali, A., Mita, S., and John, V. (2018). A tutorial on au-
tonomous vehicle steering controller design, simulation
and implementation. arXiv preprint arXiv:1803.03758.
Cisneros, P.S., Voss, S., and Werner, H. (2016). Efficient
nonlinear model predictive control via quasi-lpv repre-
sentation. In Decision and Control (CDC), 2016 IEEE
55th Conference on, 3216–3221. IEEE.
Corriou, J.P. (2018). Model predictive control. In Process
Control, 631–677. Springer.
Duan, G.R. and Yu, H.H. (2013). LMIs in Control
Systems: Analysis, Design and Applications. CRC press.
113
Ercan, Z., Gokasan, M., and Borrelli, F. (2017). An adap-
tive and predictive controller design for lateral control
of an autonomous vehicle. In Vehicular Electronics and
Safety (ICVES), 2017 IEEE International Conference
on, 13–18. IEEE.
Gáspár, P., Szabó, Z., Bokor, J., and Németh, B. (2016).
Robust control design for active driver assistance sys-
tems. Springer, DOI, 10, 978–3.
Gurobi, O. (2014). Inc.,“gurobi optimizer reference man-
ual,” 2015. Google Scholar.
Jiang, J. and Astolfi, A. (2018). Lateral control of an
autonomous vehicle. IEEE Transactions on Intelligent
Vehicles.
Junaid, K.M., Shuning, W., Usman, K., and Naveed, R.
(2005). Lqr autonomous longitudinal cruise control
with a minimum order state observer. In Proc. Eighth
IASTED Int. Conf., Cambridge, USA, volume 31.
Lofberg, J. (2004). Yalmip: A toolbox for modeling and
optimization in matlab. In Computer Aided Control
Systems Design, 2004 IEEE International Symposium
on, 284–289. IEEE.
Mayne, D.Q. (2014). Model predictive control: Recent
developments and future promise. Automatica, 50(12),
2967–2986.
Naeem, H.M.Y. and Mahmood, A. (2016). Autonomous
cruise control of car using lqr and h2 control algorithm.
In Intelligent Systems Engineering (ICISE), 2016 Inter-
national Conference on, 123–128. IEEE.
Ostertag, E. (2011). Mono-and Multivariable Control
and Estimation: Linear, Quadratic and LMI Methods.
control with discrete actuators: Theory and application.
Rotondo, D. (2017). Advances in Gain-Scheduling and
Fault Tolerant Control Techniques. Springer.
Sename, O., Gaspar, P., and Bokor, J. (2013). Robust
control and linear parameter varying approaches: appli-
cation to vehicle dynamics, volume 437. Springer.
Tanaka, K. and Wang, H.O. (2004). Fuzzy control systems
design and analysis: a linear matrix inequality approach.
John Wiley & Sons.
Xu, Y., Chen, B., Shan, X., Jia, W., Lu, Z., and Xu, G.
(2017). Model predictive control for lane keeping system
in autonomous vehicle. In Power Electronics Systems
and Applications-Smart Mobility, Power Transfer &
Security (PESA), 2017 7th International Conference on,
1–5. IEEE.
Yang, J., Bao, H., Ma, N., and Xuan, Z. (2017). An
algorithm of curved path tracking with prediction model
for autonomous vehicle. In Computational Intelligence
and Security (CIS), 2017 13th International Conference
on, 405–408. IEEE.