2011 IEEE/RSJ International Conference on

Intelligent Robots and Systems

September 25-30, 2011. San Francisco, CA, USA

Path tracking: combined path following and trajectory tracking for

autonomous underwater vehicles
Xianbo Xiang, Lionel Lapierre, Chao Liu and Bruno Jouvencel

Abstract— This paper proposes a novel control strategy the vehicle is required to converge to and follow a
for autonomous underwater vehicles (AUVs), named as path desired geometric path, without any temporal specifica-
tracking, which combines the conventional path following and tions. Therefore, the control objective of path following
trajectory tracking control in order to achieve smooth spatial
convergence and tight temporal performance as well. This (PF) can be considered to implement a pure spatial task.
idea is inspired by the previous work of Hindman [1] and
Encarnacao [2], however, the path tracking design herein goes A. Trajectory tracking versus path following
from path following to trajectory tracking, which indeed is an
inverse way from the previous solutions so that the complex Actually, in many practical applications, such as underwa-
projection algorithm resulting in a local stability is avoided. A ter route surveying or pipeline inspection, steering an AUV
kinematics controller is first derived by using Lyapunov direct to follow a predefined path or track a desired trajectory,
method where a virtual path parameter is introduced to bring is far more frequently adopted than stabilizing the vehicle
an extra control degree of freedom, and then it is extended
to the dynamics of AUVs based on backstepping technique.
in a fixed posture. Hence, the path following and trajectory
The resulting nonlinear control design is formally shown and tracking control attract more concerns in practice.
it yields global asymptotic convergence of the AUV to the path. In the case of trajectory tracking, the vehicle is forced to
Finally, simulation results illustrate the efficiency of the path catch up the evolving reference on the time-parameterized
tracking control design for AUVs. trajectory. As pointed out in [1], [2], [6], [7], aggressive
I. INTRODUCTION dynamic behavior and possible saturation of actuation is
prone to be happened during the trajectory tracking motion
Numerous applications related to autonomous vehicles, operation. Whereas, the path following control only focuses
including air, land and marine vehicles, are presently op- on the geometric task without any temporal restriction, the
erational in industrial, scientific and military fields and more vehicle is only required to follow the predefined speed
ambitious applications are in engineering development. In profile, and smooth spatial convergence might be achieved
order to meet various goals of different applications, vehicles with less likely saturation of actuators [1], [2], [8], [9].
must be equipped with control systems to steer them to However, as there is no temporal specification in the path
achieve various motion tasks, and there is considerable following task, there is no temporal convergence which can
interest in the development of advance methods for motion be guaranteed by path following controller. Conversely, the
control problems. Conventionally, the motion control prob- trajectory tracking controller fulfills both the spatial and
lem of autonomous vehicles addressed in the literatures can temporal requirements, such that it has the partial advantage
be classified into three basic categories: point stabilization, to implement some time-critical assignments.
trajectory tracking and path following [3], [4].
• Point stabilization B. Modified strategies
the vehicle is stabilized at a desired goal posture from a
initial configuration. As pointed out by the well-known As there are some trade-offs of spatial and temporal
Brockett’s theorem [5], there is no smooth (or even performances between trajectory tracking and path follow-
continuous) constant state-feedback law to achieve Point ing, one may expect that one controller could balance and
stabilization for nonholonomic autonomous vehicles. benefit from both the path following and trajectory tracking
• Trajectory tracking performance. later.
the vehicle is required to track a time parameterized In [1], the new concept of maneuver modified trajectory
trajectory, i.e., a geometric path with an associated tracking (MMTT) came up. The authors assume that a trajec-
timing specification. Therefore, the control objective of tory tracking controller for a given system is firstly available
trajectory tracking (TT) can be considered to implement at hand, and a Lyapunov function is known that yields
a time×space task, that is the intersection set of spatial asymptotic stability of the resulting control system about
task and temporal task. a desired trajectory. In order to execute a path following
• Path following maneuver, the vehicle should ”look at the closest point on the
path” and adopt the posture of a virtual vehicle moving along
This work was funded by the EU FreeSubNet project under grant 036186, the path at the closest point as a reference to which it should
and the first author was supported by the European Marie Curie Fellowship. converge, where the desired path is locally well defined with
The first author is with Huazhong University of Science and Technology.
Authors are also with Department of Robotics, CNRS-LIRMM-UMII, UMR the ”no sharp corners” and ”non-intersection”, such that only
5506-CC 477, 161 rue Ada, 34392 Montpellier Cedex 5, France. local stability is achieved in the MMTT control design.

978-1-61284-456-5/11/$26.00 ©2011 IEEE 3558

Authorized licensed use limited to: INDIAN INSTITUTE OF TECHNOLOGY MADRAS. Downloaded on December 21,2023 at 16:24:54 UTC from IEEE Xplore. Restrictions apply.
 In fact, this kind of maneuver modified trajectory tracking 2) AUV dynamics. Let τu and τr define the external force
(MMTT) design can be considered as going from trajectory input in surge direction and torque input about the z axis of
tracking to path following, and the MMTT strategy shows the AUV respectively, generated by the back thrusters. The
how to combine trajectory tracking and path following be- dynamics equation of the AUV can be written as
haviors into a single control law, thus achieving smooth spa- 
tial convergence to the trajectory as well as time convergence.  τu = mu u̇ − mv vr + du u
0 = mv v̇ + mu ur + dv v (3)
C. Main contributions τr = mr ṙ − muv uv + dr r


Opposite to the MMTT strategy which goes from trajec- where Iz is the moment of inertia w.r.t. the z-axis. m{.}
tory tracking to path following, a new strategy is proposed in captures the effect of mass and hydrodynamic added mass
the paper to combine trajectory tracking and path following terms, and d{.} captures the hydrodynamic damping effects
in a single controller, by designing the controller to go from [11].
path following [8], [10] to trajectory tracking, named path In these equations, and for clarity of presentation, it was
tracking in an abbreviative notation. Therefore, the path assumed that the AUV is neutrally buoyant and that the
tracking control objective can be considered to implement center of buoyancy coincides with the center of mass. We
a time+space task, that is the union set of spatial task and can clearly see that there is no control input in the second
temporal task, to primarily achieve spatial convergence while equation in (3) due to the absence of thruster on sway
approaching to the temporal specification in the end. direction, usually in an underactuated AUV system.
The organization of the paper is as following: (1) modeling
of the underactuated AUV; (2) problem formulation of path B. Problem formulation
tracking; (3) path tracking control design in kinematics stage;
(4) extension to vehicle dynamics by backstepping technique; In this section, path parameterized with time evolution
(5) path tracking control examples of different performances is introduced to formulate the path tracking problem. The
with different weighting factors. extra control degree of freedom on the virtual target related
to path parameter τ(t), results in singularity free control
II. PATH TRACKING CONTROL DESIGN law for any regular path. It is also used for tracking error
A. Modeling of underactuated AUV feedback, which enable the system robustness, and keep the
benefits from both trajectory tracking and path following
The underactuated AUV model adopted in this paper, is
behaviors. Moreover, the path variable τ(t) is generalized
based on the model and parameters of the SIRENE AUV, and
to any meaningful parameter and is not limited to arc-length
more details about this AUV can be found in [11]. According
anymore as proposed in the standard path following control
to the standard 3D motion model in [12], neglecting the
design [10].
motions in heave, roll and pitch, the simplified horizontal
As depicted in Figure 1, an underactuated AUV is required
motion equations of SIRENE AUV in surge, sway and
to track a reference path Γ(τd (t)) parameterized by path
heading directions are as follows:
parameter τd (t). The reference path can be considered as a
1) AUV Kinematics. The AUV is with generalized co-
trajectory tracking target (TT target) moving along the path
ordinates p = (x, y, ψB )T in inertial frame I. Let u and v
with desired speed profile τ̇d (t), and a path tracking target
are the longitudinal (surge) and transverse (sway) velocities,
(PT target) moving on the path is introduced to generate the
respectively. Let r is the angular speed (yaw rate). The
virtual reference path Γ(τ(t)) online. Apparently, the path of
kinematic equations of the AUV can be written as
 TT target coincides with the path of PT target if the time-
 ẋ = ucosψB − vsinψB related path parameters are the same, i.e., τ(t f ) = τd (t f ) at
ẏ = usinψB + vcosψB (1) certain time instant t f . However, the virtual path parameter
ψ̇B = r τ(t) has an extra degree of freedom in order to bridge the


Assuming u is never equal to zero. Then, the sideslip angle behaviors of the tracking vehicle and the TT target on the
β can be defined as arctan(v/u). Considering the flow frame path, such that both smooth spatial convergence and time
{W } which is obtained by rotating body frame {B} around convergence can be achieved, which are the characteristics of
the zB axis through the sideslip angle β , the kinematic path tracking control, with the performance to combine both
equations can then be re-written to yield the path following and trajectory tracking control behaviors.
 As the reference path Γ(τ(t)) is parameterized by τ(t),
 ẋ = vt cos ψW there is :
ẏ = vt sin ψW (2)    
ẋ(τ) xτ (τ)τ̇
ψ̇W = r + β̇

 ẏ(τ)  =  yτ (τ)τ̇  (4)
Where ψW = ψB + β is the global heading of vt with respect ψr (τ) τ τ
atan2(y (τ)/x (τ))
to {I}, and vt is the total speed expressed in {W }. Notice
that the control of an AUV system implies considering a where xτ (τ) = ∂ ∂x(τ) τ ∂ y(τ)
τ and y (τ) = ∂ τ .
permanent positive forward speed u > 0 (vt > 0 as well), for The total speed of the underactuated AUV vt is required
an effective control action of the rudder/fin control surfaces. to align with the tangent speed of the path vtr . Let the path

3559
Authorized licensed use limited to: INDIAN INSTITUTE OF TECHNOLOGY MADRAS. Downloaded on December 21,2023 at 16:24:54 UTC from IEEE Xplore. Restrictions apply.
 C. Controller design
In this part, the control law for underactuated AUV path
tracking is derived based on Lyapunov stability theorem and
backstepping technique [13], [14]. However, a new control
Lyapunov function is chosen in a concise pattern, where
neither the approaching angle [15] nor the LOS guidance
[16] is used. Moreover, a weighting factor λ is introduced
to flexibly tune the weighting of the path following and
trajectory tracking behaviors, which is embedded in the path
tracking control design.
1) Kinematics controller: In the kinematics stage, the
control Lyapunov function is selected as
(xe2 + y2e + 1γ ψe2 )
kτ (τ − τd )2
V = (1 − λ ) +λ (10)
Fig. 1. Path tracking of underactuated AUV 2 2
where γ, kτ are scale factors to guarantee dimensional homo-
geneity of position xe , ye , angle ψe and time τ. Meanwhile
tracking error state vector peW = [xe , ye , ψe ]T be built in the τ, τd is the virtual and desired path parameter respectively
flow frame {W }, there is with limt→∞ τ̇d 6= 0.
     (xe2 +y2e + 1 ψe2 )
xe cos ψW sin ψW 0 xr − x The first part 2
γ
in (10) concerns the spatial
peW =  ye  =  − sin ψW cos ψW 0   yr − y  tracking property, while the second part (τ − τd )2 is the
ψe 0 0 1 ψr − ψW addition of a time dependent penalty.
(5) The parameter λ ∈ (0, 1) is the weighting factor. Obvi-
where [x, y, ψW ]T is the AUV state vector with the course ously, large λ let the time specification (τ − τd ) plays more
angle ψW = ψB + β , and [xr , yr , ψr ]T is the reference state important roles in the path tracking design, which represents
vector in the inertial frame {I}. the characteristics of trajectory tracking. While small λ let
The velocity vector of the PT target on the path (vtr , ωr )T the spatial specification xe , ye , ψe plays more important roles,
can be expressed by path parameter τ as which represents the characteristics of path following.

vtr (τ)
 
v̄tr (τ)τ̇
 The time derivative of (10) along the solution of (8) is
= (6)
ωr (τ) w̄r (τ)τ̇ V̇ =(1 − λ )[−xe vt + vtr xe cos ψe + vtr ye sin ψe
where 1
( p + ψe (ωr − ω − β̇ )] + λ kτ (τ − τd )(τ̇ − τ̇d )
v̄tr (τ) = xrτ (τ)2 + yτr (τ)2 γ
2 2
xrτ (τ)yτr (τ)−xrτ (τ)yτr (τ) (7) In order to introduce the extra degree of freedom to the
w̄r (τ) = xrτ (τ)2 +yτr (τ)2
virtual path parameter τ, define an auxiliary variable τ̃ and
Hence, the corresponding error state dynamics expressed let its time derivative as
in AUV body frame {B}, can be rewritten as follows:
τ̃˙ = τ̇ − vτ (t, xe , ye , ψe ) (11)
   
ẋe ωye − vt + v̄tr τ̇ cos ψe Proposing the control input as
 ẏe  =  −ωxe + v̄tr τ̇ sin ψe  (8)   " #
ψ̇e ω̄r τ̇ − ω − β̇ vt kx xe + v̄tr vτ cos ψe
= (12)
ω ω̄r vτ − β̇ + γye v̄tr vτ sinψψe e + kψ ψe
]T
where [vt , ωW is the vehicle state vector with ωW = ψ̇B +
β̇ = ω + β̇ ,and [vtr , ωr ]T is the reference velocity vector of and choosing
the PT target.
   
vτ τ̇d
= (13)
Therefore, the control objective of path tracking of un- τ̃˙ −kv tanh Φe
deractuated AUV is to design a controller and achieve the
yields
following tasks:
kψ 2
limt→∞ kpeB k = 0 V̇ = −(1 − λ )(kx xe2 + ψ ) − kv Φe tanh Φe (14)
(9) γ e
limt→∞ |τ(t) − τd (t)| = 0, limt→∞ |τ̇(t) − τ̇d (t)| = 0
Replacing (11) into the combination of (12) and (13),
where the first term declares the spatial assignment of path yields the path tracking control laws in kinematics stage as
tracking, requiring the position and total velocity of the   
kx xe + v̄tr τ̇d cos ψe

vt
vehicle coincide with those of the PT target on the path;  ω  =  ω̄r τ̇d − β̇ + γye v̄tr τ̇d sin ψe + kψ ψe  (15)
the second term declares the dynamic assignment, requiring ψe
the speed of the PT target τ̇(t) respect that of the TT target
τ̇ τ̇d − kv tanh Φe
τ̇d (t) and force the PT target to asymptotically catch up with Note: the above path tracking controller is nonsingular, as
the TT target. the term sinψψe e is well defined and continuous at zero.

3560
Authorized licensed use limited to: INDIAN INSTITUTE OF TECHNOLOGY MADRAS. Downloaded on December 21,2023 at 16:24:54 UTC from IEEE Xplore. Restrictions apply.
 2) Dynamics controller: Let vt and ω be virtual control Replacing (17) and (18) into the derivative of Vdyn , there
inputs, αvt and αω be the corresponding virtual control laws is
shown in (15). Introduce the velocity error variables kψ 2
V̇dyn = −(1−λ )(kx xe2 + ψ )−kv Φe tanh Φe −kvt z2vt −kω z2ω
γ e
   
zvt vt − αvt
z= =
zω ω − αω That means, V̇dyn is semi-negative definite and all the
Augmenting the Lyapunov function V in (10) with the states (xe , ψe , (τ − τd ), zvt , zω ) globally converge to their
quadratic terms of zu and zr , there is equilibrium. It can be concluded that the equilibrium is
1 (xe , ye , ψe , (τ − τd ), zvt , zω ) = 06 by using Barbalat’s lemma.
Vdyn = V + zT Mz (16) Therefore, we can state the following proposition for dy-
2
  namic path tracking control of underactuated AUV.
mu 0 Proposition 2.1: (Path tracking 2: underactuated AUV)
Denote the positive definite matrix M = ,
0 mr For a predefined C 1 path given in (4) with desired speed
where mu , mr are defined in (3). τ̇d (t) for the time derivative of the path parameter τd (t), given
The time derivative of Vdyn can be written as the dynamics control inputs in (22), and the virtual control
V̇dyn = (1 − λ )[(v̄tr vτ cos ψe − αvt )xe + v̄tr vτ ye sin ψe laws in (18). The control objective (9) of path tracking is
1 achieved and the equilibrium point [xe , ye , ψe , (τ − τd )]T = 04
+ ψe (ω̄r vτ − αω − β̇ )] + zvt (mu żvt − (1 − λ )xe ) is globally asymptotically stable.
γ
1 Note that, the computation of side-slip angle is required
+ λ kτ (τ − τd )(vτ − τ̇d ) + τ̃˙ Φe + zω (mr żω − ψe ) in path tracking control, due to β̈ appearing in the control
γ
computation of torque input τr as α̇ω = f (β̈ ) shown in (18).
Let the control laws for vt and ω be chosen as The similar strategy in [9] is used to solve this problem, by
v̇t = α̇vt + ((1 − λ )xe − kvt zvt )/mu

resorting to the original dynamic model of the underactuated
(17)
ω̇ = α̇ω + ( 1γ ψe − kω zω )/mr AUV.
Firstly, recalling ω̇ in (17) and differentiating αω in (18),
where kvt , kω are positive constants, and αvt , αω , and the up- we can redefine ω̇ as
dating law for the virtual parameter τ(t) are given according
to (15) as follows: ω̇ = B1 − β̈
kx xe + v̄tr τ̇d cos ψe
   
αvt where the intermediate variable
 αω  =  ω̄r τ̇d − β̇ + γye v̄tr τ̇d sin ψe + kψ ψe  (18) sin ψe sin ψe
ψe B1 = γ(ū˙r ye τ̇d + ūr ẏe τ̇d
τ̇ τ̇d − kv tanh Φe ψe ψe
The corresponding transformation between (τvt , τβ ) and sin ψe cos ψe ψe ψ̇e − sin ψe ψ̇e
+ ūr ye τ̈d + ūr ye τ̇d )
(τu , τv ) for underactuated AUV can be written as ψe ψe2
1
+ ω̄˙ r τ̇d + ω̄r τ̈d + kψ ψ̇e + ( ψe − kω zω )/mr

τu = mu τvt / cos β
(19) γ
τβ = −τvt tan β /vt
Secondly, as the side-slip angle is β = arctan(v/u), there
Note that β = π2 is impossible, as the surge velocity u > 0 is
is assumed for underwater vehicles. β̇ = (uv̇ − vu̇)/(u2 + v2 ) = (uv̇ − vu̇)/vt2 (23)
In (19), there is
 Hence, the acceleration of side-slip angle is
 v˙t = fvt (vt , β , r) + τvt (τu , τv , vt , β )
β̇ = fβ (vt , β , r) + τβ (τu , τv , vt , β ) (20) (u̇v + uv̈ − v̇u̇ − vü)vt2 − 2(uv̇ − vu̇)vt v̇t
β̈ =
ṙ = fr (vt , β , r) + τr /mr vt4


and uv̈ − vü uv̇ − vu̇

= −2 v̇t

mv mu du 2 dv 2 vt2 vt3
 fvt = ( mu − mv )vt r sin β cos β − ( mu cos β + mv sin β )vt

mv 2 mu 2 du dv Replacing (23) to above equation, there is
fβ = −( mu r sin β + mv r cos β ) + ( mu − mv ) sin β cos β
 f = − dr +
 mur 2 uv̈ − vü v̇t
r mr mr vt sin β cos β β̈ = 2
− 2 β̇ (24)
(21) vt vt
Hence, the control inputs in dynamics stage can be written From the second term in the dynamics model (3), there
as is v̇ = −mum
ur−dv v
v
, and u = vt cos β , then β̇ in (23) can be
rewritten as


 τu = mu (v̇t − fvt )/ cos β

= [mu α̇vt + ((1 − λ )xe − kvt zvt ) − mu fvt ]/ cos β mu uvdv vu̇
β̇ = − cos2 β αr −


 − (25)
τ = mr ṙ − muv uv + dr r
mv mv vt2 vt2
r
Define the intermediate variable A1 as

1


= mr α̇ω + ( ψe − kω zω ) − muv uv + dr r



γ sin ψe
A1 = ω̄r τ̇d + γye v̄tr τ̇d + kψ ψe
(22) ψe
3561
Authorized licensed use limited to: INDIAN INSTITUTE OF TECHNOLOGY MADRAS. Downloaded on December 21,2023 at 16:24:54 UTC from IEEE Xplore. Restrictions apply.
 TABLE I
C ONTROL PARAMETERS OF AUV PATH TRACKING

kx = 1.0 kψ = 1.0 γ = 1.0 kτ = 0.5 kv = 0.3

Using (25) to represent β̇ , αω in (18) can be rewritten as

1 uvdv mu
αω = [A1 + (vu̇ + )]/(1 − cos2 β )
vt2 mv mv
Finally, as β̈ given in (24), there is
mu
ω̇ = C1 /(1 − cos2 β )
mv
where
üv v̇t u mu u̇r + mu uṙ + d˙v v
C1 = B1 + 2
+ 2 β̇ + 2 ( )
vt vt vt mv
Hence, the acceleration of side-slip angle β̈ can be com-
Fig. 2. Spatial convergence of PT: λ = 0.99
puted through these intermediate variables.
Note that the term (1 − m u 2
mv cos β ) appears in the de-
nominator of ω̇, so that the stern-dominant property of
the underactuated AUV is required, in order to let the
computation of β̈ be well posed.
III. NUMERICAL SIMULATIONS
In this simulation, the model of underactuated SIRENE
AUV [11] is used, and the underactuated SIRENE AUV is
required to track a reference sinusoid path, given by:
x(γ) = 0, y(γ) = −6 + 6cos(0.04πγ), γ̇(t) = 1.0m/s
The objective is to regulate the distance to the path and
the heading of the total speed of the vehicle to zero relative
to the given path, while keeping the time convergence.
Path tracking control law with weighting factor is adopted,
while different weighting factors are chosen to compare the
different path tracking behaviors. For each weighting factor,
all initial conditions of the AUV are the same as follows:
u(0) = 1.0ms−1 , v(0) = 0ms−1 , r(0) = 0rads−1 , x(0) =
−10m, y(0) = 10m, ψ(0) = π/2 Fig. 3. Spatial convergence of PT: λ = 0.1
For each weighting factor λ , control parameters are chosen
as the same, as displayed in Table I.
As path tracking merges the path following and trajectory TT target on the path. From Figure 4 to Figure 5, the speed
tracking behaviors, λ is the key parameter to determine how of temporal convergence is decreased. This happens because
much weights of the path following and trajectory tracking decreasing λ means decreasing the weighting of trajectory
play in the whole path tracking design respectively. From the tracking behavior, which has the disadvantage of keeping the
Lyapunov control function (10), it clearly shows that large λ temporal performance in motion control.
means more trajectory tracking weighting added in the path Hence, through this simulation, the path tracking behav-
tracking control. ior is well understood as a motion control strategy which
In the Figures 2 and 3, the spatial tracking performance combines the path following and trajectory tracking together
is shown by comparing the spatial convergence and tracking through the tunable weighting factor λ .
errors between the vehicle and the path. From Figure 2 to
Figure 3, show that the performance of spatial convergence IV. CONCLUSIONS
is increased, and the speed of tracking error tending to zero In this paper, the problem of path tracking control of au-
speeds up. This happens because decreasing λ means in- tonomous underwater vehicle is addressed, by combining the
creasing the weighting of path following behavior, such that conventional path following and trajectory tracking control
the spatial tracking convergence performance is improved. behaviors via introducing the virtual time parameter with an
In the Figures 4 and 5, the temporal performance is shown extra degree of freedom. Briefly, the idea of path tracking
by comparing the relative posture between the AUV and the can be concluded as follows. The path is parameterized by

3562
Authorized licensed use limited to: INDIAN INSTITUTE OF TECHNOLOGY MADRAS. Downloaded on December 21,2023 at 16:24:54 UTC from IEEE Xplore. Restrictions apply.
 The path tracking control is derived by Lyapunov based
nonlinear design and backstepping technique. The simulation
results illustrate the efficiency of the proposed control laws.
The tunable weighting factor clearly shows how to combine
the trajectory tracking and path following behaviors in path
tracking strategy, and the different characteristics among
these numerical examples can be perceived therein.
Future work will address the problem of path tracking con-
trol of underactuated autonomous underwater vehicles in the
presence of parameter uncertainty and external disturbances,
as well as output feedback control with reduced motion
sensor availability. By utilizing the temporal performance
of the path tracking control, coordinated paths tracking is
also of interest which can address the rendezvous problem
at the same time. Moreover, this novel control design of path
tracking proposed for underactuated autonomous underwater
vehicles in this paper, can also be extended to other kinds
of nonholonomic and underactuated autonomous vehicles,
Fig. 4. Temporal convergence of PT: λ = 0.99
including land, aerial and astronautic vehicles.
R EFERENCES
[1] R. Hindman and J. Hauser. Maneuver modified trajectory tracking. In
Proc. of Int. Symp. Mathematical Theory Networks and Systems, St.
Louis, MO, USA., pages 1473–1479, 1996.
[2] P. Encarnacao and A. Pascoal. Combined trajectory tracking and path
following: an application to the coordinated control of autonomous
marine craft. In Decision and Control, 2001. Proceedings of the 40th
IEEE Conference on, volume 1, pages 964–969, 2001.
[3] Jean-Paul P. Laumond. Robot Motion Planning and Control. Springer-
Verlag New York, Inc., Secaucus, NJ, USA, 1998.
[4] Do. Khau Duc and Pan. Jie. Control of Ships and Underwater Vehicles.
Springer-Verlag London Limited 2009, London, 2009.
[5] R. W. Brockett. Asymptotic stability and feedback stabilization.
Differential Geom. Control Theory, pages 181–191, 1983.
[6] Saif A. Al-Hiddabi and N. Harris McClamroch. Aggressive longitu-
dinal aircraft path tracking using nonlinear control. Asian Journal of
Control, 3(4):280–288, 2001.
[7] Saif A. Al-Hiddabi and N. Harris McClamroch. Aggressive longitu-
dinal aircraft trajectory tracking using nonlinear control. Jouranl of
Guidance, Control, and Dynamics, 25(1):26–32, 2002.
[8] L. Lapierre, D. Soetanto, and A. Pascoal. Nonlinear path following
with applications to the control of autonomous underwater vehicles.
In Decision and Control, 2003. Proceedings. 42nd IEEE Conference
Fig. 5. Temporal convergence of PT: λ = 0.1 on, volume 2, pages 1256 – 1261, dec. 2003.
[9] L. Lapierre and B. Jouvencel. Robust nonlinear path-following control
of an auv. IEEE Journal of Oceanic Engineering, 33(2):89–102, Apr
2008.
actual path parameter τd , pursuing the desired speed profile [10] D. Soetanto, L. Lapierre, and A. Pascoal. Adaptive, non-singular path-
predefined by τ̇d (t) which is constrained by time instant. following control of dynamic wheeled robots. In Decision and Control,
However, a virtual path parameter τ is introduced, and the 2003. Proceedings. 42nd IEEE Conference on, volume 2, pages 1765
– 1770, Dec. 2003.
evolution speed of the virtual path parameter τ̇(t) has an [11] A.P. Aguiar. Nonlinear Motion Control of Nonholonomic and Under-
extra degree of freedom, in order to comply with the status actuated Systems. PhD thesis, IST, Portugal, 2001.
of the tracking vehicle. The actual and virtual path can [12] T. I. Fossen. Guidance and Control of Ocean Vehicles. Wiley, New
York, 1994.
be considered as one trajectory tracking target (TT target) [13] I. Kanellakopoulos M. Krstic and P. Kokotovic. Nonlinear and
and another path tracking target (PT target) moving along Adaptive Control Design. Wiley, New York, 1995.
respectively. When the tracking error is large, the PT target [14] H. K. Khalil. Nonlinear Systems. Prentice-Hall, Upper Saddle River,
NJ, 3rd edition, 2002.
will slow down or speed up to collaborate with the vehicle’s [15] L. Lapierre, D. Soetano, and A. Pascoal. Nonsingular path following
movement, and path tracking is prone to path following-like control of a unicycle in the presence of parametric modelling un-
behavior; otherwise, under certain small tracking error, PT certainties. International Journal of Robust and Nonlinear Control,
16(10):485–503, Jul 10 2006.
target will catch up with the TT target, and path tracking [16] T.I. Fossen, M. Breivik, and R. Skjetne. Line-of-sight path following
acts as trajectory tracking-like behavior to keep the temporal of underactuated marine craft. In Proceedings of the 6th Conference
performance as possible as it can achieve. Therefore, the on Manoeuvring and Control of Marine Craft (MCMC 2003), Girona,
Spain, September 2003.
smoother spatial convergence is achieved and the temporal
performance is guaranteed in the end.

3563
Authorized licensed use limited to: INDIAN INSTITUTE OF TECHNOLOGY MADRAS. Downloaded on December 21,2023 at 16:24:54 UTC from IEEE Xplore. Restrictions apply.