Robust Non-singular Fast Terminal Sliding Mode Task-Space

Position Tracking Control of an Underwater

Vehicle-Manipulator System
P. S. Londhe B. M. Patre∗
SGGS Institute of Engineering & Technology,Vishnupuri SGGS Institute of Engineering & Technology,Vishnupuri
Nanded, Maharashtra 431 606 Nanded, Maharashtra 431 606
pandurangl97@gmail.com bmpatre@yahoo.com

L. M. Waghmare S. Mohan
SGGS Institute of Engineering & Technology,Vishnupuri Indian Institute of Technology
Nanded, Maharashtra 431 606 Indore, Madhya Pradesh
lmwaghmare@yahoo.com santhakumar@iiti.ac.in

ABSTRACT 1 INTRODUCTION
This paper addresses a task-space trajectory control of an under- The application of underwater robots for marine research and deep-
water vehicle-manipulator system (UVMS) employed for interac- sea autonomous underwater intervention missions involving non-
tive underwater tasks. The robust task-space tracking control is trivial underwater tasks such as shallow or deep sea pipelines
achieved by designing a non-singular fast terminal sliding mode survey pipeline, oceanographic mapping, shipwreck search and
controller (NFTSMC) with disturbance estimator and demonstrated surveillance etc. has gained huge importance in recent years. For
on a planar underwater vehicle with serial two link manipulator performing such complicated tasks, particularly those are inac-
arm attached to it. The proposed NFTSMC integrates a non-singular cessible to humans require a manipulator arm mounted on the
fast terminal sliding mode controller (NFTSMC) with a non-linear underwater vehicle, refereed as underwater vehicle-manipulator
disturbance observer. This combination not only assures finite and system (UVMS) in the literature. However, it is very difficult to
faster convergence of the systems states to the equilibrium from control the vehicle- manipulator system as a whole because of the
anywhere in the phase-plane but also overcomes the problem of following reasons:
singularity associated with conventional terminal sliding mode
• Strong dynamic coupling between the manipulator and
controller (TSMC). In addition to this, because of the disturbance
vehicle;
observer augmented in the proposed control law, the overall sta-
• Kinematically redundant systems;
bility of the closed-loop system is enhanced to a great extent. The
• Presence of model and parameter uncertainties due to
feasibility of the proposed NFTSMC is confirmed by performing
poorly known hydrodynamic parameters;
extensive numerical simulation on the UVMS for tracking a given
• Presence of un-modelled dynamics and unknown external
pre-defined task space trajectory under the influence of parameter
disturbances (including payload and other dynamic state
uncertainties, ocean current and measurement sensor noises.
variations).
KEYWORDS Therefore because of the aforementioned issues, it is very arduous
Underwater Vehicle-Manipulator System, Terminal Sliding Mode task for a control engineer to design a control system which makes
Control, Disturbance Observer, Task-Space Control the UVMS to follow a given pre-defined trajectory precisely.
In recent years, several advanced control strategies have been
ACM Reference format: proposed by many researchers for improving the trajectory track-
P. S. Londhe, B. M. Patre, L. M. Waghmare, and S. Mohan. 2016. Robust ing performance of the UVMS. Usually the manipulation tasks of
Non-singular Fast Terminal Sliding Mode Task-Space Position Tracking the UVMS are specified in the task-space in terms of desired trajec-
Control of an Underwater Vehicle-Manipulator System. In Proceedings of
tories of the end-effector, whilst control law can be derived either
AIR ’17, New Delhi, India, June 28-July 2, 2017, 6 pages.
in body-fixed or inertial (earth)-fixed reference frames. This nat-
DOI: https://doi.org/10.1145/3132446.3134870
urally leads to various control methods, the so-called task-space,
configuration-space and body-fixed frame control approaches [10].
∗ Dr. B. M. Patre is the corresponding author. As the control inputs for the manipulator arm are joint torques,
therfore mostly configuration-space control schemes are preferred.
However, being kinematically redundant structure, for performing
ACM acknowledges that this contribution was authored or co-authored by an employee,
contractor or affiliate of a national government. As such, the Government retains a any particular underwater manipulation task leads infinite number
nonexclusive, royalty-free right to publish or reproduce this article, or to allow others of configuration-space solutions. As a result, redundancy resolu-
to do so, for Government purposes only.
tion schemes are required to handle the situation [1, 4, 8, 9, 15–18].
AIR ’17, New Delhi, India
© 2017 ACM. 978-1-4503-5294-9/17/06. . . $15.00 Alternatively, control action can be achieved by deriving the control
DOI: https://doi.org/10.1145/3132446.3134870 law in the body-fixed reference frame. Besides, there are variety
AIR ’17, June 28-July 2, 2017, New Delhi, India P. S. Londhe, B. M. Patre, L. M. Waghmare, and S. Mohan

yt
of tasks where UVMS has to perform more than one objective. For
instance, consider the UVMS has to perform the end-effector po- xt
sition tracking as a primary objective and an obstacle avoidance E
End-effector
as a secondary objective. In such situations, configuration-space frame
or the body-fixed reference frame based control designs are not ym
suitable [10]. In contrast, the task-space based control could be yb
θ1 θ2
most desirable since it does not require any inverse kinematic solu- Body-fixed
xb xm
frame
tion. Furthermore, the task-space control is more appropriate when B M Manipulator-base
the online modification in the task-space of the end-effector is fre- frame

quently required [10]. The main advantage of task-space approach

is that, it utilises a feedback loop that directly minimises task-space
errors without requiring any computation of inverse kinematics
y
explicitly. Few attempts are made in this diection [12–14].
In this paper, a task-space control is designed by incorporating
a non-singular fast terminal sling mode control for ensuring finite, I x
Earth-fixed
fast state convergence to the equilibrium point and a disturbance frame
estimator for compensating the effects of un-modeled dynamics, un-
known disturbances due to ocean currents and uncertainties in the Figure 1: Coordinate frame arrangement of an UVMS
hydrodynamic parameters of the vehicle. The proposed NFTSMC
is not only giving singularity-free control but also provides a high
precision control without any prior knowledge about the upper manipulator arm; and τ ct = [τ v τ m ]T where, τ v = [X Y N ]T
bounds of the norm of lumped disturbance acting on the UVMS. is the resultant input vector of thrusters, control plane forces and
The validness of the proposed method is numerically demonstrated moments. τ m = [τm1 τm2 ]T is the vector of manipulator control
by considering a planar UVMS (consisting of a two-link serial pla- inputs while τm1 , τm2 are the joint torques of the corresponding ma-
nar manipulator mounted on an vehicle in the horizontal plane) nipulator links. τedis is the vector of external disturbances due to
for tracking a pre-defined complex trajectory in an of uncertain underwater current and/or a paylaod, sensor measurement noises.
working condition. The relation between task-space and configuration-space coordi-
The paper is structured as follows. Section 2 describes dynamic nates are related as
modeling of an UVMS. Design of proposed NFTSMC to planar
UVMS is described in Section 3. Simulation results and performance µ = f (q) (2)
analysis has been discussed in in Section 4. Finally the paper is where µ = [x t yt ]T is the vector of task-space coordinates and f (q)
concluded in Section 5. is the vector representing the forward kinematics. The derivative
of (2) gives
2 DYNAMIC MODELING OF AN UVMS
µÛ = J (q)qÛ (3)
This study uses the dynamic model of an UVMS which is derived
using the Newton - Euler and recursive Newton - Euler formulations where, J (q) is the Jacobian matrix represents mapping between
[3, 7]. The frame arrangement considered for an UVMS is shown configuration-space velocities to task-space velocities.
in Fig. 2. where, I (x, y, z) is the Earth-fixed frame, B(xb , yb , zb ) is Hence, from (3) and (1) the dynamics of an planar UVMS can be
the vehicle body-fixed frame, E(x t , yt , zt ) is the manipulator-end represented in the task-space coordinates as
effector frame and M(xm , ym , zm ) is the manipulator-base frame.
M µ µÜ + C µ µÛ + D µ µÛ + G µ + F µ = τ ct + τ edis (4)
For the sake of clarity, the motion of UVMS in the horizontal plane
is considered, in which 2-DOF manipulator arm mounted on 3-DOF where, M̂ µ , Ĉ µ , D̂ µ , Ĝ µ and F̂ µ are the known values (approx-
underwater vehicle model. Under these considerations, the dynamic imated model parameter values) of inertia matrix, Coriolis and
model of UVMS in Earth-fixed reference frame is expressed as centripetal matrix, damping effects, restoring effects and dynamic
coupling effects (forces) of the UVMS, respectively. τ ct is the vector
M(q)qÜ + C(q, q)
Û qÛ + D(q, q)
Û qÛ + G(q) +
of control torques/forces generated during task-space control of
F (q, q) = τ ct + τ edis
Û (1) the UVMS.
Considering external disturbances and parameter uncertainties in
where, q = [η ζ ]T = [x y ψ θ 1 θ 2 ]T with η = [x y ψ ]T is the
(4), we have,
vector of position and orientation vehicle in horizontal plane and
ζ = [θ 1 θ 2 ]T is the vector of joint variables while θ 1 , θ 2 are the M µ (µ) + 4M µ (µ) µÜ + C µ + 4C µ µÛ

+ D µ + 4D µ µÛ + G µ + 4G µ + F µ + 4F µ
joint positions of the underwater manipulator link, as depicted in

Fig. 2. All the matrices in (1) are defined in [13].
= τc t + τ edis (5)
M(q) is the inertia matrix; C(q, q)
Û is the matrix of Coriolis and
centripetal forces; D(q, q)
Û is the matrix of friction and hydrody- Now, by collecting unknown part including parameter uncertainties,
namic damping terms; G(q) is the vector of restoring forces; F (q, q) Û un-modeled dynamics and unknown external forces like ocean
is the vector of forces generated due to coupling between AUV and current into a single vector term as lumped uncertainty τ dis , then
Robust Non-singular Fast Terminal Sliding Mode Task-Space Position Tracking Control of an
AIR
Underwater
’17, June 28-July
Vehicle-Manipulator
2, 2017, New Delhi,
SystemIndia

(5) can be written as: The terminal sliding mode will exits if following sufficient condition
is satisfied,
M̂ µ (µ)µÜ + Ĉ µ µÛ + D̂ µ µÛ + Ĝ µ + F̂ µ = τc t + τ dis (6)
1 d 2
s < −ζi | si | (11)
where, 2 dt i
τ dis = τ edis − 4M µ (µ)µÜ − 4C µ µÛ where ζi (i = 1, · · · , n) is a positive constant.
− 4 D µ µÛ − 4G µ − 4F µ (7) According to sufficient condition for the existence of TSM, a NFTSMC
can be designed as
and 4M µ (µ)µÜ = M µ (µ)µ− Ü M̂ µ (µ)µ;
Ü 4C µ µÛ = C µ µ−
Û Ĉ µ µÛ ; 4D µ µÛ =
q
 p
Û + γsiдnγ −1 (e).
1
D µ µÛ − D̂ µ µ,Û 4G µ = G µ − Ĝ µ , and 4F µ = F µ − F̂ µ , in which M̂ µ (µ), τc t = M̂ µ (µ) µÜ d + β siдn q (e)
2−
p α
Ĉ µ , D̂ µ , and Ĝ µ , and F̂ µ are the nominal values of an planar UVMS
q p

model. τ edis is the vector of external disturbances like ocean cur- β siдn q (e)
2−
Û + Ksat(s, Φ) + N̂ µ (µ, µ)
Û (12)
rent etc. p
For the sake of simplicity, (6) can be rewritten as, where, Φ is the boundary layer thickness introduced around the
M̂ µ (µ)µÜ + N̂ µ (µ, µ)
Û = τc t + τ dis (8) sliding surface s = 0 in order to annihilate the chattering and is
defined as follows:
where, N̂ µ (µ, µ)
Û = Ĉ µ µÛ + D̂ µ µÛ + Ĝ µ + F̂ µ . ( s
i
if si > Φi
{sat(s, Φ)}i = s|sii | (13)
3 PROPOSED CONTROL SCHEME Φi if si ≤ Φi
3.1 Motivation K = diaд[K 11 , · · · , Kii , · · · , Knn ] is a positive definite matrix and
Sliding mode control (SMC) is well-known powerful control scheme Kii (i = 1, · · · , n) is a positive constant.
and extensively exercised for mutiple-input multiple-output (MIMO) When si , 0, then ei will reach the terminal sliding surface si = 0
linear as well as nonlinear systems. The main feature of SMC is its within the finite time tr i (i = 1, · · · , n), which satisfies
robustness property against parameter uncertainties and external | si (0) |
tr i 6 (14)
disturbances. However, in general conventional SMC uses linear Kii
sliding surface, which cannot guarantee finite time convergence of
the systems states to equilibrium point. To overcome this shortcom- Remark 1. When the terminal sliding mode is achieved i.e si = 0,
ing, a terminal sliding mode control (TSMC) is firstly suggested by the system dynamics is determined by the dynamics of the sliding
Venkataraman et al. [19] and Man et al. [21] based on the concept of manifold thereafter:
p
| ei |γ siдn(ei ) +
a terminal attractor. However, the conventional TSMC suffers from 1 1
ei + | eÛi | q siдn(eÛi ) = 0 (15)
a singularity dilemma and the control amplitude may go to infinity αi βi
whenever the error approaches to zero [6]. Significant research has
where ei = 0 is the terminal attractor of the system. The finite time
been done to avoid the singularity issue by non-singular terminal
tsi is taken to travel from ei (tr i ) , 0 to ei (tr i + tsi ) = 0 is given by
sliding mode control (NTSMC) [2, 5, 6, 11]. Nevertheless, NTSMC
[20]
exhibits a slower convergence rate when system states are afar
∫ 0
from the equilibrium than the conventional SMC. In this paper, a 1 1
tsi = − q q dτ (16)
control technique is proposed by adding one more nonlinear term e (t ) γ
βip i ri (τ − α i τ ) p
1
to the sliding manifold of NTSMC, termed as NFTSMC.
From the proposed NFTSMC control law in (12), it can be con-
3.2 Non-singular Fast Terminal Sliding Mode clude that, even though the proposed NFTSMC exhibits a finite and
Control fast convergence property but it require a priori information of the
The control objective is to force the states µ to track the desired upper bounds of the lumped uncertainty term, τ dis to obtain robust
position µd . Lets define the tracking error as, control performance. For improving robustness, an disturbance
estimator is proposed for online estimation of lumped uncertainty
e = µd − µ (9) term.
where, µd represents desired task-space trajectory. A non-singular
fast terminal sliding manifold for the system (8) can be defined as 3.3 Design of Disturbance Observer
p In this study, the dynamic disturbance estimation approach is pro-
siдnγ (e) + siдn (Ûe) = 0
1 1
s =e+ q (10)
α β posed to estimate the unknown perturbations from the dynam-
ics of the non-singular fast terminal sliding surface. This can be
where s = [s 1 , · · · , si , · · · , sn ]T ∈ <n is sliding surface, β = diaд[β 1 , accomplished by treating the effects of the system uncertainties,
· · · , βi , · · · , βn ] ∈ <n×n in which βi (i = 1, · · · , n) is a positive nonlinearities, unmodeled dynamics, cross-coupling effects and
constant; α = diaд[α 1 , · · · , α i , · · · , α n ] ∈ <n×n in which α i (i = external disturbances like an ocean current as a total or composite
1, · · · , n) is a positive constant; siдnγ (e) = [| e 1 |γ siдn(e 1 ), · · · , | disturbance, termed as lumped uncertainty acting on the UVMS
ei |γ siдn(ei ), · · · , | en |γ siдn(en )]T ∈ <n and p and q are positive and then estimate this unknown term using the disturbance esti-
p p
odd integers satisfying the relation 1 < q < 2 and γ > q . mation technique. Finally, its effect on the UVMS can be canceled
AIR ’17, June 28-July 2, 2017, New Delhi, India P. S. Londhe, B. M. Patre, L. M. Waghmare, and S. Mohan

Unknown torques/forces xt
yt

With some adjustment in (23) gives,

τdis ym
E

p
yb  
θ1 θ2 p
Body-fixed
xm
−1 1 −1
τÛ dis − τÛ̂ dis = −Γ M̂ η τ̃ dis + s siд q (Ûe) + τÛ dis
frame xb
M Manipulator-base

Known torques/forces B
frame

Position and/or
τct β q
velocity and /or
p
 p

AUVMS −1 1 −1
acceleration τÛ̃ dis = −Γ M̂ η τ̃ dis + s siд q (Ûe) + τÛ dis (24)
(μ , μ , μ ) β q
τdis Disturbance The assumption 3.1 holds true if
estimator
Estimate of the 1p
 p

−1 −1
lumped τÛ̃ dis = −Γ M̂ η τ̃ dis + s siд q (Ûe) . (25)
βq
disturbance term
3.4 Stability Analysis
Figure 2: Conceptual idea of proposed disturbance observer
The following assumptions are made to show the stability analysis
of the proposed NFTSMC method.
out by augmenting it in the nonlinear NFTSMC law given by (17). Assumption 3. The lumped disturbance term τ dis is bounded and
Therefore the control law in (12) is modified as there exits a constant Dmax > 0 such that 0 ≤| τ dis |≤ Dmax but

q p value of upper of bound of Dmax is not required to be known.
Û + γsiдnγ −1 (e).
1
τ µ = M̂ µ (µ) µÜ d + β siдn q (e)
2−
p α Assumption 4. The desired trajectory, µd is twice differential and
q smooth i.e. µd , µÛ d , µÜ d and available with known bounds.
p

β siдn q (e) Û + Ksat(s, Φ) − τ̂ dis + N̂ µ (µ, µ)
2−
Û (17)
p
Theorem 3.1. Consider the dynamics of an UVMS in (8), if the
Fig. 2 explains the basic idea of a proposed disturbance observer. control input vector is chosen as defined in (17), then the output
The following assumptions are made to ensure convergence of tracking error vector converges to zero in finite time with respect to s.
disturbance estimation errors.

Assumption 1. The value of the lumped disturbance is arbitrarily Proof. Consider the Lyapnunov candidate function as
large and slowly varying with time i.e. (τÛ dis ≈ 0).
V = sT s + τ̃ Tdis τ̃ dis
1 1
(26)
Assumption 2. The disturbance observer gain namely Γ is con- 2 2
stant symmetric and positive definite (SPD) matrix by design, that Differentiating (26),
is:
VÛ = sT sÛ + τ̃ Tdis τÛ̃ dis (27)
Γ = ΓT > 0. (18)
Now taking derivative of (10) and (25), we have,
A disturbance observer is suggested as,
1p
 p i
VÛ = sT
h
−1
τ̂ dis = ϒ + Γ µÛ (19) siд q (Ûe) −Ksat(s, Φ) − M η−1τ dis
βq
−1
h i
ϒÛ = −Γ M̂ µ τ ct + ϒ + Γ µÛ − N̂ µ (µ, µ) 1p
  
Û T −1 p
−1
+τ̃ dis −Γ M̂ η τ̃ dis + s siд q (Ûe) . (28)

1p p
 βq
−1
−s siдn q (Ûe) (20)
βq Simplification of (28) leads to
where, τ̂ dis is the vector of estimated lumped uncertainty; ϒ is an n
1 p p
−1
Õ
auxiliary vector which estimate the unknown perturbations from VÛ 6 − | si | | eÛi | q siдn(eÛi ) [Kii +
i=1
β i q
the dynamics of sliding surface.
T
− τ̃ Tdis ΓM η−1τ̃ dis .
i
Defining observer error as, k M η−1 k τdis i − τ̃dis (29)
i
τ̃ dis = τ dis − τ̂ dis (21) p
Since p and q are positive odd integers and 1 < q < 2, there is
Now, τ̂ dis is to be modify in a such way that the estimation error p
−1
p
−1
τ̃ dis goes to zero within a finite amount of time. | eÛi | q > 0 for eÛi , 0 and | eÛi | q = 0 for eÛi = 0. In addition, it
Differentiating (19), can be noted that the proposed control law in (17) does not contain
p p
any negative fractional power since 1 < q < 2 and γ > q . Thus,
τÛ̂ dis = ϒÛ + ΓηÜ (22) it gives singularity-free control. Hence, according to Lyapunov
stability criteria, the non-singular fast nonlinear sliding surface in
From (19) and (8), we have,
(10) converges to zero from any initial states in finite time given
1p
 p

−1 −1 by (16). On the other hand, the output tracking errors of an UVMS
τÛ̂ dis = Γ M̂ η τ̃ dis − s siд q (Ûe) (23)
βq will converge to zero in finite time. 
Robust Non-singular Fast Terminal Sliding Mode Task-Space Position Tracking Control of an
AIR
Underwater
’17, June 28-July
Vehicle-Manipulator
2, 2017, New Delhi,
SystemIndia

Table 1: Performance indices for analysis of controller per-

4 0.02
formance
0
2
xe ye

µ̃xt ,[m]
xt,[m]
Control schemes −0.02
RMS L2 norm RMS L2 norm 0
−0.04
CSMC 0.0321 0.4557 0.0225 0.3190
NTSMC 0.0024 0.0341 0.0034 0.0477 −2 −0.06
0 50 100 150 200 0 50 100 150 200
t,[s] t,[s]
NFTSMC 6.9189 ×10−4 0.0098 3.5195 ×10−4 0.0050 4 0.02

0
2
4 NUMERICAL SIMULATION AND

µ̃yt ,[m]
yt,[m]
−0.02
PERFORMANCE ANALYSIS 0
−0.04
In order to illustrate the control performance of the proposed
NFTSMC scheme, the extensive numerical simulation on an planar −2 −0.06
0 50 100 150 200 0 50 100 150 200
UVMS is performed. The vehicle physical parameters and the hydro- t,[s] t,[s]
dynamic parameters are chosen for simulation are taken from [13]. CSMC NTSMC NFTSMC

The effectiveness of the proposed controller has been evaluated

under uncertain working condition. This includes
Figure 3: Variations in the task-space positions and their cor-
• an underwater current of value (0.1, 0.2) m/s in x- and y- responding errors under uncertain working condition
axes;
• the unknown external disturbances of value [8N , 6N , 0.6Nm,
0.3Nm, 0.2Nm]T ; 3.5
• measurement sensory noises as Gaussian noises of 0.01 m 2.24
mean and 0.01 standard deviation in the position and 0.2 3 2.22
degree standard deviation in the orientation measurement 2.2

and 2.5 2.18

• 20% system parameter uncertainties. −0.62 −0.6 −0.58−0.56−0.54−0.52

Position of an
UVMS at t = 30 s
2
The comparison of the proposed method is made with conven-
tional SMC and NTSMC schemes. The parameters of the proposed
y, [m]

1.5
NFTSMC are selected as: p = 5, q = 3, γ = 1.4, β = 8I 2 and
α = 5.4I 2 . 1
The task-space position tracking performance given by all the
controllers are depicted in Figs. 3 and 4. The time history of tra- 0.5
Desired track
jectories of the UVMS in the task-space positions and their corre- CSMC
sponding errors in tracking positions under uncertain condition are 0 NTSMC
NFTSMC
shown in Fig. 3. From these results, it is confirmed that the tracking UVMS
Manipulator links
performance given by NFTSMC is quite better than CSMC. The task- −0.5 Manipulator joints

space positions tracking of an UVMS in the x − y plane is shown in −1.5 −1 −0.5 0 0.5 1 1.5 2 2.5 3 3.5
x, [m]
Fig. 4. This show that the CSMC provides steady state error in the
task-space tracking control. The position tracking performance of
an UVMS can be measured in terms of quantitative index such as L2 Figure 4: task-space position tracking of an UVMS in the x −y
norm and root mean square (RMS) of task-space errors as depicted plane under uncertain working condition
in Table 1. This table confirmed that, the proposed NFTSMC shows
minimal task-space errors in x- and y- directions as compared to
CSMC and NTSMC schemes. But, the proposed NFTSMC gives better immunity to variations in
To have an uttermost check on the robustness of the proposed the parameters uncertainties of an UVMS. This shows enhanced
NFTSMC scheme, the uncertainties in the system parameters are robustness property of the proposed NFTSMC due to the use of the
varied from 20 % to 80 % and control performance of the controller disturbance estimator in the control design.
is measured in terms of the norm of the task-space errors. The
variations in the norm of the task-space error during % variation in 5 CONCLUDING REMARKS
the system parameters of the UVMS can be observed in Fig. 5. From In this study, a fast finite time non-singular terminal sliding mode
these results, it can be concluded that the performance given by controller is designed for providing a robust prescribed task-space
NTSMC is quite acceptable with the maximum error norm of value trajectory tracking control performance of the UVMS. The effective-
0.08 for 80 % uncertainties, whereas CSMC loses its robustness ness of the proposed control scheme is tested and comparing with
property with increasing uncertainties in the system parameters. CSMC and NTSMC schemes. The proposed NFTSMC scheme offers
AIR ’17, June 28-July 2, 2017, New Delhi, India P. S. Londhe, B. M. Patre, L. M. Waghmare, and S. Mohan

[15] T. K. Podder and N. Sarkar. 2004. A unified dynamics-based motion planning

algorithm for autonomous underwater vehicle-manipulator systems (UVMS).
Robotica 22 (1 2004), 117–128. Issue 01.
0.4 [16] I. Santos, C.H.F., R. Guenther, D. Martins, and E.R. Pieri. 2006. Virtual kinematic
chains to solve the underwater vehicle- manipulator systems redundancy. Journal
0.3
of the Brazilian Society of Mechanical Sciences and Engineering 28 (2006), 354–
361. Issue 03.
|| e ||, [m]

[17] N. Sarkar and T.K. Podder. 2001. Coordinated motion planning and control of au-
0.2 tonomous underwater vehicle-manipulator systems subject to drag optimization.
Oceanic Engineering, IEEE Journal of 26, 2 (Apr 2001), 228–239.
0.1 [18] S. Soylu, B. J. Buckham, and R. P. Podhorodeski. 2010. Redundancy resolution
for underwater mobile manipulators. Ocean Engineering 37, 2fi??3 (2010), 325 –
343.
0 [19] S. T. Venkataraman and S. Gulati. 1993. Control of Nonlinear Systems Using
200
Terminal Sliding Modes. J. Dyn. Sys., Meas., Control 115, 3 (1993), 554.
150 80
[20] B. Xu, S. R. Pandian, N. Sakagami, and F. Petry. 2012. Neuro-fuzzy control of
100 60 rs
mete underwater vehicle-manipulator systems. Journal of the Franklin Institute 349, 3
para
t, [s] 50 40
in system (2012), 1125 – 1138.
0 20 ange
% ch [21] M. Zhihong, A. P. Paplinski, and H. R. Wu. 1994. A robust MIMO terminal
sliding mode control scheme for rigid robotic manipulators. IEEE Trans. Automat.
Control 39, 12 (1994), 2464–2469.
Figure 5: Variations in the norm of the task-space errors
when system parametrs are changed from 20 % to 80 % (
NFTSMC (blue color), NTSMC (violet color) and CSMC (red
color))

certain superior properties namely higher precision, fast finite time

convergence, singularity-free control and no need of any prior in-
formation of the uncertainties bounds. In addition, the proposed
control is also enhances the overall stability of the closed-loop sys-
tem due to the compensation of lumped disturbances provided by
the disturbance observer.

REFERENCES
[1] G. Antonelli and S. Chiaverini. 2003. Fuzzy redundancy resolution and motion
coordination for underwater vehicle-manipulator systems. Fuzzy Systems, IEEE
Transactions on 11, 1 (Feb 2003), 109–120.
[2] S. Y. Chen and F. J. Lin. 2011. Robust Nonsingular Terminal Sliding-Mode Control
for Nonlinear Magnetic Bearing System. IEEE Transactions on Control Systems
Technology 19, 3 (may 2011), 636–643.
[3] J. J. Craig. 1986. Introduction to Robotics: Mechanics and Control. Addison Wesley,
Boston, MA, USA.
[4] C.H. dos Santos, G. Bittencourt, and R. Guenther. 2006. Motion Coordination
for Underwater Vehicle-Manipulator Systems using a Fuzzy Hybrid Strategy. In
Intelligent Robots and Systems, IEEE/RSJ International Conference on. 3018–3023.
[5] Y. Feng, X. Yu, and F. Han. 2013. On nonsingular terminal sliding-mode control
of nonlinear systems. Automatica 49, 6 (2013), 1715 – 1722.
[6] Y. Feng, X. Yu, and M. Zhihong. 2002. Non-singular terminal sliding mode control
of rigid manipulators. Automatica 38, 12 (dec 2002), 2159–2167.
[7] T. I. Fossen. 1994. Guidance and control of Ocean Vehicles. Wiley, New York.
[8] J. Han and W. K. Chung. 2007. Redundancy resolution for underwater vehicle-
manipulator systems with minimizing restoring moments. In Intelligent Robots
and Systems, IEEE/RSJ International Conference on. 3522–3527.
[9] J. Han and W. K. Chung. 2014. Active Use of Restoring Moments for Motion
Control of an Underwater Vehicle-Manipulator System. Oceanic Engineering,
IEEE Journal of 39, 1 (Jan 2014), 100–109. DOI:http://dx.doi.org/10.1109/JOE.
2013.2241931
[10] Z. H. Ismail. 2011. Task-Space Dynamic Control of Underwater Control. Ph.D.
Dissertation. Heriot-Watt University.
[11] M. Jin, J. Lee, P. H. Chang, and C. Choi. 2009. Practical Nonsingular Termi-
nal Sliding-Mode Control of Robot Manipulators for High-Accuracy Tracking
Control. IEEE Trans. Ind. Electron. 56, 9 (sep 2009), 3593–3601.
[12] Y. Kim, S. Mohan, and J. Kim. 2014. Task space-based control of an underwater
robotic system for position keeping in ocean currents. Advanced Robotics 28, 16
(2014), 1109–1119.
[13] P. S. Londhe, M. Santhakumar, B. M. Patre, and L. M. Waghmare. 2017. Task Space
Control of an Autonomous Underwater Vehicle Manipulator System by Robust
Single-Input Fuzzy Logic Control Scheme. IEEE Journal of Oceanic Engineering
42, 1 (Jan 2017), 13–28.
[14] S. Mohan and J. Kim. 2015. Coordinated motion control in task space of an
autonomous underwater vehiclefi??manipulator system. Ocean Engineering 104,
0 (2015), 155 – 167.