Line-of-Sight Guidance for Path Following of

Marine Vehicles

Anastasios M. Lekkas and Thor I. Fossen

Abstract This paper presents an overview of the Line-of-Sight (LOS) guidance law,
which is a widely used method for generating heading reference trajectories, for
path-following applications of marine vehicles. Due to the fact that these reference
trajectories depend also on the form of the path, five path evaluation criteria are
discussed. The guidance problem is presented in a constructive manner. First, two
2-D cases are studied, therefore it is assumed that the horizontal-plane motion and
the vertical-plane motion are decoupled. The next step is to consider a 3-D path-
following scenario for a 5-DOF underwater vehicle where the motions are coupled.
The interconnection between the LOS guidance and the heading/pitch controller
is explained and a brief overview of depth control methods is given. Moreover, a
short review of integral LOS methods in order to compensate for unknown external
disturbances is presented and it is shown how course control can circumvent this
problem.

1 Introduction

Guidance systems are of utmost importance when considering the performance of

aerial, surface and underwater vehicles, regardless of the motion control scenario
involved. For more information on control motion scenarios, such as target track-
ing, path following, path tracking and path maneuvering, the interested reader is
referred to [7, 3, 42]. Path-following refers to the case where the control objective

Anastasios M. Lekkas
Department of Engineering Cybernetics, Norwegian University of Science and Technology, NO-
7491, Trondheim, Norway, e-mail: [email protected]
Thor I. Fossen
Centre for Autonomous Marine Operations and Systems, Department of Engineering Cybernetics,
Norwegian University of Science and Technology, NO-7491, Trondheim, Norway, e-mail: fos-
[email protected]

1
2 Anastasios M. Lekkas and Thor I. Fossen

is to converge to and follow a predefined path without involving any requirements

pertaining to temporal constraints. In order to achieve this, several guidance laws
have been studied and implemented in the past, with the missile community being
most likely the oldest and most active one on the field. As a matter of fact, some of
the most commonly used approaches by the marine control community have been
directly influenced by their missile community counterparts, such as the Line-of-
Sight (LOS) guidance, the Pure Pursuit (PP) guidance and the Constant Bearing
(CB) guidance. The implementation of the three aforementioned methodologies in
the case of autonomous underwater vehicles has been discussed extensively in [6].
The LOS guidance algorithm and its properties have been studied thoroughly in
the literature. Frequently, the LOS guidance is considered in connection with the
heading autopilot due to the fact that the two systems form together a cascade struc-
ture that needs to be stabilized. One of the first efforts of this kind was presented
in [21]. In [18], an implementation of the LOS guidance law for straight line path
following can be found, whereas in [4] the motion of a particle was considered and
the method was implemented for curved path following. A cascaded systems ap-
proach was also developed by [2] where the guidance system is interconnected with
a sliding mode controller in order to achieve global κ-exponential stability when the
desired path is a straight line in 3-D space. The concept of global κ-exponential sta-
bility was introduced in [37] and is equivalent to Uniform Global Asymptotic Sta-
bility (UGAS) and Uniformly Local Exponential Stability (ULES) combined. All
the aforementioned approaches assumed a constant lookahead distance. In general,
a small lookahead distance will induce more aggressive steering, hence resulting in
reaching the desired path faster, but it might also be the reason for unwanted oscil-
lations around the path. On the other hand, a large lookahead distance results in a
smoother steering which prevents unwanted oscillations, but the downside is slower
convergence to the path. The idea of combining these behaviors in a complementary
manner was investigated in recent works, such as [5],[34],[32] and [27] where the
authors discussed and implemented the possibility of using a time-varying looka-
head distance. These works, regardless of the theory each one of them was based
on, indicated that a time-varying lookahead distance can contribute to converging to
the desired path faster while at the same time reducing oscillations around the path.
Although the LOS guidance algorithm is efficient and has a very simple struc-
ture, when unknown forces (due to the wind or sea currents, for example) act on
the vehicle it is not possible (in the general case) to succeed in accomplishing the
motion control task, i.e. to converge to the desired path. A common and reasonable
strategy is then to augment the LOS guidance law by adding integral action in order
to eliminate any constant offsets induced by the constant, or slow time-varying, en-
vironmental forces. In [6] the addition of a simple integrator is discussed, whereas in
[1] a more sophisticated approach was followed resulting in a method that avoided
the integrator wind-up phenomenon.
This work presents an overview of the lookahead-based LOS guidance law in its
conventional form, when implemented for marine applications, and follows a con-
structive approach. Initially, two uncoupled 3-DOF models are considered, the first
pertains to guidance on the horizontal plane and the second pertains to guidance on
Line-of-Sight Guidance for Path Following of Marine Vehicles 3

the vertical plane. Stability proofs are given in both cases, assuming perfect heading
and pitch tracking respectively, and the result shows how the sideslip angle and the
angle of attack affect the performance. Next, a more general case in 3-dimensional
space is studied. The task is to converge to a straight line which is at a known depth
underwater. In this case the horizontal plane LOS guidance system is coupled with
the depth controller and the stability analysis is more complicated and it also de-
pends on the type of depth controller used.
In path-following applications, the shape and the properties of the path itself
have a great influence on the reference trajectories generated by the LOS algorithm.
Therefore, the path-planning problem is directly correlated to the guidance prob-
lem. It is often sufficient to design paths that consist of successive straight lines, this
selection offers simplicity since the path-tangential angle is constant for each line
and it also makes it easier to compensate for constant environmental disturbances,
such as winds and ocean currents. Stability proofs can also be more tractable when
using straight line paths. On the other hand, such paths result in a non-smooth ve-
locity function and, consequently, curvature, resulting in sudden increases of the
cross-track error when switching to the next waypoint. There is a vast literature on
the properties of the several path-planning methods and therefore a wide variety of
paths to choose from. Depending on the application and the constraints, properties
such as curvature continuity or minimum length might be more important to satisfy
than others. In this paper, paths based on straight lines are used to demonstrate the
efficiency of the LOS guidance. In addition, a section with a discussion on criteria
that can help the user evaluate the path candidates is given.
The rest of this paper is organized as follows: Section 2 presents the dynamic
model of the vessel. Section 3 discusses five path evaluation criteria than can be
considered when making a decision upon which path should be used. In Section
4, the 3-DOF LOS guidance is presented for a vessel navigating on the horizontal,
as well as the vertical, plane. Here the two cases are assumed to be uncoupled. In
Section 5, a three-dimensional scenario is studied and in this case the horizontal
LOS guidance system is coupled with the depth controller. Section 6 deals with
guidance under the influence of constant environmental disturbances and reviews
two LOS-based methods to tackle the problem. Furthermore, it gives an explanation
of why no integral action needs to be added when the course angle, instead of the
yaw angle, is controlled. Finally, Section 7 gives a brief summary of the paper.

2 Vehicle Models

This section presents the vessel model and the related assumptions that are consid-
ered in this paper. Two reference frames, namely, the North-East-Down (NED) coor-
dinate system {n} = (xn , yn , zn ) and the body-fixed reference frame {b} = (xb , yb , zb )
will be adopted in this paper to describe the motion, location and orientation of the
vehicle. The NED frame is defined as a tangent plane on the surface of the Earth
moving with the vehicle and is sufficient for local operations. Its origin is on and
4 Anastasios M. Lekkas and Thor I. Fossen

the x axis points towards the true North, the y axis points towards the true East and
the z axis points downwards, normal to the Earth’s surface. The body-fixed frame is
moving with the vehicle and its origin ob coincides with the center of gravity of the
vehicle, see also [16, Ch. 2].

2.1 Vehicle Dynamics

Similarly to [2], for the path-following task we can neglect the roll angle, hence for
an underactuated autonomous vehicle the following 5-DOF dynamic model can be
used:

η̇ = J(η)ν, (1)
Mν̇ + C(ν)ν + D(ν)ν + g(η) = τ, (2)

where M is the mass and inertia matrix, C(ν) is the Coriolis and centripetal matrix,
D(ν) is the damping matrix, g(η) describes the gravitational and buoyancy forces,
and τ includes the control forces and moments.
Accordingly, the generalized position and velocity are recognized as:

η = (x, y, z, θ , ψ)T , ν = (u, v, w, q, r)T , (3)

where (x, y, z) is the vehicle’s inertial position in Cartesian coordinates, θ is the

pitch angle and ψ is the yaw angle. In addition, u is the surge velocity, v is the sway
velocity, w is the heave velocity, q is the pitch rate and r is the yaw rate.

2.2 Kinematic Models

There are two kinematics models corresponding to (1) that are common to use. The
main difference between them is the incorporation of the effects of ocean currents.
The first model considers only absolute velocities and is the following:

ẋ = u cos (ψ) cos (θ ) − v sin (ψ) + w cos (ψ) sin (θ ), (4)

ẏ = u sin (ψ) cos (θ ) + v cos (ψ) + w sin (ψ) sin (θ ), (5)
ż = −u sin (θ ) + w cos (θ ), (6)
θ̇ = q, (7)
1
ψ̇ = r, cos (θ ) 6= 0. (8)
cos(θ )

As it is shown in [17], by incorporating relative velocities, the kinematics can be

rewritten as:
Line-of-Sight Guidance for Path Following of Marine Vehicles 5

ẋ = ur cos (ψ) cos (θ ) − vr sin (ψ) + wr cos (ψ) sin (θ ) + unc , (9)
ẏ = ur sin (ψ) cos (θ ) + vr cos (ψ) + wr sin (ψ) sin (θ ) + vnc , (10)
ż = −ur sin (θ ) + wr cos (θ ) + wnc , (11)
θ̇ = q, (12)
1
ψ̇ = r, cos (θ ) 6= 0. (13)
cos(θ )

where the relative velocities in surge, sway and heave (ur , vr , wr ) and their absolute
velocities counterparts are related as follows:

ur = u − ubc , vr = v − vbc , wr = w − wbc , (14)

with (ubc , vbc , wbc ) denoting the ocean current velocities in the body-fixed frame and
(unc , vnc , wnc ) denoting the ocean current velocities in the NED frame.
It is worth noting that, even in the presence of ocean currents, it can be easily
shown that the two kinematic models, (4)–(8) and (9)–(13) are equivalent. How-
ever, the overall analysis of the system, including the stability results of the LOS
algorithm, can change a lot depending on which one of the two is adopted. The
main criterion for choosing either one is the available measurements and what
states of the system are controlled. For instance, if Global Navigation Satellite Sys-
tem (GNSS) measurements are available, then the absolute velocities are measured.
Consequently, even if ocean current forces are present, it is possible to reduce the
uncertainty and simplify the analysis by controlling the course angle and using con-
ventional LOS guidance. For an underwater vehicle with only relative measurements
available though, the relative velocities model with yaw control and an augmented
LOS guidance law with integral action is is a more fitting option. These issues are
discussed in detail in Section 6.

3 Path Planning and Evaluation Criteria

3.1 Introduction

In the case of a path-following motion control scenario, it is important to decide

upon the path to be followed before designing and implementing a guidance algo-
rithm, this is commonly known as the path planning task. It is with respect to this
path that the guidance reference trajectories will be generated and, depending on
the overall task conditions and constraints, these paths can vary a lot. Frequently,
the first step is to introduce a given order of fixed points in space, namely the way-
points, and define the desired path as the sum of the successive straight lines that
connect these waypoints. This approach, although simple, might not be sufficient
for applications that demand high accuracy because the resulting path is not smooth.
There is a vast literature pertaining to this issue, as well as other important factors
6 Anastasios M. Lekkas and Thor I. Fossen

that arise and affect the performance in each considered case. For some missions,
for example, it is critical that the vehicle converges to and stays on the exact path,
whereas others are more concerned with finding the minimum length path, and so
on. As a consequence, many different solutions have been presented and each one
of them satisfies some desired properties that are prioritized.
Dubins showed, for instance, that for a particle with unity speed, the shortest
possible path that meets a maximum curvature bound between a starting position
with predefined orientation (starting pose) and a finishing position with predefined
orientation (finishing pose) consists of at most three pieces, each of which is either a
straight line or an arc of a circle of radius R > 0 [14]. However, the Dubins method-
ology does not result in curvature continuous paths due to the fact that a straight
line has a curvature κ = 0, whereas a circle arc has a curvature κ = 1/R. Hence,
there will be a jump in the curvature from 0 to 1/R when moving from the straight
line to the circle arc. The Euler spirals (also known as clothoids) is an alternative
approach that gives curvature continuous paths, but it is also more expensive from a
computational point of view because it has an open form solution which includes the
calculation of the Fresnel integrals. Other popular alternatives are the Pythagorean
Hodographs, first introduced in [15] and the potential field method. For a more de-
tailed and thorough treatment of path planning methods the reader is referred to [40]
and [24].

3.2 Path Evaluation Criteria

Given the variety of the methodologies available when it comes to deciding which
path is more suitable for an application, it is useful to introduce a few evaluation
criteria that can help the user make a more fitting decision. Some of the criteria
presented below have been mentioned also in [11].

3.2.1 Path Smoothness

This property is among the most important ones, due to the fact that it is directly
related to the vehicle dynamic constraints, hence it has motivated many researchers
to investigate new methodologies. Two notions can be used to describe the path
smoothness, namely, the parametric continuity and the geometric continuity. The
main difference is that parametric continuity refers to both the speed and the orien-
tation with which the parameter propagates through the path, whereas the geometric
continuity is not concerned with the parameter speed. Therefore, parametric con-
tinuity can be considered as a subset of geometric continuity. More specifically,
regarding geometric continuity:
• G0 : all subpaths are connected.
• G1 : the path tangential angle is continuous.
• G2 : the center of curvature is continuous.
Line-of-Sight Guidance for Path Following of Marine Vehicles 7

Accordingly for parametric continuity:

• C0 : all subpaths are connected.
• C1 : the velocity is continuous.
• C2 : the acceleration (hence, the curvature) is continuous.
A G1 curve looks identical to a C1 curve, however the speed magnitude is discontin-
uous for the G1 curve. Consequently, for a path-following scenario geometric con-
tinuity is equivalent to parametric continuity, but the same is not true for a motion
control scenario that includes temporal constraints, such as path-tracking.

3.2.2 Path Length

Following [30, Ch. 3], a regular arc x = x(t), a ≤ t ≤ b is rectifiable and its length
can be computed by the integral:
s
dx1 2 dx2 2 dx3 2
Z b   Z b      
dx
s=   dt = + + dt. (15)
a dt a dt dt dt
 

In many applications the goal is to design a path that will minimize (15) since this
can be associated with arriving to the destination faster, or consuming less energy.
However, this is not always true, as it can be demonstrated by the Zermelo naviga-
tion problem where the task is to find the trajectory that minimizes the travel time
when there is a constant force field (such as a steady current) present. In this case,
the straight line is not the optimal path anymore. Moreover, regarding the path length
optimality of the Dubins paths, it is worth noting that, depending on the application,
it is not always possible to find a Dubins path, as it was shown in [39].

3.2.3 Path Precision

Given a number of waypoints, path precision refers to whether the designed path
passes through all the waypoints (interpolating curve) or passes through some of
them and comes close to, but without passing through, the rest (approximating
curve). Such a property can be important in cases where the vehicle navigates in
an area with obstacles that have to be avoided.

3.2.4 Path Tractability

This property is concerned with two main issues: a) if the overall shape of the path
is of practical use, and b) what happens in the rest of the path if one of the way-
points changes location. The first branch pertains to any unnecessary zig-zagging or
wiggling the path might include between two successive waypoints. This can be the
result of satisfying another property, especially a continuous curvature. An exam-
8 Anastasios M. Lekkas and Thor I. Fossen

ple is the natural cubic spline, which can give curvature-continuous paths but their
overall resulting geometry can often be very unsuitable for vehicle navigation. On
the other hand, a monotone cubic spline methodology, such as the one presented in
[19], can give paths with a very practical shape, but without preserving curvature
continuity at the waypoint locations.

WP6
WP7a
50
x−coordinate (m)

40
WP3
30 WP4
WP2
WP7b
20
10 WP5
WP1
0
0 10 20 30 40 50 60 70 80 90
y−coordinate (m)
Fig. 1 Local control and monotonicity of the implemented path-planning method are shown. The
last waypoint of the initial path has changed, but in a way such that the lines W P6 − W P7a and
W P6 − W P7b have slopes of the same sign. This change does not affect the overall shape of the
initial path.

WP7c
x−coordinate (m)

60 WP6
WP7a

40 WP3
WP4
WP2
20
WP5
WP1
0
0 10 20 30 40 50 60 70 80 90
y−coordinate (m)
Fig. 2 Partial local control of the monotone CHSI method is shown. In this case the last waypoint
of the initial path changes in a way such that the slopes of W P6 −W P7a and W P6 −W P7c have a
different sign. This causes a small change at the segment W P5 −W P6.

The second branch can be of importance in cases where due to updated weather
data, for instance, one segment of the path should be avoided and one (or more)
waypoint should change location. The question then is what happens to the rest
of the path, and especially the part which precedes the newly-assigned waypoint
Line-of-Sight Guidance for Path Following of Marine Vehicles 9

because this might affect the vessel at its current position. Depending on the result,
three different path control behaviors can be distinguished: 1) local control, where
it is possible to change a waypoint without affecting the rest of the path, 2) global
control, where there will be changes through the whole path, and 3) partial control,
where in some cases local control is possible. An example of a local control method
is the Cubic Hermite Spline Interpolation (CHSI), while natural cubic splines is a
global control method and the monotone CHSI is a partial control method. The latter
is demonstrated in Figs. 1-2.

3.2.5 Algorithm Complexity

Naturally, each path planning method is based upon different principles and the
involved mathematics might result in algorithms that are computationally more ex-
pensive than others. Path planning is a real-time process and, although large vessels
should have enough computational power available, the same might not be true for
smaller unmanned vehicles with limited power. In dynamic environment with mov-
ing obstacles, the path generation algorithm must be able to give feasible solutions
within limited time and this can make the task a difficult one if the method is based
on open form solutions.

4 LOS Guidance Designs for the Decoupled Horizontal and

Vertical Planes

4.1 Introduction

This section deals with path-following in two dimensions by considering two inde-
pendent and decoupled 3-DOF cases, namely, the horizontal plane guidance and the
vertical plane guidance. The former is concerned with generating appropriate head-
ing reference trajectories in order for the vehicle to converge to a straight line on
the xy-plane and, similarly, the latter generates pitch reference trajectories in order
to converge to a straight line on the xz-plane. For either of them a LOS guidance
law that results in a κ-exponentially stable equilibrium point is derived. The rea-
son for studying these decoupled cases is twofold: first, the horizontal plane case
is exactly how the problem is stated and tackled for a surface vessel performing
path-following, and second, breaking the problem in two parts contributes to pre-
senting a more constructive approach and serves as a smooth transition to the 3-D
case. In this section it is assumed that no environmental disturbances are present.
Moreover, a discussion on the cascaded system formed by the LOS guidance law
and the heading, or pitch, angle controller is included at the end of the section.
10 Anastasios M. Lekkas and Thor I. Fossen

x, y , r

xk , yk d r
Horizontal LOS Heading Autopilot

⌘, v
Waypoints AUV rle;r

xk , zk ✓d q
Vertical LOS Depth Autopilot

x, z ✓, q

Fig. 3 Block diagram of the decoupled LOS guidance for the horizontal and vertical planes.

4.2 Horizontal-Plane Path Following

In the case of decoupled horizontal plane path-following we assume that θ = 0◦ ,

consequently the kinematics equation to be considered is:

ẋ = u cos(ψ) − v sin(ψ), (16)

ẏ = u sin(ψ) + v cos(ψ), (17)
ψ̇ = r. (18)

The horizontal speed Uh is given by:

p
Uh := u2 + v2 , (19)

and is assumed to be positive and bounded:

Uh,min ≤ Uh ≤ Uh,max , 0 < Uh,min . (20)

From (19)–(20) it is implied that the vessel always has at least a nonzero surge speed.
The reason for setting a minimum positive speed Uh,min is related to the stability
proof of the LOS guidance and will be explained in Section 4.2.2. The model (16)–
(18) includes only absolute velocities and describes the motion of an underactuated
vehicle since two out of three DOF’s can be controlled independently, namely the
yaw angle and the surge velocity.
Line-of-Sight Guidance for Path Following of Marine Vehicles 11

4.2.1 Path Following Objective

Assuming that the vehicle is assigned to converge to the line connecting the way-
points WPk –WPk+1 , the along-track and the cross-track error for a given vehicle
position (x, y) are given by:
   
xe > x − xk
= R (γ p ) , (21)
ye y − yk

where (xk , yk ) is the position of the k-th waypoint expressed in the NED frame, and
the rotation matrix from the inertial frame to the path-fixed reference frame is given
by:  
cos(γ p ) − sin(γ p )
R(γ p ) = ∈ SO(2). (22)
sin(γ p ) cos(γ p )
Moreover,

xe = (x − xk ) cos(γ p ) + (y − yk ) sin(γ p ), (23)

ye = −(x − xk ) sin(γ p ) + (y − yk ) cos(γ p ), (24)

where γ p is the horizontal path-tangential angle:

γ p = atan2(yk+1 − yk , xk+1 − xk ), (25)

where the two-argument function atan2 is a generalization of the arctan(y/x) that

takes into account the signs of both x and y in order to determine the quadrant of
the result, hence making it possible to distinguish between diametrically opposite
directions. Finally, the associated control objective for horizontal plane straight-line
path-following is:
lim ye (t) = 0. (26)
t→+∞

Note that the along-track error xe does not need to be minimized in a path-following
scenario, the contrary is true for applications that impose temporal constraints.

4.2.2 Horizontal LOS Guidance Law

Figure 4 depicts the geometry of the LOS guidance and some of the main variables
that are involved in the problem. Before moving on it is useful to mention that the
LOS vector is often defined differently in marine applications compared to the defi-
nition adopted by the aircraft and missile communities. According to the definition
in [42], the LOS is the line that starts at the reference point (that is, the aircraft or
the missile) and passes through the objective of the guidance (i.e the target). On
the other hand, in marine guidance applications the LOS vector starts at the vessel
and passes through a point p(xlos , ylos ) which is located on the path-tangential line
at a lookahead distance ∆h > 0 ahead of the direct projection of the vessel’s posi-
tion p(x, y) on to the path. The latter definition seems to be more congruent with
12 Anastasios M. Lekkas and Thor I. Fossen

pk+1 pn
p (xlos , ylos )
h

pk xe

U
xb
LOS
ye vector

yb
xn

yn

Fig. 4 Line-of-sight guidance geometry for straight lines in the xy plane. Here the sideslip angle is
equal to zero.

the path-following task and, as a consequence, in this paper the lookahead-based

steering method will be considered. Depending on the lookahead distance value, the
maneuvering characteristics of the vehicle can vary significantly. More specifically,
a low ∆h value will induce more aggressive steering compared to a larger value. For
this reason, algorithms proposing a time-varying ∆h have been implemented in the
past in order to obtain a more flexible behavior, see for instance [34], [32] and [27].
However, in this paper we will consider the different guidance laws with a constant
∆h in order to reduce the overall complexity of the problem and be able to compare
their performance more reliably. The lookahead-based guidance law is given by (see
[6]):
−ye
 
ψd = γ p + arctan . (27)
∆h
In the presence of external disturbances, or during turns, the heading angle ψd and
the course angle χd are not aligned anymore and are related in the following way:

χd = ψd + β , (28)

and therefore the desired heading angle is:

−ye
 
ψd = γ p + arctan −β. (29)
∆h
Line-of-Sight Guidance for Path Following of Marine Vehicles 13

By differentiating (24) with respect to time we get:

ẏe = − ẋ sin(γ p ) + ẏ cos(γ p ),

= − (u cos(ψ) − v sin(ψ)) sin(γ p )
+ (u sin(ψ) + v cos(ψ)) cos(γ p ),
=u sin(ψ − γ p ) + v cos(ψ − γ p ), (30)

and by transforming (62) in amplitude-phase form we get:

p
ẏe = u2 + v2 sin(ψ − γ p + β ),
= Uh sin(ψ − γ p + β ), (31)

where
β = atan2(v, u), (32)
which is equal to the orientation of the vehicle’s velocity vector Uh with respect to
the body-fixed frame. In other words, (32) is the angle between the vehicle’s velocity
orientation and the vehicle’s heading. This is the commonly known as sideslip, or
drift, angle.
Proposition 1. Under the assumption that the desired heading is perfectly tracked
such that ψ = ψd , the system (31) has a κ-exponentially stable equilibrium point at
ye = 0 if the desired heading angle is given by (29).

Proof. If we assume that the desired heading is perfectly tracked at all times and
choose the desired heading angle as in (29), the derivative of the cross-track error
becomes:
−ye
ẏe = Uh q . (33)
∆h2 + y2e
The Lyapunov Function Candidate (LFC)
1
V1 = y2e , (34)
2
has the time-derivative:
−y2
V̇1 = Uh q e . (35)
∆h2 + y2e
which is negative for Uh > 0. Hence, the origin ye = 0 is a UGAS equilibrium of the
system (33). Moreover, on the ball D1 = {ye ∈ R|ye | ≤ µ1 }, µ1 > 0, we have that

Uh y2e
V̇1 = − q ≤ −k1 y2e , (36)
µ12 + ∆h2
14 Anastasios M. Lekkas and Thor I. Fossen
q
for some 0 < k1 < Uh /( µ12 + ∆h2 ), which entails that the origin is a ULES equilib-
rium. The combination of UGAS and ULES implies global κ-exponential stability,
as it was shown in [26].

Discussion regarding the sideslip angle β : It is worth clarifying that the sideslip
angle that appears in (31) is not induced by any environmental forces, since in this
section they are assumed to be zero. The sideslip angle in the present case occurs
due to the nonzero sway velocity during a turn, this is what causes a difference
between the orientation ψ (heading angle) of the surge velocity u and the orientation
χ (course angle) of the total speed Uh . This is depicted in Fig. 5. When the vehicle
is moving forward in an environment without external disturbances (i.e. the vessel
following the xn axis), the total velocity Uh is equal to the surge velocity u and there
is no sideslip angle. However, during a turn (i.e the vessel moving on the curved
path), a part of the total velocity is transferred into sway velocity and the sideslip
angle is nonzero. Consequently, this is the sideslip angle that the current section
refers to. In the past there have been proofs of the LOS guidance without including
this component of β , see for instance [2]. In that case it was still possible to prove
stability because this effect is relatively small (although it can be larger for very agile
maneuvering) and the related terms can be modeled as part of the interconnecting
term of the cascade structure which was shown to satisfy the growth conditions
given in [33]. Ideally, however, incorporating (32) in the desired heading command
should increase performance.
Comment regarding the lower speed bound Uh,min : It is also important to mention
that in order for the proof above to be consistent with stability theory, it is necessary
to set a specific positive lower bound Uh,min instead of just assuming that 0 < Uh .
This necessity occurs due to the fact that if the velocity is decreasing in a way such
that it converges asymptotically to zero, then it is not possible to conclude that the
system (33) converges.

4.3 Vertical-Plane Path Following

This problem is similar to the one presented in Section 4.2 In the case of decoupled
vertical plane path-following of underwater vehicles, it is common to assume that
the yaw angle ψ = 0◦ , consequently the kinematic equations to be considered are:

ẋ = u cos(θ ) − w sin(θ ), (37)

ż = −u sin(θ ) + w cos(θ ), (38)
θ̇ = q. (39)

Apparently, for this problem the vertical speed is defined as:

p
Uv := u2 + w2 > 0, (40)
Line-of-Sight Guidance for Path Following of Marine Vehicles 15
xn

Uh
u xb

Uh = u yb

xb

yb

yn

Fig. 5 Sideslip angle during a turn. No external disturbances are present in this case.

and is assumed to be positive and bounded:

Uv,min ≤ U ≤ Uv,max , 0 < Uv,min . (41)

4.3.1 Path Following Objective

Similarly to the case for surface vessels, we assume that the vehicle is supposed to
converge to the line connecting the waypoints W Pk –W Pk+1 , the along-track and the
1
cross-track error for a given vehicle position (x, z) are given by:
   
xe x − xk
= R> (α p ) , (42)
ze z − zk

where (xk , zk ) is the position of the k-th waypoint expressed in the NED frame, and
the rotation matrix from the inertial frame to the path-tangential frame is given by:
 
cos(α p ) sin(α p )
R(α p ) = ∈ SO(2). (43)
− sin(α p ) cos(α p )

Moreover,
16 Anastasios M. Lekkas and Thor I. Fossen

xe = (x − xk ) cos(α p ) − (z − zk ) sin(α p ), (44)

ze = (x − xk ) sin(α p ) + (z − zk ) cos(α p ), (45)

where α p is the vertical path-tangential angle:

α p = atan2(−(zk+1 − zk ), (xk+1 − xk )). (46)

Consequently, the associated control objective for vertical plane straight-line path-
following is:
lim ze (t) = 0. (47)
t→+∞

4.3.2 Vertical LOS Guidance Law

The time-derivative of the vertical cross-track error gives:

że =ẋ sin(α p ) + ż cos(α p ),

= − (u cos(θ ) − w sin(θ )) sin(α p )
− (u sin(θ ) − w cos(θ )) cos(α p ),
p
= u2 + w2 cos(θ − atan2(w, u)) sin(α p )
p
− u2 + w2 sin(θ − atan2(w, u)) cos(α p ),
= −Uv [sin(θ − α) cos(α p ) − cos(θ − α) sin(α p )] (48)

and by transforming (48) according to the common trigonometric property regarding

the sinus of a difference of angles:

że = −Uv sin(θ − α − α p ), (49)

where
α = atan2(w, u) (50)
is the commonly known as the angle of attack. Similarly to the horizontal cross-
track error case, we assume that the desired pitch angle is perfectly tracked at all
times and choose the desired pitch angle as:
 
ze
θd = α p + α + arctan , (51)
∆v

with ∆v denoting the lookahead distance for the vertical-plane LOS.

Proposition 2. Under the assumption that the desired pitch angle is perfectly
tracked such that θ = θd , The system (49) has a κ-exponentially stable equilibrium
point at ze = 0 if the desired pitch angle is given by (51).

Proof. Under the assumption of perfect pitch angle tracking, adopting (51) as the
desired pitch angle gives the following derivative of the vertical cross-track error:
Line-of-Sight Guidance for Path Following of Marine Vehicles 17

−ze
że = Uv p . (52)
∆v2 + z2e

The Lyapunov Function Candidate (LFC)

1
V2 = z2e , (53)
2
has the time-derivative:
−z2
V̇2 = Uv p e . (54)
∆v2 + z2e
which is negative for Uv > 0. Hence, the origin ze = 0 is a UGAS equilibrium of the
system (52). Moreover, on the ball D2 = {ze ∈ R|ze | ≤ µ2 }, µ2 > 0, we have that

Uv z2e
V̇2 = − q ≤ −k2 z2e , (55)
µ22 + ∆v2
q
for some 0 < k2 < Uv /( µ22 + ∆v2 ), which implies that the origin is a ULES equi-
librium point and therefore a κ-exponentially stable equilibrium point.

The same comments and observations mentioned at the end of Section 4.2.2 ap-
ply in this case too for both the total speed Uv and the angle of attack α.

4.4 Cascade Structure Formed by the Horizontal/Vertical LOS

Guidance System and the Heading/Pitch Controller

In the previous sections the assumption of perfectly tracked heading (or pitch,
henceforth this clarification will be omitted), such as ψ = ψd , has been mentioned
several times and the stability results have relied on this in order to hold true. Al-
though such an assumption is not congruent with reality, it is not an oversimplifica-
tion because the guidance system is seen as one of the two systems of a cascaded
system. The other subsystem is the heading dynamics, which, apparently, includes
the heading controller. Therefore when analyzing the stability of the cascade struc-
ture, the theorems and the procedure introduced in [33] are used. Part of this pro-
cedure is to analyze the stability of each system independently before taking into
account the interconnection between them. Sections 4.2.2 and 4.3.2 dealed with
analyzing only the guidance system. The remaining steps depend on the heading
autopilot and are outside the scope of this paper, examples can be found in [2] and
[27] A depiction of the cascade systems equivalence is shown in Fig. 6.
18 Anastasios M. Lekkas and Thor I. Fossen

Vehicle Vehicle Vehicle

Heading LOS Heading LOS

controller Guidance controller Guidance

Fig. 6 Cascade system equivalence between the horizontal LOS guidance and the heading con-
troller. The total system can be studied as a cascade structure where the perturbing system (vehicle
+ heading controller) affects the convergence of the perturbed system (vehicle + LOS guidance)
via the heading error dynamics.

5 LOS Guidance Design for 3-D Coupled Motions

5.1 Introduction

This section considers a 3-D motion control scenario where an underwater vehicle is
assigned to achieve straight-line path following at a predetermined depth. Contrary
to Section 4, the horizontal LOS guidance is now coupled with the vertical motion
of the AUV. Hence 5-DOF’s are considered and the kinematics is given by (4)–(8)
The total speed in this case is:
p
U := u2 + v2 + w2 , (56)

where, similarly to the decoupled planar cases of Section 4:

Umin ≤ U ≤ Umax , 0 < Umin . (57)

Comment on the system’s structure: The task is to converge to a horizontal straight

line at a depth z = zd . Consequently, it is reasonable to consider the two following
subsystems in order to solve the problem: a) a depth controller for (6)–(7), and b)
a LOS guidance for (4)–(5) and (8). From (6)–(7) we observe that the heave-pitch
subsystem is uncoupled with respect to the rest of the states, whereas (4)–(5) and
(8) indicates that the horizontal motion is coupled with the vertical plane motion
via the pitch angle θ and the heave velocity w. A block diagram of the total system
structure can be seen in Fig. 7.

5.2 First Subsystem: Horizontal LOS Guidance Excited by the

Heave and Pitch Dynamics

The second subsystem is given by the equations:

Line-of-Sight Guidance for Path Following of Marine Vehicles 19

x, y, w, ✓ , r

xk , yk d
Horizontal LOS Heading Autopilot
r

⌘, v
Waypoints AUV rle;r

q
zd
Depth Autopilot

z, ✓, q

Fig. 7 Block diagram of the 3-D path following scenario. The horizontal plane motion (AUV +
LOS guidance) is perturbed by the vertical plane motion (AUV + depth controller). The latter,
however, is independent.

ẋ = u cos (ψ) cos (θ ) − v sin (ψ) + w cos (ψ) sin (θ ), (58)

ẏ = u sin (ψ) cos (θ ) + v cos (ψ) + w sin (ψ) sin (θ ), (59)
1
ψ̇ = r, cos (θ ) 6= 0. (60)
cos(θ )

This system is apparently different compared to the horizontal LOS solved in Sec-
tion 4.2 because the system is now coupled with the depth control system via the
pitch angle θ and the heave velocity w. The cross-track error is still given by:

ye = −(x − xk ) sin(γ p ) + (y − yk ) cos(γ p ), (61)

however, in this case it propagates differently due to the coupling with θ and w:

ẏe = − ẋ sin(γ p ) + ẏ cos(γ p ),

= − u cos (ψ) cos (θ ) sin (γ p ) + v sin (ψ) sin (γ p )
− w cos (ψ) sin (θ ) sin (γ p ) + u sin (ψ) cos (θ ) cos (γ p )
+ v cos (ψ) cos (γ p ) + w sin (ψ) sin (θ ) cos (γ p ). (62)

Using several trigonometric identities yields:

ẏe = − cos (ψ) sin(γ p )[u cos (θ ) + w sin (θ )]

+ sin (ψ) cos(γ p )[u cos (θ ) + w sin (θ )] + v cos (ψ − γ p ),
p
= u2 + w2 cos (θ − α)[sin (ψ) cos(γ p ) − cos (ψ) sin(γ p )] + v cos (ψ − γ p ),
=Uv cos (θ − α) sin (ψ − γ p ) + v cos (ψ − γ p ),
20 Anastasios M. Lekkas and Thor I. Fossen

and, finally, we have the compact form:

q
ẏe = (Uv2 cos2 (θ − α) + v2 ) sin (ψ − γ p + βv ) (63)

where:

βv = atan2 (v,Uv cos (θ − α)) . (64)

This result shows that the first subsystem (it intervenes here through θ and w) does
not prevent the minimization of the horizontal cross-track error as long as the yaw
controller can compensate for the generalized sideslip angle βv . This agrees with
intuition because if the target line is at a high depth, the horizontal LOS guidance
system will probably converge first while the AUV continues to submerge. The de-
sired yaw angle is:

−ye
 
ψd = γ p − βv + arctan . (65)
∆h

Before proceeding with the theorem and the proof pertinent to the stability of (63),
the following assumptions are made:
• A1: The desired heading ψd is perfectly tracked at all times.
• A2: The depth controller ensures that w and θ are bounded states.
π
• A3: The pitch angle satisfies the condition θ 6= ± .
2
Theorem 1. The system (63) has a κ-exponentially stable equilibrium point at ye =
0 if all the assumptions A1-A3 are satisfied and the desired heading angle is given
by (65).

Proof. This is similar to the cases proved in Section 4. The Lyapunov Function
Candidate (LFC):
1
V3 = y2e , (66)
2
has the time-derivative:
−y2
q
V̇3 = (Uv2 cos2 (θ − α) + v2 ) q e . (67)
∆h2 + y2e

which is negative. Hence, the origin ye = 0 is a UGAS equilibrium of the system

(52). In addition, on the ball D3 = {ye ∈ R|ye | ≤ µ3 }, µ3 > 0, we have that
p
(Uv2 cos2 (θ − α) + v2 )y2e
V̇3 = − q ≤ −k3 y2e , (68)
2
µ3 + ∆h2
Line-of-Sight Guidance for Path Following of Marine Vehicles 21
q
for some 0 < k3 < (Uv2 cos2 (θ − α) + v2 )/(µ32 + ∆h2 ), which implies that the ori-
gin is a ULES equilibrium point and therefore a κ-exponentially stable equilibrium
point.

5.3 Second Subsystem: Depth Control

This section discusses two popular approaches when it comes to depth control
of AUVs, namely, PID control and sliding mode control. Note that several other
methodologies have been studied in the past, such as adaptive linear controllers
[28], model predictive controllers [38] and H∞ control [20], [29], [35]. More de-
tailed information on similar control techniques for AUVs can be found in [25]. At
the end of the section we discuss the possibility of the depth reference trajectories
being generated by a vertical LOS guidance algorithm. Although such a concept
can be characterized as redundant when the overall task of the depth controller is
to achieve setpoint regulation, it constitutes an important step toward an even more
general path-following scenario where the task is to converge to any straight line in
space (i.e. γ p , α p 6= 0), hence a time-varying depth reference signal is required.

5.3.1 Conventional PID Controllers

Controlling depth using PID controllers is a common approach that has been re-
ported in the literature, see for instance [23], [41], [31]. Usually, this approach em-
ploys a linearized model for the pitch dynamics and conventional PID controllers
are tuned properly in order to stabilize the system. As it is always the case with
linearized systems, the approach might fail for large pitch angles, or time-varying
velocity. However, if the linearization assumptions are satisfied then the PID con-
troller will result theoretically in a GES equilibrium point. Experimental tests using
PID controllers for depth control have reported satisfactory performance and this
fact in combination with their simplicity makes them a more attractive approach
compared to a more complicated one like, for instance, H∞ control [25].

5.3.2 Sliding Mode Controllers

Depth control using sliding modes is also a methodology that has been implemented
extensively in the past, an early reference is that of [13], in-water tests were reported
in [22], also a well-known approach was the one presented in [21]. This controller
is based on the concept of the sliding surface to increase robustness and its goal is
to provide good performance even under the presence of modeling uncertainties and
environmental disturbances. A sliding mode controller demonstrates better perfor-
mance and is faster compared to a PID controller, but the downside is that it is more
complicated and requires a complete model. An example of a pitch and depth au-
22 Anastasios M. Lekkas and Thor I. Fossen

topilot for underwater vehicles based on sliding mode control can be found in [16,
pp. 526–528].

5.3.3 Vertical LOS for Depth Control

Adopting vertical LOS coupled with horizontal LOS is the next reasonable step in
order to achieve path following for any straight path in space, with the exception of
α p ± π/2 due to the singularity of the Euler angle representation. It is possible to
avoid this singularity by using quaternions. Due to the fact that the kinematics (6)–
(7) is uncoupled, the analysis from Section 4.3 still holds. Therefore, combining
(63) and (51) and using the property
p
cos (arctan(x)) = 1/( 1 + x2 ), (69)

gives:

Uv ∆v
ẏe = p sin (ψ − γ p ) + v cos (ψ − γ p ), (70)
∆v2 + z2e

and finally we get the compact form:

s
Uv2 ∆v2

ẏe = + v2 sin (ψ − γ + β ), (71)
p v
∆v2 + z2e

with
!
Uv ∆v
βv = atan2 v, p . (72)
∆v2 + z2e

As expected, (71) shows that the horizontal cross-track error is now a function of
the vertical cross-track error as well. Note, however, that ze does not prevent ye
from converging because the stability result for the vertical LOS guarantees that ze
is bounded and has a κ-exponentially stable equilibrium point at ze = 0. Eq. (72)
shows an generalized formulation of the sideslip angle. Similarly to the 2-D sideslip
angle case, this has an effect if the sway motion is nonzero, i.e. on a turn. Apparently,
the system (71) is stabilized with a heading command:

−ye
 
ψd = γ p − βv + arctan . (73)
∆h

This can be shown similarly to the previous proofs, for a LFC:

1
V4 = y2e , (74)
2
Line-of-Sight Guidance for Path Following of Marine Vehicles 23

we derivate V4 w.r.t time and use (73) as the desired heading command. The final
result is: s
Uv2 ∆v2 y2e
 
V̇4 = − + v2 . (75)
∆v2 + z2e ∆h2 + y2e

6 LOS Guidance Under the Influence of Sea Currents

6.1 Introduction

Section 4 considered path-following in 3-DOF under the assumption that no exter-

nal disturbances affect the vehicle. However this is rarely, if ever, true in real life
applications. In practice there is always influence due to the waves, wind and ocean
currents. The natural outcome then is that the LOS guidance will fail to succeed in
achieving the motion control objective. More specifically, when the desired path is a
straight line, these (hypothetically constant, or slow time-varying) forces will result
in a constant cross-track error from the target line. The magnitude of this error de-
pends on the magnitude of the total external force, as well as the force’s orientation
with respect to the target line. It is common practice in control systems to add an
integrator to the system in order to eliminate constant offsets and, naturally, this has
been implemented to LOS guidance in the past. The idea is to accumulate the error
with an integral term and use it to correct the heading (or pitch) reference trajectories
until the error has been eliminated. As a consequence, when the vehicle eventually
converges on the target line, its heading will not be equal with the course angle. In
other words, a sideslip angle is necessary in order for the vehicle to converge to and
follow the path. In connection with the discussion on the sideslip angle (32) due to
turning in Section 4.2.2, the sideslip angle because of the external forces is given
by:
βr = atan2(vr , ur ), (76)
see also Fig. 8. In the rest of this section, two integral LOS guidance laws pertaining
to marine vehicles are revised. Finally, it is also shown how course control (when-
ever applicable) circumvents this problem.

6.2 Conventional Integral LOS Guidance

Following [6], by rewriting (29) as

ψd = γ p + arctan(−K p ye ) −β , K p = (1/∆ ), (77)

| {z }
ψr
24 Anastasios M. Lekkas and Thor I. Fossen
xn

U
u xb
ur
v
vr

Ur yb

xb

yb

yn

Fig. 8 Sideslip angle, with a current coming from the east, hence affecting the velocity compo-
nents.

we observe that the lookahead-based steering law has the same form as a saturated
proportional control law, effectively mapping ye ∈ R into ψr (ye ) ∈ (−π/2, π/2).
Following the same line of reasoning, it is straightforward to add integral action in
order to compensate for the cross-track error caused by a constant disturbance:
 Z t 
ψd = γ p + arctan −K p ye − Ki ye dτ . (78)
t0

where Ki > 0 denotes the integral gain.

1 As is usually the case with integral control
action, careful design is necessary so as to avoid undesired effects such as wind-
up and overshooting. Wind-up refers to the case where the integral term increases
to large values and, as a consequence, has negative effects on the system’s perfor-
mance, such as very long convergence time and overshooting. Suggested method-
ologies pertaining to avoiding such phenomena can be found in [12], [8]. Equation
(78) can be rewritten as follows

ψd = γ p + arctan (−K p ye − Ki yint ) , (79)

ẏint = ye . (80)
Line-of-Sight Guidance for Path Following of Marine Vehicles 25

6.3 Integral LOS Guidance

In [1] the authors followed a different approach and presented the following modi-
fied LOS guidance law with integral action:

ψd = γ p − arctan(K p ye + Ki yint ) (81)

ye ∆
ẏint = 2 , (82)
∆ + (ye + κyint )2

where K p = (1/∆ ), Ki = K p κ and κ > 0 is a design parameter. Equation (82) has

been designed in a way such that the influence of the integrator diminishes when
the cross-track error increases, hence the wind-up risk is reduced. The authors focus
on marine surface vessels and they pair the proposed guidance law (81)–(82) with
a set of adaptive tracking controllers. After a long and extensive analysis they show
that, after explicit bounds on the parameters of the guidance law have been satisfied,
their proposed strategy results in globally asymptotic path following. The analysis
in [1] was based on the 3-DOF kinematic model (16)–(18) with absolute velocities.
In [9] the authors presented a proof of the same integral LOS guidance law using
the relative velocity kinematics (9)–(10) and (13) (with θ = 0) and showed that
this simplifies the overall analysis and results in less complicated controllers. Fig. 9
compares the performance between the conventional integral LOS and the method
developed by Børhaug et al.

50

0
ye (m)

−50

−100

−150
0 50 100 150 200
Time (sec)

Fig. 9 Cross-track error comparison between the conventional integral LOS (dotted line) and the
integral LOS proposed by Børhaug et al. (solid line).
26 Anastasios M. Lekkas and Thor I. Fossen

6.4 LOS Guidance using Course Control

In a path-following operation of a surface vehicle it is natural to assume that GNSS

(Global Navigation Satellite System) measurements are available, otherwise it is not
possible to compute the cross-track error, for instance. Consequently, it is feasible to
get absolute velocity measurements that can be used to control the course angle χ of
the ship. Especially in the case of devices using GNNS Doppler data, the achievable
accuracy can be very high, as it is reported in [10]. The ability to control χ allows
to tackle the problem in a different way by rewriting the cross-track error (31) as

ẏe = U sin(χd − γ p ), (83)

which, similarly to the proof in Section 4, can be easily shown to have a κ-

exponentially stable equilibrium point at ye = 0 by choosing the desired course angle
as:
−ye
 
χd = γ p + arctan . (84)
∆
Interestingly, it can be concluded by comparing (84) with (81) that, by controlling
the course angle instead of the heading angle, the integral term is not necessary any-
more. This is a natural outcome of expressing the cross-track error in the form (31)
because in that way it becomes entirely known how the sideslip angle affects ye . As
a result, the necessity to deal with the current effect uncertainty is circumvented be-
cause the law (84) attempts to lead the overall velocity vector orientation to the right
value, hence there is no need for an extra state that will accumulate the cross-track
error so as to compensate for the heading control deviation from that orientation.
It is worth noting that in previous works the desired course command fed into the
course controller included integral action as well, see for instance [36] and [7]. In
practice, however, it will probably still be useful to include integral action because
there are always model uncertainties and other unknown factors present that have to
be compensated for. Fig. 10 compares the performance between the course control
case and the integral LOS guidance law developed by Børhaug et al. As it was ex-
pected, the fact that there is no uncertainty from the beginning in the course control
method leads to better results.

7 Conclusions

This paper has given an overview of the LOS guidance law for marine vehicles.
Proofs in 3-DOF for horizontal and vertical plane navigation incorporating the
sideslip angle and the angle of attack were given. Then, the more general task of
converging to a horizontal line at a predetermined depth underwater was studied.
The effect of constant environmental forces were discussed and two methods from
the literature were revised and compared. Moreover, it was shown that controlling
the course instead of the yaw angle removes all the uncertainty and leads to even
Line-of-Sight Guidance for Path Following of Marine Vehicles 27

30

20
ye (m)
10

−10
0 50 100 150 200
Time (sec)

Fig. 10 Cross-track error comparison between the course control case (solid line) and the integral
LOS proposed by Børhaug et al. (dashed line).

better performance. The cascade structure between the guidance system and the
heading dynamics was discussed. The importance of choosing a suitable path de-
pending on the application was stressed and, in addition, a section proposing path
evaluation criteria was presented.

8 Acknowledgments

The first author wishes to acknowledge useful discussions with Andreas Reason
Dahl and Dr. Morten Breivik regarding the path evaluation criteria. This work was
supported by the Centre for Ships and Ocean Structures, the Centre for Autonomous
Marine Operations and Systems, NTNU, and the Norwegian Research Council.

References

1. Børhaug, E., Pavlov, A., Pettersen, K.Y.: Integral LOS control for path following of under-
actuated marine surface vessels in the presence of constant ocean currents. In: 47th IEEE
Conference on Decision and Control, pp. 4984–4991. Cancun, Mexico (2008)
2. Børhaug, E., Pettersen, K.Y.: Cross-track control for underactuated autonomous vehicles. In:
Proceedings of the 44th IEEE Conference on Decision and Control, and the European Control
Conference, pp. 602–608. Seville, Spain (2005)
3. Breivik, M.: Topics in guided motion control of marine vehicles. Ph.D. thesis, Norwegian
University of Science and Technology (2010)
4. Breivik, M., Fossen, T.I.: Path following for marine surface vessels. In: Proceedings of the
OTO’04, pp. 2282–2289. Kobe, Japan (2004)
5. Breivik, M., Fossen, T.I.: Principles of guidance-based path following in 2D and 3D. In:
Proceedings of the 44th IEEE Conference on Decision and Control, and the European Control
Conference, pp. 627–634. Seville, Spain (2005)
28 Anastasios M. Lekkas and Thor I. Fossen

6. Breivik, M., Fossen, T.I.: Guidance laws for autonomous underwater vehicles, chap. 4, pp.
51–76. INTECH Education and Publishing (2009)
7. Breivik, M., Hovstein, V.E., Fossen, T.I.: Straight-line target tracking for unmanned surface
vehicles. Modeling, Identification and Control 29(4), 131–149 (2008)
8. Burger, M.: Disturbance rejection using conditional integrators. Ph.D. thesis, Norwegian Uni-
versity of Science and Technology (2011)
9. Caharija, W., Candeloro, M., Pettersen, K.Y., Sørensen, A.J.: Relative velocity control and
integral LOS for path following of underactuated surface vessels. In: 9th IFAC Conference on
Manoeuvring and Control of Marine Craft. Arenzano, Italy (2012)
10. Chalko, T.J.: High accuracy speed measurement using GPS (Global Positioning System). Sci-
entific Engineering Research P/L (2002)
11. Dahl, A.R.: Overview and evaluation of path planning and guidance algorithms. Tech. rep.,
Department of Engineering Cybernetics, NTNU (2012)
12. Davidson, M., Bahl, V., Moore, K.L.: Spatial integration for a nonlinear path tracking control
law. In: Proceedings of the 2002 American Control Conference., pp. 4291–4296. IEEE (2002)
13. Dougherty, F., Woolweaver, G.: At-sea testing of an unmanned underwater vehicle flight con-
trol system. In: Proceedings of the (1990) Symposium on Autonomous Underwater Vehicle
Technology, pp. 65–73. IEEE (1990)
14. Dubins, L.E.: On curves of minimal length with a constraint on average curvature, and with
prescribed initial and terminal positions and tangents. American Journal of Mathematics
79(3), 497–516 (1957)
15. Farouki, R.T., Sakkalis, T.: Pythagorean hodographs. IBM Journal of Research and Develop-
ment 34(5), 736–752 (1990)
16. Fossen, T.I.: Handbook of Marine Craft Hydrodynamics and Motion Control. John Wiley and
Sons Ltd. (2011)
17. Fossen, T.I.: How to incorporate wind, waves and ocean currents in the marine craft equations
of motion. In: 9th IFAC Conference on Manoeuvring and Control of Marine Craft. Arenzano,
Italy (2012)
18. Fossen, T.I., Breivik, M., Skjetne, R.: Line-of-sight path following of underactuated marine
craft. Proceedings of the 6th IFAC MCMC, Girona, Spain pp. 244–249 (2003)
19. Fritsch, F.N., Carlson, R.E.: Monotone piecewise cubic interpolation. SIAM Journal on Nu-
merical Analysis 17(2), 238–246 (1980)
20. Fryxell, D., Oliveira, P., Pascoal, A., Silvestre, C.: Integrated design of navigation, guidance
and control systems for unmanned underwater vehicles. In: Proceedings OCEANS ’94, vol. 3,
pp. III–105. IEEE (1994)
21. Healey, A.J., Lienard, D.: Multivariable sliding mode control for autonomous diving and steer-
ing of unmanned underwater vehicles. IEEE Journal of Oceanic Engineering 18(3), 327–339
(1993)
22. Hills, S.J., Yoerger, D.R.: A nonlinear sliding mode autopilot for unmanned undersea vehicles.
In: Proceedings OCEANS’94., vol. 3, pp. III–93. IEEE (1994)
23. Jalving, B.: The NDRE-AUV flight control system. IEEE Journal of Oceanic Engineering
19(4), 497–501 (1994)
24. LaValle, S.M.: Planning algorithms. Cambridge university press (2006)
25. Lea, R.K., Allen, R., Merry, S.L.: A comparative study of control techniques for an underwater
flight vehicle. International Journal of Systems Science 30(9), 947–964 (1999)
26. Lefeber, E.: Tracking control of nonlinear mechanical systems. Ph.D. thesis, University of
Twente (2000)
27. Lekkas, A.M., Fossen, T.I.: A time-varying lookahead distance guidance law for path follow-
ing. In: 9th IFAC Conference on Manoeuvring and Control of Marine Craft. Arenzano, Italy
(2012)
28. Li, J.H., Lee, P.M.: Design of an adaptive nonlinear controller for depth control of an au-
tonomous underwater vehicle. Ocean engineering 32(17), 2165–2181 (2005)
29. Liceaga-Castro, E., van der Molen, G.M.: Submarine H∞ depth control under wave distur-
bances. IEEE Transactions on Control Systems Technology 3(3), 338–346 (1995)
Line-of-Sight Guidance for Path Following of Marine Vehicles 29

30. Lipschutz, M.M.: Differential Geometry. McGraw-Hill (1969)

31. McEwen, R., Streitlien, K.: Modeling and control of a variable-length AUV. Proc 12th UUST
(2001)
32. Oh, S.R., Sun, J.: Path following of underactuated marine surface vessels using line-of-sight
based model predictive control. Ocean Engineering 37(2-3), 289–295 (2010)
33. Panteley, E., Lorı́a, A.: On global uniform asymptotic stability of nonlinear time-varying sys-
tems in cascade. Systems & Control Letters 33(2), 131–138 (1998)
34. Pavlov, A., Nordahl, H., Breivik, M.: MPC-based optimal path following for underactuated
vessels. In: 8th IFAC International Conference on Manoeuvring and Control of Marine Craft,
pp. 340–345. Guaruja, Brazil (2009)
35. Silvestre, C., Pascoal, A., Healey, A.J.: AUV control under wave disturbances. In: Interna-
tional Symposium on Unmanned Untethered Submersible Technology, pp. 228–239. Univer-
sity of New Hampshire-Marine Systems (1997)
36. Skejic, R., Breivik, M., Berg, T.E.: Investigating ship maneuvers around a floating structure
under the influence of a uniform current in deep and calm water. In: Second International
Conference on Ship Manoeuvring in Shallow and Confined Water: Ship to Ship Interaction,
pp. 339 – 350 (2011)
37. Sørdalen, O.J., Egeland, O.: Exponential stabilization of nonholonomic chained systems.
IEEE Transactions on Automatic Control 40(1), 35–49 (1995)
38. Steenson, L.V., Phillips, A.B., Turnock, S.R., Furlong, M.E., Rogers, E.: Effect of measure-
ment noise on the performance of a depth and pitch controller using the model predictive
control method. In: Autonomous Underwater Vehicles (AUV), 2012 IEEE/OES, pp. 1–8.
IEEE (2012)
39. Techy, L., Woolsey, C.A.: Minimum-time path planning for unmanned aerial vehicles in steady
uniform winds. Journal of guidance, control, and dynamics 32(6), 1736–1746 (2009)
40. Tsourdos, A., White, B., Shanmugavel, M.: Cooperative path planning of unmanned aerial
vehicles. Wiley Online Library (2011)
41. Williams, S.B., Newman, P., Dissanayake, G., Rosenblatt, J., Durrant-Whyte, H.: A decou-
pled, distributed AUV control architecture. In: International Symposium on Robotics, vol. 31,
pp. 246–251. unknown (2000)
42. Yanushevsky, R.: Guidance of Unmanned Aerial Vehicles. CRC Press (2011)