Professional Documents
Culture Documents
Adaptive Sliding Mode Output Feedback Control For Dynamic Positioning Ships With Input Saturation
Adaptive Sliding Mode Output Feedback Control For Dynamic Positioning Ships With Input Saturation
Ocean Engineering
journal homepage: www.elsevier.com/locate/oceaneng
A R T I C L E I N F O A B S T R A C T
Keywords: In this paper, a finite-time observer-based adaptive sliding mode output feedback controller is developed for
Dynamic positioning system dynamic positioning ship while the Coriolis matrix, hydrodynamic nonlinear damping matrix, input saturation,
Input saturation unmeasured states, unknown time-varying environment disturbances and transient performance are considered
Continuous finite-time observer
simultaneously. A continue finite-time state observer is proposed to estimate the unmeasured states of dynamic
Adaptive sliding mode control
Unmeasured state
positioning ship in finite time and an auxiliary system is constructed to handle with the control saturation. A
terminal sliding mode surface with the signal generated by the auxiliary system is designed while the good
transient performance is achieved under multiple constraints and the nominal output feedback controller pro
posed in this paper is chatter-free guaranteeing the steady performance. In additional, an adaptive update law is
introduced that the prior knowledge of disturbance is no need and the finite-time convergence of loop-system is
proved strictly. Finally, simulations are given to illustrate the effectiveness and advantages of the proposed
control scheme.
1. Introduction the DP controller are possibly constrained by the maximum forces and
moment that the propulsion system can produce. In (Du et al., 2016), Du
With the increasing reliance on oceans exploitation in a spectrum of Jialu et al. proposed a robust dynamic surface control strategy where an
human activities, the research for advance marine mechatronic systems auxiliary dynamic system to handle input saturation. Perez and Donaire
is attracting more and more attention. Dynamic positioning (DP) system proposed DP proportional-integral control, where disturbances and
is a sort of typical marine mechatronic system from a control perspective input saturation were handled by the integral action with anti-windup
for surface vessels and is essential for many marine applications, such as scheme(Perez and Donaire, 2009).
drilling, pipe-laying, and diving support(Shi et al., 2017). With the ad Considering the constraint of sensors in accuracy and cost, only the
vances of the nonlinear control, the DP nonlinear control has gradually positions of ships can be measured in practice while the positions and
gained much attention and various approaches have been used for DP velocities of ships are necessary in the control design for dynamic
systems to achieve better performance, such as robust control (Wang positioning ship (Fossen and Strand, 1999). Considering the dynamic
et al., 2018), hybrid control (Nguyen et al., 2007), fault-tolerant control positioning in the presence of unknown dynamic parameters of the ship,
(Chen M et al., 2016), etc.(Hu et al., 2015; Hu and Du, 2018; Do, 2011; unmeasured velocities and unknown time-varying disturbances, an
Veksler et al., 2016; Værnø et al., 2019; Zhang et al., 2018b). However, adaptive robust output feedback controller(Du et al., 2015) was devel
highly coupled control channels, unknown time-varying disturbances oped for the DP system by merging adaptive RBF neural networks and
and strong nonlinearity increase the design difficulty of the robust high-gain observer into the vectorial backstepping method. But
control for DP systems which makes it still an open topic (Sørensen, high-gain observer can only guarantee the boundedness of estimation
2011). error and can’t be analyzed in the frame of Lyapunov theory which is not
Input saturation, giving rise to degraded performance and even good to analysis the stability of closed-loop systems. In addition, if the
instability of the DP control system, is a potential problem for the DP Coriolis matrix and the nonlinear damping term are considered, the
system due to the fact that the commanded control inputs calculated by error system of observer will contain the nonlinear term related to the
* Corresponding author.
E-mail addresses: drliangkun@126.com (K. Liang), 2541070@sina.com (X. Lin), chenyu@zua.edu.cn (Y. Chen), Lijuan041@163.com (J. Li), dingfuguang@hrbeu.
edu.cn (F. Ding).
https://doi.org/10.1016/j.oceaneng.2020.107245
Received 22 August 2019; Received in revised form 9 March 2020; Accepted 10 March 2020
Available online 15 April 2020
0029-8018/© 2020 Elsevier Ltd. All rights reserved.
K. Liang et al. Ocean Engineering 206 (2020) 107245
Table 1
I Paremeters of CyberShip II
Xu 0.7225 Yv_ 10.0000
Yv 0.8612 Nr_ 1.0000
Yr 0.1079 Iz 1.7600
Nr 0.5000 Xu_ 2.0000
Nv 0.1052 Yr_ 0.0000
2
K. Liang et al. Ocean Engineering 206 (2020) 107245
3
K. Liang et al. Ocean Engineering 206 (2020) 107245
Where τmax i denotes the maximum control forces or moments that the Fðx1 ; x2 Þ ¼ FC ðx1 ; x2 Þx2 þ FD ðx1 ; x2 Þ (9)
ship’s propulsion system can produce, respectively. Let defining x1 ¼ η,
x2 ¼ Jυ, J ¼ Jðψ Þ and the Equations (1) and (2) can be rewritten as Without loss of generality, some assumptions are introduced first:
Where Fðx1 ; x2 Þ ¼ ðSðrÞ JM 1 ðCðJT x2 Þ þ DðJT x2 ÞÞJT Þx2 dðtÞ ¼ FC ðx1 ; x2 Þy ¼ FC ðx1 ; yÞx2 and kFC ðx1 ; x2 Þk � FCM kx2 k
JM 1 τd , Here, k⋅k stands for the determinant of a matrix. dM1 ,dM2 and FCM are
4
K. Liang et al. Ocean Engineering 206 (2020) 107245
unknown constants.
Assumption 2. Defining dðtÞ ¼ JM 1 τd and it’s unknown time-varying
yet bounded, that is, there exists unknown positive constants dj such that
kdðtÞk � dM
5
K. Liang et al. Ocean Engineering 206 (2020) 107245
and system states are bounded, the control input needed is bounded. unique in forward time for t 2 ½0; Tx ðx0 ÞÞ and lim xðt; x0 Þ ¼ 0. Here
t→Tx ðx0 Þ
Thus, the Assumption is bounded is reasonable. To satisfy this assump
Tx ðx0 Þ is called the convergence time (of the initial state x0 ). The equi
tion, the parameter σ can be large.
librium of (10) is finite-time stable if it is Lyapunov stable and finite-
The following lemmas and definition are given here to express the
time convergent. If U ¼ D ¼ <n , the origin is a globally finite-time
approaches to design finite-time controller and observer for dynamic
stable equilibrium.
positioning ship.
Lemma 1. (Bhat and Bernstein, 2000). Considering system (10), suppose
Definition 1. (Hong et al., 2002). Consider the following system
there exists a continuous function VðxÞ such that the following conditions
xðtÞ
_ ¼ f ðxðtÞÞ (10) hold:
6
K. Liang et al. Ocean Engineering 206 (2020) 107245
Vðx0 Þ1 α
T�
cð1 αÞ
Lemma 3. (Zhu et al., 2010). Consider system (10), and suppose that
there exist continuous function VðxÞ, scalars c、α 2 ð0; 1Þ and 0 < η < ∞
such that
_ �
VðxÞ cVðxÞα þ η
Then, the trajectory of system (10) is practical finite-time stable. In
addition, the time needed to reach such a neighborhood is bounded as.where
Vðx0 Þ is the initial value of VðxÞ.
Remark 2. It is clear from Lemma 3 that any asymptotically stable system
is practically finite-time stable. However, for some systems with absolutely
unknown disturbance such as system (1)~(2) in the presence of unknown
external disturbance, the asymptotical stability cannot be obtained easily.
Since the sysytem states can be only forced to converge into a neighborhood of
the sliding surface in finite time, so the finite-time convergeness is called
Fig. 14. The desired trajectory and actual trajectory between Scenario 4 and practical finite-time stablity.
Scenario 2.
7
K. Liang et al. Ocean Engineering 206 (2020) 107245
Fig. 15. The comparison of trajectory errors between Scenario 4 and Scenario 2.
Fig. 16. Comparison of positions’ estimation errors between Scenario 4 and Scenario 2.
Lemma 4. (Hu and Jiang, 2018). Consider the system (10). Suppose there 3. Main results
is a Lyapunov function VðxÞ, positive constants p1 2 ð0; 1Þp2 < p1 ,α >
0、β > 0 and γ > 0. Let V0 be the initial value of VðxÞ. Then, the following In this subsection, a finite-time observer-based adaptive sliding mode
statements hold. output feedback controller is designed to achieve the control object
aforementioned. In particular, the controller designed here can be
_ <
VðxÞ αVðxÞp1 þ βVðxÞp2 divided into four parts: an auxiliary is introduced to handle with input
Then the trajectory of the system (10) is finite-time uniformly ultimately saturation, an adaptive state feedback controller is designed assuming
bounded.The settling time is bounded as follows. all the states are known, a continuous finite-time state observer is con
. structed to estimate the state in finite time and an adaptive sliding mode
T � V01 p1 ½ðα θÞð1 p1 Þ� output feedback controller is integrated based the previous steps.
8
K. Liang et al. Ocean Engineering 206 (2020) 107245
Fig. 17. Comparison of velocities’ estimation errors between Scenario 4 and Scenario 2.
z2 ¼ x 2 x2d ξ (13)
Where x1d and x2d are desired trajectory and velcoties. In this paper, the
Fig. 18. The desired trajectory and actual trajectory in Scenario 5. sliding mode surface (Shao et al., 2017) is designed as following.
Fig. 19. Comparison of velocities’ estimation errors between Scenario 4 and Theorem 2.
9
K. Liang et al. Ocean Engineering 206 (2020) 107245
Where τ1 is a feedback term, τ2 is a compensated term for saturation and In this section, combing with the auxiliary system (11), adaptive
τ3 is an robust adaptive compensated term. k1 , k2 are design matrix and sliding mode state feedback controller(16)~(20), continuous finite-time
kd ,εd are design parameters. τ3 can compensate the unknown time- observer (21) and (22), an adaptive sliding mode controller is proposed
varying external disturbance and the upper bound information of here. Firstly, the modified sliding mode surface can be construted as.
disturbance is not required, which not only reduces the conservative bs ¼ bz 2 þ b
kβðbz 1 Þ (25)
problem caused by designing too high upper bound, but also can
effectively weaken the chattering. Where the state x1 , x2 in sliding mode surface (14) are replaced by
In order to prove the finite-time convergence of controller we x1 , b
estimation b x 2 in (21) and (22). At this point, the adaptive sliding
designed, three steps are provided here in Theorem 1: Firstly, the uni mode output feedback controller and adaptive law are designed as.
formly ultimate bounded of closed-loop system is proved so that the
boundedness of adaptive estimated error is garanteed; the finite-time bτ ¼ MJ T ðbτ 1 þ bτ 2 þ bτ 3 Þ (26)
convergence of sliding mode surface is achieved; finally, we will prove
that the position of dynamic positioning ship can be converged into the bτ 1 ¼ Fðx1 ; b
x2Þ _ z1Þ
kβðb k1 bs k2 sigr1 ðbs Þ þ x_2d (27)
small region around the desired trajectory.
bτ 2 ¼ Aξ Bξr0 σGsgnðξÞ (28)
Theorem 1. Considering the dynamic positioning ship model (5) and (6),
the proposed controller (16)~(19) and adaptive law (20) guarantees that x1 bs b
and x2 converge to small regions around the equilibrium points in finite time. bτ 3 ¼ dM (29)
kbs k
Proof. The proof of Theorem 1 is shown in the Appendix.
_
b
d M ¼ kd ð εd b
d M ðtÞ þ kbs kÞ (30)
3.3. Continuous finite-time state observer design In order to prove the finite-time ability, Theorem 3 is provided here.
Theorem 3. Under Assumption 1~Assumption 3, considering the dy
In most cases, measurements of the ship’s velocities are not avail
namic positioning ship (5) and (6) with unmeasured state, input saturation
able. Therefore, estimations of the velocities must be computed from
and unknown time-varying external disturbance, the proposed adaptive
noisy position and heading measurements through a state observer (Hu
sliding mode output feedback controller (26)~(29), finite-time observer (21)
et al., 2018; Hu and Niu, 2017; Liu et al., 2016). In the sequel, a
,(22) , auxiliary system (11) and adaptive law (30) can guarantee that x1
continuous finite-time observer is designed for dynamic positioning ship
and x2 converge to small regions around the equilibrium points in finite time.
to eatimate the velocities based on the position in finite time and the
following conclusion can be given. Proof. The proof of Theorem 3 is shown in the Appendix.
Theorem 2. Considering the dynamic positioning ship (5) and (6), the Remark 5. When auxiliary systems or anti-windup control is introduced to
finite-time observer is proposed. handle with the input saturation, a nominal controller is constructed without
r2 þ1
considering the input constraint. Since the time of saturation is short, it’s
x_ 1 ¼ bx 2
b L1 sigðb
x1 x1 Þ 2
(21) important to establish a nominal control approach such that good steady-
performance can be achieved. In this paper, a nominal adaptive sliding
x_ 2 ¼ Fðx1 ; b
b x 2 Þ þ Jðψ ÞM 1 τ L2 sigðb
x1 x1 Þr2 (22) mode output feedback controller is designed as.
_
~x_ 2 ¼ Fðx1 ; b
x 2 Þb
x2 Fðx1 ; x2 Þx2 dðtÞ L2 sigð~x1 Þr2 (24) b
d M ¼ kd ð εd b
d M ðtÞ þ kbs kÞ (34)
10
K. Liang et al. Ocean Engineering 206 (2020) 107245
are defined in (26)~(30). In addition, the estimation states are measured by 2011) and the observer can be designed as.
finite-time observer design in Theorem 2.
x_ 1 ¼ b
b x2 x1
L1 ðb x1 Þ (35)
Remark 6. It should point out that the nominal adaptive sliding mode
output feedback controller consisting the controller (31)~(34) and the x_ 2 ¼ Fðx1 ; b
b x 2 Þ þ Jðψ ÞM 1 τ L2 ðb
x1 x1 Þ (36)
observer (21), (22) is inherently continuous which means that it’s chatter–
free. The sign function is replaced by continuous function sig and the adaptive The parameters of controller and observer are chosen as
update law is introduced to estimate the upper bound of lumped disturbance, L1 ¼ diagf1:5; 1:5; 1:5g, L2 ¼ diagf200;200;200gr1 ¼ 0:7Θ ¼ 0:5k ¼
which means the chattering caused by the discontinuous function and the diagf30;30;30g,k1 ¼ diagf8;9;2g ,k2 ¼ diagf7;7;2g ,A ¼ diagf5;3;5g ,B ¼
unknown disturbance is effectively reduced. diagf2;2;5g.
Scenario 5. Adaptive sliding mode output feedback controller in
Remark 7. In this paper, the finite-time output feedback controller is Theorem 3 and Scenario 4 with model uncertainties.
designed for dynamic positioning with input constraint and unknown To prove the robust of the controller we designed, the disturbance of
disturbance. For further work, it will be interesting to consider the finite- model uncertainties (Fu et al., 2018) is introduced without adjusting the
time control with dynamic uncertainties and state constraint (Maryam, parameters of controller and observer in Scenario 2 and Scenario 4
2018; He et al., 2017, 2019; Xiao and Yin, 2016; Li et al., 2019). where the lumped disturbance can be design as:
0 1
4. Simulations 2 cosð0:5tÞcosðtÞ þ 0:3cosð0:5tÞsinð0:5tÞ 3
τd ¼ @ 0:01 sinð0:1tÞ A
In order to present the proposed control strategy in the previous 0:6 sinð1:1tÞcosð0:3tÞ
0 3 1
section work effectively, numerical simulations tested on a supply vessel ν þ 0:6u þ 0:1 sinðtÞ
called CyberShip II(Skjetne et al., 2005), which is a 1:70 scale-replica of þ @ 10ur þ u þ 0:1 sinðtÞ A
a supply ship whose main parameters is shown in Table 1, are performed 4ur þ 10v2 þ 0:1 sinðtÞ
and the results are shown in this section. Figs. 1-6 show the performance of nominal controller in Scenario 1
The actuator input limitation is set as jτimax j ¼ 50. The unknown for dynamic positioning ship without input saturation. The actual tra
external disturbances (Yu et al., 2018) τd with significantly large jectory and desired trajectory are showed in Fig. 1. During the initial
magnitude and high frequency, which can be roughly considered as stage, the nominal adaptive sliding mode controller we designed can
complex environment including ocean winds, waves, and currents, are ensure that the dynamic positioning ship track to the desired trajectory
governed by quickly with small overshoot which means that good transient perfor
0
2 cosð0:5tÞcosðtÞ þ 0:3cosð0:5tÞsinð0:5tÞ 3
1 mance is achieved. In addition, the tracking error in other stages can be
τd ¼ @ 0:01 sinð0:1tÞ A guaranteed even if the gradient of desired trajectory changes signifi
0:6 sinð1:1tÞcosð0:3tÞ cantly. The track errors of trajectory and velocities are showed in Figs. 2
and 3, respectively. It can be found that the trajectory error of the dy
The initial state of dynamic positioning ship is ηð0Þ ¼ ð0m0m0:5radÞT namic positioning ship can be maintained within 5 � 10 3 m while the
and the desired trajectory is designed as. heading angle can track the desired value quickly and the error can
�
xd ¼ 0:2t converge into 0:02radwithout any chattering phenomenon. The veloc
yd ¼ 10 sinð0:1tÞ ities errors can be maintained smoothly within 0:02m=s,0:01m=s and
0:02rad=s, respectively. Fig. 4 shows the control input without
This simulation contain three parts: the simulation for nominal constraint. Figs. 5 and 6 show the observer error of finite-time observer
controller, the simulation for controller considering the input saturation we proposed. It shows that the observer errors fall into zero approxi
and the contrasting case which shows the superiority of our scheme mately in a short time period. It can also be observed that x~1 converge to
comparing with the existing results. the acceptable accuracy of 1 � 10 4 mðradÞ within 1 s while the esti
Scenario 1. Nominal controller in Remark 5. mation error in velocities can be maintained within 5 �
Based on the controller proposed in Remark 5 and the parameters 10 3 m=s、5 � 10 3 m=s and 8 � 10 3 rad=s, respectively. In addition, no
can be chosen as L1 ¼ diagf10; 20; 20g,L2 ¼ diagf150; 50; 100g,r2 ¼ knowledge of external disturbance is required while only the bound
0:6,r1 ¼ 0:7Θ ¼ 0:2k ¼ diagf1:5; 0:8; 0:8gk1 ¼ diagf8; 9; 2g,k2 ¼ edness is requested to be guaranteed for the observer we designed.
diagf7; 7; 2g. From Figs. 7–12, the performance of nominal controller in Scenario
Scenario 2. Adaptive sliding mode output feedback controller in 2 for dynamic positioning ship in the present of input saturation are
Theorem 3. showed. The actual trajectories under controller in Theorem 3 are shown
Choosing the parameters for controller developed in Theorem 3 as in Fig. 7, from which we can see that the proposed controller can achieve
following L1 ¼ diagf10; 20; 20gL2 ¼ diagf150; 50; 100gr2 ¼ 0:6r1 ¼ fast transient response and high steady-state accuracy despite the pres
0:7r0 ¼ 0:6,Θ ¼ 0:5 k ¼ diag f30; 30; 30gk1 ¼ diagf8; 9; 2gk2 ¼ ence of external disturbances and input saturation, and lack of velocity
diagf7; 7; 2gA ¼ diagf5; 3; 5g,B ¼ diagf2; 2; 5g. measurements. In particular, Fig. 8 shows the track errors in surge, sway
Scenario 3. Adaptive sliding mode output feedback controller in and yaw direction, respectively, where the steady performance can be
Theorem 3 with worse ocean environment. seen. Figs. 9 and 10 show the input signal with constraint and signal
To prove the capability of anti-disturbance, the disturbance is generated by auxiliary system. Due to the fact that the disturbance is
enhanced as large as 10 times without regulating the parameters of non-zero whereas the control input is zero at the initial moment, the
controller and observer in Scenario 2.That is to say, the disturbance can saturation is activated and the auxiliary system handle with the effect
be designed as τ d ¼ 10*τd
_
caused by input saturation timely. When control input jτi j � 50NðN:mÞ ,
Scenario 4. Adaptive sliding mode output feedback controller based the nominal controller proposed in Remark 5 which guarantees the good
on a traditional observer in (Wondergem et al., 2011). performance can be achieved. Figs. 11 and 12 show the observer error
To illustrate the superiority of our approach, a controller based on a considering the input constraint, from which we can find that it doesn’t
conventional observer proposed in (Wondergem et al., 2011) is pro affect the estimation of observer and the observer still maintains accu
vided. Due to the input constraint and the rate of convergence aren’t rate estimation performance while the estimation error converge into
considered in (Wondergem et al., 2011), the controller is based on the 2 � 10 3 m=s, 1 � 10 3 m=s and 1 � 10 2 rad=s, respectively.
Theorem 3 while the observer is established as in (Wondergem et al., From Fig. 13, anti-disturbance performance of controller we
11
K. Liang et al. Ocean Engineering 206 (2020) 107245
designed is showed. Compared with the Scenario 2, the disturbance in this paper has a better trasient performance. In addition, the controller
this simulation is enhanced as larger as 10 times than that in Scenario 2 and observer provded in Theorem 3 have robust against the uncertainty
while the parameters of controller and observer are not changed. As while there is no need to change the parameters.
shown in Fig. 13, since the initial disturbance is very large while the
control input is constrained so that the overshoot is more obvious than 5. Conclusion
Fig. 7. Then, the auxiliary system is activated and the controller we
designed can make the ship following the desired trajectory within short In this paper, an adaptive sliding mode output feedback controller is
time. It’s illustrated that the controller provided in Theorem 3 can investigated for dynamic positioning ship considering the Coriolis ma
withstand worse environment while the parameters adjusting aren’t trix, hydrodynamic nonlinear damping matrix, input saturation, un
required. measured states, unknown time-varying environment disturbances and
From Figs. 14-17, a comparison simulation is provided to better show transient performance simultaneously. A continue finite-time state
the convergence speed and robustness of the proposed controller and observer is proposed to estimate the unmeasured states of dynamic
observer comparing with the existing works in (Wondergem et al., positioning ship in finite time and an auxiliary system is constructed to
2011). It can be found in Fig. 14 that the actual trajectory in Scenario 4 handle with the control saturation. A terminal sliding mode surface
has a significant oscilation in the initial stage and the convergence time consisting the signal generated by the auxiliary system is designed while
is longer than the resluts in Scenario 2 which is more obvious in Fig. 15. the fast convergence is achieved under multiple constraints. In addition,
In the case that the controllers are the same and the steady performance an adaptive update law is introduced that the prior knowledge of
are similar, the transient porformance is based on the observer. From disturbance is no need. Fast convergence rate of the closed-loop system
Fig. 16 and Fig. 17 we can find that a better trasient proformance is is ensured by analyzing system stability in light of finite-time control
achieved by the observer we designed in which the overshoot is less and theory. Numerical simulation results and comparison have shown the
the converbence time is shorter. Summarizing all the cases considered, advantages and improvements of the proposed control scheme. Further
the fast convergence rate of the proposed controller and observer have work includes low computational complexity algorithm to adjust pa
been verified. rameters and finite-time control design for the dynamic positioning with
From Fig. 18 and Fig. 19, a comparison with model uncertanties state constraints and dynamic uncertainties.
while the parameters aren’t adjusted between controller in Scenario 2
and controller based on traditional observer in Scenario 4 is showed. It Declaration of competing interest
can be found from Fig. 18 that the actual trajectory based on traditional
observer has a significant oscilation in the initial stage while the tra We declare that we do not have any commercial or associative in
jectory based on the Theorem 3 is more smooth. It’s obvious that the terest that represents a conflict of interest in connection with the work
controller and observer we designed have better robustness against to submitted.
uncertainties. In addition, the estimation errors comparsion between
two observers with uncertainties is provided from Fig. 19. We can find CRediT authorship contribution statement
that the observer we designed can estimate the accurate state within
shorter time more smoothly. In this thesis, the controller and observer Kun Liang: Conceptualization. Xiaogong Lin: Resources. Yu Chen:
we desinged have good robustness. Software. Juan Li: Data curation. Fuguang Ding: Investigation.
In conclusion, when the input saturation isn’t activated, the nominal
adaptive sliding mode controller we designed can ensure the good Acknowledgements
control performance of the dynamic positioning ship. When the control
input of the dynamic positioning ship exceeds the input constraint, the This research was partially supported by the Science and Technology
designed auxiliary system can quickly compensate for the control per Project of Henan Province (Project nos. 202102210137,
formance degradation caused by the saturation constraint. Even in the 192102210109), the Aeronautical Science Fund (Project nos.
worese environment, the ship can maintain good steady performance 20170155001) and the National Science Technology Support Program
and parameters adujsting is not required. At the same time, compared of China (Project nos. 51609046).
with the conventional state observer, the control method designed in
Appendix
A. Proof of Lemma 5
X
3 X
3
¼ ξT Aξ bi ξir0 þ1 ξT σ gi sgnðξi Þ ξT JM 1 Δτ
i¼1 i¼1
(38)
X
3
� λmin ðAÞkξk2 λmin ðBÞkξkr0 þ1 σ gi jξi j ξT JM 1 Δτ
i¼1
r0 þ1 r0 þ1
� 2λmin ðAÞV1 2 2 λmin ðBÞV1 2
12
K. Liang et al. Ocean Engineering 206 (2020) 107245
Where ξ0 is the initial state of ξ. From Lemma 1, the finite-time convergence of V1 and ξ can be guaranteed. Besides, the convergence time ta is
estimated by.
1 r0 r0 þ1
1 2λmin ðAÞV1 2 ðξ0 Þ þ 2 λmin ðBÞ
(39)
2
ta � ln r0 þ1
λmin ðAÞð1 r1 Þ 2 2 λmin ðBÞ
Thus this Proof is complete.
B. Proof of Theorem 1
Where d
~ M ¼ dM d M . Taking the time derivative for V2 .
b
1
V_ 2 ¼ sT s_ þ d~M d~_ M
kd
_
_ 1 ÞÞ
¼ sT ðz_2 þ k⋅βðz d~M b
dM (41)
X
3
¼ sT ðFðx1 ; x2 Þ þ τi þ dðtÞ _ 1 ÞÞ
x_2d þ Aξ þ Bξr0 þ σ GsgnðξÞ þ k⋅βðz kskd~M þ εd b
d M d~M
i¼1
Where α1 ¼ minf2λmin ðk1 Þ; kd εd g,β1 ¼ 12εd d2M . According to boundedness theorem (Zhou and Yang, 2015), s and d
~M are uniformly ultimate bounded.
Therefore, there exists a positive constant dM such that dM � dM holds.
~
Then, in order to prove the finite-time stability, we define a new Lyanupov function candidate.
1
V3 ¼ s T s (45)
2
The corresponding time derivative is
X
3
V3 � λmin ðk1 Þksk2 λmin ðk2 Þ jsi jr1 þ1 þ kskdM _
i¼1
X
3
r þ1 1 1 2
� λmin ðk1 Þksk2 λmin ðk2 Þ jsi j 1 þ ksk2 þ dM
i¼1
2 2
0 1
� � B r1 þ1 dM C
2
1 B 2 C r12þ1
� λmin ðk1 Þ ksk2 B2 λmin ðk2 Þ r1 þ1 CV3 (46)
2 @ 2 A
2V3
� �� !r 2þ1
2 1
dM
From Lemma 2, when the parameter satisfies λmin ðk1 Þ 1
2
0, V3 will converge to V3 � 12 2λmin ðk2 Þ
, so the sliding manifold will converge to
!r 1þ1
2 1
dM
ksk � 2λmin ðk2 Þ
in finite time.
Finally, combining with the Lemma 5, the finite-time stability for dynamic positioning ship in the present of input saturation can be guaranteed.
The Lyapunov function is chosen as V4 ¼ V1 þ V2 and the corresponding time derivative is.
13
K. Liang et al. Ocean Engineering 206 (2020) 107245
X
3
r þ1
V_ 4 � λmin ðk1 Þksk2 λmin ðk2 Þ jsi j 1 þ εd b
d M d~M
i¼1
(47)
λmin ðAÞkξk2 λmin ðBÞkξk r0 þ1
εd d~M b
d M ¼ εd d~M ðdM d~M Þ
� �
2 1 ~2 δd
� εd d~M þ d þ d2
2δd M 2 M (48)
εd ð2δd 1Þ~2 εd δd 2
¼ dM þ d
2δd 2 M
Hence, the inequality of (47) can be rewritten as.
X
3
V_ 4 � λmin ðk2 Þ jsi jr1 þ1 λmin ðBÞkξkr0 þ1
(49)
i¼1
εd ð2δd 1Þ~2 ε d δd 2
dM þ d
2δd 2 M
Defining rc ¼ minfr0 ; r1 g,σmin ¼ fλmin ðk2 Þ; λmin ðBÞg and V_ 4 can be rewritten as.
X
3
V_ 4 � λmin ðk2 Þ jsi jrc þ1 λmin ðBÞkξkrc þ1 þ εd b
d M d~M
i¼1
X
3 � �rc2þ1
εd ð2δd 1Þ~2
� λmin ðk2 Þ jsi jrc þ1 λmin ðBÞkξkrc þ1 dM
i¼1
2δd
(50)
� �rc2þ1
εd ð2δd 1Þ~2
þ dM þ εd b
d M d~M
2δd
� �rc2þ1
rc þ1 εd ð2δd 1Þ~2
� γ0 V4 2 þ dM þ εd b
d M d~M
2δd
2
rc þ1 rc þ1
Where kd ¼ εdδð2δ
dγ
d 1Þ,γ 0 ¼2 2 σ min .
εd ð2δd 1Þ~2
If 2δd dM > 1, we have.
� �rc2þ1
rc þ1 εd ð2δd 1Þ~2
V_ 4 � γ 0 V4 2 þ d M þ εd b
d M d~M
2δd
�
rc þ1
γ0 V4 2 þ
εd ð2δd 1Þ~2
d M þ εd b
d M d~M (51)
2δd
rc þ1 εd δd 2
� γ0 V4 2 þ dM
2
εd ð2δd 1Þ~2
If 2δd
dM � 1, we obtain.
�
� �rc þ1 �
εd ð2δd 1Þ~2 2 ��
dM �
2δd �
�εd ð2δd 1Þ~2
2δd dM �1
�
� �rc þ1 �
εd ð2δd 1Þ~2 2 ��
� dM � (52)
2δd �
�εd ð2δd 1Þ 2
2δd d~M >1
C. Proof of Theorem 2
14
K. Liang et al. Ocean Engineering 206 (2020) 107245
Where FCM1 and FCM2 are unknown constants. According the Assumption 1 we obtain.
kFD ðx1 ; x2 Þ FD ðx1 ; b
x 2 Þk � ðdM1 þ dM2 kx2 b x 2 kÞkx2 x2k
b
(56)
¼ dM1 k~x2 k þ dM2 k~x2 k2
Substituting (55) and (56) into (54) we have.
x2k
kFðx1 ; x2 Þx2 Fðx1 ; bx 2 Þb
� kFC ðx1 ; x2 Þx2 FC ðx1 ; bx 2 Þb
x 2 k þ kFD ðx1 ; x2 Þ FD ðx1 ; b
x 2 Þk
� 2FCM1 k~x2 k þ FCM2 k~x2 k2 þ dM1 k~x2 k þ dM2 k~x2 k2
¼ dM1 k~x2 k þ dM2 k~x2 k2 (57)
Where dM1 ¼ 2FCM1 þ dM1 ,dM2 ¼ FCM2 þ dM2 are unknown constants.
In order to analysis the stability of the observer error system, the following Lyapunov candidate function is constructed.
V5 ¼ X T PX (58)
00 1T 1T
� �
B r2 þ1
T p11 � I3 p12 � I3
Where error bector is X
~ ¼ @@sigðx
~1 Þ 2 A ; x ~2 C
A and P ¼ is a positive define matrix which means p11 > 0 and p11 p22 p212 > 0
p12 � I3 p22 � I3
0 0 10 1 1
r2 þ 1 @
r2 1
2 A@~
r2 þ1
A C
B diag j~x1 j x2 L1 sigð~x1 Þ 2
⋅B
@ 2 C
A
Fðx1 ; b
x 2 Þb
x2 Fðx1 ; x2 Þx2 dðtÞ L2 sigð~x1 Þr2
0 1
B r2 1 C
B jx
1 B 0 0
0 ~11 j 2 C
C
r2 1 B C
~1 j 2 A ¼ B C
r2 1
Where diag@jx B 0 x12 j
j~ 2 0 C .λmin ðL1 Þ, λmin ðL2 Þ are the minimal eigenvalues of L1 and L2 . Based on the defination of sigð ⋅Þ,
B C
B r2 1
C
B C
@ 0 0 jx
~13 j 2 A
we have.
r2 1 r2 þ1
sigð~x1 Þr2 ¼ j~x1 jr2 sgnð~x1 Þ ¼ j~x1 j 2
j~x1 j 2
sgnð~x1 Þ
r2 1 r2 þ1
(60)
¼ j~x1 j 2
sigð~x1 Þ 2
X3 r2 1 r2 þ1 B sigð~
x 1i Þ 2
C �
V_ 5 � j~x1i j 2 @sigð~x1i Þ 2 ~x2i A QB
@
C
A
~ dM1 k~x2 k þ dM2 k~x2 k2 þ dM pM kXk
þ pM kXk ~ (61)
i¼1
~x2i
Where pM ¼ maxðp12 ; p22 ÞQ11 ¼ r22þ1λmin ðL1 Þp11 þ λmin ðL2 Þp12 , Q22 ¼ r2 þ1
2 p12 , Q12 ¼ Q21 ¼ r24þ1λmin ðL1 Þp12 þ 12λmin ðL2 Þp22 r2 þ1
4 p11 ,
15
K. Liang et al. Ocean Engineering 206 (2020) 107245
� � � �
Q11 � I3 Q12 � I3 Q11 Q12
Q ¼ ,Q ¼ .
Q21 � I3 Q22 � I3 Q21 Q22
Then, the following inequalities hold.
1 ~ 2 � V5 � 1λmax ðPÞkXk
λmin ðPÞkXk ~ 2 (62)
2 2
1 1
V2 ~ � pffiffiffiffiffiV
2
r2 1 r2 1 r2 1
k~x1 k 2
� ð2=λmin ðPÞÞ 2
V5 2 (66)
Thus, from (61)–(66), the time derivative of the Lyapunov function (61) becomes.
r2 1 r2 1
V_ 5 � ð2=λmin ðPÞÞV5 2 λmin ðQÞkXk ~ 2 þ dM2 pM kXk
~ 2 þ dM1 pM kXk2 ~ 3 þ dM pM kXk
~
� � 1 � � 3 � �1
r2 þ1 r2 þ1
λmin ðPÞ λmin ðPÞ 2 32 λmin ðPÞ 2 12
� λmin ðQÞð2=λmin ðPÞÞ 2 V5 2 þ dM1 pM V5 þ dM2 pM V5 þ dM2 pM V5
2 2 2
r2 þ1
(67)
3 1
� γ1 V5 2 þ γ2 V5 þ γ3 V52 þ γ 4 V52
� � 1 � � 3
2
� � 1
2
r2 þ1
Where γ 1 ¼ λmin ðQÞð2=λmin ðPÞÞ 2 , γ 2 ¼ dM 1 p M λmin ðPÞ
2 , γ3 ¼ dM2 pM λmin ðPÞ
2 , γ4 ¼ dM pM λmin ðPÞ
2 . Since 0 < r2 < 1,we have r2 þ 1 < 2, that is to
1 r2 þ1
say < 2 2 < 1. According Lemma 1 and Lemma 4, one further obtains.
0 1
1
r2 þ1 B r2 γ 2 V52 þ γ 3 V5 þ γ 4 C
B C
(68)
1
V_ 5 � ðγ 1 θÞV5 2 θV52 BV52 C
@ θ A
1
r2 þ1
γ V 2 þγ V þγ
Here, let us define that X 2 Ω1 when V5 2 2 5 3 5
θ
4
> 0 and V5 ð0Þ is the initial value of V5 . In the light of Lemma 1 and Lemma 5, it implies that
when X 2 Ω1 , the erro system (23) and (24) is locally fast-finite-time uniformly ultimately bounded and the setting time is.
1 r2
V5 ð0Þ 2
To ¼ � � (69)
ðγ1 θÞ 1 2r2
D. Proof of Theorem 3
Proof. In order to analysis the stability of the observer error system, we define a Lyanupov function candidate V6 ¼ V4 þ V5 and the corresponding
time derivative is.
0 1
rc þ1 εd δd r2 þ1
B C 1
(70)
1
V_ 6 � γ 0 V4 2 þ 2
dM γ1 V5 2 þ @γ2 V52 þ γ 3 V5 þ γ 4 AV52
2
When X 2 Ω1 holds and choosing an arbitrary constant θ belonging to ð0; αÞ, the inequality (70) can be rewritten as.
rc þ1 εd δd r2 þ1
V_ 6 � γ 0 V4 2 þ 2
dM ðγ1 θÞV5 2 (71)
2
� �
rc þ1 r2 þ1
Defining γ 5 ¼ minfγ0 ; γ 1 θg,r3 ¼ min 2
; 2 and Δ ¼ 2
dM , so the following inequality holds.
εd δd 2
V_ 6 � γ4 V6r3 þ Δ (72)
According Lemma 3, the Theorem 3 is thus proved. This completes the Proof.□
16
K. Liang et al. Ocean Engineering 206 (2020) 107245
References Maryam, S.K., 2018. Anti-disturbance dynamic surface control scheme for a class of
uncertain nonlinear systems with asymmetric dead-zone nonlinearity. ISA (Instrum.
Soc. Am.) Trans. 81, 86–95.
Bhat, S.P., Bernstein, D.S., 2000. Finite-time stability of continuous autonomous systems.
Nguyen, T.D., Sørensen, A.J., Quek, S.T., 2007. Design of hybrid controller for dynamic
SIAM J. Contr. Optim. 38 (3), 751–766.
positioning from calm to extreme sea conditions. Automatica 43 (5), 768–785.
Chen, M., Ren, B.B., Wu, Q.X., et al., 2015. Anti-disturbance control of hypersonic flight
Perez, T., Donaire, A., 2009. Constrained control design for dynamic positioning of
vehicles with input saturation using disturbance observer. Sci. China Inf. Sci. 58 (7),
marine vehicles with control allocation. Model. Ident. Contr. 30 (2), 57–70.
1–12.
Sørensen, A.J., 2011. A survey of dynamic positioning control systems. Annu. Rev. Contr.
Chen, M., Jiang, B., Cui, R.X., 2016. Actuator fault-tolerant control of ocean surface
35 (1), 123–136.
vessels with input saturation. Int. J. Robust Nonlinear Control 26 (3), 542–564.
Shao, S., Zong, Q., Tian, B., et al., 2017. Finite-time sliding mode attitude control for
Do, K.D., 2011. Global robust and adaptive output feedback dynamic positioning of
rigid spacecraft without angular velocity measurement. J. Franklin Inst. 354,
surface ships. J. Mar. Sci. Appl. 10, 325–332.
4656–4674.
Du, J.L., Hu, X., Liu, H., Chen, C.L.P., 2015. Adaptive robust output feedback control for
Shen, Y.J., Xia, X.H., 2008. Semi-global finite-time observer for nonlinear systems.
a marine dynamic positioning system based on a high-gain observer. IEEE Trans.
Automatica 44 (12), 3125–3156.
Neural Netw. Learn. Syst. 26 (11), 2775–2786.
Shi, Y., Shen, C., Fang, H.Z., Li, H.P., 2017. Advanced control in marine mechatronic
Du, J.L., Hu, X., Krsti�c, M., Sun, Y., 2016. Robust dynamic positioning of ships with
systems: a survey. IEEE/ASME Trans. Mech. 22 (3), 1121–1131.
disturbances under input saturation. Automatica 73, 207–214.
Skjetne, R., Fossen, T.I., Petar, V., Kokotovi�c, 2005. Adaptive maneuvering, with
Fossen, T.I., Strand, J.P., 1999. Passive nonlinear observer design for ships using
experiments, for a model ship in a marine control laboratory. Automatica 41 (2),
Lyapunov methods: full-scale experiments with a supply vessel. Automatica 35 (1),
289–298.
3–16.
Veksler, A.V., Johansen, T.A., Borrelli, F., et al., 2016. Dynamic positioning with model
Fossen, T.I., 2012. Handbook of Marine Craft Hydrodynamics and Motion Control. Wiley,
predictive control. IEEE Trans. Contr. Syst. Technol. 24 (4), 1340–1353.
Hoboken, NJ, USA.
Værnø, Svenn, A., Brodtkorb, A.H., et al., 2019. Compensation of bias loads in dynamic
Fu, M.Y., Yu, L.L., 2018. Finite-time extended state observer-based distributed formation
positioning of marine surface vessels. Ocean Eng. 178, 484–492.
control for marine surface vehicles with input saturation and disturbances. Ocean
Wang, N., Qian, C., Sun, J.C., et al., 2016a. Adaptive robust finite-time trajectory
Eng. 159, 219–227.
tracking control of fully actuated marine surface vehicles. IEEE Trans. Contr. Syst.
He, W., Yin, Z., Sun, C., 2017. Adaptive neural network control of a marine vessel with
Technol. 24 (4), 1454–1462.
constraints using the asymmetric barrier Lyapunov function. IEEE Trans. Cybernetics
Wang, Y., Tuo, Y., Yang, S.X., et al., 2018. Reliability-based robust dynamic positioning
47 (1), 1641–1651.
for a turret-moored floating production storage and offloading vessel with unknown
He, W., Kong, L., Dong, Y., et al., 2019. Fuzzy tracking control for a class of uncertain
time-varying disturbances and input saturation. ISA (Instrum. Soc. Am.) Trans. 78,
MIMO nonlinear systems with state constraints. IEEE Trans. Syst. Man Cybernetics:
66–79.
Systems 49 (3), 543–554.
Wang, Y.Y., Gu, L., Gao, M., et al., 2016b. Multivariable output feedback adaptive
Hong, Y.G., Xu, Y.S., Huang, J., 2002. Finite-time control for robot manipulators. Syst.
terminal sliding mode control for underwater vehicles. Asian J. Contr. 18 (1),
Contr. Lett. 46 (4), 243–253.
247–265.
Hu, Q.L., Niu, G.L., 2017. Attitude output feedback control for rigid spacecraft with
Wondergem, M., Lefeber, E., Pettersen, K.Y., Nijmeijer, H., 2011. Output feedback
finite-time convergence. ISA Trans. 70, 173–186.
tracking of ships. IEEE Trans. Contr. Syst. Technol. 19 (2), 442–448.
Hu, Q., Jiang, B., 2018. Continuous finite-time attitude control for rigid spacecraft based
Xiao, B., Yin, S., 2016. Velocity-free fault-tolerant and uncertainty attenuation control
on angular velocity observer. IEEE Trans. Aero. Electron. Syst. 54 (3), 1082–1092.
for a class of nonlinear systems. IEEE Trans. Ind. Electron. 63 (7), 4400–4411.
Hu, Q., Zhang, J., Zhang, Y., 2018. Velocity-free attitude coordinated tracking control for
Yu, L., Guoqing, Z., Zhijian, S., et al., 2018. Adaptive cooperative formation control of
spacecraft formation flying. ISA (Instrum. Soc. Am.) Trans. 73, 54–65.
autonomous surface vessels with uncertain dynamics and external disturbances.
Hu, X., Du, J.L., Shi, J., 2015. Adaptive fuzzy controller design for dynamic positioning
Ocean Eng. 167, 36–44.
system of vessels. Appl. Ocean Res. 53, 46–53.
Zhang, G.Q., Huang, C.F., Wu, X.X., Zhang, X.K., 2018. Adaptive finite-time dynamic
Hu, X., Du, J.L., 2018. Robust nonlinear control design for dynamic positioning of marine
positioning control of full-actuated ship with servo system uncertainties. Acta
vessels with thruster system dynamics. Nonlinear Dynam. 94, 365–376.
Autom. Sin. 44 (10), 181–186.
Li, Hongyi, Zhao, Shiyi, He, Wei, Lu, Renquan, 2019. Adaptive finite-time tracking
Zhang, G., Huang, C., Zhang, X., et al., 2018. Practical constrained dynamic positioning
control of full state constrained nonlinear systems with dead-zone. Automatica 100,
control for uncertain ship through the minimal learning parameter technique. IET
99–107.
Control Theory & Appl. 12 (18), 2526–2533.
Liu, H., Zhang, T., Tian, X., 2016. Continuous output-feedback finite-time control for a
Zhou, J., Yang, J., 2015. Smooth sliding mode control for missile interception with finite-
class of second-order nonlinear systems with disturbances. Int. J. Robust Nonlinear
time convergence. J. Guid. Contr. Dynam. 38 (7), 1311–1318.
Control 26 (2), 218–234.
Zhu, Z., Xia, Y.Q., Fu, M.Y., 2010. Attitude stabilization of rigid spacecraft with finite-
time convergence. Int. J. Robust Nonlinear Control 21 (6), 686–702.
17