Professional Documents
Culture Documents
˙ 0 1 0 0 0 0
ξ1
−ω02n −2λwn ω0n
ξ˙˙2 0 0 0 0
ξ3 = 0 0 0 1 0 0
ξ˙4 2
0 0 −ω0e −2λwe ω0e 0 0
˙ 0 0 0 0 0 1
ξ5
Fig. 1. Vessel Reference frames ξ˙6 0 0 0 0 −ω02 −2λwψ ω0ψ
ψ
very slow for the class of offshore supply vessels, thus | {z }
ξ̇
| {z
Aw
}
is referred as Low Frequency (LF) dynamics); ξ1
0 0 0
ξ1
• wave-induced motion where the wave frequency ξ2 Kwn 0 0 ξ
ξ23
0 1 0 0 0 0
oscillatory motion induced by first-order waves is
ξ3 0
+ 0 0
w ; ηW F = 0 0 0 1 0 0
0 w
described by a seakeeping model, this dynamics is ξ4 0 Kwe
0 0 0 0 0 1
ξ4
ξ5 0 0 0 ξ5
referred as Wave Frequency (WF) dynamics. 0 0 Kwψ
| {z }
ξ6 Cw
ξ6
| {z } |
2.1 Manoueuvering Model
{z }
ξ Ew
(4)
The motion of a surface vessel for the development of a where the components of η W F are three linear ship wave
dynamic positioning system is described by a model which response models in surge, sway and yaw with state vector
is based on the common assumption that only horizontal ξ ∈ <6 , the terms Kwi = 2λwi ω0i σwi with i ∈ {n, e, ψ}
forces must be counteracted, as stated in Fossen [2011]. represent constant gains, where σwi are the wave intensi-
In this hypotheses the generalized velocity vector ν , ties, λwi are damping coefficients and ω0i are the domi-
T
[u, v, r] in the ship body-fixed frame {b} is considered, nating wave frequency for i ∈ {n, e, ψ}. These parameters
where u is the surge velocity, v is the sway velocity are estimated using the technique described in Fossen and
and r the yaw rate, see Fig. 1. Ship position is referred Perez [2009] when there are significant changes in heading
to the local geographical inertial North-east-down frame and at regular intervals of 20 minutes, which is the time
{n}, fixed to the Earth and described by the generalized period for which the sea state can be considered to be
T stationary. Assuming that the vessel is in position control
position η , [n, e, ψ] , where n and e are the ship position
mode the measured motion η m is recorded and detrended
in the {n} frame and ψ is the ship orientation referred to
to provide an off-line estimation of the wave induced mo-
the n frame.
tion η̂ W F . The white noise ww = [wwn , wwe , wwψ ]T ∼
Under the assumption of low-speed manoeuvring the N (0, Qw ) has with zero mean and covariance matrix Qw is
quadratic terms of velocity, Coriolis terms and non-linear estimated from the sample covariance of prediction errors
damping terms are neglected. With the recalled notation, η = η̂ W F − η W F , Fossen [2011].
the DP model is described by the following dynamics:
M ν̇ LF + Dν LF = τ c + τ env , (1) 2.3 Thruster Allocation
where ν LF is the low-frequency generalised velocity vector,
M is the rigid body generalised mass, which includes Marine vessels with n DOF are characterized by n gener-
inertia and added mass, D is the linear damping com- alized control forces τ c ∈ <n which are distributed among
ponent, τ c and τ env are the generalized control force and the r thrusters in terms of control inputs u ∈ <r :
environmental disturbances, respectively. The kinematic
equation has the following form: τ c = T (α)u (5)
cos ψ −sinψ 0 where u is the thrust force vector. The thruster configura-
η̇ LF = R(ψ)ν LF = sin ψ cosψ 0 ν LF , (2) tion matrix T (α) depends on the location and orientation
0 0 1 of the thrusters. The considered 3DOF offshore supply
where ν LF is the low-frequency ship position and the vessel has three azimuth thrusters, as shown in Figure 1.
matrix R(ψ) describes rotation from {b} frame to {n} Thruster configuration matrix T = T (α) follows:
frame. c(α ) 1 c(α ) c(α )2
3
T (α) = s(α1 ) s(α2 ) s(α3 )
2.2 Environmental Disturbances −ly1 c(α1 ) − lx1 s(α1 ) ly2 c(α2 ) − lx2 s(α2 ) lx3 s(α3 )
(6)
Environmental disturbances include both slowly varying Where c(αi ) and s(αi ) stand for cos(αi ) and sin(αi )
and high frequency forces. The slowly varying disturbances respectively. Referring to the thruster configuration matrix
include second order wave drift, ocean currents and wind in (6), Figure 1 shows that the allocation of the 3 azimuthal
forces. These effects are modelled in the control plant by thrusters is symmetrical with respect to the longitudinal
a bias b, described by the following dynamic equation: axis of the vessel. The input disturbances d are modelled as
slowly varying bias, described by the dynamics equation
ḃ = wb , (3)
ḋ = wd , where wd ∼ N (0, Qd ) is a white noise source.
where wb ∼ N (0, Qb ) is a white noise with zero mean and Therefore the inclusion of input disturbances transforms
covariance matrix Qb Fossen [2011]. This environmental the thruster allocation as follows:
disturbance force, referred to ship body frame {b} has the
form (Fossen [2011]): τ c = T (u + d), (7)
τ env = R(ψ)T b.
56
IFAC CAMS 2013
September 17-20, 2013. Osaka, Japan
2.4 Dynamic Positioning Continuous Control Model x(k) = AD (x(k − 1)) + B D u(k − 1) + E D w(k − 1)
y(k) = Hx(k) + wη (k)
Under the considered hypotheses above the considered with
dynamic positioning (DP) model has the following form: AD = e
A c Ts
Ts
Z
ξ̇ = Aw ξ + E w ww ,
Ac τ −1
η̇ LF = R(ψ)ν LF , BD = e dτ B = Ac (AD − I)B
0
ḃ = wb , Z Ts
(8) Ac τ −1
ḋ = wd , ED = e dτ E = Ac (AD − I)E
T 0
˙ = −Dν LF + T (u + d) − R(ψ) b + wνLF ,
M ν LF Z Ts
η m = η LF + η W F + wη . Ac τ AT
Ts2 Ts3
QD = e Qe c τ T
dτ = QTs + Ac Q + QAc
T
+ AQA
2 3
where wνLF ∼ N (0, Qν ) is a white noise process repre- 0
(11)
senting the model inaccuracies. Defining the state vector
Then the discrete time EKF has the form given in Jetto
x = [ξ T , η TLF , bT , dT , ν TLF ]T ∈ <21 and the measured ship et al. [1999] and Benetazzo et al. [2012].The implemen-
position η m ∈ <3 in the {n} frame as the superposition tation of the Kalman filter requires the estimation of the
of LF and WF dynamics, the DP model in (8) has the parameter of the model (11). For more details see Fossen
following non linear state space form: [2011] and Fu et al. [2010]. The ship motion η m in Eq. (8)
is measured by devices (GPS and compass). These sensors
do not work at the same frequency and not all measures are
A 0 0 0 0
0
w
57
IFAC CAMS 2013
September 17-20, 2013. Osaka, Japan
position and orientation has the form: Stable internal dynamics of this sliding surface is ensured
∆η = η ∗ − η m . (18) choice of the parameters Λ1 = diag (λ1,u , λ1,v , λ1,r ), Λ2 =
diag (λ2,u , λ2,v , λ2,r ) according to the following condition:
In order to solve the DP problem the dynamic posi-
2
tioning error defined in the {n} is conveniently refor- {{λ1,i , λ2,i }/ Roots p(δ) = δ + λ1,i δ + λ2,i < 1} ∀i ∈ {u, v, r}
mulated in the {b} frame in terms of velocity track- (24)
ing error. The velocity tracking error is defined as
the difference between the ship low frequency velocity
T Computation of the control law Given the DP discrete
ν LF (k) = [uLF (k), vLF (k), rLF (k)] and the reference
∗ ∗ ∗ ∗ T time model (15) and two-steps sliding surfaces (23) satis-
velocity ν LF (k) = [u (k), v(k) , r (k)] , obtained using fying (24), the following Discrete-Time Variable Structure
the inverse kinematic equation of (2) and the imposed ref- Control Law:
erence track (18) in the {n} frame and it has the following
form: τ ∗ (k) = τ ∗eq (k) + τ ∗n (k) (25)
∆ν LF (k) = ν LF (k) − ν ∗LF (k) (19)
58
IFAC CAMS 2013
September 17-20, 2013. Osaka, Japan
satisfies the Discrete-Time Sliding Mode Existence Condi- irregular waves. JONSWAP spectrum has been used for
tion (22), where the equivalent control term τ eq is given simulating the different sea conditions for the experiments,
by: stated in Table 1. Results of experiments are compared
τ ∗eq (k) = ν LF (k)− F̄ ν LF (k)−Λ1 ∆ν LF (k)−Λ2 ∆ν LF (k−1) (26) with Matlab/Simulink simulations, using the same setup
T and model, with MCSim software.
and the discontinuous control term τ n = τn∗u τn∗v τn∗r
is given by The performance index is considered for a quantitative
comparison between results of various simulations. The
ϑi (|si (k)| − %i ) if |si (k)| > %i
τn∗i (k) = i ∈ {u, v, r} performance index used here is the Integral of Squared
−si (k) + τn∗i (k − 1) if |si (k)| ≤ %i
Error ISE. For the DP system index depend on the
(27) error between position and orientation simulated and
with parameters |ϑi | < 1 with i ∈ {u, v, r} and % = position and orientation measured by sensors, namely
T
[%u %v %r ] given by: e(t) = ηdes (t) − η(t). The index expression is ISE =
−1 R tf
% = ∆F V + ∆Gb + ∆Gd + %∗ (28) (tf − t0 ) t0
T
e(t) e(t) dt.
Moreover, integral action for drift forces compensation is In the following, we show the comparison between the sim-
used to minimize the steady state position error ∆η, as ulated and experimental results, considering the following
stated in Loria et al. [2000]. case scenarios:
τ I (k) = τ I (k − 1) + KI Ts (η ∗LF (k − 1) − η LF (k − 1)) (29) Case 1 The Cybership III holds position in calm sea
In order to obtain zero steady-state errors, it is possible to for 1000 seconds after the waves are full developed and
sum the integral action (29) to DTVSC in the control law, then sails on a short straight movement.
see Fossen [2011]. The final control law is given combining Case 2 The Cybership III holds position in moderate
the DTVSC and Integral actions: sea for 1000 seconds after the waves are full developed
τ c (k) = τ ∗ (k) + τ I (k) (30) and then sails on a short straight movement.
Case 3 The Cybership III sails in moderate sea condi-
5. EXPERIMENTAL RESULTS tions, with two changes of reference position and head-
ings.
The 1:30 scale model of a naval surveillance vessel op- Case 4 The Cybership III holds position sails in the
erating in the North Sea,the CyberShip 3 (CS3), was same condition as in CASE 1 for a robustness test.
used for testing. This ship is located in a 40m x 6.45m CASE 4 refers to the following loading conditions: the
x 1.5m towing tank of the Marine Cybernetics Labora- vessel mass is increased by a 3.8kg load placed at the
tory (MCLab), at the Norwegian University of Science center of buoyancy to test the robustness of the approach
and Technology. The ship is equipped with two azimuth in different load conditions. Reference positions in the {n}
thrusters located at the aft and one at the bow, as shown in frame are shown in Table 2. Environmental conditions
Fig. 1. Thrusters positions are Ly1 = 0.11m, Ly2 = 0.11m, during tests, in terms of significant wave height Hs , modal
Lx1 = Lx2 = 0.789m and Lx3 = 0.636m. An azimuth frequency w0 are shown in Table 1. Results in term of ISE
thruster is installed in the bow. It has a mass of m = 75 kg, performance index are in Table 3. Results from CASE 1
length of L = 2.27 m and breadth of B = 0.4 m. Referring experiments and simulations are shown in Fig. 2. Results
to Fig. 1, the maximum thrust of the main thrusters is from the CASE 3 experiments simulations are shown in
21.9 N and the maximum thrust of the fore thruster is Fig. 3.
10 N. Ship devices and motor drives are controlled by an
onboard computer which uses a QNX real-time operating Table 1. Sea Conditions
system. The control system presented in this paper is Sea State Hs (m) ω0 (rad/s)
developed under Simulink/Opal RT using rapid proto- Calm [0 0.1] [1.11 1.8]
typing techniques and automatic code generation under Moderate [0.1 1.69] [0.74 1.11]
Matlab/Simulink and Opal on the host PC, built and High [1.69 6] [0.53 0.74]
downloaded to the onboard PC using a wireless Ethernet.
Internal variable are sent to the host PC by wireless com- Table 2. Reference positions and headings
munication, in a master/slave SW architecture. A Real- Case Time (s) n (m) e (m) ψ (deg)
Time Positioning System provides 6DOF real-time motion 1 0-1000 0 0 0
of the vessel, thus Earth-fixed position and heading. The 1 1000-2000 2 2 20
positioning system software processes the 500fps video 2 0-1000 0 0 0
stream from 3-cameras mounted on the towing carriage, 2 1000-2000 2 2 20
which is placed to the ship and the markers mounted on 3 0-800 2.5 1.5 arctan (n/e)
the vessel in the line of sight of the cameras. The cameras 3 800-1400s 3 1.5 arctan (n/e)
3 1400-2000 3 3.5 arctan (n/e)
emit infrared light and receive the light reflected from
the low-mass optical targets. To simulate the different sea Table 3. ISE performance index
conditions a wave maker system, produced by the Danish
Hydraulic Institute (DHI), is used. The wave maker is a 6 n (surge position) e (sway position) ψ (yaw rotation)
Sim. Exp. Sim. Exp. Sim. Exp.
meter width single paddle and is operated with a electrical Case 1 0.023 0.037 0.021 0.043 0.151 0.232
servo actuator. The system is equipped with a Active Wave Case 2 0.073 0.113 0.047 0.079 0.364 0.418
Absorption Control System (AWACS 2). The system has Case 3 0.057 0.078 0.052 0.072 0.153 0.214
a DHI Wave Synthesizer which can produce regular and Case 4 0.029 0.041 0.029 0.048 0.162 0.249
59
IFAC CAMS 2013
September 17-20, 2013. Osaka, Japan
REFERENCES
Jens G Balchen, Nils A Jenssen, Eldar Mathisen, and
Steinar Sælid. A Dynamic Positioning System Based
on Kalman Filtering and Optimal Control. Modeling,
Identification and Control, 1(3):135–163, 1980.
Flavia Benetazzo, Gianluca Ippoliti, Sauro Longhi, and
Paolo Raspa. Discrete Time Variable Structure Control
for the Dynamic Positioning of an Offshore Supply
(a) Simulated and experimented (b) Reference and experimented
Vessel. In Preprints of the 2012 IFAC Workshop on
path yaw rotation angle
Automatic Control in Offshore Oil and Gas Production,
2012.
Fig. 2. CASE 1 - Simulation (red dashed line) and experi- M.L. Corradini, G. Ippoliti, S. Longhi, and G. Orlando.
mental (blue line) position and headings using DTVS A quasi-sliding mode approach for robust control and
controller. speed estimation of PM synchronous motors. IEEE
Transactions on Industrial Electronics, 59(2):1096 –
1104, Feb. 2012.
Thor I. Fossen. Handbook of Marine Craft Hydrodynamics
and Motion Control. John Wiley and Sons, Ltd, Chich-
ester, UK, April 2011. ISBN 9781119994138.
T.I. Fossen and T. Perez. Kalman filtering for positioning
and heading control of ships and offshore rigs. IEEE
Control Systems Magazine, CST-29(6):32–46, 2009.
(a) Simulated and experimental (b) Simulated and experimental Mingyu Fu, Fuguang Ding, Meng Li, and Pengfei Yi.
path yaw rotation angle A nonlinear estimate filter designed for ship dynamic
positioning. In 8th IEEE International Conference on
Fig. 3. CASE 3 - Simulation (red dashed line) and experi- Control and Automation, pages 914–919, June 2010.
mental (blue line) position and headings using DTVS Katsuhisa Furuta. Sliding mode control of a discrete
controller. system. Systems and Control Letters, 14(2):145 – 152,
Experimental results show that the DTVS controller guar- 1990. ISSN 0167-6911.
antees a satisfactory performance with respect to the sim- L. Jetto, S. Longhi, and G. Venturini. Development
ulations controller in every case considered. Results of the and experimental validation of an adaptive extended
CASE 1 simulations, depicted in Fig. 2(b), show that kalman filter for the localization of mobile robots. IEEE
the DTVSC controller rapidly brings the system state Transactions on Robotics and Automation, 15:219–229,
nearby the sliding surfaces, while the integral action coun- 1999.
teract the 2-nd order slowly varying drift forces. While A. Loria, T.I. Fossen, and E. Panteley. A separation
Case 3 simulation results in Fig. 3(b) show that DTVS principle for dynamic positioning of ships: theoretical
control system performs smooth during reference position and experimental results. Control Systems Technology,
changes, notably for the yaw rotation angle. Results in IEEE Transactions on, 8(2):332 –343, mar 2000. ISSN
Table 3 show that the worse environmental conditions of 1063-6536.
CASE 2 respect to CASE 1 are reflected by far larger ISE M.C. Mora and J. Tornero. Path planning and trajectory
performance indices, but in all the cases the ISE index generation using multi-rate predictive artificial poten-
value is the same order of magnitude both for simulations tial fields. In IEEE/RSJ International Conference on
and experimental tests, corroborating the validity of the Intelligent Robots and Systems, pages 2990 –2995, sept.
approach. Robustness is tested comparing CASE 1 and 2008.
CASE 4: Table 3 shows that a change (5%) in the loading T. Perez, T.I. Fossen, and A.J. Sørensen. A discussion
conditions affects ISE indices by a 11.62% in sway position- about seakeeping and manoeuvring models for surface
ing, 10.81% in yaw positioning and 7.32% in yaw rotation. vessels. Mss-tr-001, Centre for Ships and Ocean Struc-
tures (CESOS), Norwegian University of Science and
6. CONCLUSION Technology NTNU, Trondheim, Norway, 2004.
Asgeir J. Sørensen. Structural issues in the design and
operation of marine control systems. Annual Reviews in
The problem of dynamic positioning plays a key role
Control, 29(1):125 – 149, 2005. ISSN 1367-5788.
in all those cases where it is not possible to anchor
Asgeir J. Sørensen. A survey of dynamic positioning
the ship at the seabed, or where the ship’s position
control systems. Annual Reviews in Control, 35(1):123
is bound to a specific point on the bottom. In this
– 136, 2011. ISSN 1367-5788.
paper an architecture for the dynamic positioning control
E.A. Tannuri and A.C. Agostinho. Higher order sliding
using sliding mode control and discrete-time models is
mode control applied to dynamic positioningsystems.
tested in an experimental setup. Simulations confirm the
pages 132 – 137, Rostock-Warnemünde, Germany, sep
robustness of the control scheme for calm and moderate
2010.
seas, corroborating simulation results. ISE performance
Vadim Utkin. Sliding modes in control and optimiza-
index value is selected to classify test results, same order
tion. Springer-Verlag, Berlin Allemagne, 1992. ISBN
of magnitude are found for simulations and experimental
9783540535164.
tests, corroborating the validity of the approach.
60