Ocean Engineering 220 (2021) 108257

Contents lists available at ScienceDirect

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

Robust depth control of a hybrid autonomous underwater vehicle with

propeller torque’s effect and model uncertainty
Huy Ngoc Tran a, b, *, Thanh Nguyen Nhut Pham a, Sik Hyeung Choi c
a
Faculty of Electrical & Electronics Engineering of, Ho Chi Minh City University of Technology (HCMUT), 268 Ly Thuong Kiet Street, District 10, Ho Chi Minh City,
Vietnam
b
Vietnam National University Ho Chi Minh City, Linh Trung Ward, Thu Duc District, Ho Chi Minh City, Vietnam
c
Department of Mechanical Engineering, Korea Maritime and Ocean University, Busan, Korea

A R T I C L E I N F O A B S T R A C T

Keywords: This paper presents a study of depth tracking controller design for a hybrid AUV in the presence of model un­
Autonomous underwater vehicle (AUV) certainty and propeller torque’s effect. Firstly, the six degrees of freedom (6-DOF) nonlinear equations of motion,
Depth control as well as the operating mechanisms and specific characteristics of the hybrid AUV, are described. Subsequently,
Backstepping
the model for depth-plane is extracted by decoupling and linearizing the 6-DOF AUV model. Furthermore, a
nonlinear disturbance observer (NDO) is constructed to deal with the linearization errors and uncertain com­
ponents in the depth-plane model. A depth tracking controller is then designed based on the backstepping
technique to guarantee the tracking error converges to an arbitrarily small neighborhood of zero. Besides, the
robust stability of the proposed controller concerning the propeller torque’s effect and the model uncertainty is
analyzed. To ensure the objectivity and feasibility of the proposed method, the depth controller is applied to the
6-DOF model of AUV so that it maintains the coupling between roll, yaw, and pitch motion. Finally, the nu­
merical simulation is carried out via MATLAB/SIMULINK to verify the controller’s effectiveness, feasibility, and
stability.

1. Introduction have a sharply reduced performance when either changing the model
parameters or applying in practice. This can be seen through the com­
Nowadays, autonomous underwater vehicles (AUV) have become a parison of simulation and experimental results in Naik and Singh (2007);
major tool for activities in deep sea because of the significant Prestero (2001); Mahapatra and Subudhi (2018). Therefore, stability
improvement in their performance. Specifically, in the scientific field is analysis and quality characterization surveys of the controller under the
surveys, data collection and environmental sampling for hydrological effect of model uncertainty is essential.
research (Wynn et al., 2014; Eichhorn et al., 2018); in the military is The control problem for AUV is a complex issue, so it is usually
clearing sea mines planted by enemies, conducting war activities with divided into three sub-problems for steering, diving and speed control
the navy (Hagen et al., 2003; Mondal et al., 2019); in the commercial (Jalving, 1994). In each sub-problem, it will be linearized to reduce
field, AUVs are used to make detailed maps and surveys of the seafloor interaction between each other, thereby simplifying the design of the
(oil and gas industries) before building subsea infrastructure (Mondal controller. Consequently, AUV usually has three main controllers that
et al., 2019; Zhang et al., 2015). Therefore, research and development of are speed, steering, and diving controllers. In this paper, we only focus
controllers for AUV are essential. on studying the depth control problem (diving control problem) for
In general, the controller design bases on the assumption that the AUV. For the depth control problem, in recent years, there have been
model parameters are accurate. However, the model of AUV is a set of many research projects showing the different algorithms applied to
highly coupled nonlinear equations (Newman, 1977; Fossen, 1994) that design the depth controller for AUV. For instance, with the simplest and
lead to identifying completely accurate models (through estimation or most common algorithm, PID, there are Jalving (1994), Prestero (2001),
empirical methods, etc.) is a difficult task. Besides, many control algo­ Tanakitkorn et al. (2017). In Jalving (1994), after decomposing the AUV
rithms based on the linearization methods around the working point will control into three subsystems, the PID controller was built for each

* Corresponding author. Faculty of Electrical & Electronics Engineering of, Ho Chi Minh City University of Technology (HCMUT), 268 Ly Thuong Kiet Street,
District 10, Ho Chi Minh City, Vietnam

https://doi.org/10.1016/j.oceaneng.2020.108257
Received 6 June 2020; Received in revised form 2 October 2020; Accepted 18 October 2020
Available online 31 December 2020
0029-8018/© 2020 Elsevier Ltd. All rights reserved.
H.N. Tran et al. Ocean Engineering 220 (2021) 108257

system. For the depth controller, the mathematical model of depth-plane Additionally, there exist other methods which are applied to control
is established under the state-space form by linearizing and decoupling the depth of AUV. For example, a gain scheduling controller is proposed
the motion model of AUV in the vertical plane. The pitch transfer by Silvestre and Pascoal (2007) using a reduced output feedback tech­
function and the depth transfer function are then extracted from the nique for INFANTE AUV or the state-dependent Riccati equation (SDRE)
state-space model to analyze. Then the parameters are selected to sta­ technique developed in Naik and Singh (2007). Further, a constrained
bilize for the closed-loop transfer function of the whole system. In self-tuning controller based on Nonlinear Auto-Regressive Moving
following the same approach as Jalving (1994), Prestero (2001) has Average eXogenous (NARMAX) is utilized by Rout and Subudhi (2017)
developed a simple pitch-and-depth PD controller based on transfer to track a given reference for a torpedo-shaped AUV. Besides, Yu et al.
functions derived from the linear state-space model. In Tanakitkorn (2018) explores a sliding mode fuzzy control method combines with
et al. (2017), the PI-D scheme has been used to design the depth line-of-sight (LOS) guidance law to apply to a bioinspired robotic dol­
controller for Delphin2 AUV. This PI-D scheme has the derivative term phin. Moreover, various H∞ control methods have been developed for
fed by an actual state of the system instead of the tracking error to avoid underwater vehicles by Liceaga-Castro and van der Molen (1995),
the effect of the sudden change of setting value. The PI-D controller will Moreira and Soares (2008), Mahapatra et al. (2016), Mahapatra and
make the system works more smoothly but this controller will not be Subudhi (2018). Similarly, the H2 optimal control methods applied to
able to process the time-varying desired signal. Moreover, the controller AUV in Moreira and Soares (2008), Wadoo et al. (2012), Qiao et al.
gain values are identified using a heuristic approach based on the (2018).
Ziegler-Nichols method. In general, the papers mentioned above often use linearizing and
Besides linear controllers, nonlinear algorithms have also been decoupling equations (3-DOF or 4-DOF) in the vertical motion of AUV
studied and applied to AUV. Specifically, Hong et al. (2010), Joe et al. for controller design, stability analysis, and simulation. It means that the
(2014), Yan et al. (2016), Tran et al. (2019) have developed various coupling between roll, pitch, and yaw motion is neglected. Note that a
sliding mode control (SMC) algorithms for the depth-plane motion of torpedo-shaped AUV has a small moment of inertia and drag in the roll
AUVs. Hong et al. (2010) presented how to build an SMC controller and motion. Thus, the roll dynamic easily suffers oscillatory when the AUV is
an experimental verification of it. This paper also addresses the effect of affected by propeller torque, unknown disturbances, yaw and pitch
positive buoyancy in the dynamic model. Nevertheless, the author did motion. Moreover, the roll motion is dynamically coupled into pitch
not analyze the stability nor pointed out the advantages of the proposed motion so unwanted roll motion can reduce the diving performance and
control algorithm. In Joe et al. (2014), a second-order SMC was designed provide no clear opportunities for suppressing roll motion (Petrich and
based on the 4-DOF diving model of AUV (ignore roll and pitch). The Stilwell, 2011). Besides, due to hydrodynamic characteristics, the mo­
simulation and experiment were conducted on Cyclops AUV to verify the tion equations of AUV are highly nonlinear and strongly coupled, so it is
stability of the controller in the presence of parameter uncertainties and inevitable to have model uncertainties in the linear process. Therefore, it
external disturbances. However, the internal stability of the closed-loop is necessary to investigate the robust stability and performance of the
system was not guaranteed. An integral-fast terminal sliding mode controller in the presence of propeller torque’s effect and model un­
control (IFTSMC) was explored by Yan et al. (2016) to control the AUV certainties. On the other hand, the simulation model of AUV should be
under the effect of parameter perturbations and wave disturbances. 6-DOF to ensure objectivity as well as the accuracy of the designed
Notwithstanding, this paper only proves the convergence on the 3-DOF controller.
linear motion model of AUV in the vertical plane. Tran et al. (2019) Motivated by the aforementioned analysis, this paper focuses on
proposed an indirect strategy to control depth for the hybrid AUV designing the nonlinear tracking controller for the vertical motion of a
through pitch angle control. The simulation results show the simplicity hybrid AUV. The control laws are produced by employing the back­
and effectiveness of the proposed strategy. But the paper does not stepping technique to ensure stability and force the depth tracking error
analyze the stability and performance of the controller when there are to an arbitrarily small neighborhood of zero, regardless of propeller
parameter uncertainties. torque’s effect and model uncertainties. The designed controller is
Moreover, several studies on backstepping techniques applied to applied to a 6-DOF nonlinear model to ensure objectivity and feasibility.
depth control have been developed by Lapierre (2009), Cao et al. The simulation results will be specific evidence showing the efficiency,
(2011), Wei et al. (2015), Gharesi et al. (2017), Mahapatra and Subudhi disturbances rejection ability, robustness, and superiority of the pro­
(2018). Lapierre (2009) proposed the adaptive controller and switching posed method.
schemes to resolve external parameter uncertainty. In this paper, the The remnant of this paper is organized as follows. Section 2 presents
adaptive laws were developed instead of the disturbance rejection to the AUV 6-DOF models, the operating mechanisms and specific char­
ensure the robustness. The simulation results show that the controller acteristics of the hybrid AUV, and formulates control problem. Section 3
can handle when there are parameter uncertainties. However, the con­ proposes the design of nonlinear disturbance observer (NDO) and
trol performance isn’t high, such as the response time is slow, the control backstepping controller for the depth-plane motion of the AUV and also
signal is not smooth. Another backstepping controller combined with analyzes the robust stability of the proposed controller under propeller
adaptive law to solve the parameter uncertainty and external distur­ torque’s effect and model uncertainties. In Section 4, the numerical
bance was also discovered by Cao et al. (2011). Nevertheless, the simulations and discussions are carried out to validates the previous
controller design and verification simulation were based only on the analysis and design. Finally, the conclusions of this work are presented
3-DOF model in depth-plane. Wei et al. (2015) formulated a linear in Section 5.
state-space equation for vertical motion of AUV, then used backstepping
techniques to build controllers. This paper addressed model un­ 2. AUV modelling and problem formulation
certainties and developed a nonlinear disturbance observer (NDO) to
deal with it. Similarly, a backstepping controller combines with a linear 2.1. AUV modelling
extended state observer (LESO) was design by Gharesi et al. (2017).
Further, Mahapatra and Subudhi (2018) explored a depth control using The model will be considered in this paper is the AUV2000
the backstepping technique and nonlinear H-infinity technique. In this mentioned in (Tran et al., 2019), AUV2000 shown in Fig. 1 is a hybrid
paper, random perturbations and uncertain hydrodynamic parameters AUV designed to integrate the outstanding characteristics of conven­
were used to investigate the robustness of the proposed controller. The tional AUV and underwater glider (a type of AUV that employs
simulation and experiment results demonstrated the controller’s effec­ variable-buoyancy propulsion instead of traditional propellers or
tiveness and feasibility, however, the control performance decline is thrusters). Therefore AUV2000 can operate in two separate modes,
proportional to the uncertainty magnitude. specifically without using the thruster (Glider mode) and using the

2
H.N. Tran et al. Ocean Engineering 220 (2021) 108257

thruster (AUV mode). However, in this paper, we only focus on 2.1.2. Dynamic model of hybrid AUV
designing the depth controller for AUV2000 in AUV mode. Following to (Prestero, 2001), the 6-DOF nonlinear dynamic equa­
tions of motion of AUV2000 can be expressed as
2.1.1. Kinematics model [ ( ) ] ∑
In accordance with SNAME (SNAME, 1950), the 6-DOF nonlinear m u̇ − vr + wq − xG q2 + r2 + yG (pq − ṙ) + zG (pr + q̇) = Xext (2a)
kinematic and dynamic equations of motion of AUV are described in the [ ( ) ] ∑
earth-fixed frame {e} and the body-fixed frame {b} as shown in Fig. 2. m v̇ − wp + ur − yG r2 + p2 + zG (qr − ṗ) + xG (qp + ṙ) = Yext (2b)
According to Fossen, we have coordinate transform relating the trans­
[ ( ) ] ∑
lational velocities and the rotational velocities between {b} and {e} for m ẇ − uq + vp − zG p2 + q2 + xG (rp − q̇) + yG (rq + ṗ) = Zext (2c)
various underwater vehicles in general and AUV in particular as follows:
( ) ∑
⎧ Ixx ṗ + Izz − Iyy qr + m[yG (ẇ − uq + vp) − zG (v̇ − wp + ur)] = Kext (2d)
⎪ ⎡ ⎤ ⎡ ⎤
⎪
⎪ ẋ u ∑
⎪
⎪ (2e)
⎣ ⎦
⎨ ẏ = J1 (η2 )⎣ v ⎦ Iyy q̇ + (Ixx − Izz )rp + m[zG (u̇ − vr + wq) − xG (ẇ − uq + vp)] = Mext
⎡ ż ⎤ ⎡w ⎤ (1) ∑
⎪ φ̇ p ( )
⎪
⎪ ⎣ θ̇ ⎦ = J2 (η2 )⎣ q ⎦ Izz ṙ + Iyy − Ixx pq + m[xG (v̇ − wp + ur) − yG (u̇ − vr + wq)] = Next (2f)
⎪
⎪
⎩ ψ̇ r
where m is the AUV’s mass, (xG , yG , zG ) is the center of gravity, and Ixx ,
Iyy , Izz respectively are inertial moments about the x, y, and z axes. The

⎡ ⎤
cos ψ cos θ − sin ψ cos φ + cos ψ sin θ sin φ sin ψ sin φ + cos ψ sin θ cos φ
⎢ ⎥
J1 (η2 ) = ⎢ ⎥
⎣ sin ψ cos θ cos ψ cos φ + sin ψ sin θ sin φ − cos ψ sin φ + sin ψ sin θ cos φ ⎦
− sin θ cos θ sin φ cos θ cos φ
⎡ ⎤
1 sin φ tan θ cos φ tan θ
⎢ ⎥
⎢ ⎥
⎢0 cos φ − sin φ ⎥
J2 (η2 ) = ⎢ ⎥
⎢ ⎥
⎣ sin φ cos φ ⎦
0
cos θ cos θ

where right side of equations (2a)-(2f) are the sum of the components of the
force and moment acting on the vehicle:
∑
Xext = X u̇ u̇ + Z ẇ wq + Z q˙ q2 − Y v˙vr − Y r˙r2 + Xu|u| u|u|
(3a)
− (W − B)sin θ + Xprop
η1 = [ x y z ]T : the position of the vehicle with respect to {e}
∑
η2 = [ φ θ ψ ]T : the orientation (roll, pitch, yaw) of the vehicle Yext = Y v˙v̇ + Y r˙ṙ + X u̇ ur − Z ẇ wp − Z q̇ pq + Yv|v| v|v| + Yr|r| r|r| + Yuvl uv
with respect to {e} +(W − B)cos θ sin φ + Yrud
υ1 = [ u v w ]T : the translational velocities of the vehicle with (3b)
respect to {b}
∑
υ2 = [ p q r ]T : the rotational velocities of the vehicle with respect Zext = Z ẇ ẇ + Z q˙ q̇ − X u˙ uq + Y v˙vp + Y r˙rp + Zw|w| w|w| + Zq|q| q|q|
( ) (3c)
to {b} + Zuwl + Zuwf uw + Zuqf uq + (W − B)cos θ cos φ

Fig. 1. AUV2000 prototype.

3
H.N. Tran et al. Ocean Engineering 220 (2021) 108257

∑ ( ) ( ) ( ) ( ( )
Kext = K p˙ ṗ + Z ẇ − Y v˙ wv + Z q˙ + Y r˙ vq − Z q̇ + Y r˙ wr + N r˙ Vvar = Sc . lc − xp (4)
)
− M q̇ qr + Kp|p| p|p| + (yG W − yB B)cos θ cos φ where Sc is the cross section of cylinder, lc is the length of cylinder, and
− (zG W − zB B)cos θ sin φ + Kprop (3d) xp is the piston position.
From Fig. 3 the formula for determining the buoyancy center of AUV
∑ ( ) ( ) can be expressed as
Mext = Z q˙ ẇ + M q˙ q̇ −Z ẇ − X u̇ uw − Y r˙vp + K p˙ − N r˙ rp − Z q˙ uq
( ) ⎡ ⎤T
+ Mw|w| w|w| + Mq|q| q|q| + Muwl + Muwf uw + Muqf uq c l +x
p
⎢Vvar 2 + Vfix lfix ⎥
̅→
rcb = [xB , 0, 0]T = ⎣ , 0, 0⎦ (5)
− (zG W − zB B)sin θ − (xG W − xB B)cos θ cos φ Vvar + Vfix
(3e)
∑ ( ) ( ) and the buoyancy of AUV is:
Next = N v˙v̇ + N r˙ṙ − X u̇ − Y v˙ uv + Z q˙ wp − K p˙ − M q˙ pq + Y r˙ur ( )
B = ρg Vfix + Vvar (6)
+ Nv|v| v|v| + Nr|r| r|r| + Nuvl uv + (xG W − xB B)cos θ sin φ
+ (yG W − yB B)sin θ + Nrud (3f) where ρ is the density of fluid, g is the gravitational acceleration.

In (3a)-(3f), X(.) , Y(.) , Z(.) , K(.) , M(.) , N(.) are the added mass, damping
2.1.3.2. Analysis of the counterweight system. Denote the total weight of
and lift terms, (xB , yB , zB ) is the center of buoyancy, W and B respectively
are weight and buoyancy of AUV. The propeller force and torque are AUV without counterweight (Fig. 4) as m , and let →h r = [l , l , l ]T be
h hx hy hz

Xprop and Kprop , and the rudder force and moment are Yrud and Nrud . In this the mass center of mh in the body-fixed frame. Because the counter­
paper, since we only concentrate on depth control, we will reduce Yrud , weight’s mass mm is evenly distributed around the x-axis, the mass
Nrud to zero in subsequent calculations. r = [x , 0, 0]T
center of the counterweight in the body-fixed frame is → m m
where xm is the counterweight position. As a result, the center of gravity
2.1.3. The novel characteristic and operating mechanisms of AUV2000 of AUV can be determined by the following formula:
As mentioned above, AUV2000 is a hybrid AUV that is designed to [ ]T
mh →
rh + mm →
rm mh lhx + mm xm mh lhy mh lhz
operate without using the thruster. Therefore, many structural elements ̅→
rcg = = [xG , yG , zG ]T = , ,
mh + mm mh + mm mh + mm mh + mm
resemble those of a typical underwater glider, the most significant of
which are counterweight system and ballast system. In this section, the (7)
dissection of these operating mechanisms of AUV2000 is presented,
and the weight of AUV is:
thereby paving the way for the construction of the controller in the next
section. W = mg = (mh + mm )g (8)

2.1.3.1. Buoyancy engine dynamics analysis. AUV2000 ballast system is

a piston-cylinder which is arranged as shown in Fig. 3. Let Vfix be the Remark 1. Note that because →rh is constant, it is evident from (7) that
volume occupied by AUV when the ballast system is full of water and moving the counterweight only alters xG . In other words, yG and zG are
r = [l , 0, 0]T be the current position of the center of buoyancy of AUV
̅→ constant.
fix fix
considered in the body-fixed frame. Moreover, denote Vvar as the volume Remark 2. Like REMUS in (Prestero, 2001), to deal with propeller
of water that the cylinder emits, then its relation to the position of the torque’s effect, AUV2000’s internal devices are arranged to obtain yG , zG
piston is formularized as so that the hydrostatic roll moment matches propeller torque, thus
compensating for the roll offset. Therefore, by setting the number of
revolutions (RPM) for the propeller to achieve the operating speed of
AUV, the roll angle will be small and bounded during movement.

2.2. Problem formulation

In this paper, the objective is to develop a controller to maneuver the

AUV tracking the desired depth in the vertical plane, notwithstanding of
highly coupled model nonlinearities, propeller torque’s effect, and
model uncertainties. The objective can be divided into three sub­
problems, as shown below.

Fig. 2. AUV2000 body-fixed and earth-fixed coordinate systems.

Fig. 3. Structure of the ballast system on AUV2000.

4
H.N. Tran et al. Ocean Engineering 220 (2021) 108257

Fig. 4. Structure of the counterweight system on AUV2000.

+ Nonlinear disturbance observer (NDO) is employed to deal with where

linearization errors and model uncertainties in the depth-plane ( )
[( ) ]
model. X u̇ − Z ẇ + Muwl + Muwf uw + Muqf − Z q˙ uq + Mw|w| w|w|
+ Given a desired depth zd , design a control law that determines the +Mq|q| q|q| − mzG wq − zG W sin θ + xB B cos θ
position of the counterweight based on the backstepping technique fq =
Iyy − M q̇
to guarantee the depth tracking error converges to an arbitrarily
small neighborhood around zero as t goes to infinity, despite model muq + W cos θ
uncertainties and propeller torque’s effect. gq = −
Iyy − M q˙
+ The numerical simulations of the proposed controller applied to
the 6-DOF AUV model are carried out to verify the disturbance d is the sum of the uncertain components of the model and the lineari­
resistance, robustness, and effectiveness of the developed control zation errors.
law. Thereby, the depth-plane model can be expressed as
⎧
3. Controller design ⎨ ż = − u sin θ + v cos θ sin φ + w cos θ cos φ
θ̇ = q cos φ − r sin φ (13)
⎩
q̇ = fq + gq xG + d
There exists various methods to build a depth controller for AUV,
some of which are based on a state-space equation linearized from the
depth-plane model ((Silvestre and Pascoal, 2007), (Naik and Singh, 3.1. NDO design
2007), (Wei et al., 2015), (Gharesi et al., 2017)); transfer function
((Jalving, 1994), (Prestero, 2001), (Qiao et al., 2018)); LOS in the ver­ To estimate d, we define the new variable F = d − Lq, where L > 0 is
tical plane ((Tran et al., 2019), (Yu et al., 2018), (Rout and Subudhi, the observer gain which needs to be selected later. Then the nonlinear
2017)). This paper presents a different methodology to build a depth disturbance observer for d is constructed as follows:
controller using the backstepping technique. {
̂ ̂ + Lq
d (= F
Before starting the design of the controller, it is noticeable that when ) (14)
̂˙ = − L f q̇ + gq˙ xm + ̂
the AUV operates without thruster, the ballast system no longer affects F d
too much to the diving and floating process of AUV. Therefore, for
convenience and simplicity, the paper will consider the ballast system where ̂d and ̂F are the estimations of d and F, respectively. The estimated
with the volume to balance the buoyancy with the weight of AUV. error is defined:
Utilizing (6), (8) and (4) we can solve the position of piston to satisfy
̃
d=d − ̂
condition B = W. d
From (1) we derive two differential equations for depth and pitch as
follows: and its derivative is:
{ ( ) ( )
˙
̃ ˙ ̂˙ − Lq̇ = ḋ + L f q̇ + gq˙ xm + ̂
ż = − u sin θ + v cos θ sin φ + w cos θ cos φ (9a) d = ḋ − ̂
d = ḋ − F d − L f q˙ + gq̇ xm + d = ḋ − L̃
d
θ̇ = q cos φ − r sin φ (9b)
(15)
Further, combining (2e) and (3e), the nonlinear dynamic equation of
pitch can be expressed as: 3.2. Backstepping controller design
Iyy q̇ + (Ixx − Izz )rp + m[zG (u̇ − vr + wq) − x(G (ẇ − uq)+ vp)]
( ) From here, we start building control laws by the backstepping
= Z q̇ ẇ + M q̇ q̇ − Z ẇ − X u˙ uw − Y r˙vp + K p˙ − N r˙ rp − Z q̇ uq
( ) (10) technique and Lyapunov functions as follows:
+Mw|w| w|w| + Mq|q| q|q| + Muwl + Muwf uw + Muqf uq
− (zG W − zB B)sin θ − (xG W − xB B)cos θ cos φ Step 1. Let the desired depth be zd , then the depth error can be defined
as:
Performing linearization for equation (10) with u̇ = ẇ = p = v = r =
φ = 0 and note that zB = 0 from (5), we can obtain: e = z − zd , (16)
( ) ( ) ( )
Iyy − M q˙ q̇ = Muwl + Muwf − Z ẇ + X u̇ uw + Muqf − Z q̇ uq + Mw|w| w|w| Take the derivative of the depth error and utilizing (9a) yields:

+ Mq|q| q|q| − mzG wq − zG W sin θ + xB B cos θ ė = − u sin θ + v cos θ sin φ + w cos θ cos φ − żd (17)
− (muq + W cos θ)xG
Define the first control Lyapunov function (CLF) as: V1 := 12e2 whose
(11)
time derivative is
Now, from (10) and (11), we can derive pitch dynamic equation as ( )
following form: V̇ 1 = eė = e − u sin θ + v cos θ sin φ + w cos θ cos φ − żd
q̇ = fq + gq xG + d (12)
The first virtual control law is selected as follows:

5
H.N. Tran et al. Ocean Engineering 220 (2021) 108257

⎧
⎨ α = θ − θd (18a)
⎪

⎪ k1 e − żd Step 3. From (22a) and (12), the derivative of the second virtual
⎩ θd = (18b)
U0 control law can be derived as

β̇ = q̇ − q̇d = fq + gq xG + d − q̇d (25)

where U0 is desired operating speed of AUV and k1 is the control gain
which needs to be determined. Then the derivative of the first CLF can be
Consider the third CLF candidate: V3 = V2 + 12β2 and its derivative is
expressed as:

( ) ( )
V̇ 1 = e u(θ − sin θ) − uθ + v cos θ sin φ + w cos θ cos φ − żd = ue(θ − sin θ) + e − u(α + θd ) + v cos θ sin φ + w cos θ cos φ − żd
( ) (
k1 e − żd u
= ue(θ − sin θ) − ueα + e − u + v cos θ sin φ + w cos θ cos φ − żd =− k1 e2 − ueα + e u(θ − sin θ) + v cos θ sin φ + w cos θ cos φ
U0 U0
( ))
u
+ żd − 1
U0
u
=− k1 e2 − ueα + pe (19)
U0

( )
u
( ( )) V̇ 3 = V̇ 2 + ββ̇ = − k1 e2 − k2 α2 + pα + αβ cos φ + β fq + gq xG + d − q̇d
u U0
where pe = e u(θ − sin θ) + v cos θ sin φ + w cos θ cos φ + żd − 1 ( )
U0 u
= − k1 e2 − k2 α2 + pα + β fq + gq xG + d − q̇d + α cos φ
(20) U0
The control law is then designed as

Step 2. Differentiating (18a) with respect to time and utilizing (9b), − k3 β − fq + q̇d − α − ̂
d
xG = (26)
we can obtain: gq

α̇ = θ̇ − θ̇d = q cos φ − r sin φ − θ̇d (21)

where k3 > 0 and ̂ d is the estimations of d which is feedback from NDO.
1 2 Utilizing (25) and (7), the counterweight position can be determined.
Define the second CLF as: V2 := V1 + 2α whose time derivative is
With the chosen control law, we obtain:
( ) ⎛ ⎞
u
V̇ 2 = V̇ 1 + αα̇ = − k1 e2 − ueα + pe + α q cos φ − r sin φ − θ̇d u ⎜ − k3 β− fq + q̇ − α − ̂
d ⎟
U0 V̇ 3 = − k1 e2 − k2 α2 +pα +β⎝fq +gq d
− q̇d + α cosφ+d⎠
( ) U0 gq
= − k1 e2 + α q cos φ − r sin φ − θ̇d − ue
u
=− k1 e2 − k2 α2 − k3 β2 +pα + αβ(cos φ− 1)+β̃
d
U0
The second virtual control law is chosen as follows:
{
β = q − qd (22a) Let pβ = pα + αβ(cos φ − 1) (27)
qd = − k2 α + θ̇d + ue (22b)
Then:

where k2 > 0 is control gain, hence V̇ 3 = −

u
k1 e2 − k2 α2 − k3 β2 + pβ + β̃
d (28)
( ) U0
u
V̇ 2 = − k1 e2 + pe + α (β + qd )cos φ − r sin φ − θ̇d − ue
U0
2
u
( ) Step 4. Define the fourth CLF as: V4 = V3 + 12̃ d
=− k1 e2 + pe + αβ cos φ + α qd cos φ − sin φr − θ̇d − ue Differentiating the fourth CLF and utilizing (15) yields
U0
u ( )
(( ) ) ˙
u V̇ 4 = V̇ 3 + ̃
d̃
d= − k1 e2 − k2 α2 − k3 β2 + pβ + β̃
d+̃
d ḋ − L̃
d
=− k1 e2 + pe + αβ cos φ + α − k2 α + θ̇d + ue cos φ − r sin φ − θ̇d − ue U0
U0 u 2
= − k1 e2 − k2 α2 − k3 β2 − L̃
d + pβ + β̃ d +̃ dḋ
(( ) ) U0
u
=− k1 e2 + pe + αβ cos φ − k2 α2 + α θ̇d + ue (cos φ − 1) − r sin φ
U0

(( ) )
Set pα = pe + α θ̇d + ue (cos φ − 1) − r sin φ (23) Continue to set pd = pβ + β̃
d+̃
dḋ (29)

We get the final result as

The second CLF can be rewritten as follows:
u 2
u V̇ 4 = − k1 e2 − k2 α2 − k3 β2 − L̃
d + pd (30)
V̇ 2 = − k1 e2 − k2 α2 + pα + αβ cos φ (24) U0
U0

6
H.N. Tran et al. Ocean Engineering 220 (2021) 108257

Then we can choose the control gains kj (j = 1, 2, 3) that satisfy the

3.3. Stability analysis
following conditions:
⎧ ( )
To facilitate the analysis of stability in the next step, the following u 1 ⎪ 1 U0
k1 − − ε2 umax Δφ > 0 ⎪ k1 > + ε2 umax Δφ
assumptions are required: U0 4ε1 ⎪
⎪
⎪
⎪ 4ε 1 u
⎪
⎪
⎪
Assumption 1. The model uncertainties and the linearization errors k2 −
umax Δφ
−
1
−
Δφ ⎪ u
> 0 ⎨ k2 > max φ +
Δ 1
+
Δ φ
⃒ ⃒
⃒ ⃒ 4ε2 4ε3 4ε4 ⇔ 4ε2 4ε3 4ε4 (34)
are bounded, so there exists a positive constant Δḋ that satisfies ⃒ḋ⃒ ≤ Δḋ . ⎪
⎪
k3 − ε4 Δφ − ε5 > 0 ⎪
⎪ k > ε Δ + ε
⎪
⎪ 3 4 φ 5
⎪
Assumption 2. The predefined desired depth is finite, such that |zd |, 1 1 ⎪
⎪
⎪ 1 1
⃒ ⃒ ⃒ ⃒ ⃒ ⃒ L− − >0 ⎩ L> +
⃒ ⃒ ⃒ ⃒ ⃒ ⃒ 4ε5 4ε6 4ε5 4ε6
⃒żd ⃒, and ⃒z̈d ⃒ are bounded. Thus, from (18b), it can be inferred that ⃒θ̇d ⃒ is
⃒ ⃒ ⃒ ⃒ ⃒ ⃒
⃒ ⃒
⃒ ⃒
⃒ ⃒
⃒ ⃒ Denote
bounded, and we can set ⃒⃒żd ⃒⃒ ≤ Δzd and ⃒⃒θ̇d ⃒⃒ ≤ Δθ̇d . ⎡( u ) ( )
⎧ 1 umax Δφ 1 Δφ ⎤
⎪
⎪ k − − ε u Δ , k − − − ,
⎪ ⎢ U0 1 4ε1 2 max φ 2
4ε2 4ε3 4ε4 ⎥
Assumption 3. Due to the mechanical properties of AUV2000 when ⎪
⎪
⎪
⎪
⎢ ⎥
⎨ ρ := min⎢ ⎥>0
designed, the pitch angle during operation will be bounded, specifically, ⎢
⎣ ( ) ⎥
⎦
( ) 1 1
⎪ (k − ε Δ − ε ), L − −
|θ| ≤ 700 . Therefore, |θ − sin θ| ≤ 18
7π 7π
− sin 18 . ⎪
⎪
⎪
⎪
3 4 φ 5
4ε5 4ε6
⎪
⎪
⎩ ( )2
Assumption 4. The surge velocity u of AUV is non-negative and will
2
μ := ε1 Δe + ε3 Δθ̇ Δφ + Δr + ε6 Δd˙ 2
converge to the vicinity of the desired operating speed U0 during (35)
movement (Jalving, 1994; Fossen, 1994; Prestero, 2001; Qiao et al.,
From (33) and (35), it can be inferred that
2018). Let the maximum speed of AUV be umax , we have |u| ≤ umax and
⃒ ⃒
⃒u ⃒
⃒ − 1⃒ ≤ 1. V̇ 4 ≤ − 2ρV4 + μ (36)
⃒U 0 ⃒
Considering the tracking error vector that contains the error states of
Assumption 5. The linear and angular velocities of AUV are bounded both the control and observer systems as follows:
during operation. Hence, there exist the small positive constants Δv , Δw , [ ]T
and Δr , such that the following conditions is satisfied: |v cos θ sin φ| ≤ Δv , Ω : = e, α, β, ̃
d
|w cos θ cos φ| ≤ Δw and |r sin φ| ≤ Δr .
The fourth CLF can be further derived as V4 = 12‖Ω‖2 , which
Moreover, from Remark 2, it can be indicated that |cos φ − 1| ≤ Δφ
employing the Comparison Lemma in Khalil (2002), yields.
where Δφ is a very small positive constant. From (20), (23), (27) and
V4 (t) ≤ V4 (0)e− 2ρt + 2μρ for. t ∈ [0, tfinal )
(29), we have:
( ( )) Thus,
u √̅̅̅
pd = e u(θ − sin θ) + v cos θ sin φ + w cos θ cos φ + żd − 1 μ [ )
U0 ‖Ω(t)‖ ≤ ‖Ω(0)‖e− ρt
+ , t ∈ 0, tfinal (37)
(( ) ) ρ
+α θ̇d + ue (cos φ − 1) − r sin φ + αβ(cos φ − 1) + β̃ d+̃dḋ equation (37) means that the tracking error vector remain in a
bounded set around zero, which can be reduced by increasing ρ and
Utilizing these assumptions above, we can obtain: decreasing μ, in other words, by tuning the control and observer gain
( (
7 7
) ) kj (j = 1, 2, 3), L and choosing εi (i = 1, 2, 3, 4, 5, 6). Obviously, it can be
pd ≤ |e| umax π − sin π + Δv + Δw + Δzd + |α|((Δθ̇d + umax |e|)Δφ + Δr ) concluded that the tracking error vector converges to a specified
18 18
⃒ ⃒ ⃒ ⃒
⃒ ⃒ ⃒⃒̃ compact set around the origin with the estimated law in (14) and control
+|αβ|Δφ + ⃒β̃ d⃒ + ⃒d⃒Δd˙ laws in (18), (22), (26).
(31) Remark 3. It is worth noting that the control laws (18), (22), and (26)
( )
do not contain the roll angle, thereby the control signal is independent of
Let Δe = umax 7
18
7
π − sin 18 π + Δv + Δw + Δzd and using the Young the variation of the roll rate and the roll angle during maneuvering. This
inequality, we get: makes the control signal smoother and more accessible to apply in
⃒ ⃒ ⃒ ⃒ practice.
⃒ ⃒ ⃒⃒̃
pd ≤ |e|Δe + umax |α||e|Δφ + |α|(Δθ̇d Δφ + Δr ) + |αβ|Δφ + ⃒β̃ d⃒ + ⃒d⃒Δd˙
e 2 ( 2
α
)
α 2 ( 2
α 4. Simulation results
≤ + ε1 Δ2e + + ε2 e2 umax Δφ + + ε3 (Δθ̇d Δφ + Δr )2 +
4ε1 4ε2 4ε3 4ε4
) This section presents the numerical simulations of the proposed
1 ̃2 1 ̃2
+ ε4 β2 Δφ + ε5 β2 + d + d + ε6 Δd˙ 2 controller apply to AUV2000 on Matlab/Simulink. By analyzing the
4ε5 4ε6
simulation results, the effectiveness, feasibility, and stability of the
(32) proposed method are clarified. In this study, the desired operating speed
of the vehicle is considered to be U0 = 1.28(m /s). The control and
where εi (i = 1, 2, 3, 4, 5, 6) are positive designed constants. Substituting
observer gains selected for the controller are k1 = 3, k2 = 7.5, k3 = 0.15,
(32) into (30) yields:
L = 5. To ensure objectivity and completeness, in these simulations, the
( ) ( )
u 1 umax Δφ 1 Δφ 2 vehicle is expected to track different desired paths in the vertical plane.
V̇ 4 ≤ − k1 − − ε2 umax Δφ e2 − k2 − − − α − (k3
U0 4ε1 4ε2 4ε3 4ε4 Besides, the initial condition of the AUV considered for all simulation is
( ) as follows:
1 1 ̃2 ( )2
− ε4 Δφ − ε5 )β2 − L − − d + ε1 Δ2e + ε3 Δθ̇ Δφ + Δr {
4ε5 4ε6
[x, y, z, φ, θ, ψ ]T = [0, 0, 0, 0, 0, 0]T
+ ε6 Δd˙ 2 [u, v, w, p, q, r]T = [0, 0, 0, 0, 0, 0]T
(33)

7
H.N. Tran et al. Ocean Engineering 220 (2021) 108257

Fig. 5. The depth response when the desired depth is constant by the proposed controller (blue solid line) and the controller in Tran et al. (2019) (red solid line) with
desired depth (black dotted line). (For interpretation of the references to colour in this figure legend, the reader is referred to the Web version of this article.)

Fig. 6. The state response of AUV2000 when the desired depth is constant by the proposed controller (blue solid line) and the controller in Tran et al. (2019) (red
solid line). (For interpretation of the references to colour in this figure legend, the reader is referred to the Web version of this article.)

Fig. 7. The depth response with the composite desired depth by the proposed controller (blue solid line) and the controller in Tran et al. (2019) (red solid line). (For
interpretation of the references to colour in this figure legend, the reader is referred to the Web version of this article.)

8
H.N. Tran et al. Ocean Engineering 220 (2021) 108257

Fig. 8. The state response of AUV2000 with the composite desired depth by the proposed controller (blue solid line) and the controller in Tran et al. (2019) (red solid
line). (For interpretation of the references to colour in this figure legend, the reader is referred to the Web version of this article.)

Fig. 9. The depth response with the composite desired depth by the proposed controller with NDO (blue solid line) and without NDO (red solid line) in the presence
of the model uncertainties. (For interpretation of the references to colour in this figure legend, the reader is referred to the Web version of this article.)

Fig. 10. The state response of AUV2000 with the composite desired depth by the proposed controller with NDO (blue solid line) and without NDO (red solid line) in
the presence of the model uncertainties. (For interpretation of the references to colour in this figure legend, the reader is referred to the Web version of this article.)

9
H.N. Tran et al. Ocean Engineering 220 (2021) 108257

4.1. Simulations in the absence of the model uncertainties 5. Conclusion

Because of the similarity when using the nonlinear algorithm to build The paper provides a new approach to building the depth tracking
the controller for AUV2000, the comparison of control performances are controllers for AUV2000 - a hybrid AUV type. The proposed controller
constructed between the proposed backstepping controller and the has been applied to the 6-DOF AUV model to ensure nonlinearity and
controller developed in Tran et al. (2019). strongly coupled in the motion equations as well as a feasibility study for
practical applications. The paper also considered the effect of roll angle
4.1.1. The desired depth is constant and model uncertainty on the controller and proved the stability of the
Fig. 5 shows that both the Tran’s controller and the backstepping developed controller. The numerical simulations have been carried out
controller perform consistently with the 6-DOF model, not only speeding and validated to verify the effectiveness and robustness of the proposed
up the convergence process of the vehicle to the desired depth but also method. Moreover, the simulations results demonstrate the proposed
stabilizing the equilibria. However, the proposed backstepping control law has superior performance to previous ones and solves the
controller has better control performance, such as no overshoot and depth control problem more fully for the depth tracking problem in the
faster convergence. Because Fig. 6 shows the pitch angle during presence of the model uncertainties and propeller torque’s effect.
maneuvering is always within the limit (smaller than 70◦ ), this ensures
that the AUV will not fall into a singular point. Moreover, the roll angle Declaration of competing interest
overshoots when counterweight displaces abruptly then recovers to an
equilibrium. Therefore, it indicates that the proposed controller can The authors declare that they have no known competing financial
operate consistently, even when including the involvement of the roll interests or personal relationships that could have appeared to influence
angle, roll rate, sway velocity, yaw rate. the work reported in this paper.

4.1.2. The desired depth is not constant (żd ∕

= 0) Acknowledgement
In this case study, we examine the composite desired depth, which is
a combination of three scenarios are as follows: 1) zd = 5 + 0.3t, This research is funded by Vietnam National University Ho Chi Minh
(0 ≤ t < 50); 2) zd = 30 − 0.2t, (50 ≤ t < 100); 3)zd = 30 − city (VNU-HCM) under grant number B2018-20b-01. We acknowledge
20 cos(0.005π t − 0.5π), (t ≥ 100). The results in Fig. 7 illustrate the the support time and facilities from Ho Chi Minh University of Tech­
disadvantage of most classical controllers (Prestero, 2001; Tanakitkorn nology (HCMUT), VNU-HCM; Laboratory of Advance Design and
et al., 2017; Tran et al., 2019; Qiao et al., 2018) that it is incapable of Manufacturing Processes, HCMUT; National Key Lab. of Digital Control
making the vehicle converge on the desired depth when the desired and System Engineering (DCSELAB), HCMUT for this study.
depth is not constant. Meanwhile, the proposed controller not only en­
sures excellent tracking performance for AUV but also keeps AUV’s roll References
angle in a bounded set around zero, as shown in Fig. 8.
Through all the simulation results above, the proposed method has Cao, J., Su, Y., Zhao, J., 2011. Design of an adaptive controller for dive-plane control of a
torpedo-shaped AUV. J. Mar. Sci. Appl. 10 (3), 333–339. https://doi.org/10.1007/
proved its superiority compared to the previous ones because it enables s11804-011-1077-y.
AUV to track straight lines, even with the curve paths. Besides, the Eichhorn, M., Ament, C., Jacobi, M., Pfuetzenreuter, T., Karimanzira, D., Bley, K.,
tracking error always converges to zero and has fast convergence speed Boer, M., Wehde, H., 2018. Modular AUV system with integrated real-time water
quality analysis. Sensors 18 (6), 1837. https://doi.org/10.3390/s18061837.
(about 20 s) with many different simulation conditions, thus the effi­ Fossen, T., 1994. Guidance and Control of Ocean Vehicles. Wiley, NY, USA.
ciency and stability of the proposed control law are verified. Moreover, Gharesi, N., Ebrahimi, Z., Forouzandeh, A., Arefi, M.M., 2017. Extended state observer-
it is easily observed that the velocity u is changed when the pitch angle based backstepping control for depth tracking of the underactuated AUV. In: 2017
5th International Conference on Control, Instrumentation, and Automation (ICCIA),
alters, thereby using U0 in the controller design instead of using u in 2018-Janua, pp. 354–358. https://doi.org/10.1109/ICCIAutom.2017.8258706.
formula (18b) delivers the control signal smoother without affecting the Hagen, P.E., Storkersen, N., Vestgard, K., Kartvedt, P., 2003. The HUGIN 1000
system stability. autonomous underwater vehicle for military applications. Oceans 2003. Celebrating
the Past …. Teaming Toward the Future (IEEE Cat. No.03CH37492) 2 (2027),
1141–1145. https://doi.org/10.1109/OCEANS.2003.178504, 2.
Hong, E.Y., Soon, H.G., Chitre, M., 2010. Depth Control of an Autonomous Underwater
4.2. Simulations in the presence of the model uncertainties Vehicle, STARFISH. OCEANS’10 IEEE SYDNEY, pp. 1–6. https://doi.org/10.1109/
OCEANSSYD.2010.5603566.
Jalving, B., 1994. The NDRE-AUV flight control system. IEEE J. Ocean. Eng. 19 (4),
To evaluate the controller’s robustness, Joe et al. (2014) varied the 497–501. https://doi.org/10.1109/48.338385.
hydrodynamic coefficients by 20% to their actual values, Qiao et al. Joe, H., Kim, M., Yu, S., 2014. Second-order sliding-mode controller for autonomous
(2018) and Mahapatra and Subudhi (2018) considered the uncertainties underwater vehicle in the presence of unknown disturbances. Nonlinear Dynam. 78
(1), 183–196. https://doi.org/10.1007/s11071-014-1431-0.
of the model parameters by 20–25% and 30% of its nominal value, Khalil, H., 2002. Nonlinear Systems. Prentice Hall PTR.
respectively. In this paper, to verify the disturbance resistance and Lapierre, L., 2009. Robust diving control of an AUV. Ocean. Eng. 36 (1), 92–104. https://
robustness of the proposed controller, the parameters of fq and gq in doi.org/10.1016/j.oceaneng.2008.10.006.
Liceaga-Castro, E., van der Molen, G.M., 1995. Submarine H/sup ∞/depth control under
formula (12) deviate by 30–40% from their accurate value. wave disturbances. IEEE Trans. Contr. Syst. Technol. 3 (3), 338–346. https://doi.
In Fig. 9, the desired depths of several scenarios are as follows: 1) org/10.1109/87.406981.
zd = 20, (0 ≤ t < 60); 2) zd = 35 − 0.25t, (60 ≤ t < 100); 3) zd = 30 − Mahapatra, S., Subudhi, B., 2018. Design and experimental realization of a backstepping
nonlinear H ∞ control for an autonomous underwater vehicle using a nonlinear
20 cos(0.005π t − 0.5π), (t ≥ 100). It is noticeable from Fig. 9 that the matrix inequality approach. Trans. Inst. Meas. Contr. 40 (11), 3390–3403. https://
proposed controller still works without NDO, yet it suffers from a small doi.org/10.1177/0142331217721315.
steady-state error. Besides, Fig. 10 shows that there is hardly any dif­ Mahapatra, S., Subudhi, B., Rout, R., Kumar, B.V.S.S.K., 2016. Nonlinear H∞ control for
an autonomous underwater vehicle in the vertical plane. IFAC-PapersOnLine 49 (1),
ference in the state variables of the two controllers. These simulation 391–395. https://doi.org/10.1016/j.ifacol.2016.03.085.
results sufficiently validate the strong robustness of the backstepping Mondal, K., Banerjee, T., Panja, A., 2019. Autonomous underwater vehicles: recent
controller. In addition, the proposed controller is combined with NDO to developments and future prospects. Int. J. Res. Appl. Sci. Eng. Technol. 7 (11),
215–222. https://doi.org/10.22214/ijraset.2019.11036.
compensate for the systematical parametric uncertainties, thereby
Moreira, L., Soares, C., 2008. $H_{2}$ and $H_{\infty}$ designs for diving and course
eliminating the steady-state error and maintaining the control perfor­ control of an autonomous underwater vehicle in presence of waves. IEEE J. Ocean.
mance. In conclusion, the outstanding characteristics of the proposed Eng. 33 (2), 69–88. https://doi.org/10.1109/JOE.2008.918689.
controller are preserved despite the presence of the model uncertainties
and propeller torque’s effect.

10
H.N. Tran et al. Ocean Engineering 220 (2021) 108257

Naik, M.S., Singh, S.N., 2007. State-dependent Riccati equation-based robust dive plane Tran, N.H., Huynh, T.D., Ton, T.P., Huynh, T.H., 2019. Design of depth control for hybrid
control of AUV with control constraints. Ocean. Eng. 34 (11–12), 1711–1723. AUV. In: Proceedings of the 6th International Conference on Advanced Engineering -
https://doi.org/10.1016/j.oceaneng.2006.10.014. Theory and Applications 2019. Bogotá, Colombia.
Newman, J.N., 1977. In: Marine Hydrodynamics, first ed. MIT Press, Cambridge. Wadoo, S.A., Sapkota, S., Chagachagere, K., 2012. Optimal control of an autonomous
Petrich, J., Stilwell, D.J., 2011. Robust control for an autonomous underwater vehicle underwater vehicle. In: 2012 IEEE Long Island Systems, Applications and
that suppresses pitch and yaw coupling. Ocean. Eng. 38 (1), 197–204. https://doi. Technology Conference (LISAT), vol. 3, pp. 1–6. https://doi.org/10.1109/
org/10.1016/j.oceaneng.2010.10.007. LISAT.2012.6223100.
Prestero, T., 2001. Verification of a six-degree of freedom simulation model for the Wei, C., Yanhui, W., Jianhui, Z., 2015. Back-stepping control of underactuated AUV’s
REMUS autonomous underwater vehicle. Ph.D. Thesis. In: Massachusetts Institute of depth based on nonlinear disturbance observer. In: 2015 34th Chinese Control
Technology. Conference (CCC), 2015-Septe, pp. 6061–6065. https://doi.org/10.1109/
Qiao, L., Ruan, S., Zhang, G., Zhang, W., 2018. Robust H2 optimal depth control of an ChiCC.2015.7260587.
autonomous underwater vehicle with output disturbances and time delay. Ocean. Wynn, R.B., Huvenne, V.A.I., Le Bas, T.P., Murton, B.J., Connelly, D.P., Bett, B.J.,
Eng. 165, 399–409. https://doi.org/10.1016/j.oceaneng.2018.07.019. Ruhl, H.A., Morris, K.J., Peakall, J., Parsons, D.R., Sumner, E.J., Darby, S.E.,
Rout, R., Subudhi, B., 2017. NARMAX self-tuning controller for line-of-sight-based Dorrell, R.M., Hunt, J.E., 2014. Autonomous Underwater Vehicles (AUVs): their
waypoint tracking for an autonomous underwater vehicle. IEEE Trans. Contr. Syst. past, present and future contributions to the advancement of marine geoscience.
Technol. 25 (4), 1529–1536. https://doi.org/10.1109/TCST.2016.2613969. Mar. Geol. 352, 451–468. https://doi.org/10.1016/j.margeo.2014.03.012.
Silvestre, C., Pascoal, A., 2007. Depth control of the INFANTE AUV using gain-scheduled Yan, Z., Yu, H., Hou, S., 2016. Diving control of underactuated unmanned undersea
reduced order output feedback. Contr. Eng. Pract. 15 (7), 883–895. https://doi.org/ vehicle using integral-fast terminal sliding mode control. J. Cent. S. Univ. 23 (5),
10.1016/j.conengprac.2006.05.005. 1085–1094. https://doi.org/10.1007/s11771-016-0358-7.
Sname, 1950. Nomenclature for Treating the Motion of A Submerged Body through A Yu, J., Liu, J., Wu, Z., Fang, H., 2018. Depth control of a bioinspired robotic dolphin
Fluid. Technical Report Bulletin 1-5. Society of Naval Architects and Marine based on sliding-mode fuzzy control method. IEEE Trans. Ind. Electron. 65 (3),
Engineers, NY, USA. 2429–2438. https://doi.org/10.1109/TIE.2017.2745451.
Tanakitkorn, K., Wilson, P.A., Turnock, S.R., Phillips, A.B., 2017. Depth control for an Zhang, F., Marani, G., Smith, R.N., Choi, H.T., 2015. Future trends in marine robotics [TC
over-actuated, hover-capable autonomous underwater vehicle with experimental spotlight]. IEEE Robot. Autom. Mag. 22 (1), 14–122. https://doi.org/10.1109/
verification. Mechatronics 41 (3), 67–81. https://doi.org/10.1016/j. MRA.2014.2385561.
mechatronics.2016.11.006.

11