Anda di halaman 1dari 6

10th Conference

10th Conference on
on Manoeuvring
Manoeuvring and
and Control
Control of
of Marine
Marine Craft
Craft
10th Conference
August
August 24-26, on Manoeuvring
24-26, 2015.
2015. Copenhagen, and
Copenhagen, Control of Marine Craft
Denmark
Denmark
10th Conference on Manoeuvring andAvailable online atCraft
Control
August 24-26, 2015. Copenhagen, Denmark of Marine www.sciencedirect.com
August 24-26, 2015. Copenhagen, Denmark

ScienceDirect
IFAC-PapersOnLine 48-16 (2015) 184189
Visual Image Based Dynamical Positioning
Visual
Visual Image
Image Based
Based Dynamical
Dynamical Positioning
Positioning
Using Control Laws with Multipurpose Structure
Using
Using Control Laws with Multipurpose Structure
Control Laws with Multipurpose Structure
Evgeny
Evgeny I.
I. Veremey,
Veremey, Margarita
Margarita V.
V. Sotnikova
Sotnikova
Evgeny I. Veremey, Margarita V. Sotnikova
Evgeny I. Veremey, Margarita V. Sotnikova
Faculty
Faculty of
of Applied
Applied Mathematics
Mathematics and
and Control
Control Processes,
Processes,
Faculty of Applied
Saint-Petersburg Mathematics
State University, and Control Processes,
Saint-Petersburg, Russia.
Saint-Petersburg
Faculty of StateMathematics
Applied University, Saint-Petersburg,
and Control Russia.
Processes,
Saint-Petersburg State University,
(e-mail: Saint-Petersburg,
e_veremey@mail.ru) Russia.
Saint-Petersburg State University,
(e-mail: Saint-Petersburg, Russia.
e_veremey@mail.ru)
(e-mail: e_veremey@mail.ru)
(e-mail: e_veremey@mail.ru)
Dynamic positioning
Abstract: Dynamic
Abstract: positioning systems
systems are
are widely
widely used
used now
now toto support
support marine
marine operations
operations of
of various
various
Abstract:
types. Dynamic positioning systems are widely used using
now tononlinear
support marine operationscontrol
of various
types. In
In this
Abstract: this paper
Dynamic the
the problem
paper positioning of
of dynamic
problemsystems positioning
are widely
dynamic used using
positioning now tononlinear visual
visual feedback
support marine operationscontrol
feedback with
of various
with
types.
special In this paper the problem of dynamicThe positioning using nonlinear visual the
feedback control with
specialInmultipurpose
types. this paper the
multipurpose structure
problem
structure is
is considered.
of dynamicThe
considered. control
control objective
positioning is
is to
to stabilize
using nonlinear
objective visual the
stabilize difference
feedback between
control
difference with
between
specialand
given multipurpose
actual structure is considered. Thepoints
control objective is to stabilize the difference between
special
given and actual position
multipurpose of
of the
structure
position theisobserved visual
considered.
observed Thepoints
visual in
control the
the image
imageisplane.
in objective The
The additional
to stabilize
plane. complexity
the difference
additional of
between
complexity of
given
the and actual
problem is position ofbythepresence
determined observedofvisual points
external in the image
disturbances, plane.
such as The additional
waves and wind. complexity
In of
given and actual
the problem position ofbythepresence
is determined observedofvisual points
external in the image
disturbances, plane.
such The additional
as waves In this paper
this
and wind.complexity of
paper
the problem
joint is determined
mathematical model by
of presence
aa ship of external disturbances, such as waves and wind. In thisunified
paper
the
jointproblem is determined
mathematical model by ship and
of presence and screen
screen image
of external dynamic
disturbances,
image dynamic is
such obtained.
is as waves and
obtained. Using
wind.
Using special
In thisunified
special paper
joint mathematical
multipurpose model
structure of of a
control ship and screen image dynamic is obtained. Using special unified
joint mathematical
multipurpose model
structure of of a shiplaw,
control andmentioned
law, screen image
mentioned problem is
is decoupled
dynamic
problem is obtained.
decoupled into particular
intoUsing specialseparated
particular unified
separated
multipurpose
problems. structure of control law, mentioned problem is decoupled into particular separated
problems. The
multipurpose resulting
resultingof nonlinear
Thestructure control law,
nonlinear controller provides
mentioned
controller asymptotic
problem
provides stability
is decoupled
asymptotic and
stabilityinto desirable
and particular dynamic
desirable separated
dynamic
problems.
properties The
of the resulting
closed-loop nonlinear
system controller
regarding provides
external asymptotic
disturbances. stability and desirable dynamic
problems. The resulting nonlinear controller provides
properties of the closed-loop system regarding external disturbances. asymptotic stability and desirable dynamic
properties of the closed-loop system regarding external disturbances.
properties
2015, IFACof the closed-loopFederation
(International system regarding external
of Automatic disturbances. by Elsevier Ltd. All rights reserved.
Control)
Keywords:
Keywords: dynamical
dynamical positioning, control
positioning, control law, visual
law, visual control,Hosting
control, stability, external
stability, external disturbances.
disturbances.
Keywords: dynamical positioning, control law, visual control, stability, external disturbances.
Keywords: dynamical positioning, control law, visual control, stability, external disturbances.

an
1. INTRODUCTION
INTRODUCTION an integral
integral action
action of of the
the controller
controller and and aa notch
notch filtering
filtering effect.
effect.
1. an integral action of the controller and a notch filtering effect.
1. INTRODUCTION an integral action of the controller and a notch filtering effect.
1. INTRODUCTION The
The dynamical
dynamical correctorcorrector synthesis
synthesis is is based
based on on thethe ideas
ideas of of
Dynamical positioning
Dynamical positioning (DP) (DP) systems
systems are are widely
widely usedused forfor the
the The dynamical corrector synthesis is based onmathematical
the ideas of
Dynamical positioning (DP) systems are widelyareas usedsuchfor the H
The -optimization
dynamical corrector
H -optimization theory.
theory.synthesisCorrespondent
is based onmathematical
Correspondent the ideas of
various types of marine
marine vessels in different
different as H -optimization
Dynamical
various
various types of
positioning
types of (DP)
marine vessels
systems
vessels in are widelyareas
in different
usedsuch
areas such
for the
as
as background
H -optimization
background for LTItheory.
for LTI Correspondent
marine autopilot
theory.
marine autopilot design was
Correspondent
design wasmathematical
discussed in
mathematical
discussed in
hydrography,
various types of
hydrography, inspection of
of marine
marine vessels
inspection construction,
in different
marine areas such
construction, wreck
wreckas background for LTI marine autopilot design was discussed in
hydrography, inspection of marine construction, wreck Veremey,
background
Veremey, 2012.
for LTI marine autopilot design was discussed in
2012.
investigation,
investigation, underwater
hydrography, inspection of
underwater cable
cable laying,
marine
laying, and
and so so on.
construction, The
on.wreck
The Veremey, 2012.
investigation,
exhaustive underwater cablesystems
laying,is and so on. The Veremey, 2012.
exhaustive survey
investigation, survey of
of DP
DP control
underwater cablesystems
control done
laying,is and
done so in
in Srensen,
on. The
Srensen, In this
In this paper
paper we we design
design nonlinear
nonlinear visual
visual feedback
feedback control
control law law
exhaustive survey of DP control systems is done in Srensen, In
2011.
2011. Some
exhaustive
Somesurveycentral
centraloftheoretical
DP controland
theoretical and practical
systems
practical issues
is done
issues providing
in Srensen,
providing forthis
for
In DP paper
DP
this problem.
problem.
paper
we design
we We consider
We
design
nonlinear
consider
nonlinear thatvisual
that the DP
the
visual DPfeedback
vessel is
vessel
feedback
control
is equipped
equipped
control
law
law
2011.
basic Some
backgroundcentraloftheoretical
DP control and are
control practical issuesinproviding
presented Fossen, for DPa problem. We consider that the DP vessel is equipped
2011.
basic Some
backgroundcentraloftheoretical
DP and are
practical issuesinproviding
presented Fossen, with
for camera.
withDPa problem.
camera. We The control
Theconsider
controlthat objective is
the DP vessel
objective stabilizing
is equipped
is stabilizing the
the
basic
1994 andbackground
and 2011, Srensen,of
Srensen,DP control
2012. are presented in Fossen, with a camera. The control objective is ofstabilizing the
basic
1994 background
2011, of DP 2012.control are presented in Fossen, difference
with a camera.
difference between
betweenThe actual
actual and
control given
and given positions
objective
positions is of the observed
stabilizing
the observed the
1994 and 2011, Srensen, 2012. difference between actual and given positions of the observed
1994 and 2011, Srensen, 2012. visual
visual 3D
difference points
points in
3D between the
the image
inactual and given
image plane. The
The proposed
positions
plane. control
of the observed
proposed control
The mathematical
The mathematical validation
validation forfor the
the special
special structure
structure of of DP
DP visual 3D points is in based
the image plane. The proposed control
The mathematical validation design
visual
design 3Dalgorithm
points is
algorithm in based
the imageon
on the multipurpose
theplane. The proposed
multipurpose structure
structure controland
and
nonlinear
nonlinear
The control laws
control
mathematical laws on
validation thefor
on the for
the of
base
base
the
special
of nonlinear
nonlinear
special
structure of DP
asymptotic
asymptotic
structure of DP design algorithm is based on the multipurpose structure and
nonlinear control laws on the base of and nonlinear asymptotic includes
includes two stages.
design algorithm
two stages. At first
is based
At first stage
on stage we find
find visual
the multipurpose
we visual feedback
structure
feedback and
observers
nonlinear
observers controlis
is given laws on
given in
in theFossen
base of and
Fossen nonlinearStrand, 1999,
asymptotic
Strand, 1999, includes two stages. At first stage we find visual feedback
observers is and given in 2000.
FossenIn theseand papersStrand,sufficient
1999, control for
includes
control fortwo thestages.
the image At
image plane
plane and
firstand obtain
stage
obtain correspondent
wecorrespondent
find visual feedback desired
desired
Loria,
Loria, Fossen, Panteley, control for the
observers
Loria,
Fossen,
Fossen,
is and
and
given in 2000.
Panteley,
Panteley,
FossenIn these
2000. In
and papers
these
Strand,sufficient
papers
1999,
sufficient camera and
control
camera and
for the DP image
DP vessel plane
image
vessel motion.
plane
motion.
andAtobtain
andAtobtain
the
correspondent
the second
second stage the
correspondent
stage the desired
control
desired
control
conditions
Loria, of
of the
Fossen,
conditions andglobal
the asymptotic
Panteley,
global 2000. Instability
asymptotic are
are derived
these papers
stability and
sufficient
derived and camera and DP vessel motion. At the second stage the control
conditions of the global asymptotic stability are derived and law for
camera
law for and
given
given camera
DP camera velocityAttracking
vessel motion.
velocity tracking
the second by stage
by DP platform
DP platform
the control is
is
the possibility
conditions
the possibilityof the of independent
of global tuning
asymptotic
independent for
tuningstability observers and
are derived
for observers state
and
and state law for given camera velocity tracking by DP closed-loop
platform is
the possibility of independent tuning for observers and state designed.
designed.
law for given The resulting
The camera
resulting controller
controller
velocity trackingprovides
provides
by DP closed-loop
platform is
space
the
space control laws
possibility
control laws is proven
proven by
of independent
is by analogy
tuning with aa separation
for observers
analogy with separation
and state designed. The resulting controller provides closed-loop
space control laws is proven by analogy with a separation system
designed.
system asymptotic
asymptotic stability
stability and
The resulting and desirable
controller
desirable system
system dynamics
provides closed-loop
dynamics in
in
principle
space for LTI
control
principle for LTIlawssystems.
is proven by analogy with a separation
systems. system asymptotic stability and desirable system dynamics in
principle for LTI systems. the
the presence
system asymptotic
presence of
of external disturbances.
stability
external and desirable system dynamics in
disturbances.
principle for LTI systems. the presence of external disturbances.
The alternative
The alternative approach
approach to to DP
DP control
control systems
systems design
design is is the
the the presence of external disturbances.
The alternative
theory approach to
of multi-purposes
multi-purposes DP control
control systems design
laws synthesis,
synthesis, whichis the
was 2.
2. MODELS
MODELS OF OF SHIP
SHIP AND AND SCREEN
SCREEN IMAGE IMAGE
The alternative
theory of approach to DP control
control laws systems design which iswas
the 2. MODELS OF SHIP AND SCREEN IMAGE
theory
presented of in
multi-purposes
Veremei and control
Korchanov, laws 1989.
synthesis, which
Central ideas was
of 2. MODELS OF SHIP AND SCREEN IMAGE
theory
presented of in
multi-purposes
Veremei and control Korchanov, laws 1989.
synthesis, which
Central ideaswasof
presented
this approach in Veremei
are and Korchanov,
reflected on the modern1989. Central
level in ideas of
Veremey, Let
Let introduce
introduce two two coordinate
coordinate frames.
frames. First
First frame
frame Ox Ox s yy s zz s isis
presented
this approach in Veremei and Korchanov,
are reflected on the modern 1989. Central
level ideas of
in Veremey, Let introduce two coordinate frames. First frame Ox ss y ss z ss is
this
2010 approach
and areSotnikova
reflected on the modern level in Veremey, Let introduce
attached to to DP two
DP vesselcoordinate
vessel and and its frames.
its axes
axes haveFirst frame
have usual
usual directions
directions is
as
Ox ss y ss z ss as
this and 2013,
2010approach 2013,areSotnikova
reflected on and
andtheVeremey,
modern level
Veremey, 2013.
2013.inThe
The multi-
Veremey,
multi- attached
attached to DP vessel and its axes have usual directions as
2010
purpose and 2013,
control Sotnikova
laws and
synthesis Veremey,
is based on 2013.
the The multi-
optimization describedtoin
attached
described inDPFossen,
Fossen, 1994.
vessel 1994.
and itsTheThe second
axessecond
have usualcoordinate
directions
coordinate frame
frame as
2010
purpose andcontrol
2013, lawsSotnikova and isVeremey,
synthesis based on2013. The multi-
the optimization described in Fossen, 1994. Theand second coordinate frame
purpose
methods control
with thethelaws
goalsynthesis
to raise is based
raise the
the on the optimization
effectiveness and quality
quality OXYZ is
OXYZ
described is attached
in Fossen,
attached to
to the camera
1994.
the cameraTheand its
its axes
second are
are directed
coordinate
axes directedframe as
as
purpose
methods control
with laws
goal synthesis
to is based on the optimization
effectiveness and OXYZ is attached to the camera and its axes are directed as
methods
level of thewith
the the
systems goal
to beto
be raise
designed.the effectiveness and quality follows:
OXYZ isaxis
follows: axis OZ
attached coincide
to the camera
OZ coincide with optical
with optical axis
and itsaxis ofare
axesof aa camera
camera
directedand andas
methods
level of with the goal
systems to todesigned.
raise the effectiveness and quality follows: axis OZ OXcoincide OY with optical axis of a camera and
level of the systems to be designed. axis Ox
follows: ,
Ox saxis axes and
coincide with are in
optical the
axisnormalized
of a camera image and
level of the systems to be designed. axis , OZ OX
axes and OY are in the normalized image
The special
The special multipurpose
multipurpose structure
structure includes
includes somesome basic
basic part
part axis Ox ss , axes OX and OY are in the normalized image
The special separate
multipurpose structure includes for some axis
plane
planeOx of , axes
ofss the cameraOX and andtheir are in theare
OY directions normalized
the same image
the same as
as for
and several
and
The several
special separate
multipurpose elements
elements to be
to
structurebe adjusted
adjusted
includes for
somethebasic
the
part
specified
specified
basic part plane
the
of the
camera
camera
and their directions
andcorrespondently
are
their directions(Szeliski,
are the same as We
for
for
and several of
conditions separate
a elements
vessel to be These
motion. adjustedelements
for the specified
can the
planeaxes
the axesof the
Oy and
Oy ss camera
and Ozand
Oz their directions(Szeliski,
ss correspondently are the same 2011).
2011). as We for
and several of
conditions separate elements
a vessel to be These
motion. adjustedelements can be
for the specified be the axes Oy s and Oz s correspondently (Szeliski, 2011). We
conditions
switched of aoffvessel motion. These the elements can be will consider
the axes and the
that
Oy ss that the correspondently
origin
Oz ss origin for both (Szeliski, 2011).
both coordinates
coordinates framesWe is
switched on
conditions onofor as
as needed
or aoffvessel to
to provide
motion.
needed These the
provide best
best dynamical
elements can be
dynamical will
will
consider
consider
for frames is
switched
behaviour.
switched
on or main
behaviour.onTheThe
off asadjustable
or main
needed toelement
off asadjustable
provideofthe
needed toelement
provideofthe the best
thebest
dynamical
structure is
is so
dynamical
structure so the same.
will
the same.
consider The that
The camera
that
camera
the is
the is
origin
fixedfor
origin
fixed on both
for
on boardcoordinates
both
board of the
the DP
coordinates
of
framesand
DP vessel
vessel
framesand
is
is
behaviour. The main adjustable element of the structure is so the
can same.
rotate TheThe camera
only camera is
togetheriswith fixed
with on board
theonvessel.
vessel. of the DP vessel and
called dynamical
behaviour. The main
called dynamical corrector. This
adjustable
corrector. term
This element of
term of of control law
the structure
control provides
law provides is so the same.
can rotate only together fixedthe board of the DP vessel and
called dynamical corrector. This term of control law provides can rotate only together with the vessel.
called dynamical corrector. This term of control law provides can rotate only together with the vessel.
2405-8963
Copyright
Copyright 2015,
IFAC IFAC (International Federation of Automatic Control)
IFAC 2015
2015 184Hosting by Elsevier Ltd. All rights reserved.
184
Copyright
Peer reviewunder
IFAC responsibility
2015 184Control.
of International Federation of Automatic
Copyright IFAC 2015
10.1016/j.ifacol.2015.10.278 184
IFAC MCMC 2015
August 24-26, 2015. Copenhagen, Denmark
Evgeny I. Veremey et al. / IFAC-PapersOnLine 48-16 (2015) 184189 185

Let consider a traditional nonlinear 3DOF model of a DP point a corresponding to some object point A X a , Y a , Z a .
vessel in the following form The coordinates x a , y a and X a , Y a , Z a are related by
M D d ( t ), formula (3). Taking into account state vector (6), the
(1) dynamics of image points are given by:
R ( ) ,

where the vector u v p T represents velocities in a c L c , Z c , (7)


vessel-fixed frame, x y is a joint position x, y
T
where
and a heading angle vector relative to an Earth-fixed 2
xp 1
1 xp
frame. Vector E 3 implies a control action generated by Zp Zp
the propulsion system, vector d R 3 is provided by the yp Za
external disturbances of any nature. Matrices M and D with L 0 x p y p , Z c .

Zp Zc
the constant elements are positive definite, M M T . xa 1
2
1 xa
The only nonlinearity of the system (1) is determined by the Za Za

yaw angle orthogonal rotation matrix
Equations (7) must be considered together with the dynamics
cos sin 0 of depth coordinates Z c . From (4) we get

R ( ) R ( ) sin cos 0 . (2) L , Z ,
0 Z (8)
1
c z c c
0
1 0 xpZ p
Let us take into account that usually measurements of the where L c . As a result, equations (1),(2)
vessel velocities are not available for DP automatic system, 1 0 x a Z a
so any control law must be designed only on the base of and (7),(8) considered together represents a full mathematical
position and heading measurements. model in the dynamical positioning problem with visual
information. So, the mathematical model is given by
Now let describe the dynamic in the image plane. To this
end, consider a 3D point P X p , Y p , Z p representing, for M D d ( t ),
(9)
example, some point on the surface of the observed landmark R ( ) ,
or object of interest. The corresponding image plane point p c L c , Z c d c ( t ),
coordinates is given by: (10)
Z c L z c , Z c .
X p Yp
xp , yp . (3) Here d R 3 is an additional external disturbances of any
Z p Z p
nature. The measurement output is represented by vectors
Such point can be extracted and tracked in video sequence , c , Z c . The depth Z c is considered to be known from
using SIFT detector (Lowe, 2004). Using theoretical stereo vision or using nonlinear asymptotic observers
mechanics (Chaumette, 2006) we get following equations for (Szeliski, 2011).
the relative motion between camera (DP vessel) and 3D space
point: The goal is the construction of a nonlinear dynamic control
law of the form
X p v rZ p , Y p 0 , Z p u rX p . (4)
f ( , , , c , Z c , d ),
From (1) and (2) we can obtain: (11)
g ( , , , c , Z c , d ).
xp 1 2
1 x p u where E k is a state space vector of the controller,
x p Z Z
p p
. (5) 3
y yp v d E is the desired given position of the 3D points P and
p 0 x p y p r A in the image plane, that is the vector d represents desired
Z
p
object of interest position in the image plane of the camera.
Equations (5) represent a mathematical model describing The controller (11) must provide desirable equilibrium
image points motion induced by the camera (DP vessel) position, that is
motion. lim c t d . (12)
t
Notice that its not enough to use only one point, representing
The controller (11) must provide global asymptotical stability
object of interest, for DP visual control problem. Let consider
(GAS) for equilibrium position (12) and to ensure certain
the extended state vector:
desirable features of a closed-loop dynamics. Here we will
c xp yp xa T , (6) consider an integral action of control with respect to slow
varied components of the external disturbances d (t )
where x a is a horizontal coordinate of the second additional

185
IFAC MCMC 2015
186 Evgeny I. Veremey et al. / IFAC-PapersOnLine 48-16 (2015) 184189
August 24-26, 2015. Copenhagen, Denmark

and d (t ) , determined by the drift, current and wind loads. z v H ,


(21)
Besides that we will consider a filtering effect of the control F ( p ) , p d / dt , v Kz .
law regarding to wave disturbances.
Here z E 3 is a state vector of asymptotic observer,
3. DP-CONTROL BASED ON VISUAL IMAGE e z is an estimation error and K is the same matrix as
in the controller (19). The first equation in (21) represents
The stated control problem (9)-(12) is quite difficult for direct asymptotic observer, which stability is evidently guaranteed
solution. In this paper we propose two stages control design for any positive definite matrix H . Second equation
procedure. At first stage we use the velocity vector as a describes the transfer function model of dynamical corrector.
control input and design the local feedback control law Let state space model of this corrector is given by equations
d c , Z c , d (13) p p ,
(22)
for the subsystem (10). This controller must provide given p ,
equilibrium position (12) and its global asymptotic stability.
where matrixes , , , are related by equation
At the second stage we design a velocity tracking controller
d , d (14)
En s
c
1 F ( s ) , n c dim p .
in order to provide desired equilibrium point d for The last third equation in (21) is a control input, which is
closed loop system (9),(14) and its global asymptotic equal to the sum of basic control law of the form (19) and
stability. Besides that we will demand integral action and dynamical corrector output (22). It is important to note that
filtering properties for closed-loop systems (10),(13) and the asymptotic observer in control structure (21) is necessary
(9),(14) with respect to disturbances d (t ) and d (t ) . to use for integral action and filtering properties of the
control system.
Let now consider the first stage in details. So, we have
mathematical model Let investigate the stability, integral action and filtering
properties for the closed-loop system (18), (21). Taking into
c L c , Z c d c ( t ) , (15) account (18) and (21) we have following error dynamics:
where is considered as a control input. Let introduce error H d (t ). (23)
vector e ( t ) c t d . Then, from (15) we get Combining (21)-(23) we obtain closed loop system equations
e L c , Z c d c ( t ) . (16) H 0 0 E

Also, let consider an auxiliary control variable z H K z 0 d ( t ),
p (24)
v L c , Z c . (17) 0 p 0

v p .

If matrix L c , Z c is nonsingular for each time instant t K z

then for known vector v we can obtain corresponding The characteristic polynomial of this system is equal to
control input L c , Z c v . Taking into account (17) the
1
( s ) o ( s ) a ( s ) ( s ) ,
error dynamics (16) can be rewritten in the form
where o ( s ), a ( s ) are stable characteristic polynomials of
e v d c (t ) . (18) the closed loop system (18), (19) and asymptotic observer in
In the simplest way we can design linear feedback controller structure (21) correspondently. The last polynomial ( s ) is a
in the form characteristic polynomial of the dynamical corrector (22). As
v Ke (19) a result, the system (24) is stable if the polynomial ( s ) is
with any positive definite matrix K . In this case zero stable, that is matrix is Hurwitz.
equilibrium position of the closed-loop system (18), (19) is Now let consider the integral action of the control law (21)
asymptotically stable. Taking into account (17) the with respect to the constant disturbance vector d c ( t ) d c 0 .
corresponding feedback control law is given by
To this end, from (24) we can deduce corresponding steady-
L
1
c , Z c K c d . (20) state equations
Notice that the obtained controller (20) is a particular case of 0 v 0 H 0 , v 0 Kz 0 F ( 0 ) 0 , (25)
the general form (13). Its evident that the controller (20)
provides desired equilibrium position (12) in the absence of where v 0 , 0 , z 0 equilibrium position and 0 H 1d c 0 is
constraints. But, such simplest controller cant counteract to a static estimation error due to bias disturbance d c ( t ) d c 0 .
wind and wave disturbances. So, let us propose more From (25) the error vector e 0 0 z 0 is derived as follows
comprehensive controller represented by the following
multipurpose structure: e0 K
1
H K K 0 , K F ( 0 ), (26)
As a result, from (26) we can conclude that the zero steady

186
IFAC MCMC 2015
August 24-26, 2015. Copenhagen, Denmark
Evgeny I. Veremey et al. / IFAC-PapersOnLine 48-16 (2015) 184189 187

state error is provided if following condition is satisfied control law (31) provide desired equilibrium position and
guaranteed its global asymptotic stability for any diagonal
K K H .
positive definite matrix K d .
Using (22) this condition takes the form
But, the problem is how to implement control law (31)
1
K . (27) because the velocity vector can't be measured. In order to
So, if matrix is stable and matrix , , , satisfy estimate this vector let us introduce asymptotic observer and
control law in the following multipurpose form:
condition (27), then multipurpose control law (21) provide
closed-loop system stability and integral action relative to M z v Dz
T
R ( ) K 1 ( z ),
v
constant disturbance d c ( t ) d c 0 . (32)
z R ( ) z v K 2 ( z ),
Let now investigate the filtering effect which can be provided K d ( z v K v d ) F 2 ( p )( z ). (33)
by dynamical corrector. To begin with, we firstly transform
closed loop system (24) into transfer function form with the Here p d / dt , F 2 ( p ) is a transfer matrix of the LTI
external disturbance d c (t ) as input signal and velocity vector dynamical corrector. The equations (32) form a nonlinear
3 3
v as output one. The resulting model is given by asymptotic observer, where z E and z E are the
v P1 ( s ) P 2 ( s ) F ( s ) , (28) estimations of the vectors and correspondently. For the
where transfer matrices is defined by formulas full construction of the control law with the given structure
(32), (33), it is necessary to determine matrices K 1 , K 2 and
1 1
P1 ( s ) K ( E s K ) H , P2 ( s ) K ( E s K ) E. also a transfer matrix F 2 ( p ) with the aim to achieve the
Using mathematical model (28) we directly derive the mentioned design objective. The dynamical corrector must
following filtering condition also provide integral action and some filtering properties of
the control system.
F c ( j 0 ) 0 , F c ( s ) P1 ( s ) P 2 ( s ) F ( s ). (29)
Here 0 is a particular given frequency, which must be First of all let us consider an asymptotic observer and form
differential equations for error dynamics with respect to the
filtered. Consequently, from (29) we have equation
errors of observation (t ) (t ) z v (t ) and
1
F ( j 0 ) P2 ( j 0 ) P1 ( j 0 ) R I j . (30) (t ) (t ) z (t ) . On the base of (9) and (32) we have
Here R and I are real and imaginary part of the matrix T
M v D v R ( ) K 1 d ( t ),
F j 0 with the complex components. So, in order to (34)
R ( ) v K 2 .
provide filtering effect at the given frequency 0 the
dynamical corrector matrix F s must satisfy equation (30). Observe that with the assumption d ( t ) 0 system (34) has
So, the integral action and filtering effect is provided if desirable zero equilibrium position and the choice of the
dynamical corrector matrix F s satisfy to constraints (27) matrices K 1 and K 2 must provide it GAS and GES
and (30). In order to find specific conditions and explicit (globally exponentially stable). As it was proven in Fossen
formulas for the components of matrix F s it's necessary to and Strand, 1999, on the base of passivity property and in
fix the order of the dynamical filter and to solve system of accordance with Kalman-Yakubovich-Popov lemma, a
equations derived from (27) and (30). sufficient condition of desirable stability is a diagonal
structure and positive definiteness of the matrices K 1 and
4. MULTIPURPOSE SHIPS SPEED CONTROL K 2 . Let suppose, that the matrices with mentioned properties
Now let consider the mathematical model (9) of the DP are selected by the some way. So, the control algorithm (32),
vessel. The objective is to design control law in the form (33) provide desired vessel velocity d .
(14), which provide given velocity d of the vessel
Let consider the problem of stability for the equilibrium
produced by the algorithm (21) and (20).
position d in the closed-loop system (9),(32),(33). To
In accordance with the multipurpose structure control design begin with, we introduce the matrices 2 , 2 , 2 , 2 of a
procedure, the first question is a construction of the basic
state feedback control low of the form corrector such that F 2 ( s ) 2 ( E l s 2 ) 1 2 2 , which
allows to present a corrector model in normal form
*
K d ( K d ). (31) (7)
p 2 p 2 , 2 p 2 . (35)
It can be noted that if matrix K is such that
1 Here p E l is a state space vector of the corrector, E 3 is
K K d ( D K d ) , then the steady-state velocity in the
its output vector, l is a corrector order, E l is correspondent
closed-loop system (9), (31) is equal to the desired value in
the absence of disturbances, that is 0 d . As a result, the identity matrix. Assume that all the mentioned matrices have
constant components and that the matrix is Hurwitz.

187
IFAC MCMC 2015
188 Evgeny I. Veremey et al. / IFAC-PapersOnLine 48-16 (2015) 184189
August 24-26, 2015. Copenhagen, Denmark

Now remark that the variables z v , z can be excluded from At the end of the discussion let consider the integral action of
the equation (33) of the control signal , using observation dynamical corrector. Let us obtain a requirement to the
errors as follows transfer matrix F 2 ( s ) of dynamical corrector, which
guaranties the astatic property of a controller.
K d ( z v K v d ) F2 ( p ) K d ( K v d )
*
(36) To this end, suppose that there exists equilibrium position of
K d v F 2 ( p ) K d v p . the closed loop system (38) for certain external
Taking into account the observation error dynamics (34), disturbance d ( t ) d 0 . First of all, consider the observation
representation (35) and identity (36), the equations of the error equations for equilibrium position:
controller (32), (33) can be transformed to the form T
0 D v R ( 0 ) K 1 d 0 , 0 R ( 0 ) v K 2 , (40)
p 2 p 2 ,
where 0 is equilibrium value of a heading angle. Under the
T
M v D v R ( ) K 1 d ( t ),
(37) condition that
R ( ) v K 2 ,
D R ( 0 )K 1
T
*
K d v 2 p 2 . det 0
R ( ) K2
0
Now let consider the full system of differential equations for
closed loop system (9), (37), assuming d (t ) 0 . But, before the system (40) has a unique solution 0 h 1d 0 , 0 h 2 d 0 .
that let notice that the equation of velocity dynamic (9) can Here h 1 , h 2 are determined from equations (40).
be transformed to the form
Now let consider the controller equations (32), (33) at
M ( D K d )( d ) m , equilibrium position:
where m K d v 2 p 2 . If d is constant or slowly T
0 Dz v R ( 0 ) K 1 0 ,
varying vector then from the last equation we get
0 R ( 0 ) z v K 2 0 ,
1 1
d M ( D K d )( d ) M m .
K d ( z v K v d ) F ( 0 ) 0 .
Now the full system of equations for closed loop dynamics is
given by An evident transformation of these equations implies

1 1 T 1
v M D v M R ( ) K 1 M d ( t ),
D K d e D K d v 0 F2 ( 0 ) 0 R T ( 0 ) K 1 0 . (41)

R ( ) v K 2 , Taking into account the equilibrium values for vectors


0 , 0 , we can derive from (42)
p 2 p 2 , (38)

D K d e D K d h 1 F2 ( 0 ) h 2 R T ( 0 ) K 1h 2 d 0 .
1 1
e M ( D K d ) e M ( K d v 2 p 2 ), (43)
R ( ) . For any disturbance d ( t ) d 0 the linear system (43) has
Here e d is a velocity tracking error. It is convenient unique trivial solution e 0 if following requirement is
to introduce the following new vectors: fulfilled :

x1 p T
e
T T
, x 2 T T
T D K d h 1 F2 ( 0 )h 2 R T ( 0 ) K 1h 2 0.

As a result, the corrector matrix F 2 ( s ) must satisfy following


and to rewrite equations (38) as follows:
condition:
1 : x 2 A 0 ( ) x 2 , 2 : x 1 A 1 x 1 A 2 x 2 . (39)
F 2 ( 0 ) D K d h 1h 2
1 T
R ( 0 ) K 1 K 2 . (44)
Here matrices A 0 , A 1 and A 2 is given by
It is easy to see, that the simplest way to realize a requirement
2 0 0 2 (44) is to use corrector with no dynamics, so F 2 ( p ) K 2 .
A 1 1 , A 2 1


M D K d M 2
1 1
M 2 M K d
5. EXAMPLE OF VISUAL IMAGE BASED DP
M 1D M
1
R ( )K 1
T
A o ( ) .
R T ( ) K2 Let illustrate practical implementation of proposed idea on

the base of the DP control system for the vessel "Northern
System (39) has a zero equilibrium position. Its GAS in the Clipper" with the model (1), taken from Fossen and Strand,
absence of corrector is proved in Loria A., T. I. Fossen, 1999. Length of the vessel is L 76 . 2 m and mass is
and E. Panteley, 2000. If the dynamical corrector is used then 6
m 4 . 59 10 kg. Let accept following desired image points
matrix 2 must be Hurwitz as a sufficient condition for GAS
(Veremey E.I., 2013). coordinates d 20 30 25 T and wave disturbances with the
main frequency 0 0 . 455 rad / s . Correspondent transient

188
IFAC MCMC 2015
August 24-26, 2015. Copenhagen, Denmark
Evgeny I. Veremey et al. / IFAC-PapersOnLine 48-16 (2015) 184189 189

processes are shown on Fig. 1 (image points trajectories special constraint which guaranty offset-free steady state
(t ) ), Fig. 2 (linear velocity u and v of the vessel) and Fig. position. Nevertheless, wave filtering effect for this control
3 (control input ). As can be seen from figures, desired law must be investigated in the future work. In this study we
positions is achieved while the control signal is not represent object of interest in 3D space using only two points.
oscillating due to the dynamical corrector filtering effect. Of course, future research work should be related with the
generalization of the proposed approach for more number of
50
3D points and for under actuated vessels.

40
The reported study was partially supported by Russian
Foundation for Basic Research (RFBR), research project No.
30 14-07-00083.
(t)

20
REFERENCES
10 Srensen, A. J. (2011) A survey of dynamic positioning
control systems. Annual Reviews in Control, 35, pp. 23
0
0 200 400 600 800 1000 1200 136.
t (sec) Fossen T. I. (1994) Guidance and Control of Ocean Vehicles.
Fig. 1. Image points trajectories (t ) . John Wiley & Sons. New York.
Fossen T. I. (2011) Handbook of Marine Craft
4 Hydrodynamics and Motion Control. John Wiley &
u(t) Sons, Ltd.
linear speed (m/s)

2 v(t) Srensen, A. J. (2012) Lecture notes on marine control


systems. Technical Report UK-12-76, Norwegian
0 University of Science and Technology, 2012.
Fossen, T. I. and J. P. Strand. (1999) Passive Nonlinear
-2 bserver Design for Ships Using Lyapunov Methods:
Experimental Results with a Supply Vessel. Automatica,
-4
0 200 400 600 800 1000 1200 Vol. (35), No. (1), pp. 3-16.
t (sec) Loria A., T. I. Fossen, and E. Panteley. (2000) A Separation
Fig. 2. Linear speed of the vessel. Principle for Dynamic Positioning of Ships: Theoretical
7 and Experimental Results. IEEE Transactions of Control
x 10 Systems Technology, Vol. 8, No. 2, pp. 332-343.
5
1(t) Veremei E.I., Korchanov V.M. (1989) Multiobjective
stabilization of a certain class of dynamic systems.
0 2(t) Automation and Remote Control, 49 (9 pt 2), pp. 1210
(t)

3(t) 1219.
-5 Veremey E.I. (2010) Synthesis of multiobjective control laws
for ship motion. Gyroscopy and Navigation, 1 (2), pp.
119 125.
-10
0 100 200 300 400 500 Veremey E.I. (2013) Dynamical Correction of Positioning
t (sec) Control Laws // Proceedings of the 9th IFAC Conference
on Control Applications in Marine Systems (CAMS
Fig. 3. Transient process for control input . 2013). Osaka, Japan, September 17-20, 2013. P. 31-36.
Sotnikova M.V., Veremey E.I. (2013) Dynamic Positioning
6. CONCLUSIONS Based on Nonlinear MPC // Proceedings of the 9th
The problem of visual dynamic positioning was considered. IFAC Conference on Control Applications in Marine
Two stages control design scheme is proposed. At first stage Systems (CAMS 2013). Osaka, Japan, September 17-
we design control law for image plane trying to stabilize error 20, 2013. P. 37-42.
between actual and desired observed points position. At the Veremey E.I. (2012) H -Approach to Wave Disturbance
second stage the control law synthesis is performed in order Filtering for Marine Autopilots. Proceedings of 9th IFAC
to track desired vessel velocity. In both steps we use special Conference on Maneuvering and Control of Marine
multipurpose structure which allow us not only achieve Craft. Arenzano, Italy, September 1921.
design objective, but also provide integral action and wave Szeliski R. (2011) Computer Vision: Algorithms and
filtering effect for closed-loop system. In details we discuss Applications. Springer, 2011. 812 p.
tuneable parameters of this structure and found special Lowe D.G. (2004) Distinctive image features from scale-
requirements for the elements of this structure. invariant keypoints // International Journal of Computer
Vision. 2004. Vol. 2, No. 60, pp. 91-110.
In this paper we propose control law for image plane with Chaumette F., Hutchinson S. (2006) Visual Servo Control:
integral action and wave filtering. We also designed control Basic Approaches // IEEE Robotics & Automation
law for tracking given velocity. In this case we found a Magazine. 2006. Vol. 13, No. 4, pp. 82-90.

189

Anda mungkin juga menyukai