2014 IEEE 28-th Convention of Electrical and Electronics Engineers in Israel

Non-linear Controller for Non-linear Model of

Hovering Autonomous Underwater Vehicles
Boris Braginsky and Hugo Guterman
Department of Electrical and Computer Engineering
Ben-Gurion University of the Negev
Beer-Sheva, Israel
borisbr@bgu.ac.il, hugo@ee.bgu.ac.il

Non-linear controllers for attitude, altitude, and position of a while the water weight is almost neutral. The seven thrusters
Hovering Autonomous Underwater Vehicle (HAUV) non-linear placed on the hull provide six degrees of freedom (DOFs).
model are presented. The controllers allow the vehicle to reach any
desired point in the horizontal and vertical planes. The controllers
were applied to a small-sized, torpedo-shaped HAUV with six
Communication
degrees of freedom. The coupling between attitude and position
makes the trajectory tracking control problem particularly hard. module
However, at least in the case of set-point control, it is possible to
adopt a two decouple control strategy. A non-linear controller is
applied for achieving attitude regulation, while a non-linear control Thrusters
strategy is used for achieving smooth motion back to the desired
position coordinates. The performance of the proposed control
Image Sonar
scheme is evaluated numerically.
Index Terms—HAUV; Non-linear Control; Robotic arm;
Robotic arm
I. INTRODUCTION
The field of Unmanned Underwater Vehicles (UUVs) is of
increasing interest to the scientific community due to its many
exciting applications. These vehicles are capable of performing
complex missions in spite of the many limitations of embedded object
sensors, processing, and control [1]. UUVs can be divided into
two groups: Remotely Operated Vehicles (ROVs), underwater
vehicles physically linked, via a tether, to an operator; and
Autonomous Underwater Vehicles (AUVs), which are able to
navigate autonomously. These vehicles have civilian and
military applications and perform specific tasks such as search Fig. 1. Torpedo-shaped HAUV with robotic arm.
and rescue in high risk areas, autonomous sensing for weather
forecasting, maintenance and fault detection of marine The coupling between the attitude and the position makes
platforms and pipelines (oil and gas), and underwater the trajectory tracking control problem particularly hard.
archeology [2]. Generally, locomotion of the AUV is linear, However, at least in the case of set-point control, it is possible
while the ROV has hovering capabilities. Lately, the Hovering to adopt a two decouple control strategy. A non-linear
AUVs (HAUV) have been proposed. The HAUV has several controller is applied for achieving attitude regulation and an
advantages over regular AUVs, such as maneuverability (it is independent non-linear control strategy is applied for achieving
possible move in any direction) and hovering [3] [4] [5]. smooth motion back to the desired position coordinates. The
In this study, non-linear controllers for the attitude and proposed control scheme performance is evaluated
position of an HAUV (see Fig. 1) non-linear model are numerically.
presented. These controllers allow the vehicle to reach any II.MODELING
desired point in the horizontal and the vertical planes. The
platform is equipped with state-of-the-art sensors and In this section, the dynamic equations of the AUV’s motion are
embedded processors that are used to control the HAUV. outlined. In general, any movement of a vehicle in a 3D space
Propulsion is generated by seven brushless DC motors, two for involves six DOFs. It is convenient to define two coordinate
each of the X, Y directions and three for the Z axis (Fig. 2). frames, as shown in Fig. 2.
The HAUV dimensions are 2 meters in length and 30
centimeters in diameter, and the air weight is about 80 kg,

Authorized licensed use limited to: Central Mechanical Engineering Research Institute. Downloaded on October 21,2022 at 07:47:05 UTC from IEEE Xplore. Restrictions apply.
 !
Euler vector rate of change 𝜁 = 𝜙, 𝜃, 𝜓 . Assuming that
! !
𝜃 ∈ − ,  , Eq. (3) is obtained and since det 𝐿!" = 1/ cos 𝜃
! !
is the relationship between the angular velocity and the Euler
angle, rates may be inverted provided that 𝜃 ≠ 𝜋/2. Assuming
that this is the case,

𝜁 = 𝐿!" 𝜁 𝜔! (3)
where
1 sin 𝜙 tan 𝜃 cos 𝜙 tan 𝜃
𝐿!" = 0 cos 𝜙 − sin 𝜙 (4)
0 sin 𝜙 sec 𝜃 cos 𝜙 sec 𝜃

As the vehicle motion does not normally approach the

Fig. 2. The HAUV body frame and the motor thrust forces.
The body frame mechanical system motion is given by Eq.
(1).
𝑚𝑣! + 𝜔! ×𝑚𝑣! = 𝐹!
𝐽𝜔! + 𝜔! ×𝐽𝜔! = 𝑀!
where J is the inertia tensor with respect to the body frame.
The above equations describe how the forces and
moments affect the translational and rotational velocity of the
rigid body. The kinematic equations that allow relating
quantities defined in terms of the body coordinate system to
quantities in the inertial system, and vice versa, must be stated.
Roll, pitch, and yaw are the rotations of the body about the x,
y, and z axes, respectively. The transformation 𝑅!! (described
in Eq. (2)) was calculated by cascading the 3 separate angular
transformation matrices. The order of the rotations from the
inertial frame of reference was: rotation about the z-axis with
the yaw angle 𝜓, then rotation about the y-axis with the pitch
angle 𝜃, and, finally, rotation about the x-axis with the roll
angle  𝜙. The following coordinate transform relates a vector
in body-fixed coordinates with a vector in inertial or earth-
fixed coordinates [6].
𝑐𝜃𝑐𝜓 𝑐𝜓𝑠𝜃𝑠𝜙 − 𝑐𝜙𝑠𝜓 𝑐𝜓𝑠𝜃𝑐𝜙 + 𝑠𝜙𝑠𝜓
𝑅!! = 𝑐𝜃𝑠𝜓 𝑐𝜙𝑐𝜓 + 𝑠𝜃𝑠𝜙𝑠𝜓 −𝑠𝜙𝑐𝜓 + 𝑠𝜃𝑐𝜙𝑠𝜓 (2)
−𝑠𝜃 𝑐𝜃𝑠𝜙 𝑐𝜃𝑐𝜙
where 𝑠𝛼 = sin 𝛼 and 𝑐𝛼 = cos 𝛼.
Let the Euler angles vector be 𝜁 = 𝜙, 𝜃, 𝜓 ! , the rotation
speed vector be 𝜔! = 𝑝, 𝑞, 𝑟 ! , and the linear velocity in
body axes be 𝑣! = 𝑢, 𝑣, 𝑤 ! . We wish to express the
relationship between the body angular velocity 𝜔! and the
singularity condition this is not a serious constraint. If this
situation occurs, then it would become necessary to model the
vehicle motion using extreme pitch angles, and the analysis
could then resort to an alternative kinematics representation,
such as quaternions.
Applying Eq. (1) and Eq. (3), the motion of the vehicle is
described by Eq. (5).
𝐹!
𝑣! = −𝜔! ×𝑣! +
𝑚
𝜁 = 𝐿 𝜁 𝜔!
𝜔! = −𝐽 𝜔! ×𝐽𝜔! + 𝐽!! 𝑀!
!! (5)
(1) Since the water vehicle is considered a rigid body, the
force F! and the moment 𝑀! are due to the action of the
hydrodynamic, propulsive, gravitational, and buoyancy field
forces. In the current analysis the gravitational and buoyancy
forces are neglected, as the vehicle is almost neutral in the
water. The result of the combined external forces and
moments is described as follows [7]:
𝐹! = [𝐹!"#$%&'( + 𝐹!"#$#%&' + 𝐹!!!"#$ +   𝐹!"# ]
𝑀! = [𝑀!"#$%&'( + 𝑀!"#$#% + 𝑀!!!"#$ + 𝑀!"# ] (6)
In what follows, the cross product for 𝑐, 𝑑 ∈ 𝑅 ! is
expressed as a matrix operator, that is 𝑆 𝑐 𝑑 = 𝑐×𝑑 = −𝑑×𝑐,
where  𝑆 ∙ is a 3×3 skew symmetric matrix. Applying the
rotation matrix 𝑅!! in Eq. (2), the following state-space model
of the considering system is obtained:
𝜒! = 𝜒!
𝑅!! 𝜁 𝐹!
𝜒! =
𝑚
𝜁 = 𝐿!" 𝜁 𝜔!
𝜔! = 𝐽!! 𝑆 𝐽𝜔! 𝜔! + 𝐽!! 𝑀! (7)
where χ! = 𝑥, 𝑦, 𝑧 !  is the position vector of the vehicle
center of mass in terms of the inertial frame, 𝜒! = 𝑥, 𝑦, 𝑧 ! .

Authorized licensed use limited to: Central Mechanical Engineering Research Institute. Downloaded on October 21,2022 at 07:47:05 UTC from IEEE Xplore. Restrictions apply.
 With regard to the thrust produced by the 7 motors, the III.STABILIZING CONTROLLER FOR THE ATTITUDE SUBSYSTEM
force and torque vectors of thrust in Eq. (6) are given by:
A. Attitude regulation
𝑀!!!"#$ = 𝑟!"## 𝐹6 − 𝐹7 , 𝑟! 𝐹5 − (𝐹6 + 𝐹7) , 𝑟! 𝐹1 − 𝐹2 In the controller design it is assumed that all state
variables are measured. It is important to state that both the 𝜁
+ 𝑟! 𝐹3 − 𝐹4
and 𝜔! subsystems in Eq. (7) are independent of the position
𝐹!!!"#$ = 𝑏 𝐹1 + 𝐹2 , 𝑏 𝐹3 + 𝐹4 , 𝑏 𝐹5 + 𝐹6 + 𝐹7 (8) and velocity vectors 𝜒! , 𝜒! , respectively, while the attitude 𝜁
vectors are highly coupled with the vector 𝜔! .
where 𝑟! ,  𝑟! , 𝑎𝑛𝑑  𝑟! are the distances from the motors to the We concentrate now on the attitude subsystem, defined
body center of mass, 𝛼 is the angle of flipper for controlling by:
the roll, and 𝑏 > 0  is the thrust factor. The force and torque
vectors produced by changing balance while using the robotic 𝜁 = 𝐿!" 𝜁 𝜔!
arm in Eq. (6) are given by moment 𝑀!"# and force 𝐹!"# . 𝜔! = 𝐽 𝑆 𝐽𝜔! 𝜔! + 𝐽!! 𝑀!
!! (11)
Vehicle body lift results from the vehicle moving through
the water at an angle of attack, causing flow separation and a Following the approach and results in [10] the following
subsequent drop in pressure along the aft, upper section of the attitude controller is defined:
vehicle hull. The pressure drop is modeled as a point force
applied at the center of pressure. As this center of pressure 𝑀!!!"#! = − 𝐿!" 𝜁 𝐾𝜁 + 𝐵𝜔! + 𝑀!"#$%&'(
(12)
does not line up with the origin of the vehicle-fixed coordinate + 𝑀!"#$#% + 𝑀!"#
system, the force also leads to a pitching moment about the
origin. In our model, Hoerner’s estimation of body lift was where 𝐾 = 𝐾 ! , 𝐵 = 𝐵 ! > 0 are arbitrarily selected constant
used [8]. In vector form, the HAUV lift force and lift moments matrices. Using Eq. (11) and the Lyapunov condition, the
are as follows: stability of the system can be proved.
!
𝐹!"#$%&'( = 0, 𝑌!" 𝑢𝑣, 𝑍!" 𝑢𝑤
1 !
𝑀!"#$%&'( = 0, 𝑀!" 𝑢𝑤, 𝑁!" 𝑢𝑣 !
(9) 𝑉 𝜁, 𝜔! = 𝜁 𝐾𝜁 + 𝜔!! 𝐽𝜔! (13)
2

In vector form, the HAUV damping forces and moments Substituting 𝑀!!!"#$ Eq. (12) into Eq. (11), the derivative
are [9]: of 𝑉  along the trajectories of the resulting closed-loop system,
namely, 𝑉 = 𝜁 ! 𝐾𝜁 + 𝜔!! 𝐽𝜔! , satisfies:
𝐹!"#$#%&' = 𝑋! ! 𝑢 𝑢 , 𝑌! ! 𝑣 𝑣 + 𝑌! ! 𝑟 𝑟 , 𝑍! ! 𝑤 𝑤
+ 𝑍! ! 𝑞 𝑞   𝑉 = 𝜁 ! 𝐾𝐿!" 𝜁 𝜔! + 𝜔!! 𝑆 𝐽𝜔! 𝜔!
(14)
𝑀!"#$#% = 𝐾! ! 𝑝 𝑝 , 𝑀! ! 𝑤 𝑤 + 𝑀! ! 𝑞 𝑞 , 𝑁! ! +𝑤!! 𝑀!"#$%&'( + 𝑀!"#$#% + 𝑀!!!"#$ + 𝑀!"#
(10)
+ 𝑁! ! 𝑟 𝑟
when 𝑆 𝐽𝜔! is skew symmetric, then 𝜔!! 𝑆 𝐽𝜔! 𝜔! = 0
The variables from Eqs. (9) and (10) are described in Table 1. and 𝜁 ! 𝐾𝐿!" 𝜁 𝜔! = 𝜔!! 𝐿!!" 𝐾𝜔!  . Therefore Eq. (14) can be
written as:
Table I: Non-linear forces and moment coefficients.
 𝑉 = 𝜁 ! 𝐾𝐿!" 𝜁 𝜔! + 𝜔!! 𝑀!"#$%&'( + 𝑀!"!"!# + 𝑀!"#
𝑋! ! Axial drag [kg/m] 𝐾! ! Rolling Resistance
−𝜔!! 𝐿!" 𝜁 𝐾𝜁 + 𝐵𝜔! + 𝑀!"#$%&'( + 𝑀!"#$#% + 𝑀!"#
[kg*m2/rad2]
𝑌! ! Cross flow Drag 𝑀! ! Cross flow Drag [kg] =   −𝜔!! 𝐵𝜔! − (𝑀!"# − 𝑀!"# ) (15)
[kg/m]
𝑌! ! Cross flow Drag 𝑀! ! Cross flow Drag where 𝑀!"#  is the estimated moment that guarantees system
[kg*m/rad2] [kg*m2/rad2] stability. The subsystem will be stable if  𝑉 < 0, then we can
𝑌!" Body lift Force 𝑀!" Body lift Moment choose 𝑀!"# >   𝑀𝑎𝑥𝑖𝑚𝑢𝑚  𝑀!"# . Since the raise weight of
[kg/m] [kg] the arm is limited, the stability is guaranteed.
𝑍! ! Cross flow Drag 𝑁! ! Cross flow Drag [kg]
[kg*m/rad2] B. Position regulation
𝑍! ! Cross flow Drag 𝑁! ! Cross flow Drag Due to the non-linear coupling between the attitude and the
[kg/m] [kg*m2/rad2] position variables, the process of attitude regulation is
𝑍!" Body lift Force 𝑁!" Body lift Moment associated with drift in the 𝜒! = 𝑥, 𝑦, 𝑧 ! coordinates. To
[kg/m] [kg] reduce the resulting drift, an additional non-linear controller
for position regulation is proposed.

Authorized licensed use limited to: Central Mechanical Engineering Research Institute. Downloaded on October 21,2022 at 07:47:05 UTC from IEEE Xplore. Restrictions apply.
 The position subsystem is defined by: The results of the attitude controller are shown in Fig. 5, while
the behavior of the position controller appears in Fig. 6.
𝜒! = 𝜒!
𝑅!! 𝜁 𝐹!
𝜒! = (16)
𝑚

In order to stabilize the position subsystem, 𝐹!!!"# is

defined as:

𝐹!!!"#$ = [−𝐹!"#$%&'( − 𝐹!"#$#%&' − 𝐹!"#

(17)
− 𝑅!! 𝜁 !! (𝜒! 𝑚 + 𝜒! )]

where 𝐴 = 𝐴! > 0 are arbitrarily selected constant matrices. If

det 𝑅!! 𝜁 = 1  for all 𝜃  𝑎𝑛𝑑  𝜓, then 𝑅!! 𝜁 !! will always
exist. Using the Lyapunov stability condition the stability of
the system can be probed. The following Lyapunov function is
proposed: Fig. 3. Attitude set point tracking.

1 ! (18)
𝑉! = 𝜒 𝐴𝜒 + 𝜒!! 𝐴𝜒!
2 ! !

Substituting 𝐹!!!"#$ from Eq. (17) into Eq. (16), the

derivative of 𝑉!  along the trajectories of the resulting closed-
loop system, namely 𝑉! = 𝜒!! 𝐴𝜒! + 𝜒!! 𝐴𝜒! , satisfies:

𝜒!! 𝐴𝑅!! 𝜁 𝐹! 𝜒!! 𝐴(−𝜒! 𝑚 − 𝜒! )

𝑉! = 𝜒!! 𝐴𝜒! + = 𝜒!! 𝐴𝜒! +
𝑚 𝑚
= −𝜒!! 𝐴𝜒! − (𝐹!"# + 𝐹!"# )

where 𝐹!"#  is the estimated force that guarantees system

stability and 𝐹!"# is the arm’ force. The subsystem is stable if
 𝑉 < 0; then we can choose 𝐹!"# >   𝑀𝑎𝑥𝑖𝑚𝑢𝑚  𝐹!"# . Since
the raise weight of the arm is limited, the stability is
Fig. 4. Position drift during the attitude and position regulation.
guaranteed.
IV. SIMULATION RESULTS
Example 1: This example demonstrates the action of the
two controllers (attitude and position). The considered initial Arm moving
! !
condition is 𝜁 0 = 0, , 𝜋 (angles in radians) and the initial
!
position is 𝜒! = 0,0,0 . The attitude non-linear controller
Regular
tries to achieve zero attitude and position. The non-linear
controller
controller’s objective is to keep the HAUV in its initial
position. The results of the attitude controller are shown in
Fig. 3, while the behavior of the position controller appears in
Fig. 4.
Example 2: This example demonstrates the action of the
two controllers (attitude and position) while using a robotic
arm. The considered initial condition is 𝜁 0 =
! !
0, , 𝜋 (angles in radians) and the initial position is
!
𝜒! = 0,0,0 . The arm starts to move at t=40 sec. The attitude Fig. 5. Attitude set point tracking while using a robotic arm.
non-linear controller tries to achieve zero attitude and position.

Authorized licensed use limited to: Central Mechanical Engineering Research Institute. Downloaded on October 21,2022 at 07:47:05 UTC from IEEE Xplore. Restrictions apply.

Fig. 6. Position drift during the attitude and position regulation while using a
robotic arm.
V.CONCLUSIONS
The main goal of this research was to develop non-linear
controllers for stabilizing the attitude and the position of an
HAUV during hovering and while using a robotic arm. A
relatively simple controller for regulating the attitude and
position subsystems for the highly non-linear HAUV system
has been presented. Simulation results demonstrate the
controller performance.
VI. REFERENCES
[1] G. Antonelli and G. Antonelli, Underwater robots.
Springer, 2014.
Authorized licensed use limited to: Central Mechanical Engineering Research Institute. Downloaded on October 21,2022 at 07:47:05 UTC from IEEE Xplore. Restrictions apply.
[2] E. Desa, R. Madhan, and P. Maurya, "Potential of
autonomous underwater vehicles as new generation
ocean data platforms," Current science, vol. 90, pp.
1202-1209, 2006.
[3] I. Torres, J. Torres, and R. Lozano, "A new AUV
configuration with four tilting thrusters: Navigation
and hover tasks," in Cce, 2012, pp. 1-6.
[4] B. M. Ferreira, J. Jouffroy, A. C. Matos, and N. A.
Cruz, "Control and guidance of a hovering AUV
pitching up or down," in Oceans, 2012, 2012, pp. 1-
7.
[5] D. Maalouf, V. Creuze, and A. Chemori, "State
feedback control of an underwater vehicle for wall
following," in Control & Automation (MED), 2012
20th Mediterranean Conference on, 2012, pp. 542-
547.
[6] B. Etkin and L. D. Reid, Dynamics of flight: stability
and control: Wiley New York, 1982.
[7] R. McEwen and K. Streitlien, "Modeling and control
of a variable-length auv," Proc 12th UUST, 2001.
[8] S. F. Hoerner, Fluid-dynamic lift: Eigenverl., 1985.
[9] M. E. Rentschler, "Dynamic simulation modeling and
control of the Odyssey III autonomous underwater
vehicle," Massachusetts Institute of Technology,
Department of Mechanical Engineering, 2003.
[10] I. Zohar, A. Ailon, and H. Guterman, "Simple
controllers for the nonlinear model of quadrotor
helicopters," in Itzhack Y. Bar-Itzhack Memorial
Symposium, Israel, 2012.