Ocean Engineering 197 (2020) 106864

Contents lists available at ScienceDirect

Ocean Engineering
journal homepage: www.elsevier.com/locate/oceaneng

Modeling and controller design of an active motion compensated gangway

based on inverse dynamics in joint space
Liang Lihua, Le Zhiwen ∗, Zhang Songtao, Li Jianfeng
College of Automation, Harbin Engineering University, Harbin 150001, China

ARTICLE INFO ABSTRACT

Keywords: An active motion compensated gangway is important and efficient equipment for personal transferring from
Active motion compensation a supply vessel to wind turbines or other structures, vice versa. This paper aims to explore the kinematics
Gangway and dynamics of a serial motion compensated gangway, and to design a controller of it. In this paper, the
Inverse dynamics control
prototype of a serial manipulator is taken as the research object. The kinematics and dynamics of the system
Serial manipulator
are established, respectively, and the Euler-Lagrangian equation is adopted. The supply vessel is regarded as
six degrees of freedom Dynamic Positioning (DP) model, and motions in seaway are analyzed. An improved
dynamics controller is adopted in this paper. Compared with PD controller with gravity, simulations are
carried out both in situations that the gangway tip moves with desired trajectories and docking situations.
Both simulation results reveal that the controller adopted in this paper perform well, and the steady state
errors are decreased significantly, which are desirable in practice.

1. Introduction the increasing complexity of the work, 1DOF of heave motion compen-
sation cannot solve the problem very well. Thus, Yingguang Chu et al.
Offshore turbine and other platforms have been widely used during presented an effective method of hoisting motion compensation and
the past decades. While suffering from the ocean waves, winds and anti-swaying flexible control in offshore crane operation based on the
currents, the working environments are usually worse than that of kinematics and energy dissipation principle of manipulator (Chu et al.,
mainland. Ships or offshore structures will move in six degrees of 2014). They established the integrated multi-domain system model,
freedom(6DOF) under the influences of ocean waves, winds and other
and model of accumulator in active heave motion compensation is also
disturbances, namely, sway, surge, heave, roll, pitch, yaw (Perez and
given. The simulation results show that the algorithm is effective both
Blanke, 2012; Liang et al., 2017b). As a consequence, it is a tough
in the case of motion compensation and anti-sway motion (Chu et al.,
work for transferring crews or other devices from a supply vessel to
the offshore structures, and vice versa. 2015, 2016).
Motion compensation technologies can be used to compensate the While recent years, personal transferring is becoming more and
vessel motions caused by the ocean disturbances. Umesh A. Korde more popular. Though the heave motion compensated crane performs
has studied a feasible method for heave compensation of deep water well in practice, it is not an advisable choice for personal transferring
drilling vessel under irregular wave excitation (Ulrich and Kumar, between supply vessel and wind turbines or other platforms. Researches
1991). The system makes use of the good interaction of the coupled on multi-DOF active motion compensated gangways are good solutions
oscillator to achieve the desired effect, and performs well in a large for the problem presented above. In order to compensate all 6DOF
frequency band. The dynamic model is used to study the performance of motions of a vessel in seaway, Ampelmann has developed an active
the system under certain operating conditions. The frequency domain motion compensation gangway system with six degrees of freedom
calculation of simple mathematical parameters and irregular spectrum motion platform as motion compensator, which known as the Stewart
shows that the heave compensation system is effective within the linear platform, as shown in Fig. 1. As stated by D.J.C. Salzmann in Salzmann
range. Nonlinearities of a system are common in reality, K.D. Do and J.
(2010) , the 6DOF motions of a vessel caused by ocean waves can be
Pan proposed a method to design a nonlinear controller for active heave
compensated by a 6DOF Stewart platform, and it can be accomplished
compensation system based on Lyapunov direct method and distur-
by adjusting the motion directions of the top platform with respect to
bance observer to improve the compensation effects. And the stability
the vessel motion, namely moving on the contrary directions of vessel
of closed-loop system is analyzed in details (Do and Pan, 2008b,a). With

∗ Corresponding author.
E-mail address: [email protected] (Z. Le).

https://doi.org/10.1016/j.oceaneng.2019.106864
Received 4 June 2019; Received in revised form 1 November 2019; Accepted 14 December 2019
Available online 6 January 2020
0029-8018/© 2019 Elsevier Ltd. All rights reserved.
L. Liang et al. Ocean Engineering 197 (2020) 106864

a 6DOF Stewart platform, and it is accomplished by adjusting the

motion directions of the top platform with respect to the vessel motion,
namely moving on the contrary direction of vessel motion. This is a
parallel motion compensation system. Although it can compensate the
motions of a vessel in six degrees of freedom, the drawback is that the
installation space required is very large. Since it is a parallel system,
the workspace is usually small. If the ship moves with large amplitude,
the corresponding 6DOF motion platform becomes a larger one, which
will lead to the increase of the costs and difficulties of maintenance.
Compared with the parallel motion compensation system, the serial
one is more flexible. One of the serial motion compensation systems is
as shown in Fig. 2. Raúl Guanche, Michele Martini and Alfonso Jurado
et al. proposed a feasible method for personnel transferring during
offshore operation (Guanche et al., 2016). The numerical analysis of
Fig. 1. A parallel active motion compensated gangway. docking operation between a vessel and wind turbine is carried out
by Wu (2014). The modeling of active motion compensation gangway
system and the design of PID controller are studied by Yu (2017).
However, the ship motions due to ocean waves are treated as regu-
lar, which is not universally applicable. Therefore, the active motion
compensation gangway system is taken as the research object in this
paper. This paper is organized as the following, firstly, system models
are established in Section 2, which consist of kinematic and dynamic
models, then the system Jacobian matrix and dynamic equation of
motion of the system are derived. Because the gangway system is
installed on the deck of the vessel, the 6-DOF model of the ship is
given, and the ITTC two-parameter spectrum is taken as the wave
disturbance in this part. The controller based on inverse dynamics is
designed in Section 3. Then the simulation models are built and carried
out in Matlab/Simulink and simulation results are given in Section 4.
According to the simulation results, the controller adopted in this paper
performs well both in docking and moving situations.

Fig. 2. A serial active motion compensated gangway.

2. Mathematics models of the active motion compensated gang-
way system

2.1. Reference frames

Before establishing the kinematics model of the system, we need to

establish the corresponding reference frames, as shown in Fig. 3. The
main frames of the whole systems studied in this paper are as follows:
( Inertial ) frame: The inertial frame is presented as {𝑁𝐸𝐷} =
𝑥𝑛 , 𝑦𝑛 , 𝑧𝑛 . Usually, the origin of inertial frame is fixed on earth, of
which the 𝑥𝑛 axis points to the true north, 𝑦𝑛 points to the east,
according to the right hand rule, 𝑧𝑛 is vertical downwards. Though
the {𝑁𝐸𝐷} frame is not a static frame to some degree which means
it moves with the earth rotation, it can be regarded as an inertial
frame with respect to maneuver and control of marine craft in limited
area (Fossen, 2011).
Body attached
( frame:
) The body attached frame can be presented as
{𝐵𝑜𝑏𝑦} = 𝑥𝑏 , 𝑦𝑏 , 𝑧𝑏 , it is a moving frame. The origin 𝑜𝑏 is attached at
the moving body. For a vessel, 𝑥𝑏 is from aft to bow, 𝑦𝑏 points to the
starboard, and 𝑧𝑏 is downwards based on the right hand rule.
Link attached frame: In order to facilitate the modeling of this
Fig. 3. Reference frames. gangway system, the link attached frame 𝑖 is established following the
rules below. 𝑧𝑖 is along to the joint axis, 𝑥𝑖 is parallel to the common
perpendicular, in case of intersecting of joint axis 𝑖 and 𝑖+1, then the 𝑥𝑖
motion. This is a parallel motion compensation system. Although it is perpendicular to the plane decided by the joint axis 𝑖 and 𝑖+1 (Craig,
can compensate the motions of a vessel in six degrees of freedom, the 2005).
drawback is that the installation space required is very large. Since it is And in this paper, the inertial frame is assumed coincide with the
a parallel system, the workspace is usually small. If the ship moves with body attached frame, and the body attached frame is moving with the
large amplitude, the corresponding 6DOF motion platform becomes a vessel while the inertial frame not.
larger one, which will lead to the increase of the costs and difficulties Based on the mechanical model shown in Fig. 3, the whole block
of maintenance. scheme of this gangway system with ship motion and sea waves can
It is well known that a 6DOF Stewart platform can move in 6DOF, be described as Fig. 4. The reference model is to calculate gangway tip
which has been widely used as flight simulator in flight simulations. position with ship motions due to sea waves. In Fig. 4, the reference
Therefore, as stated by D.J.C. Salzmann in Salzmann (2010), the 6DOF model is to calculate the gangway tip position with motions of the
motions of a vessel caused by ocean waves can be compensated by vessel.

2
L. Liang et al. Ocean Engineering 197 (2020) 106864

Fig. 4. Scheme of simplified active motion compensated gangway system.

Table 1 The relationship of velocities between gangway tip and joints can be
DH parameters of the active motion compensated gangway. obtained by the Jacobian matrix. For the whole gangway system in this
Link 𝑎𝑖−1 𝛼𝑖−1 𝑑𝑖 𝜃𝑖 paper, joint 1 and joint 2 are revolute joints, and joint 3 is prismatic
1 𝐿𝑥 0 𝐿1 + 𝐿𝑧 𝜃1∗ one, therefore, the gangway tip velocities can be obtained according to
2 0 𝜋/2 0 𝜃2∗ the Jacobians of the whole gangway system, and it can be expressed
3 0 3𝜋/2 𝑑3∗ 0
by:
4 0 0 𝐿4 0
𝝃 = 𝑱 𝒒̇ (4)
[ ]𝑇 [ ]𝑇
where 𝝃 = 𝝂 𝑏𝑝 𝝎𝑏𝑝 , and 𝑱 = 𝑱 𝜈 𝑱 𝜔 , and
2.2. Mathematical model of the mechanical system
( ) ( )
⎡ 𝑠1𝑠2 (𝐿4 + 𝑑3 ) −𝑐1𝑐2 (𝐿4 + 𝑑3 ) −𝑐1𝑠2⎤
The study of robot dynamics can be divided into two parts: kine- ⎢−𝑐1𝑠2 𝐿4 + 𝑑3 −𝑠1𝑐2 𝐿4 + 𝑑3 −𝑠1𝑠2⎥
⎢ ( ) ⎥
matics and dynamics. The former focused only on the geometrical of 0 −𝑠2 𝐿 + 𝑑 𝑐2 ⎥
𝑱 =⎢ 4 3
motion while the latter is the analysis of the forces causing the motion. ⎢ 0 𝑠1 0 ⎥
A serial active motion compensation gangway is shown in Fig. 2. ⎢ 0 −𝑐1 0 ⎥⎥
⎢
According to its structural characteristics, there are two revolute joints ⎣ 1 0 0 ⎦
and one prismatic joint in this gangway system, a revolute–revolute- Inverse kinematics is the opposite process of the forward. Thus, ac-
prismatic(RRP) spherical type manipulator model can be adopted for cording to Eq. (3), the tip position is derived with forward kinematics,
the gangway in this paper, which is shown in Fig. 3. Under this thus the joints variables can be obtained by the inverse solutions, which
simplification, the kinematics of the gangway system can then be can be expressed by:
derived. Four parameters are used to describe the motion compensation ( )
𝑦𝑏𝑡𝑖𝑝
gangway system, namely, Denavit–Hartenberg(DH) parameters (Craig, 𝜃1 = arctan
2005), and the joint variables are 𝜃𝑖 for revolute joint 𝑖 and 𝑑𝑖 for 𝑥𝑏𝑡𝑖𝑝 − 𝐿𝑥
prismatic one respectively, and the DH parameters can be obtained with ⎛ ⎞
respect to the link attached frames, and it can be stated in Table 1, ⎜ 𝑥𝑏𝑡𝑖𝑝 − 𝐿𝑥 ⎟
𝜃2 = arctan ⎜ ( )⎟ + 𝜋 (5)
where: 𝑳𝑥 and 𝑳𝑧 are the distances from origin of frame {1} to the
⎜ cos 𝜃1 𝑧𝑏𝑡𝑖𝑝 − 𝐿𝑧 − 𝐿1 ⎟
origin of {𝐵𝑜𝑑𝑦} in 𝑥𝑏 direction and 𝑧𝑏 direction, respectively, 𝑳1 and ⎝ ⎠
𝑳4 represent the lengths of link 1 and link 3, respectively. 𝜃1∗ , 𝜃2∗ and 𝐿𝑥 − 𝑥𝑏𝑡𝑖𝑝
𝑑3∗ are the joint variables of the gangway system. 𝑑3 = − 𝐿4
cos 𝜃1 sin 𝜃2
Thus the homogeneous transformation matrix between two con-
And the joints velocities can be derived using inverse Jacobian, and
nected links, and it can be expressed as the following, which is the
it can be expressed as:
homogeneous transformation matrix from link 𝑖 to 𝑖 − 1 :
⎡𝜃̇ 1 ⎤ ⎡𝜈𝑡𝑖𝑝𝑥 ⎤
⎡ 𝑐𝜃𝑖 −𝑠𝜃𝑖 0 𝑎𝑖−1 ⎤ ⎢𝜃̇ ⎥ = 𝐽 −1 ⎢𝜈 ⎥
⎢ ⎥ (6)
𝑠𝜃 𝑐𝛼 𝑐𝜃𝑖 𝑐𝛼𝑖−1 −𝑠𝛼𝑖−1 −𝑑𝑖 𝑠𝛼𝑖−1 ⎥ ⎢ ̇2 ⎥ 𝜈 ⎢ 𝑡𝑖𝑝𝑦 ⎥
𝑻 𝑖−1 = ⎢ 𝑖 𝑖−1 (1) ⎣𝑑3 ⎦ ⎣𝜈𝑡𝑖𝑝𝑧 ⎦
𝑖 ⎢𝑠𝜃𝑖 𝑠𝛼𝑖−1 𝑐𝜃𝑖 𝑠𝛼𝑖−1 𝑐𝛼𝑖−1 𝑐𝛼𝑖−1 𝑑𝑖 ⎥
⎢ 0 0 0 1 ⎥ The joints velocities are derived using the inverse Jacobian, singu-
⎣ ⎦
larity can be encountered. Thus, the pseudo inverse is adopted, and it
where 𝜃𝑖 , 𝛼𝑖−1 , 𝑎𝑖−1 and 𝑑𝑖 are the four DH parameters of the gangway can be described as:
system. ( )−1
𝑱+ = 𝑱𝑇 𝑱𝑱𝑇
Thus, based on the parameters in Table 1, the homogeneous trans-
formation matrix from gangway tip to {𝐵𝑜𝑑𝑦} frame can be obtained, In this case, the joints velocities can be obtained by the following:
and it can be described as: ( )
( ) 𝒒̇ = 𝑱 + 𝝃 + 𝑰 − 𝑱 + 𝑱 𝒃
⎡𝑐1𝑐2 −𝑐1𝑠2 𝑠1 𝐿𝑥 − 𝐿4 + 𝑑3 𝑐1𝑠2 ⎤
⎢ ( ) ⎥ where 𝒃 is a vector that can be chosen arbitrary.
𝑠1𝑐2 −𝑠1𝑠1 −𝑐1 − 𝐿4 + 𝑑3 𝑠1𝑠2 ⎥
𝑻 𝑏𝑝 = ⎢ ( ) (2) Lagrange dynamic equation is a method that based on system
⎢ 𝑠2 𝑐2 0 𝐿𝑧 + 𝐿1 + 𝐿4 + 𝑑3 𝑐2⎥
⎢ 0 0 0 1 ⎥ energy, which consist of potential and kinematic energy. The Lagrange
⎣ ⎦
equation and its differential form can be expressed as the following:
where 𝑐(∗) and 𝑠(∗) are 𝑐𝑜𝑠(∗) and 𝑠𝑖𝑛(∗) respectively.
𝑳=𝑲 −𝑷
Therefore, the gangway tip position expressed in {𝐵𝑜𝑑𝑦} is as the
𝑑 𝜕𝑳 𝜕𝑳 (7)
following: − = 𝝃̇
𝑑𝑡 𝜕 𝒒̇ 𝜕𝒒
⎡𝑥𝑏 ⎤ ⎡ 𝐿 − (𝐿 + 𝑑 ) 𝑐1𝑠2 ⎤ where 𝑳 is the Lagrangian, 𝑲 is the kinematic energy, and 𝑷 is the
⎢ 𝑡𝑖𝑝 ⎥ 𝑥 ( 4 )3 ( )𝑇
𝒐𝑏𝑝 = ⎢𝑦𝑏𝑡𝑖𝑝 ⎥ = ⎢ − 𝐿4 + 𝑑3 𝑠1𝑠2 ⎥ (3) potential energy, 𝒒 = 𝑞1 , 𝑞2 , … , 𝑞𝑛 is the joint variables, and 𝝃 =
( ) ( )𝑇
⎢𝑧𝑏 ⎥ ⎣𝐿𝑧 + 𝐿1 + 𝐿4 + 𝑑3 𝑐2⎥⎦
⎢
𝜉1 , 𝜉2 , … , 𝜉𝑛 is the generalized force.
⎣ 𝑡𝑖𝑝 ⎦

3
L. Liang et al. Ocean Engineering 197 (2020) 106864

For this 3 links gangway system, the total kinetic energy is then It is obviously from Eq. (14) that 𝑴 ̇ − 2𝑪 = 𝑼 𝑀 − 𝑼 𝑇 , while
𝑀
obtained by the following: 𝑼 𝑀 − 𝑼 𝑇𝑀 is a skew-symmetric matrix, which implies that 𝑴
̇ − 2𝑪 is a
skew-symmetric matrix.
1 ∑∑
3 3
1
𝑲= 𝑚 (𝒒) 𝑞̇ 𝑖 𝑞̇ 𝑗 = 𝒒̇ 𝑇 𝑴 (𝒒) 𝒒̇ (8)
2 𝑖=1 𝑗=1 𝑖𝑗 2 2.3. Ship motions in 6DOF
Then the total inertial matrix of this gangway system is obtained by
In reality, the motion compensated gangway system is installed on
combining the inertial matrix of each link, and it is derived as follows:
the deck, as shown in Fig. 2, the motions of the vessel should be
studied. The vessel in this paper is a surface vessel without considering
∑
3
( ) control forces, thus the seakeeping model is adopted here. As discussed
𝑴 (𝑞) = 𝑴(𝑞)𝑖 = 𝑑𝑖𝑎𝑔 𝑚11 , 𝑚22 , 𝑚33 (9)
in Fossen (2011), the seakeeping equations of motion are expressed in
𝑖=1
the following form:
where: 𝑑𝑖𝑎𝑔(∗) is the diagonal matrix, and the components are as
following: 𝜼̇ = 𝑱 (𝜼) 𝝂
(16)
( ) 𝑴 𝑠 𝝂̇ + 𝑪 𝑠 (𝝂) 𝝂 + 𝑫 𝑠 (𝝂) 𝝂 + 𝒈𝑠 (𝜼) = 𝝉 𝑤
𝑚11 = 𝐼1𝑧 + 𝐼2𝑦 + 𝐼3𝑧 (𝑐2)2
[ ( )2 ] where: 𝝂 = [𝑢, 𝑣, 𝑤, 𝑝, 𝑞, 𝑟]𝑇 is the velocity vector, and 𝜼 =
+ 𝐼2𝑥 + 𝐼3𝑥 + 𝑚𝑙2 𝑌𝑐2 + 𝑚𝑙3 𝑍𝑐3 + 𝑑3 (𝑠2)2 [𝑥, 𝑦, 𝑧, 𝜙, 𝜃, 𝜓]𝑇 represents the position vector, 𝑴 𝑠 = 𝑴 𝑅𝐵 + 𝑴 𝐴 is the
2
( ) (10) inertial matrix, 𝑪 𝑠 = 𝑪 𝑅𝐵 + 𝑪 𝐴 is the Coriolis-centripetal matrix, 𝑫 𝑠 ,
2
𝑚22 = 𝐼2𝑧 + 𝐼3𝑦 + 𝑚𝑙2 𝑌𝑐2 + 𝑚𝑙3 𝑍𝑐3 + 𝑑3 𝒈𝑠 and 𝜏𝑤 are the damping coefficient, restoring forces and wave forces
2
respectively.
𝑚33 = 𝑚𝑙3
The homogeneous transformation matrix between {𝐵𝑜𝑑𝑦} and
The whole system potential energy is obtained in the {𝐵𝑜𝑑𝑦} frame, {𝑁𝐸𝐷} frame can be expressed as the following:
for this 3 links gangway system, the whole potential energy can be
⎡𝑐𝜓𝑐𝜃 −𝑠𝜓𝑐𝜃 + 𝑐𝜓𝑠𝜃𝑠𝜙 𝑠𝜓𝑠𝜙 + 𝑐𝜓𝑐𝜙𝑠𝜃 𝑥⎤
obtained from the summation of each link, and it can be stated as: ⎢ ⎥
𝑠𝜓𝑐𝜃 𝑐𝜓𝑐𝜃 + 𝑠𝜃𝑠𝜓𝑠𝜙 −𝑠𝜃𝑐𝜓 + 𝑠𝜃𝑠𝜓𝑐𝜙 𝑦⎥
∑
3
( ) ∑
3
( ) 𝑻 𝑛𝑏 = ⎢ (17)
⎢ −𝑠𝜃 𝑐𝜃𝑠𝜙 𝑐𝜃𝑐𝜓 𝑧⎥
𝑃 = 𝑃𝑙 𝑖 = − 𝑚𝑙𝑖 𝑔0𝑇 𝑝0𝑐𝑖 ⎢ 0
𝑖=1 𝑖=1 ⎣ 0 0 1⎥⎦

where 𝑝0𝑐𝑖
represents the gravity center of link 𝑖 expressed in frame 𝑖. Further in Fossen (2011), for an arbitrarily point P in {𝐵𝑜𝑑𝑦} frame,
From the differential Lagrange equation (7), the system dynamics can the system’s velocity transformation matrix can be derived from:
be obtained as the following: 𝝂 𝑏𝑝 = 𝝂 + 𝝎 × 𝒓𝑏𝑝
( ) (18)
∑
3 ∑
3 ∑
3
= 𝝂 + 𝑺 𝑇 𝒓𝑏𝑝 𝝎
𝑚𝑖𝑗 (𝑞) 𝑞̈𝑗 + ℎ𝑖𝑗𝑘 (𝑞) 𝑞̇ 𝑘 𝑞̇ 𝑗 + 𝑔(𝑞)𝑖 = 𝜉𝑖 (11)
𝑗=1 𝑗=1 𝑘=1 where: 𝑺(∗) is the skew symmetric matrix. And the system transforma-
𝜕𝑚𝑖𝑗 tion matrix is defined as:
1 𝜕𝑚𝑗𝑘 [ ( )] [ ( )]
where: ℎ𝑖𝑗𝑘 (𝑞) = − , (𝑖, 𝑗, 𝑘 = 1, 2, 3), which are termed the ( ) ( )
𝜕𝑞𝑘 2 𝜕𝑞𝑖 𝑰 3×3 𝑺 𝑇 𝒓𝑏𝑝 𝑰 3×3 𝑺 𝒓𝑏𝑝
𝑯 𝒓𝑏𝑝 = , 𝑯 −1 𝒓𝑏𝑝 =
first type Christoffel symbols. 𝟎3×3 𝑰 3×3 𝟎3×3 𝑰 3×3
The system dynamic equation expressed in Eq. (11) can be rewritten
Therefore, transformation of linear and angular velocity vectors
in a common vector form and it can be stated as below:
between two points with respect to in can be expressed by:
𝑴 (𝑞) 𝑞̈ + 𝑪 (𝑞, 𝑞)
̇ 𝑞̇ + 𝒈 (𝑞) + 𝑭 = 𝝃 (12) [ 𝑏] [ ]
𝝂𝑝 ( ) 𝝂𝑏
𝑏 𝑜
where: 𝑴 (𝒒) is the system inertial matrix, which is positive definite = 𝑯 𝒓𝑝
𝝎𝑏𝑛𝑏 𝝎𝑏𝑛𝑏
symmetric matrix; 𝑪 (𝒒, 𝒒)̇ is the Coriolis-centripetal matrix, 𝒈 (𝒒) is
the system restoring term, F is the disturbance, and 𝜉 represents the This is a useful property for computing the gangway tip velocities
generalized force. In Eq. (12), the restoring term 𝒈(𝒒) can be acquired due to wave disturbances.
by differential equation of system potential energy, and it can be
described as the following: 2.4. Random ocean waves

⎡ 0( ⎤ The ocean waves are random not only in time, but in space, and
∑ ⎢ ( )) ⎥
3
𝜕𝑃 the wave elevation presents irregular characteristics in any particular
𝒈 (𝒒) = = ⎢𝑔 𝑚𝑙2 𝑌𝑐2 + 𝑚𝑙3 𝑞3 + 𝑍𝑐3 𝑠2⎥ (13)
𝑖=1
𝜕𝑞𝑖 ⎢ ⎥ sea state (Pérez and Blanke, 2002). However, it is commonly used and
⎣ −𝑔𝑚𝑙3 𝑐2 ⎦
convincing to describe sea waves by their Power Spectral Density(PSD)
As stated in Lebret et al. (1993), the Coriolis-centripetal matrix in under the assumption that the sea elevation performs as a realization of
Eq. (12) is an 𝑛 × 𝑛 matrix that can be derived using the following a stationary and homogeneous, zero mean Gaussian stochastic process.
equation: The ITTC recommended 2-parameter spectrum is adopted in this paper.
1[ ̇ ] For fully developed wave, the PSD can be expressed by:
𝑪= 𝑴 (𝒒) + 𝑼 𝑇𝑀 − 𝑼 𝑀 (14)
2 173ℎ21∕3 ( )
where: 𝑼 𝑀 is a matrix that can be obtained by the following equation: −691
𝑆𝑤 (𝜔) = 𝑒𝑥𝑝 (19)
𝑇 4 𝜔5 𝑇 4 𝜔4
( ) 𝜕𝑴 𝑇
𝑼 𝑀 = 𝑰 𝑛 ⊗ 𝒒̇ 𝑇 where: 𝑆𝑤 is the PSD of random wave, 𝜔 is the angular frequency, 𝑇 is
𝜕𝒒
the average period of the wave, and ℎ1∕3 is the significant wave height.
where: ⊗ represents the Kronecker product, and satisfied the following Thus, a long-crested irregular sea wave elevation can be written as the
equations: sum of harmonic components (Fossen, 2011):
[ ]𝑇
𝜕𝑴 𝑇 𝜕𝑴 𝜕𝑴 𝜕𝑴 ∑
𝑁
( )
= ∈ 𝑅(𝑛⋅𝑛)×𝑛
𝜕𝒒 𝜕𝑞1 𝜕𝑞2 𝜕𝑞3 𝜀= 𝜀𝑘 cos 𝜔𝑘 + 𝜙𝑘
(15) 𝑘=1
𝒒̇ 𝑇
) ⎡
0 0⎤ (20)
( ∑𝑁 √
( ) ( )
𝑰 𝑛 ⊗ 𝒒̇ 𝑇 = ⎢ 0 𝒒̇ 𝑇 0 ⎥ ∈ 𝑅𝑛×(𝑛⋅𝑛)
⎢ ⎥ = 2𝑆 𝜔𝑘 𝛥𝜔 cos 𝜔𝑘 + 𝜙𝑘
⎣0 0 𝒒̇ 𝑇 ⎦ 𝑘=1

4
L. Liang et al. Ocean Engineering 197 (2020) 106864

Fig. 5. Scheme of the inverse dynamics controller.

Fig. 6. Scheme of the whole gangway system with controller.

where: 𝜀𝑘 is the random phase angle of wave component 𝑘. which are vital important for system modeling. In the real cases,
these unknown parameters are hard to acquire due to some nonlinear
3. Controller design characteristics of the system. And as a result, the control efficiency will
decreased or even becomes worse. Therefore, to avoid these drawbacks
3.1. Inverse dynamics control above, the system control input in Eq. (22) can be reelected as the
following:
The inverse dynamics control has been used in the parallel robot
̂ (𝒒) 𝒖 + 𝒏̂ (𝒒, 𝒒)
𝝃=𝑴 ̇ (24)
(Khudher and Powell, 2016; Liao et al., 2017). In this work, we applied
it into the serial manipulator. where: 𝑴̂ and 𝒏̂ are the dynamic estimation terms of 𝑴 and 𝒏 of the
According to Eq. (12), the gangway system dynamics can be rewrit- system, respectively. And the estimation errors can be expressed as the
ten as following: following:
𝑴 (𝒒) 𝒒̈ + 𝒏 (𝒒, 𝒒)
̇ =𝝃 (21) ̃ =𝑴
𝑴 ̂ −𝑴 𝒏̃ = 𝒏̂ − 𝒏
where: 𝒏 (𝒒, 𝒒) ̇ + 𝒈 (𝒒) + 𝑭 is an integral term of the system.
̇ = 𝑪 (𝒒, 𝒒) Combining Eqs. (21) and (24), we can obtain:
Thus, the control input can be chosen as:
𝑴 (𝒒) 𝒒̈ + 𝒏 (𝒒, 𝒒) ̂ (𝒒) 𝒖 + 𝒏̂ (𝒒, 𝒒)
̇ =𝑴 ̇ (25)
𝝃 = 𝑴 (𝒒) 𝒖 + 𝒏 (𝒒, 𝒒)
̇ (22)
It is obviously that 𝑴 is a reversible diagonal inertial matrix, thus,
By substituting Eq. (22) into Eq. (21), we can obtain:
Eq. (25) can be rewritten in the following form:
𝑴 (𝒒) (𝒒̈ − 𝒖) = 0 ( )
𝒒̈ = 𝑴 −1 𝑴𝒖 ̂ + 𝒏̃ − 𝒏
( )
Since 𝑴(𝑞) above is a diagonal matrix, and it is reversible, which = 𝒖 + 𝑴 −1 𝑴 ̂ − 𝑰 𝒖 + 𝑴 −1 𝒏̃ (26)
results in 𝒒̈ = 𝒖. Here, the gangway system input 𝝃 is linear and =𝒖−𝜸
decoupled under the new input 𝒖, which implies that each component ( )
in 𝒖 is only influenced by the joint 𝑞𝑖 . Therefore, with the desired joint where: 𝜸 = 𝑰 − 𝑴 −1 𝑴 ̂ 𝒖 − 𝑴 −1 𝒏.
̃ Thus, the new system control input
[ ]𝑇 of 𝒒̈ can be chosen as:
variables 𝒒 𝑑 = 𝑞1𝑑 ... 𝑞𝑛𝑑 , the system control input can be chosen
as: 𝒖 = 𝒒̈ 𝑑 + 𝑲 𝐷 𝒆̇ + 𝑲 𝑃 𝒆 (27)
𝒖 = 𝒒̈ 𝑑 + 𝑲 𝐷 𝒆̇ + 𝑲 𝑃 𝒆 (23) In this case, the system error equation is obtained and it can be
According to Spong and Ortega (1990), 𝑲 𝑃 and 𝑲 𝐷 in Eq. (23) are expressed as:
chosen as positive definite diagonal matrices, which ensures that the
𝒆̈ + 𝑲 𝐷 𝒆̇ + 𝑲 𝑃 𝒆 = 𝜸 (28)
closed-loop system is linear, decoupled, and exponentially stable, and
the closed-loop damping ratio and natural frequency can be arbitrarily In the system error equation, 𝜸 is nonlinear and coupled. The
assigned. convergence of system errors depends not only on the left part of
Eq. (28), but also the uncertain term 𝜸. Assuming that the desired joint
3.2. Improved inverse dynamics control is 𝒒 𝑑 , then the system error can be stated as:
𝒆 = 𝒒𝑑 − 𝒒
Although the inverse dynamics control is easy to guarantee that
the closed-loop system is linear, decoupled, and exponentially stable 𝒆̇ = 𝒒̇ 𝑑 (29)
by choosing proper positive definite diagonal matrices 𝑲 𝑃 and 𝑲 𝐷 , it 𝒆̈ = 𝒒̈ 𝑑 − 𝒒̈ = 𝒒̈ 𝑑 − 𝒖 + 𝜸
is difficult to accomplish since unknown parameters exist in reality,

5
L. Liang et al. Ocean Engineering 197 (2020) 106864

In order to get 𝜀 in Eq. (32), the direct Lyapunov method is adopted

here. Take the Lyapunov function as the following:

𝑽 (𝒙) = 𝒙𝑇 𝑸𝒙 > 0 (∀𝒙 ≠ 0) (33)

where: 𝑸 ∈ 𝑅2𝑛×2𝑛 is a positive definite symmetric matrix. As a result,

the first order differential equation of Lyapunov function Eq. (33) is
shown as following:
𝑉̇ (𝒙) = 𝒙̇ 𝑇 𝑸𝒙 + 𝒙𝑇 𝑸𝒙̇
[ ] [ ]
= 𝒙̇ 𝑇 𝑨 ̃ 𝑇 + (𝑩 (𝜸 − 𝜺))𝑇 𝑸𝒙 + 𝒙𝑇 𝑸 𝑨 ̃ 𝑇 𝒙 + 𝑩 (𝜸 − 𝜺) (34)
( )
= 𝒙̇ 𝑇 𝑨 ̃ 𝑇 𝑸 + 𝑸𝑨 ̃ 𝒙 + 2𝒙𝑇 𝑸𝑩 (𝜸 − 𝜺)

̃ is a Hurwitz matrix, and 𝑸 is positive definite, the following

Since 𝑨
equation holds:
̃ 𝑇 𝑸 + 𝑸𝑨
𝑨 ̃ = −𝑷 (35)
Fig. 7. Wave spectrum.
where 𝑷 is a positive definite symmetric matrix. Assuming that 𝒛 =
𝑩 𝑇 𝑸𝒙, thus, the first order differential equation of Lyapunov function
can be rewritten as:
𝑉̇ (𝒙) = −𝒙̇ 𝑇 𝑷 𝒙 + 2𝒙𝑇 𝑸𝑩 (𝜸 − 𝜺)
(36)
= −𝒙̇ 𝑇 𝑷 𝒙 + 2𝒛𝑇 (𝜸 − 𝜺)
In order to make the differential Lyapunov function negative, the
second term of Eq. (36) on the right should be set negative, since the
former is always negative for all 𝒙 ≠ 0. Let:
𝒛
𝜀=𝛽 (𝛽 > 0)
‖𝒛‖
‖𝜀‖ = 𝛽
which implies:
𝒛𝑇 𝒛
𝒛𝑇 (𝜸 − 𝜺) = 𝒛𝑇 𝜂 − 𝛽
‖𝒛‖
(37)
≤ ‖𝒛‖‖𝜸‖ − 𝛽‖𝒛‖
= ‖𝒛‖ (‖𝜸‖ − 𝛽)
Fig. 8. Real time sea state realization.
If ‖𝜸‖ ≤ 𝛽, then the differential Lyapunov function 𝑉̇ (𝒙) ≤ 0. A way
to seek for 𝛽 is described as the followings:
[ ]𝑇 ( )
For the system error equation (29), assuming that 𝒙 = 𝒆 𝒆̇ is the ‖𝜸‖ = ‖ 𝑰 − 𝑴 −1 𝑴 ̂ 𝒖 − 𝑴 −1 𝒏‖̃
( )
new state variable. Then the system state–space form can be expressed ≤ ‖𝑰 − 𝑴 −1 𝑴‖ ̂ ‖𝒒̈ 𝑑 + 𝑲𝒙 + 𝜀‖ + ‖𝑴 −1 ‖‖𝒏‖
̃
by: (38)
≤ 𝛼‖𝒒̈ 𝑑 ‖ + 𝛼‖𝑲‖‖𝒙‖ + 𝛼𝛽 + ‖𝑴 −1 ‖‖𝒏‖ ̃
( )
𝒙̇ = 𝑨𝒙 + 𝑩 𝒒̈ 𝑑 − 𝒖 + 𝜸 (30) ≤ 𝛼𝑄𝑀 + 𝛼‖𝑲‖‖𝒙‖ + 𝛼𝛽 + 𝑀𝑚𝑎𝑥 𝜙

where, Thus, by setting:

[ ] [ ]
𝑨=
0 𝑰
∈ 𝑹2𝑛×2𝑛 𝑩=
0
∈ 𝑹2𝑛×𝑛 1 ( )
0 0 𝐼 𝛽≥ 𝛼𝑄𝑀 + 𝛼‖𝑲‖‖𝒙‖ + 𝑀𝑚𝑎𝑥 𝜙
1−𝛼
Though the exact parameters of uncertain term 𝜸 are hardly to the differential Lyapunov function is negative definite, namely 𝑉̇ (𝒙) ≤
obtain, but of which the estimation range is assumed already known. 0. And differential Lyapunov function can be expressed by:
( )
Under the assumption, the system is asymptotically stable with control 𝒛𝑇 𝒛
input 𝒖 while 𝜸 within the estimation range. 𝑉̇ (𝒙) = −𝒙̇ 𝑇 𝑷 𝒙 + 2𝒛𝑇 𝜸 − 𝛽 < 0 (∀𝒙 ≠ 0) (39)
‖𝒛‖
Then the system control input can be expressed by: On conclusion, the block scheme of the controller adopted above in
𝒖 = 𝒒̈ 𝑑 + 𝑲 𝐷 𝒆̇ + 𝑲 𝑃 𝒆 + 𝜀 (31) this paper can be expressed as Fig. 5:
Therefore, the whole gangway system block scheme as shown in
where 𝜀 is added to ensure the system’s robustness under the effect of Fig. 4 can be expressed as Fig. 6 with the improved inverse dynamics
uncertain term 𝜂. control derived above.
By substituting Eq. (31) into Eq. (30), we can get:
4. Simulation results and discussion
̃ + 𝑩 (𝜸 − 𝜀)
𝒙̇ = 𝑨𝒙 (32)
The aim of the active motion compensation system is to keep the
where:
gangway tip still in inertial frame, namely {𝑁𝐸𝐷} frame. This implies
[ ]
̃ = 𝑨 − 𝑩𝑲 = 0 𝐼 that the gangway tip motion should be opposite to that caused by the
𝑨
−𝑲 𝑃 −𝑲 𝐷 ocean waves. While the ocean wave effects directly on the vessel other
than gangway tip. Thus, the motions of the vessel in all 6DOF should be
[ ]
𝑲 = 𝑲𝑃 𝑲𝐷 studied at first. While as studied in Liang et al. (2017a), roll motion of a

6
L. Liang et al. Ocean Engineering 197 (2020) 106864

Table 2
Main parameters of the simulation vessel.
Notation Description Value
𝐿𝑝𝑝 Vessel length 82.8 m
𝑇 Vessel draft 6 m
𝐵 Vessel beam 19.2 m
𝐺𝑀𝐿 Longitudinal metacenter 103.6 m
𝐺𝑀𝑇 Transverse metacenter 2.14 m
𝑀𝑠 Total mass of the vessel 6.36 ∗ 106 kg

Table 3
Main parameters of the simulation vessel.
Notation Description Value
𝐿0 Length of 𝐿0 5 m
𝐿4 Length of 𝐿4 5 m
𝑌𝑐2 COG in 𝑦 axis of frame {2} 3.5 m
𝑍𝑐3 COG in 𝑧 axis of frame {3} 3 m
𝑚1 Mass of link 1 928 kg
𝑚2 Mass of link 2 647 kg
𝑚3 Mass of link 3 412 kg
𝐼1 Inertial matrix of link 1 𝑑𝑖𝑎𝑔(412, 336, 525)
𝐼2 Inertial matrix of link 2 𝑑𝑖𝑎𝑔(2123, 214, 1807)
𝐼3 Inertial matrix of link 3 𝑑𝑖𝑎𝑔(958, 902, 175)

vessel is reduced by the ordinary fin stabilizers when the ship moves at
conventional speed, and decreased by using zero speed fin stabilizers
and Magnus rotating roll stabilizer at zero speed (Liang et al., 2018,
2017). In this paper, a DP model with zero speed is adopted from Fossen
and Perez (2004).
The significant wave height is set to 2.5 m, with wave spread
direction 𝜋∕4, and the water depth is assumed to be infinity. Main
parameters of the simulation vessel is as shown in Table 2, where the
parameters are acquired from Fossen and Perez (2004).
Thus, the sea PSD and real time state realization can be obtained,
the significant wave height is 2.5 m. As shown in Fig. 7, wave PSD is
displayed, and the real time sea state realization is shown in Fig. 8. Ship
motions in 6DOF can be simulated as shown in Fig. 9 with different
wave directions.
As can be seen from Fig. 9(a)∼(c), the main degrees of freedom that
influenced by the ocean waves are surge, sway and heave, which are
the main degrees of freedom to be compensated in this motion compen-
sation gangway system. The motions of the vessel can be compensated
by driving the gangway tip moves in the opposite directions both in
amplitudes and velocities.
In this paper, the base of the motion compensation gangway system
is installed on the deck, and can be expressed by (𝐿𝑥 , 𝐿𝑦 , 𝐿𝑧 ) in {𝐵𝑜𝑑𝑦}
for simplification, the origin of base frame is located at the 𝑜𝑏 𝑥𝑏 𝑧𝑏 plane
of {𝐵𝑜𝑑𝑦}, which implies that 𝐿𝑦 = 0, and it is the same as in Table 3,
𝐿𝑦 vanished. Therefore, the coordinate for the origin of base frame can
be expressed by (𝐿𝑥 , 0, 𝐿𝑧 ). And in this paper, 𝐿𝑥 = 20, 𝐿𝑧 = 3.
The motion compensation gangway system is accomplished in both
the gangway tip stand still and keep moving, and the models are
established using Matlab/Simulink. For both simulations, the gangway
initial states are set at (𝑞1 , 𝑞2 , 𝑞3 ) = (0, 3𝜋∕2, 4.3) to avoid singular values.
And the parameters of the gangway system is displayed in Table 3.
Therefore, the gangway tip position can be obtained using Eq. (3), and
can be expressed as (0, 8, 30)𝑇 in {𝐵𝑜𝑑𝑦} frame. And the parameters of
the controller are chosen as the following:

⎡100 0 0 ⎤
𝑲𝑃 = ⎢ 0 150 0 ⎥
⎢ ⎥
⎣ 0 0 400⎦ Fig. 9. Ship motions of 6DOF in different wave directions. (a) 0, (b) 𝜋/4 and (c) 𝜋/2.
⎡15 0 0⎤
𝑲𝐷 =⎢0 5 0⎥
⎢ ⎥
⎣0 0 15⎦

7
L. Liang et al. Ocean Engineering 197 (2020) 106864

Fig. 10. Gangway tip position errors in x, y, z directions of docking state, wave Fig. 11. Gangway tip position errors in x, y, z directions of docking state, wave
direction 0. direction 𝜋/4.

8
L. Liang et al. Ocean Engineering 197 (2020) 106864

Fig. 13. Tip position with and without motion compensation.

Besides, to satisfy Eq. (35), 𝑸 can be selected as:

⎡262.5 0 0 −0.5 0 0 ⎤
⎢ 0 262.5 0 0 −0.5 0 ⎥
⎢ ⎥
0 0 262.5 0 0 −0.5 ⎥
𝑸=⎢
⎢ −0.5 0 0 2.525 0 0 ⎥
⎢ 0 −0.5 0 0 2.525 0 ⎥⎥
⎢
⎣ 0 0 −0.5 0 0 2.525⎦
PD control with gravity compensation can easily ensure that the
origin system is globally asymptotically stable if 𝐾𝑃 and 𝐾𝐷 are positive
definite (Kelly et al., 2006). Thus, in the following parts, simulation
results of these two controllers are given.
The motion compensation task is divided into two parts: the docking
state and moves with desired trajectories.
Case 1: The docking state aims to keep the gangway tip motionless,
which is also known as regulation (Kelly et al., 2006). For motion
compensation purpose, the motions of gangway tip should be on the
contrary directions against to the vessel motions caused by the ocean
disturbances.
It is obviously from Fig. 10∼12, in the situation of non-compensated,
gangway tip moves far away with time increasing both in 𝑥 and 𝑦
directions, while in 𝑧 direction oscillates around its initial state. This is
mainly caused by the fact that restoring force and moment do exist in
ship heave motion, while the other two do not.
Simulations of the adopted controller in this paper compared with
PD control with gravity compensation are as shown in Figs. 10–12.
Where 𝑒𝑥𝑅 , 𝑒𝑦𝑅 and 𝑒𝑧𝑅 are gangway tip position error in 𝑥, 𝑦 and
𝑧 direction with the controller adopted here, respectively, while 𝑒𝑥𝐺 ,
𝑒𝑦𝐺 and 𝑒𝑧𝐺 are the results of PD controller with gravity compensation.
According to Figs. 10 to 12, we can conclude that the vessel motions
are compensated with both controllers, and the gangway tip keeps its
position of the initial state with time passing by, while the inverse
dynamics controller adopted in the paper performs well in 𝑥 direction
(see Fig. 13).
As a consequence, the compensated gangway tip positions are as
shown in Fig. 16, where 𝑥𝑑 , 𝑦𝑑 and 𝑧𝑑 are the desired positions, 𝑥𝑢𝑛 , 𝑦𝑢𝑛
and 𝑧𝑢𝑛 are positions without motion compensation, while 𝑥𝑎𝑐𝑡 , 𝑦𝑎𝑐𝑡 and
𝑧𝑎𝑐𝑡 are the compensated tip position in 𝑥, 𝑦, 𝑧 directions, respectively.
According to Figs. 10∼12, gangway tip position errors decreased
Fig. 12. Gangway tip position errors in x, y, z directions of docking state, wave in 𝑥, 𝑦 and 𝑧 directions, while in 𝑦 direction, the inverse dynamics
direction 𝜋/2. controller performs better both in tuning time and system overshoot
than the PD control with gravity compensation.
Case 2: Different from the docking case, which is to keep the
gangway tip still at an exact point, the second situation to be studied in

9
L. Liang et al. Ocean Engineering 197 (2020) 106864

Fig. 14. Gangway tip position errors in x, y, z directions of moving state, wave Fig. 15. Gangway tip position errors in x, y, z directions of moving state, wave
direction 0. direction 𝜋/4.

10
L. Liang et al. Ocean Engineering 197 (2020) 106864

Fig. 17. Actual tip position with motion compensation.

this paper is to keep the gangway tip moves with desired trajectories,
which are expressed as the following:
2𝜋
𝑥𝑑 = 1 ∗ 𝑠𝑖𝑛( 𝑡) + 30.7
10
2𝜋
𝑦𝑑 = 1.2 ∗ 𝑠𝑖𝑛( 𝑡)
14
2𝜋
𝑧𝑑 = 0.8 ∗ 𝑠𝑖𝑛( 𝑡) + 8
8
where 𝑥𝑑 , 𝑦𝑑 and 𝑧𝑑 are all given in inertial frame.
It is the same as the first situation that in 𝑧 direction, the restoring
force and moment exist, which can be obtained from the tip position
errors shown in Figs. 14∼16. And with the affections of ocean waves,
the gangway tip moves in irregular trajectories rather than the desired
without motion compensation. And the tracking errors of gangway tip
with control affections of motion compensation are shown in Figs. 14–
16, in which 𝑒𝑥𝑢 , 𝑒𝑦𝑢 and 𝑒𝑧𝑢 are the position errors without motion
compensation, 𝑒𝑥𝑅 , 𝑒𝑦𝑅 and 𝑒𝑧𝑅 are positions without motion compen-
sation, while 𝑒𝑥𝐺 , 𝑒𝑦𝐺 and 𝑒𝑧𝐺 are the compensated tip position in 𝑥, 𝑦, 𝑧
directions, respectively.
By comparing Figs. 14∼16, we can find out that the tracking er-
rors decreased in significant amplitudes in 𝑥, 𝑦 and 𝑧 directions with
the controller adopted in this paper. While using the PD controller
with gravity compensation, the tracking errors decreased, the system
oscillates, as a result, the system tends to instability easily.
The real gangway tip positions in 𝑥, 𝑦 and 𝑧 directions are shown
as Fig. 17. Where 𝑥𝑑 , 𝑦𝑑 and 𝑧𝑑 are the desired positions, 𝑥𝑢𝑛 , 𝑦𝑢𝑛 and
𝑧𝑢𝑛 are the positions without motion compensation, 𝑥𝑎𝑐𝑡 , 𝑦𝑎𝑐𝑡 and 𝑧𝑎𝑐𝑡
are the positions with motion compensation.
From the simulation results of both cases above, we can conclude
that in the docking state, both controllers are capable in the motion
compensation system. While in the later case, the inverse dynamics
controller performs better, which is also corresponding to the fact that
the PD control with gravity compensation does works in cases that the
desired position are constants, which has been studied in Kelly et al.
(2006).

5. Conclusion

An active motion compensation gangway system is analyzed in

this paper, both kinematics and dynamics of the gangway system are
Fig. 16. Gangway tip position errors in x, y, z directions of moving state, wave derived. In order to compensate vessel motions due to the ocean waves,
direction 𝜋/2. a DP model and ITTC 2-parameter PSD are adopted in this paper.
Different wave directions that effect the ship motions are simulated.
And based on the inverse dynamics, the robust controller is designed
and simulated using Matlab/Simulink. According to the simulation

11
L. Liang et al. Ocean Engineering 197 (2020) 106864

results of ship motions in 6DOF, restoring force and moment exist in Do, K.D., Pan, J., 2008b. Nonlinear control of an active heave compensation system.
heave for translational motion. As a result, compared with the other Ocean Eng. 35 (5–6), 558–571.
Fossen, Thor I., 2011. Handbook of Marine Craft Hydrodynamics and Motion Control.
two directions, gangway tip position is periodicity in z direction. From
John Wiley & Sons Ltd..
the simulations results, we can conclude that in docking case, both Fossen, T.I., Perez, T., 2004. Marine systems simulator (MSS). URL: https://github.com/
controllers are capable. While in the second case, the inverse dynamic cybergalactic/MSS.
controller is preferable. PD control with gravity is not a optional one. Guanche, R., Martini, M., Jurado, A., Losada, I.J., 2016. Walk-to-work accessibility
Besides, motions are compensated not only in the DOF with restoring assessment for floating offshore wind turbines. Ocean Eng. 116, 216–225.
Kelly, R., Davila, V.S., Perez, J.A.L., 2006. Control of Robot Manipulators in Joint
forces and moments, but also the other two. Space. Springer Science & Business Media.
Khudher, D., Powell, R., 2016. Quadratic programming for inverse kinematics control
Declaration of competing interest of a hexapod robot with inequality constraints. In: 2016 International Conference
on Robotics: Current Trends and Future Challenges (RCTFC). pp. 1–5.
Lebret, G., Liu, K., Lewis, F.L., 1993. Dynamic analysis and control of a stewart platform
The authors declare that they have no known competing finan-
manipulator. J. Robot. Syst. 10 (5), 629–655.
cial interests or personal relationships that could have appeared to Liang, L., Sun, M., Luan, T., 2017a. Design fuzzy input-based adaptive sliding mode
influence the work reported in this paper. control for vessel lift-feedback fin stabilizers with shock and vibration of waves.
Shock Vib. 2017, 1–13.
Acknowledgments Liang, L., Zhao, P., Zhang, S., et al., 2017. Simulation and analysis of Magnus rotating
roll stabilizer at low speed. Ocean Eng. 142, 491–500.
Liang, L., Zhao, P., Zhang, S., et al., 2017b. Simulation and analysis of Magnus rotating
This work is supported by the Fundamental Research Funds for the roll stabilizer at low speed. Ocean Eng. 142, 491–500.
Central Universities under Grants HEUCFM170404 and Harbin Science Liang, L., Zhao, P., Zhang, S., et al., 2018. Simulation and experimental study on control
and Technology Innovation Talent Research Special Fund Project under strategy of zero-speed fin stabilizer based on disturbance and compensation. PLoS
One 13 (10), e0204446.
Grant 2017RC2017XK009006.
Liao, W., Liang, W., Wang, A., 2017. Inverse dynamics control of a parallel robot
based on RBF neural network. In: 2017 International Conference on Advanced
References Mechatronic Systems (ICAMechS). pp. 372–375.
Pérez, Tristan, Blanke, Mogens, 2002. Simulation of Ship Motion in Seaway. Technical
Chu, Y., Æsøy, V., Bunes, Ø., Pedersen, E., 2016. Modeling and simulation of the accu- Report EE02037.
mulator during active heave compensation operations. In: ASME 2016 International Perez, T., Blanke, M., 2012. Ship roll damping control. Annu. Rev. Control 36 (1),
Conference on Ocean, Offshore and Arctic Engineering. V001T01A007. 129–147.
Chu, Y., Æsøy, V., Ehlers, S., Zhang, H., 2015. Integrated multi-domain system Salzmann, D.J.C., 2010. Ampelmann, development of the access system for offshore
modelling and simulation for offshore crane operations. Ship Technol. Res. 62 (1), wind turbines.
36–46. Spong, Mark W., Ortega, Romeo, 1990. On adaptive inverse dynamics control of rigid
Chu, Y., Sanfilippo, F., Asoy, V., Zhang, H., 2014. An effective heave compensation robots. IEEE Trans. Automat. Control 35 (1), 92–95.
and anti-sway control approach for offshore hydraulic crane operations. In: IEEE Ulrich, N., Kumar, V., 1991. Passive mechanical gravity compensation for robot
International Conference on Mechatronics and Automation. pp. 1282–1287. manipulators. In: Proceedings. 1991 IEEE International Conference on Robotics and
Craig, J.J., 2005. Introduction to Robotics: Mechanics and Control. Pearson/Prentice Automation, Vol. 2, pp. 1536–1541.
Hall, Upper Saddle River, NJ, USA. Wu, M., 2014. Numerical analysis of docking operation between service vessels and
Do, K.D., Pan, J., 2008a. High performance control of an active heave compensation offshore wind turbines. Ocean Eng. 91, 379–388.
system. IFAC Proc. Vol. 41 (2), 5059–5064. Yu, F., 2017. Modeling, Simulation and Control of Motion Compensated Gangway
in Offshore Operations (Master’s degree). Norwegian university of science and
technology.

12