You are on page 1of 16

This article has been accepted for inclusion in a future issue of this journal.

Content is final as presented, with the exception of pagination.

IEEE JOURNAL OF OCEANIC ENGINEERING 1

Task Space Control of an Autonomous Underwater


Vehicle Manipulator System by Robust Single-Input
Fuzzy Logic Control Scheme
Pandurang S. Londhe, M. Santhakumar, Member, IEEE, Balasaheb M. Patre, Senior Member, IEEE,
and Laxman M. Waghmare, Member, IEEE

Abstract—In this paper, a robust single-input fuzzy logic control the potential impact to enhance ability of humans to explore
Robust Single Input Fuzzy Logic Controller (RSIFLC) scheme is the ocean. However, for nontrivial underwater tasks, ROVs are
proposed and applied for task-space trajectory control of an au- not suitable because of being not only expensive but also ex-
tonomous underwater vehicle manipulator system (AUVMS) em-
ployed for underwater manipulation tasks. The effectiveness of the tremely difficult to control and requiring two or more skilled
proposed control scheme is numerically demonstrated on a planar operators for their operations [2]. To overcome these issues, au-
underwater vehicle manipulator system [consisting of an underwa- tonomous underwater vehicles (AUVs) are mostly preferred for
ter vehicle and a two link rotary (2R) serial planar manipulator]. underwater intervention tasks since they provide high speed and
The actuator and sensor dynamics of the system are also incorpo- precision in addition to being less expensive and easy to operate
rated in the dynamical model of an AUVMS. The proposed control
law consists of a feedforward term to exaggerate the control activ- [3]–[6]. The key element in underwater intervention performed
ity with immoderation from the known desired acceleration vector with autonomous vehicles is autonomous manipulation, usu-
and an estimated perturbed term to compensate for the unknown ally completed through manipulator systems [7]. Generally, the
effects namely external disturbances and unmodeled dynamics as AUVs with attached manipulator systems are widely used in
a first part and a single-input fuzzy logic control as a feedback por- the fields of scientific research and ocean systems engineering
tion to enhance the overall closed-loop stability of the system as a
second part. The primary objective of the proposed control scheme for performing interactive tasks such as opening and closing of
is to track the given end-effector task space trajectory despite of valves, cutting in coordination, drilling, sampling, coring, etc.
external disturbances, system uncertainties, and internal noises [8]–[10].
associated with the AUVMS. To show the efficacy of the proposed An AUV when equipped with a manipulator system becomes
control scheme, comparison is made with conventional fuzzy logic
a kinematically redundant system, i.e., it has more degrees of
control (CFLC), sliding mode control (SMC), and proportional–
integral–derivative (PID) controllers. Simulation results confirmed freedom (DOF) than is required to perform a task in its op-
that with the proposed control scheme, the AUVMS can success- erational space (i.e., m-dimensional). This leads to an infinite
fully track the given desired spatial trajectory and gives better and number of joint-space solutions for specific task space coordi-
robust control performance. nates and thus requires redundancy resolution schemes to handle
Index Terms—Autonomous underwater vehicle manipulator the situation [2], [11]. A number of algorithms have been pro-
system (AUVMS), Lyapunov stability, proportional–integral– posed over a period of time to resolve the problem of redundancy
derivative (PID) control, robust control, single-input fuzzy logic resolution [2], [11]–[19]. However, many of these approaches
control, task space control. require sophisticated optimization techniques [20] and the mo-
tions of the autonomous underwater vehicle manipulator system
I. INTRODUCTION (AUVMS) are coordinated through joint-space control schemes.
Besides, there are a variety of underwater tasks, where control
N recent years, the application of underwater robotic vehi-
I cles in the field of marine research for the explorations of
ocean resources and preservation of oceanic environments has
of the end-effector motion is highly desirable and needs to be
very precise. In such situations, joint-space control schemes
may not be suitable [21]. The task space control scheme can be
gained significant momentum [1]. Underwater robotic vehicles used for more than one control objectives, i.e., position tracking
such as remotely operated vehicles (ROVs) have demonstrated as a primary objective and low control activity, such as obsta-
cle avoidance, etc., as secondary objectives [22]. Also, online
Manuscript received June 4, 2015; revised December 1, 2015, February 2, modification of motion of the end-effector task space can be
2016, and March 7, 2016; accepted March 23, 2016. This work was supported easily accommodated by task space control schemes in contrast
by the Naval Research Board, Directorate of Naval Research and Development,
DRDO, Government of India under Sanction NRB-23B/SC/11-12.
to joint-space control schemes [21].
Associate Editor: B. Englot. Considering the above facts, a robust single-input fuzzy logic
P. S. Londhe, B. M. Patre, and L. M. Waghmare are with the Shri Guru Gobind control scheme is proposed and applied for the task space con-
Singhji Institute of Engineering and Technology, Vishnupuri, Nanded 431606,
India (e-maiil: pandurangl97@gmail.com; bmpatre@ieee.org; lmwaghmare@
trol problem of an AUVMS. The control of AUV becomes more
yahoo.com). difficult when the manipulator is attached to it. This leads to
M. Santhakumar is with the Mechanical Engineering Department, Indian highly nonlinear, coupled, and high-dimensional nature of an
Institute of Technology, Indore, Madhya Pradesh 453441, India (e-maiil:
santhakumar@iiti.ac.in).
AUVMS. In addition, uncertainties in the hydrodynamic param-
Digital Object Identifier 10.1109/JOE.2016.2548820 eters and influence of the external disturbances, such as ocean

0364-9059 © 2016 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See http://www.ieee.org/publications standards/publications/rights/index.html for more information.
This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

2 IEEE JOURNAL OF OCEANIC ENGINEERING

currents, make the task of controlling an AUVMS even more proposed control scheme is evaluated with a characteristic tra-
challenging. Several modeling and control methods applied on jectory tracking control problem of an AUVMS. The obtained
underwater robotics systems can be found in the literature [4], results show better robust control performance of the proposed
[8], [23]–[36]. However, due to unstructured properties of inter- control scheme.
active underwater work, a good understanding of the dynamics The paper is organized as follows. Section II describes dy-
of a robotic manipulator mounted on a moving underwater ve- namic modeling of an AUVMS. Section III presents the pro-
hicle is required for autonomous manipulation tasks. Therefore, posed robust single-input fuzzy logic controller design scheme.
the aforementioned control technique requires an exact math- The stability analysis of the closed-loop system is discussed
ematical model of the underwater robotic system to achieve in Section IV. Performance analysis of the proposed control
better control performance. It is well known that obtaining an scheme to an AUVMS is presented in Section V, followed by
exact mathematical model of an AUVMS is extremely arduous the conclusion in Section VI.
since it involves uncertainties in hydrodynamic parameters, the
dynamics are highly coupled and nonlinear, and also it is as- II. DYNAMIC MODELING OF AN AUVMS
sociated with time-varying dynamical characteristics. In such
In this work, dynamic modeling of an AUVMS is derived
situations, intelligent control techniques such as fuzzy logic
using Newton–Euler and recursive Newton–Euler formulation
control (FLC) can be useful. The main feature of FLC is that it
schemes [43], [44], [3]. The frame assignment of the AUVMS
can be applied to the plants that are not so well defined math-
is presented in Fig. 6, where, I (x, y, z) is the Earth-fixed (in-
ematically. For such plants, FLC can be designed to emulate
ertial) frame, B (xb , yb , zb ) is the vehicle body-fixed (moving)
human deductive thinking, i.e., embody human-like thinking
frame, E (xt , yt , zt ) is the manipulator-end effector frame, and
into an automatic control system [37], [38]. The conventional
M (xm , ym , zm ) is the manipulator-base frame. The dynamic
FLCs came from the concepts of linear proportional–derivative
equations of motion in the inertial (Earth-fixed reference) frame
(PD) and proportional–integral (PI) control schemes [39], [40].
of an AUVMS can be expressed as follows:
Based on experience of a human expert, heuristics fuzzy con-
trol rules can be constructed to model the behavior of the sys- M (q) q̈ + C (q, q̇) q̇ + D (q, q̇) q̇ + G (q) + F (q, q̇)
tem. The nonlinearities of the system can be handled by the
= τct + τedis (1)
appropriate choice of input and output membership functions
(MFs). Another feature of an FLC is that it provides a high de- where
gree of robustness and immunity to external disturbances and  
Mv H T (ζ)
can be configured to be self-learning and adaptive. However, it M (q) = (2)
requires substantial computational power, due to the complex H (ζ) Mm (ζ)
decision-making processes. For example, conventional FLC in-  
Cv (η, η̇) 0 
volves fuzzification, rule-base storage, inference mechanism, C (q, q̇) = (3)
and defuzzification operations [37]. To have better accuracy in 0 Cm ζ, ζ̇
control, a larger set of rules is required, which leads to longer  
Dv (η, η̇) 0 
computational time. But this may not be suitable for practical D (q, q̇) = (4)
applications since there are many implementation issues that 0 Dm ζ, ζ̇
must be addressed, namely real-time response, communication  
bandwidth, computational capacity, and onboard battery [41], Fv (q, q̇)
F (q, q̇) = (5)
[42]. Then, the design of an FLC becomes very difficult due to Fm (q, q̇)
   
the increased number of fuzzy control rules as well as tuning Gv (η) τ
parameters. Therefore, it is necessary to design an FLC that G (q) = , τct = v (6)
Gm (ζ) τm
has simple control structure and computational efficiency when
compared to the conventional FLC. and q = [η ζ]T with η = [x y z φ θ ψ]T is the vector of ab-
In this paper, a new robust single-input fuzzy logic control solute positions and Euler angles (roll, pitch, and yaw). x is
scheme is designed and applied for complex trajectory track- the surge position, y is the sway position, z is the heave po-
ing control in task space of an AUVMS. The proposed control sition, φ is the roll angle, θ is the pitch angle, and ψ is the
scheme is divided into two parts, namely, a feedforward term yaw angle. Mv (q) q̈ is the vector of inertial forces and mo-
along with a disturbance estimator and a single-input fuzzy ments of the vehicle (including added mass effects). Cv (ν, η) η̇
logic control law. The first part of the proposed control scheme is the vector of Coriolis and centripetal effects of the vehi-
is used to enhance the control activity with indulgence from a cle (including added mass effects). Dv (ν, η) η̇ is the vector
known desired acceleration vector and an estimated perturbed of damping effects of the vehicle (both linear and quadratic
term to compensate for the unknown effects and unmodeled damping terms). Gv (η) is the vector of restoring forces and
dynamics, while the second part acts as a feedback portion to moments acting on the vehicle. τv = [X Y Z K M N ]T is
enhance the overall closed-loop stability of the system. The the resultant input vector of thrusters, control plane forces, and
proposed scheme enables overcoming the difficulty due to pa- moments. ζ = [θ1 θ2 · · · θn ]T is the vector of joint variables
rameter uncertainties, external disturbances such as buoyancy, while θ1 , θ2 , ..., θn are the joint positions of the corresponding
reaction forces, payload variations, etc. The effectiveness of the underwater manipulator links, where n is the number of joints.
This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

LONDHE et al.: TASK SPACE CONTROL OF AN AUTONOMOUS UNDERWATER VEHICLE MANIPULATOR SYSTEM 3

Mm (ζ) ζ̈ is the vector of inertial forces and moments of the where


manipulator (including added mass effects). Cm (ζ, ζ̇)ζ̇ is the τdis = τedis − δMμ μ̈ − δCμ μ̇ − δDμ μ̇ − δGμ − δFμ
vector of Coriolis and centripetal effects of the manipulator (in-
cluding added mass effects). Dm (ζ, ζ̇)ζ̇ is the vector of damping δMμ μ̈ = M̂μ μ − Mμ μ̈
effects of the manipulator (both linear and quadratic damping δCμ μ̇ = Ĉμ μ̇ − Cμ μ̇
terms). Gm (ζ) is vector of the restoring forces and moments
for the manipulator. H T (ζ) ζ̈ and H(ζ, ζ̇)η̈ are the vectors of δDμ μ̇ = D̂μ μ̇ − Dμ μ̇
interaction (dynamic coupling) forces and moments between δGμ = Ĝμ − Gμ
the vehicle and the manipulator due to the manipulator and the
vehicle inertial effects, respectively. Fv (q, q̇) and Fm (q, q̇) are δFμ = F̂μ − Fμ . (12)
the vectors of the coupling dynamic effects (other than inertial
M̂μ , Ĉμ , D̂μ , Ĝμ , and F̂μ are the known values (approximated
effects) due to the manipulator inclusion and its motion on the
model parameter values) of the inertia matrix, the Coriolis and
vehicle, and the interaction effects due to the vehicle motion on
centripetal matrix, damping effects, restoring effects, and dy-
the manipulator, respectively. τm = [τm 1 τm 2 · · · τn m ]T is the
namic coupling effects (forces) of the AUVMS, respectively.
vector of manipulator control inputs while τm 1 , τm 2 , ..., τn m are
τdis is the lumped uncertainties vector consisting of internal
the joint torques of the corresponding manipulator links. τedis
(parametric uncertainties) (τidis ) and external (τedis ) distur-
is the vector of external disturbances due to underwater current
bances acting on the AUVMS. The following assumptions and
and/or a paylaod, sensor measurement noise, etc. The AUVMS
properties are considered to ensure the asymptotic convergence
consists of an n-DOF manipulator and a 6-DOF vehicle, which
of disturbance and trajectory tracking response in the overall
means the system has more DOF than the dimension of the task
closed-loop system.
space (kinematically redundant system), required to perform
Assumption 1: The single-input FLC (SIFLC), feedforward
the given task. The task coordinates are related with the system
controllers, and disturbance estimator gains, namely, Kc , Λ, and
states given as
Γ, are constant symmetric-positive-definite (SPD) matrices by
design, that is
μ = f (q) (7)
Kc = KcT > 0, Λ > 0, and Γ = ΓT > 0. (13)
T
where μ = [xt yt zt ] is the vector of task space coordinates
Assumption 2: The rate of change of the disturbance acting
and f (q) is the vector representing the forward kinematics.
on the AUVMS is negligible in comparison with the error dy-
Differentiating (7) with respect to time gives
namics of the uncertainty estimator, i.e., slowly varying distur-
bances (τ̇dis ≈ 0), and this assumption is not overly restrictive
μ̇ = J (q) q̇ (8)
and is commonly made in the AUVMS literature [22], [23],
[45]–[47].
where J (q) is the Jacobian matrix, mapping system (joint and
The following properties are observed for the AUVMS with
body fixed) velocities to task space velocities.
respect to the inertial (fixed-ground base) frame.
The dynamic equations of motion of the AUVMS in the task
Property 1: The inertia matrix of the proposed AUVMS with
space can be expressed as follows:
respect to the inertial frame is always positive definite [23], [44],
that is
Mμ μ̈ + Cμ μ̇ + Dμ μ̇ + Gμ + Fμ = τct + τedis (9)
M (q) > 0 ∀ q ∈ Rn . (14)
where Mμ is the inertia matrix, Cμ is the Coriolis and centripetal
Property 2: For the AUVMS, the following property exists
matrix, Dμ is the damping matrix, Gμ is the vector of restoring
with respect to the inertial reference (base) frame [44]:
effects, and Fμ is the vector of dynamic coupling of the AUVMS
in task space. τct is the vector of task space control inputs of the ξ T (Ṁ (q) − 2C(q, q̇))ξ = 0, ∀ ξ ∈ Rn , q ∈ Rn , q̇ ∈ Rn (15)
AUVMS. τedis is the vector of external disturbances in the task
space. for example, Ṁ (q) − 2C (q, q̇) is a skew-symmetric matrix.
Under parametric uncertainties, the AUVMS dynamic equa-
tion of motion in (9) can be stated as III. ROBUST SINGLE INPUT FUZZY LOGIC CONTROL SCHEME
A. Single-Input Fuzzy Logic Controller
(Mμ + δMμ )μ̈ + (Cμ + δCμ )μ̇ + (Dμ + δDμ )μ̇ + Gμ
In conventional FLC, with two controlled inputs, namely, the
+ δGμ + Fμ + δFμ = τct + τedis . (10) error (e) and the rate of change of error (ė), a 2-D Toeplitz
structure rule table can be constructed on a phase-plane (e, ė)
By gathering uncertainties along with external disturbances into [48] as shown in Table I, where e = μd − μ, and μd and μ are
a single term, which is known as lumped uncertainty τdis , then the desired and actual task space position vectors, respectively.
(10) takes the following form: The main feature of this rule base structure is that it has the same
output membership in a diagonal direction. In addition to that,
M̂μ μ̈ + Ĉμ μ̇ + D̂μ μ̇ + Ĝμ + F̂μ = τct + τdis (11) each point on the particular diagonal line has a magnitude that
This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

4 IEEE JOURNAL OF OCEANIC ENGINEERING

Fig. 1. Coordinate frame arrangement of an AUVMS.

TABLE I
FLC RULE TABLE WITH THE TOEPLITZ STRUCTURE

is proportional to the distance from its main diagonal line on


the normalized input space. This property also holds in the case
of the proportional–integral–derivative (PID)-type FLC which
uses the error, the sum of the error, and the change of the er- Fig. 2. Rule table with infinitesimal quantization levels.
ror as fuzzy input variables [48]. By considering the consistent
patterns of the output memberships in Table I, there is scope
to simplify the table considerably. Instead of using two-variable
input sets (e, ė), it is possible to obtain the corresponding output, It can be noted that the main diagonal line called the switching
using a single variable input only with the help of a signed dis- line can be represented as a straight line function and given as
tance method, which was first realized by Choi et al. [48]. Due
to the skew-symmetric property of an FLC rule table shown in s : ė + χe = 0 (16)
Table I, if the quantization level of the independent variables be-
comes infinitesimal, then the boundaries become straight lines,
as shown in Fig. 2. Also, note that the absolute magnitude of where χ is the slope of the main diagonal line LZ , as shown
the control input is proportional to the signed distance d from in Fig. 3. The distance between points A (e1 , e˙1 ) and B (e0 , e˙0 )
the main diagonal line. To find distance d, let B (e0 , e˙0 ) be an can be calculated as
intersection point of the main diagonal line and the line per- 
pendicular to it from a known operating point A (e1 , e˙1 ), as | e˙1 + χe1 |
d1 = (e0 − e1 )2 + (ė0 − e˙1 )2 = . (17)
illustrated in Fig. 3. 1 + χ2
This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

LONDHE et al.: TASK SPACE CONTROL OF AN AUTONOMOUS UNDERWATER VEHICLE MANIPULATOR SYSTEM 5

Fig. 4. CFLC input (e and ė) and output (U C F L C ) membership functions.

Fig. 3. Derivation of a signed distance variable (d).

TABLE II
REDUCED RULE TABLE OBTAINED USING THE SIGNED DISTANCE METHOD
Fig. 5. SIFLC with PWL control surface.
ds LN L LN M LN S LZ LP S LP M LP L

u PL PM PS Z NS NM NL
and SIFLC, the “equivalent” triangular membership functions
can be used for input and output variables, as shown in Fig. 4.
Generalization of (17) can be rewritten as follows: This is to be compared with the SIFLC rule table shown in
Table I. From Table II, it can be seen that the control rules are
| ė + χe | significantly reduced which is the main feature of the SIFLC.
d= . (18)
1 + χ2 For a two-input CFLC with a fuzzification level p, the number of
rules to be inferred is p2 whereas an equivalent SIFLC requires
For any arbitrary point A (e, ė), the signed distance ds is defined
only p rules. The simplicity of the control structure of SIFLC
as
allows for very rapid computation because the fuzzification, rule
| ė + χe | ė + χe s inference, and defuzzification processes are no longer required.
ds = sgn (s) = = (19)
1+χ 2 1+χ 2 1 + χ2 The structure of an SIFLC based on the signed distance method
and its corresponding rule table can be translated as the block
where

diagram shown in Fig. 5. With the 1-D array of the rule table (see
1, for s > 0 Table II), the control surface of the SIFLC can be approximated
sgn (s) = (20)
−1, for s < 0. as a 1-D piecewise linear (PWL) function. This can be achieved
As the sign of the control input is negative for s > 0 and positive with the following conditions [49]:
for s < 0 and its magnitude is proportional to the distance from 1) the input membership function (MF) is triangular in shape;
the line s = 0, hence 2) the output membership function is a singleton;
3) the fuzzification and defuzzification processes use the cen-
USIFLC ∝ −ds . (21) ter of gravity (CoG) method.
Hence, the control action can be exclusively determined by ds Fig. 5 shows a linear PWL control surface obtained with
only. Therefore, it is appropriate to call it the SIFLC. With asymmetrical MFs, denoted as ψc , which has a constant slope
the help of this derived distance ds , the 2-D rule table can be throughout the universe of discourse (UoD). Such a surface
converted into the 1-D rule table, as depicted in Table II, where is achieved when the triangular peaks of the input MF and
LP L , LP M , LP S , LZ , LN S , LN M , and LN L are the diagonal the spacing in-between the singletons are equal. Under these
lines of Table I. These diagonal lines correspond to the new conditions, the control output generated by Table II is linear
input, while NL, NM, NS, Z, PS, PM, and PL represent the [40], [49] and given as
output of the corresponding diagonal lines, as shown in Table II.
USIFLC = −Gd ds . (22)
Therefore, the input to the fuzzy controller is distance ds , while
its output is the change in the control output USIFLC . The final Substituting the value of ds from (19) in (22), we have
output is obtained by multiplying USIFLC with an output scaling
factor, say Gu . To compare the control structures between CFLC USIFLC = − [K1 e + K2 ė] (23)
This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

6 IEEE JOURNAL OF OCEANIC ENGINEERING

Fig. 6. Concept of disturbance estimator.

where Gd > 0 is a constant obtained by linear FLC. Here, application. Roughly speaking, the idea behind the disturbance
ds is bounded with ds1 ≤ ds ≤ dsN , where ds1 and dsN are, estimator is to lump all the internal (parametric uncertainties)
respectively, the minimal and maximum values of a signed and external unknown torques/forces (ocean current) acting on
distance. N is the number
of fuzzy sets for a signed
dis- the AUVMS into a single disturbance term called lumped uncer-
tance. K1 = (Gd χ) /( 1 + χ2 ) and K2 = (Gd ) /( 1 + χ2 ) tainty and then estimate this unknown term using the disturbance
are positive gain constants. The main advantage of using a PWL estimation technique.
function is that the control surface can be constructed using a A disturbance estimator is proposed as
simple lookup table that results in a much faster computation
τ̂d is = Υ + Γμ̇ (27)
time. The nonlinear control surface can be constructed with the  
peak locations of the input and output MFs are arranged in an Υ̇ = −ΓM̂μ−1 τc t + Υ + Γμ̇ − Ĉμ μ̇ + D̂μ μ̇ + Ĝμ + F̂μ
unequal space. This results in a PWL with multiple linear re-
gions with linear lines of different slopes, which requires more +M̂μ−1 U S IF L C (28)
MFs [50].
where τ̂dis is the estimated vector of disturbances/uncertainties.
To ensure the asymptotic stability of the overall closed-loop
Υ is an auxiliary vector to estimate the unknown disturbance
control system, the control law in (24) is augmented with edt
vector.
such that
  Define the estimation error
U SIFLC = − ė + 2Λe + Λ 2
edt (24) τ̃dis = τ̇dis − τ̂˙ dis . (29)

where Λ = (χ/2) and χ = (K1 /K2 ). U SIFLC is the augmented Now, τ̂dis is to be updated in a such way that the estimation error
SIFLC. This arrangement guarantees the asymptotical stability τ̃dis goes to zero asymptotically.
of the overall closed-loop system. Differentiating (27), we have
Remark 1: The control structure of SIFLC is shown in Fig. 5, τ̂˙ dis = Υ̇ + Γμ̈. (30)
where ds and u are the input and the output of the signed distance
FLC, respectively. Let d˜s and ŨSIFLC be the fuzzy variables Substituting the value of Υ̇ from (28) and known acceleration μ̈
of ds and u, respectively. The UoDs of d˜s and ŨSIFLC are from (11), we get
normalized from −1 to 1. Thus, the range of nonfuzzy variables
τ̂˙ dis = ΓM̂μ−1 τ̃dis + M̂μ−1 U SIFLC . (31)
ds and u must be scaled to fit the UoD of fuzzified variable d˜s
and ŨSIFLC with scaling factors k1 and k2 , respectively, as Some adjustment in (31) gives
d˜s = k1 ds (t) (25) τ̇dis − τ̂˙ dis = −ΓM̂μ−1 τ̃dis − M̂μ−1 U SIFLC + τ̇dis (32)
ŨSIFLC (t) = k2 u. (26) τ̃˙ dis = −ΓM̂μ−1 τ̃dis − M̂μ−1 U SIFLC + τ̇dis . (33)

B. Disturbance Estimator Since, in general, there is no prior information about the deriva-
tive of the disturbance, it is reasonable to suppose that τ̇dis = 0,
A basic idea in the design of an estimator is to modify the which implies that the disturbance varies slowly relative to the
estimation by the difference between the estimated output and estimator dynamics. Hence
the actual output. Fig. 6 shows the conceptual diagram of a typ-
ical disturbance estimator that is used in an underwater robotic τ̃˙ dis = −ΓM̂μ−1 τ̃dis − M̂μ−1 U SIFLC . (34)
This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

LONDHE et al.: TASK SPACE CONTROL OF AN AUTONOMOUS UNDERWATER VEHICLE MANIPULATOR SYSTEM 7

Fig. 7. Proposed robust nonlinear controller block diagram.

Hence, (34) holds assumption 2 and the convergence of estima- Kc and Λ are the gain matrices of SIFLC and the feedfor-
tion errors has been proved in the overall stability proof of the ward controller, respectively. They are chosen as symmetric
closed-loop system explained in Section IV. Equation (34) indi- positive-definite metrics. μ̈r is the virtual reference (desired)
cates that the rates of convergence of estimating tracking errors task space acceleration vector. μd , μ̇d , and μ̈d are the desired
are decided by the value of Γ and may get affected due to a time- task space position, velocity, and acceleration vectors, respec-
varying nature of M̂μ−1 , which consists of hydrodynamic coef- tively. Γ is a positive diagonal matrix. U SIFLC is the centralized
ficients that varies with change in environmental disturbances augmented SIFLC input vector. τ̂dis is the estimated vector of
in practical situations. However, inclusion of U SIFLC in the Υ disturbances/uncertainties. Υ is an auxiliary vector to estimate
auxiliary vector will improve the performance of the uncertainty the unknown disturbance vector. The proposed controller  con-

disturbance estimation. Therefore, to improve adaptability to- sists of four integral components, namely, the SIFLC U SIFLC ,
ward large parameter variations during the operation, a robust the feedforward controller (M̂μ̈ (μ) μ̈r ), the model-based (feed-
SIFLC has been designed and explained in Section III-C. back linearization) controller (known or estimated matrices of
the model), and the uncertainty estimator (τ̂dis ). The corre-
C. Robust-Single Input Fuzzy Logic Controller sponding block diagram of the proposed control scheme can be
Now, let us consider a complete control law design including visualized in Fig. 7. The model-based controller in the proposed
the augmented SIFLC law along with the feedforward control control scheme helps in faster learning of the plant dynamics
term and the uncertainty estimator, which gives robust control by an SIFLC. However, in practice, it is very difficult to ob-
performance against uncertainties in the model parameters and tain or estimate the exact values of the model matrices of the
environmental disturbances. With this, a proposed robust control AUVMS; still little knowledge about these model matrices is
scheme can be given as sufficient for obtaining the desirable control performance from
the SIFLC. This is due to plant dynamics being modeled using
  seven fuzzy control rules, as given in Table II. The same is used
τct = M̂μ μ̈r − Kc U SIFLC − τ̂dis + Ĉμ μ̇ + D̂μ μ̇ + Ĝμ + F̂μ
for developing a control structure of an SIFLC (see Fig. 5). The
(35)
block diagram flow starts from the user inputs given in terms of
where
the desired task space variables based on the manipulation tasks
(missions) and time, namely, μd and t. Based upon the above
μ̈r = μ̈d + 2Λė + Λ2 e
user inputs, the trajectory planner provides the desired task space
τ̂dis = Υ + Γμ̇ coordinates, namely, time trajectories of the task space position,
  velocity, and acceleration vectors [μd (t), μ̇d (t), and μ̈d (t)],
Υ̇ = −ΓM̂μ−1 τct + Υ + Γμ̇ − Ĉμ μ̇ + D̂μ μ̇ + Ĝμ + F̂μ
respectively.
+M̂μ−1 U SIFLC The proposed AUVMS has the measurement systems which
give joint space positions and velocities of the manipulator
ė = μ̇d − μ̇, e = μd − μ. (36)
This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

8 IEEE JOURNAL OF OCEANIC ENGINEERING

system, which can be converted into the corresponding task of (11) can be rewritten as follows:
space positions and velocities by forward kinematics. The sensor 
dynamics are incorporated into the AUVMS dynamics. There- μ̈ = M̂μ−1 τct + τdis − Ĉμ μ̇ − D̂μ μ̇ − Ĝμ − F̂μ . (40)
fore, from the actual and desired task space states, the tracking
Now, differentiating the value of U SIFLC in (24), we have
errors, i.e., the position error, the velocity error, and the inte-
 
gral position error, are calculated and fed to the robust SIFLC U˙ SIFLC = − ë + 2Λė + Λ2 e
block. Based upon the magnitude of the tracking errors, the  
proposed robust control scheme generates a control signal as = − μ̈d − μ̈ + 2Λė + Λ2 e . (41)
stated in (35) and actuates the actuator in such a way so as to Substituting the value of μ̈ from (40) and the control law from
have the minimum tracking errors. The proposed robust control (35) in (41), we get
scheme enhances the closed-loop stability of the system and
improves the transient performance of the system. The feed- U˙ SIFLC = −Kc U SIFLC + M̂μ−1 τ̃dis . (42)
forward controller gives the desired input values theoretically T
based on the desired acceleration and inertia matrix. However, Multiplying (42) by U SIFLC , we have
in the actual case, these desired input values may vary con-
U SIFLC U˙ SIFLC = −U SIFLC Kc U SIFLC + U SIFLC M̂μ−1 τ̃dis .
T T T
siderably due to the system dynamics. Therefore, the virtual
(43)
reference acceleration vector (μ̈r ) and the theoretical input val-
Putting the value of (43) and the value of (34) in (39), we have
ues are calculated based on the revised controller scheme in
T
the feedforward controller. The model-based controller uses an V̇ = −U SIFLC Kc U SIFLC − τ̃dis
T
ΓM̂μ−1 τ̃dis . (44)
inverse dynamic model output with some known inaccuracy for
approximating joint torques/forces of the manipulator system. Hence
A model-based control term linearizes the feedback scheme, V >0 and V̇ < 0, ∀ U SIFLC = 0, ∀ τ̃dis = 0
making the system approximately linear and decoupled, which
results in a straightforward and simple structure of the system. therefore, also
The uncertainty estimator has an efficient estimation approach V =0 and only if U SIFLC = 0, τ̃dis = 0.
V̇ = 0,
to estimate the perturbation from the dynamics of the SIFLC (45)
and the known inverse dynamic model. This helps to compen- Based on the Lyapunov direct method and the Barbalat lemma
sate for the model uncertainties, unknown external disturbances, [45], [51], the closed-loop system is globally asymptotically
and time-varying parameters (this estimator restricted with the stable, i.e., the system task space position and velocity tracking
bounded and slowly varying disturbances). errors and the disturbance/uncertainty estimating tracking errors
are converging to zero asymptotically [45], [51], i.e.,
IV. STABILITY ANALYSIS OF A ROBUST SIFLC lim U SIFLC (t) = 0, lim τ̃dis (t) = 0
t→∞ t→∞
Lemma 1: The closed-loop system is globally asymptoti-
cally stable if and only if the following conditions are satisfied: lim e (t) = 0, lim ė (t) = 0. (46)
t→∞ t→∞

Therefore, the AUVMS follows the given desired task space


V > 0 and V̇ < 0, ∀ U SIFLC = 0, ∀ τ̃dis = 0
trajectory with minimal errors. 
V = 0 and V̇ = 0, only if U SIFLC = 0, τ̃dis = 0. (37)
V. PERFORMANCE ANALYSIS
Proof: Consider the system in which the governing equations
are given by (9) and let us define a positive–definite Lyapunov A. Description of the System
candidate function as The performance of the system is gauged by performing ex-
1 T 1 T tensive numerical simulations involving positioning and trajec-
V = U U SIFLC + τ̃dis τ̃dis . (38) tory tracking of the end effector. For ease of understanding and
2 SIFLC 2
clarity, the AUVMS horizontal plane motion is considered. In
Choosing a control input of the form as given in (35) along with that, a 2-DOF manipulator system (MS) equipped with a 3-DOF
an uncertainty estimator function as given in (36) will lead to horizontal plane AUV model is considered for the numerical
the asymptotic convergence to zero of the AUVMS tracking task simulation purpose. With this consideration, vector q in (1) is
space position, velocity, and uncertainty in the error estimations, taken here as q = [η ζ] = [x y ψ θ1 θ2 ] with η = [x y ψ]
i.e., the AUVMS will follow the given desired trajectory. and ζ = [θ1 θ2 ]. μ = [xt yt ] is the task space position vector
Differentiating with respect to time along its state trajectories considered in the proposed system. The AUVMS considered for
will give the analysis consists of a planar serial manipulator which has
two active rotary joints. The shape of the vehicle is rectangular
V̇ = U SIFLC U˙ SIFLC + τ̃dis
T T ˙
τ̃ dis (39) prism, and it has four thrusters which help in maneuvering all
motions of the vehicle. The shapes of the manipulator links have
where U˙ SIFLC is the time derivative of the output of the SIFLC been assumed to be cylindrical, though the performance of the
controller. The AUVMS equations of motion given in the form control scheme is least affected by this fact as these variations
This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

LONDHE et al.: TASK SPACE CONTROL OF AN AUTONOMOUS UNDERWATER VEHICLE MANIPULATOR SYSTEM 9

Fig. 8. Schematic diagram of the planar AUVMS.

would be accommodated with ease. The inertial parameters of ing the starting point with respect to time. This complex task
the vehicle and the manipulator links have been found from space trajectory can be represented in the polynomial form as
the solid model of the system and hence emulate the actual pa- given by (47) and (48), shown at the bottom of the page, where
rameters. The hydrodynamic parameters are obtained by using x (t) and y (t) are in meters. The initial velocity vectors of the
empirical relations [43]. The schematic diagram of the AUVMS links were set to be zero, and random values were assigned to
considered for the numerical study is depicted in Fig. 8. The the estimated system disturbance vectors, but the intended and
vehicle physical and hydrodynamic parameters are given in the actual orientation of the end effector is assumed to be almost
Appendix. equal.

B. Description of the Task C. Results and Discussions


The task space vector consists of the end effector positions, The results of the predefined complex task space position
i.e., [xt = xend yt = yend ]. The test case chosen here demands tracking control of the planar AUVMS are presented, and these
the AUVMS start from a user-defined initial position and return can be used as a reference to the performance analysis of the
to the same position after traversing a predefined trajectory. A proposed control scheme. The underwater effects that are bound
complex task space trajectory considered here is a loop com- to act on the AUVMS, such as the underwater currents, damping
prising a vertical rise and a circular path defined by two different effects, etc. have been included in the disturbance vector which
arcs, namely, concave up and concave down preceded by a ramp has been exclusively appended to the dynamics of the system
and a vertical drop and ended with a slow rising ramp, thus reach- to simulate the underwater conditions that the AUVMS will be



⎪ 2.5,   0s ≤ t ≤ 30s

⎪ 2.5 + 0.8 cos 900 + 93 (t − 30) , 30s ≤ t ≤ 60s

⎪  
⎨ 1.7 + 0.8 cos − 9 (t − 60) − 0.8, 60s ≤ t ≤ 90s
x (t) = 3 (47)
2 −4 3

⎪ 0.9 − 0.0080(t − 90) + 1.778 × 10 (t − 90) , 90s ≤ t ≤ 120s



⎪ −1.5, 120s ≤ t ≤ 150s

−1.5 + 0.0133(t − 150)2 − 2.96 × 10−4 (t − 150)3 , 150s ≤ t ≤ 180s

⎪ 0.5 + 0.0067t2 − 1.4815 × 10−4 3
t , 0s ≤ t ≤ 30s



⎪ 2.5 + 0.8 sin 90 0
+ 9
(t − 30) − 0.8, 30s ≤ t ≤ 60s
⎨ 1.7 + 0.8 sin − 9 (t −
⎪ 3
60)

, 60s ≤ t ≤ 90s
3
y (t) = 2 −4 3 (48)

⎪ 0.9 + 0.0070(t − 90) − 1.555 × 10 (t − 90) , 90s ≤ t ≤ 120s

⎪ 2 −4 3

⎪ 3 − 0.0117(t − 120) + 2.592 × 10 (t − 120) , 120s ≤ t ≤ 150s

−0.5 + 0.0033(t − 150)2 − 7.4074 × 10−4 (t − 150)3 , 150s ≤ t ≤ 180s
This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

10 IEEE JOURNAL OF OCEANIC ENGINEERING

TABLE III bances are considered as a vector (for the simulations [10 N, 5
REDUCED SIFLC RULE TABLE
N, 0.5 Nm, 0.2 Nm, 0.1 Nm]T ) and the unknown internal dis-
turbances like the sensory noises in the position and orientation
ds 2 -0.7 -0.466 -0.23 0 0.233 0.466 0.7
measurement are introduced as Gaussian noises of 0.01-m mean
-u S I F L C 1 -0.99 -0.66 -0.33 0 0.33 0.66 0.99 and 0.01 standard deviation in the position 0.2◦ standard devi-
ation in the orientation measurement. The system uncertainties
are considered as 20% (i.e., the system model consists of only
subjected to while performing any of the positioning or tracking 80% of actual values). All the results which have been obtained
operations. Considerations such as disturbances, parameter un- are depicted in Figs. 9–14, which illustrate the tracking perfor-
certainties, and sensor noises have all been incorporated in the mance of the controllers while tracking a trajectory in the task
numerical model ensuring the usage of the proposed controller space. Here, all the controllers are tuned to the best parameter
in an actual prototype without compromising either performance settings as mentioned above for performing the space control
or effectiveness. task. The tracking performance of the proposed scheme is quite
The control torque force, i.e., τc , generated by the proposed acceptable and adequate for underwater applications.
control law as given in (35), used to manipulate the joint vari- The control performance of all the controllers is measured in
ables of the AUVMS, namely, θ1 and θ2 , in such a way that terms of the norm of the vector. The L2 -norm (Euclidean) is
the proposed planar AUVMS tracked the desired task space tra- considered for the analysis purpose and is defined as (consider
jectory with minimum errors. it for task space errors)
  In the proposed control law, an 
output of SIFLC U SIFLC is generated with linear PWL con-  2
 √
trol surface and signed distance (ds ) as the input, as depicted in || e ||=  e2i = eT e. (52)
Fig. 5, using (22). The PWL surface is designed using 1-D array i=1
of the rule table, as depicted in Table II. Both inputs (e, ė) and
outputs U SIFLC are normalized to [−1 1]. Hence, the value of χ The time histories of the norm of the control inputs vector and
becomes unity. Using the signed distance method and input MFs the task space position tracking errors under uncertain working
as shown in Fig. 4 and singleton MFs for output, the Toeplitz conditions are depicted in Figs. 9 and 10, respectively. From
rule in Table I is reduced to the 1-D rule table as shown in Ta- these results, it can be observed that the norm given by the pro-
ble III. The designed control surface can be generated using a posed controller immediately falls down to zero and provides
simple lookup table, which leads to simplification of the control smooth control efforts as compared to the CFLC, SMC, and PID
structure of SIFLC. Also, the control performance by CFLC controllers. The PID controller shows poor control performance
is similar to SIFLC. However, CFLC requires a trial-and-error as compared to the CFLC and the SMC. The CFLC requires
approach to fine tune, because it has more tuning parameters fine tuning of controller parameters to have better control per-
than when compared to SIFLC. For the sake of clarity, a com- formance, which is a time-consuming process. Fig. 11 shows
parison of the proposed controller is also made with CFLC. the variations in the AUV motion in the horizontal plane, i.e.,
As mentioned above, a comparison of the proposed scheme is the position along x- and y-axes and their corresponding an-
made with the other controllers like CFLC and linear PID and gular velocities, yaw angle (ψ), and yaw angle rate (r) during
conventional sliding mode control (SMC) given as follows: the task space trajectory tracking control. It can be observed
that the proposed controller provides smooth change in forward
τct = Kc
UCFLC : CFLC (49) velocity and sway velocity while performing trajectory tracking
in the task space. Time trajectories of the joint-space angles
de
τct = KP e + KI edt + KD : PID (50) during the uncertain condition can be seen in Fig. 12. During
dt
the predefined complex task space trajectory tracking control,
τct = M̂μ (μ̈d + Ks1 σ + Ks2 sign (σ)) variations in the task space positions and their corresponding
errors are depicted in Fig. 13. It can be concluded that the pro-
+ Ĉμ μ̇ + D̂μ μ̇ + Ĝμ + F̂μ : SMC (51)
posed control scheme converges the task space position errors
where Kc
is the output scaling gain of the CFLC controller and within a short duration of time which is quite acceptable when
UCFLC is the control torque given by CFLC. KP , KI , and KD compared to other controllers since adequate oscillation in the
are the proportional, integral, and derivative gains of the linear task space errors is produced by the PID controller during the
PID controller, respectively. Ks1 and Ks2 are the positive di- task, while CFLC and SMC controllers provide steady state er-
agonal gain matrices of the SMC controller. σ = ė + λe is the rors in the range of 10−3 . Fig. 14 shows the task space motion
sliding surface, where λ > 0 is a positive constant. The nomi- trajectory tracking of the AUVMS in the (x-y)-plane given by
nal controller settings are considered for the above controllers all the controllers. It can be seen in the zoomed scale version
given by (35), (49), and (50), and their values are as follows: for of Fig. 14 that the PID controller shows adequate errors in the
CFLC, Kc = 5I2 ; for PID, KP = 2.5, KI = 3.5, and KD = 1; task space trajectory tracking when compared to the CFLC con-
for SMC, Ks1 = 2 and Ks2 = 3; and the for proposed RSIFLC, troller followed by SMC, while the proposed one shows the
Kc = 5I2 , Λ = 5I2 , and Γ = 2I2 . For the uncertain condition, tight tracking performance. It can be seen that the control per-
the underwater current is considered with a velocity vector of formance given by CFLC and RSIFLC is quite similar. How-
(0.2, 0.1) m/s in x- and y-axes. The unknown external distur- ever, CFLC requires more parameters to tune along with the
This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

LONDHE et al.: TASK SPACE CONTROL OF AN AUTONOMOUS UNDERWATER VEHICLE MANIPULATOR SYSTEM 11

Fig. 9. Time trajectories of the norm of the vector of control inputs (an uncertain working condition).

Fig. 10. Time trajectories of the norm of the vector of task space tracking errors (an uncertain working condition).

trail-and-error approach for designing large fuzzy control rules To have an utmost check on the robustness of the proposed
to obtain a better control performance. In contrast to this, the control scheme, variations in the controller gain matrices (i.e.,
proposed RSIFLC has a simple control structure and is easy to Kc , Λ, and Γ) as well as inertia matrix (Mμ̈ (μ)) of the system
tune and design. were performed and the control performance of the controller
A quantitative analysis of the tracking performance of an was observed in terms of the norm of the task space errors. The
AUVMS given by all the controllers in terms of the maximum change in the norm of the task space errors during the variation of
error (ME), the root mean square error (RMS), and the L2 -norm the Kc gain of the proposed controller can be seen in Fig. 15(a).
(Euclidean) in the task-space trajectory of the end effector was It is observed from these results that the performance of the
performed and can be found in Table IV. From this table, it is proposed controller shows similar kind of the control perfor-
confirmed that the proposed control scheme gives minimum mance for values of Kc changing from 0.05 to 80. This suggests
tracking errors in x- and y-directions as compared to other that the proposed controller has ability to adapt the variations
controllers. Among all the controllers, the PID controller gives in the controller parameters. Similarly, variations in the second
the maximum error in the task space position tracking control parameter Γ are carried out and the control performance is ob-
of the value 0.0385 m. served and depicted in Fig. 15(b). Here, with the variation in Γ
This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

12 IEEE JOURNAL OF OCEANIC ENGINEERING

Fig. 11. Time trajectories of AUVMS motion in the horizontal plane during the underwater manipulator task position control.

Fig. 12. Time trajectories of the joint space positions during the task space trajectory control (uncertain condition).

from 0.05 to 50, the control performance given by the proposed proposed controller gives satisfactory control performance for
controller remains almost the same. Also, changes in the norms the Λ changes from 2 to 10. From the above discussion, one can
of the task space errors with respect to variations in the third conclude that the proposed controller provides a wide range of
gain parameter Λ can be observed in Fig. 16(a). In this case, choices for the section of controller parameters and omits the
if the value of Λ increases beyond 10, then the magnitude of optimal selection. This indicates that there are no strong rela-
the norm of the task space errors also increases, which results tions between these control gains. Therefore, individual tuning
in the degradation of the control performance. However, the of controllers can be applied.
This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

LONDHE et al.: TASK SPACE CONTROL OF AN AUTONOMOUS UNDERWATER VEHICLE MANIPULATOR SYSTEM 13

Fig. 13. Time trajectories of the errors in task space tracking positions (uncertain condition).

Fig. 14. Task space motion trajectories tracking of the AUVMS in the (x-y)-plane.

TABLE IV
QUANTITATIVE ANALYSIS OF THE TASK SPACE TRACKING CONTROL ALONG x- AND y-DIRECTIONS

xe ye
ME RMS L 2 norm ME RMS L 2 norm

CFLC 0.0042 0.0012 0.0176 0.0023 6.7245 ×10 −4 0.0095


PID 0.0385 0.0029 0.0408 0.0148 0.0023 0.0324
SMC 0.0154 0.0028 0.0395 0.0033 7.5061 ×10 −4 0.0106
RSIFLC 3.9181 ×10 −4 8.7681 ×10 −4 0.0124 2.003 ×10 −4 4.7376 ×10 −4 0.0067
This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

14 IEEE JOURNAL OF OCEANIC ENGINEERING

Fig. 15. Change in the norm of the task space error during variations of (a) the K c gain and (b) the Γ gain of the proposed controller.

Fig. 16. Change in the norm of the task space error during variations of (a) the Λ gain of the proposed controller and (b) the percentage change in the mass
matrix (M ) of the AUVMS.

To further investigate the robustness given by the proposed the conventional FLC controller. From the obtained results, the
control scheme, uncertainties in the model parameters such as an strength of the proposed control scheme can be listed as follows:
inertia matrix from 20% to 70% are performed and the control 1) the proposed controller increases the overall stability of
performance is observed. This can be seen in Fig. 16(b). From the closed-loop system;
the results it is observed that the proposed controller provides 2) conventional controllers, namely, the PID controller, pro-
good performance even for the 70% uncertainty in the inertia vide limited stability and mainly depended upon their gain
matrix of the AUVMS. values;
Consequently, it is apparent from the above discussion that 3) as the proposed controller eliminates the number of tuning
the proposed controller provides robust and efficient control parameters whereas tuning of the conventional fuzzy con-
performance while tracing a complex predefined path in the troller requires a trial-and-error procedure which is time
task space. The proposed control scheme can be easily ex- consuming and plant specific;
tended to a 3-D task space position tracking problem of an 4) poor knowledge of the system model matrices will be
AUVMS. sufficient to design the controller, whereas SMC requires
an exact mathematical model of an AUVMS to have better
control performance;
VI. CONCLUSION 5) the proposed control scheme provides great immunity
to the external disturbances and parameter uncertainties;
A robust single-input fuzzy logic control scheme was pro- hence, it is known as the robust controller;
posed and applied in the task space position trajectory tracking 6) the proposed controller has a simple control structure and
control problem of an AUVMS for underwater manipulation design method; hence, it can be used for real-time imple-
task applications. The efficacy of the controller was demon- mentation with a low-cost microprocessor.
strated with the assistance of numerical simulations on a planar This guarantees the performance of the control scheme for
type of AUVMS in the horizontal plane motion. To evaluate the usage in position and trajectory tracking applications, which
control performance, the comparison of the proposed controller is the sole motive behind developing this underwater robotic
was made with other conventional controllers such as PID and system.
This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

LONDHE et al.: TASK SPACE CONTROL OF AN AUTONOMOUS UNDERWATER VEHICLE MANIPULATOR SYSTEM 15

APPENDIX
TABLE V
VEHICLE PARAMETERS

REFERENCES [15] S. Chiaverini, “Singularity-robust task-priority redundancy resolution for


real-time kinematic control of robot manipulators,” IEEE Trans. Robot.
[1] B. Lynch and A. Ellery, “Efficient control of an AUV-manipulator system: Autom., vol. 13, no. 3, pp. 398–410, Jun. 1997.
An application for the exploration of Europa,” IEEE J. Ocean. Eng., vol. [16] T. K. Podder and N. Sarkar, “A unified dynamics-based motion plan-
39, no. 3, pp. 552–570, Jul. 2014. ning algorithm for autonomous underwater vehicle-manipulator systems
[2] N. Sarkar and T. Podder, “Coordinated motion planning and control of (UVMS),” Robotica, vol. 22, pp. 117–128, 1 2004.
autonomous underwater vehicle-manipulator systems subject to drag op- [17] C. H. F. dos Santos, R. Guenther, D. Martins, and E. R. de Pieri, “Virtual
timization,” IEEE J. Ocean. Eng., vol. 26, no. 2, pp. 228–239, Apr. 2001. kinematic chains to solve the underwater vehicle-manipulator systems
[3] S. Mohan and J. Kim, “Indirect adaptive control of an autonomous un- redundancy,” J. Brazilian Soc. Mech. Sci. Eng., vol. 28, pp. 354–361,
derwater vehicle-manipulator system for underwater manipulation tasks,” 2006.
Ocean Eng., vol. 54, pp. 233–243, 2012. [18] C. H. F. dos Santos, G. Bittencourt, and R. Guenther, “Motion coordi-
[4] H. Mahesh, J. Yuh, and R. Lakshmi, “A coordinated control of an under- nation for underwater vehicle-manipulator systems using a fuzzy hybrid
water vehicle and robotic manipulator,” J. Robot. Syst., vol. 8, no. 3, pp. strategy,” in Proc. IEEE/RSJ Int. Conf. Intell. Robots Syst., Oct. 2006, pp.
339–370, 1991. 3018–3023.
[5] S. Mohan and J. Kim, “Coordinated motion control in task space of an [19] J. Han and W. K. Chung, “Redundancy resolution for underwater vehicle-
autonomous underwater vehicle-manipulator system,” Ocean Eng., vol. manipulator systems with minimizing restoring moments,” in Proc.
104, pp. 155–167, 2015. IEEE/RSJ Int. Conf. Intell. Robots Syst., Oct. 2007, pp. 3522–3527.
[6] S. Mohan and J. Kim, “Robust adaptive tracking control of autonomous [20] C. Klein and C.-H. Huang, “Review of pseudoinverse control for use with
underwater vehicle-manipulator systems,” J. Dyn. Syst. Meas. Control, kinematically redundant manipulators,” IEEE Trans. Syst. Man Cybern.,
vol. 136, no. 5, pp. 054502–054510, 2014. vol. SMC-13, no. 2, pp. 245–250, Mar. 1983.
[7] G. Marani, S. K. Choi, and J. Yuh, “Underwater autonomous manipulation [21] Z. H. Ismail, Task-space dynamic control of underwater control, Ph.D.
for intervention missions AUVs,” Ocean Eng., vol. 36, no. 1, pp. 15–23, dissertation, Heriot-Watt Univ., Dept. Electr. Electron. Comput. Eng., Ed-
2009. inburgh, U.K., 2011.
[8] M. Santhakumar and J. Kim, “Modelling, simulation and model reference [22] S. Mohan and J. Kim, “Coordinated motion control in task space of an
adaptive control of autonomous underwater vehicle-manipulator systems,” autonomous underwater vehicle-manipulator system,” Ocean Eng., vol.
in Proc. 11th Int. Conf. Control Autom. Syst., Oct. 2011, pp. 643–648. 104, pp. 155–167, 2015.
[9] H. Shim, B.-H. Jun, P.-M. Lee, H. Baek, and J. Lee, “Workspace control [23] G. Antonelli, Underwater Robots Motion and Force Control of Vehicle-
system of underwater tele-operated manipulators on an ROV,” Ocean Manipulator Systems. Springer Tracts in Advanced Robotics, vol., Berlin,
Eng., vol. 37, no. 11-12, pp. 1036–1047, 2010. Germany: Springer-Verlag, 2014, ch. 3.
[10] L. Paull, S. Saeedi, M. Seto, and H. Li, “AUV navigation and localization: [24] P. J. From, V. Duindam, K. Y. Pettersen, J. T. Gravdahl, and S. Sas-
A review,” IEEE J. Ocean. Eng., vol. 39, no. 1, pp. 131–149, Jan. 2014. try, “Singularity-free dynamic equations of vehicle-manipulator systems,”
[11] J. Han and W. K. Chung, “Active use of restoring moments for motion Simul. Model. Practice Theory, vol. 18, no. 6, pp. 712–731, 2010.
control of an underwater vehicle-manipulator system,” IEEE J. Ocean. [25] B. Xu, S. R. Pandian, N. Sakagami, and F. Petry, “Neuro-fuzzy control of
Eng., vol. 39, no. 1, pp. 100–109, Jan. 2014. underwater vehicle-manipulator systems,” J. Franklin Inst., vol. 349, no.
[12] S. Soylu, B. J. Buckham, and R. P. Podhorodeski, “Redundancy resolution 3, pp. 1125–1138, 2012.
for underwater mobile manipulators,” Ocean Eng., vol. 37, no. 23, pp. [26] B. Xu, S. Abe, N. Sakagami, and S. Pandian, “Robust nonlinear controller
325–343, 2010. for underwater vehicle-manipulator systems,” in Proc. IEEE/ASME Int.
[13] G. Antonelli and S. Chiaverini, “Fuzzy redundancy resolution and motion Conf. Adv. Intell. Mechatron., Jul. 2005, pp. 711–716.
coordination for underwater vehicle-manipulator systems,” IEEE Trans. [27] S. Soylu, B. Buckham, and R. Podhorodeski, “Development of a coor-
Fuzzy Syst., vol. 11, no. 1, pp. 109–120, Feb. 2003. dinated controller for underwater vehicle-manipulator systems,” in Proc.
[14] G. Antonelli and S. Chiaverini, “Task-priority redundancy resolution for OCEANS Conf., Sep. 2008, DOI: 10.1109/OCEANS.2008.5151915.
underwater vehicle-manipulator systems,” in Proc. IEEE Int. Conf. Robot. [28] S. Li, X. Wang, and L. Zhang, “Finite-time output feedback tracking
Autom., May 1998, vol. 1, pp. 768–773. control for autonomous underwater vehicles,” IEEE J. Ocean. Eng., vol.
40, no. 3, pp. 727–751, Jul. 2015.
This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

16 IEEE JOURNAL OF OCEANIC ENGINEERING

[29] J. Han and W. K. Chung, “Coordinated motion control of underwater M. Santhakumar (M’15) graduated in mechanical
vehicle-manipulator system with minimizing restoring moments,” in Proc. engineering from the Government College of Engi-
IEEE/RSJ Int. Conf. Intell. Robots Syst., Sep. 2008, pp. 3158–3163. neering, Salem, Tamil Nadu, India, in 2003. He re-
[30] G. Antonelli, F. Caccavale, S. Chiaverini, and L. Villani, “Tracking control ceived the M.E. degree in manufacturing engineering
for underwater vehicle-manipulator systems with velocity estimation,” from the Government College of Technology, Coim-
IEEE J. Ocean. Eng., vol. 25, no. 3, pp. 399–413, Jul. 2000. batore, Tamil Nadu, India, in 2005 and the Ph.D.
[31] Y. Cui and N. Sarkar, “A unified force control approach to autonomous degree in robotics and control from the Indian Insti-
underwater manipulation,” Robotica, vol. 19, pp. 255–266, 2001. tute of Technology (IIT) Madras, Chennai, India, in
[32] I. Schjløberg and T. I. Fossen, “Modelling and control of underwater 2010.
vehicle-manipulator systems,” in Proc. Conf. Marine Craft Maneuv. Con- From June 2010 to March 2011, he worked as an
trol, 1994, pp. 45–57. Assistant Professor in the Department of Mechanical
[33] G. Antonelli, F. Caccavale, and S. Chiaverini, “Adaptive tracking control Engineering, National Institute of Technology Calicut (NITC), Kerala, India.
of underwater vehicle-manipulator systems based on the virtual decompo- He, then worked as World Class University (WCU) Postdoctoral Fellow at Ko-
sition approach,” IEEE Trans. Robot. Autom., vol. 20, no. 3, pp. 594–602, rean Advanced Institute of Science and Technology (KAIST), Daejeon, South
Jun. 2004. Korea. In addition to this, he received another prestigious Brain Korea 21 (BK21)
[34] T. McLain, S. Rock, and M. Lee, “Experiments in the coordinated control postdoctoral fellowship with the same institute from September 2011 to March
of an underwater arm/vehicle system,” Autonom. Robots, vol. 3, no. 2-3, 2012. In 2012, he joined the faculty of the Mechanical Engineering at the IIT
pp. 213–232, 1996. Indore, Indore, Madhya Pradesh, India. Currently, he is the Head of the Centre
[35] C. C. de Wit, O. O. Diaz, and M. Perrier, “Nonlinear control of an underwa- for Robotics and Control, IIT Indore. His active research areas include underwa-
ter vehicle/manipulator with composite dynamics,” IEEE Trans. Control ter vehicle control, underwater manipulator design and control, parallel robotic
Syst. Technol., vol. 8, no. 6, pp. 948–960, Nov. 2000. platforms, and dynamic modeling and control of dynamic systems.
[36] S. McMillan, D. Orin, and R. McGhee, “Efficient dynamic simulation of
an underwater vehicle with a robotic manipulator,” IEEE Trans. Syst. Man
Cybern., vol. 25, no. 8, pp. 1194–1206, Aug. 1995.
[37] C. C. Lee, “Fuzzy logic in control systems: Fuzzy logic controller-Part I,”
IEEE Trans. Syst. Man Cybern., vol. 20, no. 2, pp. 404–418, 1990.
[38] L. A. Zadeh, “A fuzzy-algorithmic approach to the definition of complex
or imprecise concepts,” Int. J. Man-Mach. Studies, vol. 8, pp. 249–291, Balasaheb M. Patre (M’13) was born in Basmath-
1976. nagar, India, in 1965. He received the B.E. and M.E.
[39] D. Driankov, H. Hellendoorn, and M. Reinfrank, An Introduction to Fuzzy degrees in instrumentation and control engineering
Control, New Delhi, India: Narosa Publishing House, 2004, ch. 4. from Marathwada University, Aurangabad, India, in
[40] B. J. Choi, S. Kwak, and B. K. Kim, “Design of a single-input fuzzy 1986 and 1990, respectively, and the Ph.D. degree in
logic controller and its properties,” Fuzzy Sets Syst., vol. 106, no. 3, pp. systems and control engineering from the Indian In-
299–308, Sep. 1999. stitute of Technology (IIT) Bombay, Bombay, India,
[41] K. L. F. Lewis, “Some issues about fuzzy logic control,” in Proc. 32nd in 1998.
Conf. Decision Control, 1993, pp. 1743–1748. Currently, he is a Professor of Instrumentation En-
[42] K. Ishaque, S. Abdullah, S. Ayob, and Z. Salam, “Single input fuzzy logic gineering at the Shri Guru Gobind Singhji Institute
controller for unmanned underwater vehicle,” J. Intell. Robot. Syst., vol. of Engineering and Technology, Nanded, India. He
59, no. 1, pp. 87–100, 2010. has published around 160 papers in national/international conferences/journals.
[43] T. I. Fossen, Guidance and Control of Ocean Vehicles, New York, NY, He has presented his research work at Cambridge University, U.K., and in Ger-
USA: Wiley, 1994, ch. 2 and 3. many. His areas of interest include sliding mode control and its applications,
[44] J. J. Craig, Introduction to Robotics: Mechanics and Control, Boston, MA, robust control, large-scale systems, interval arithmetic applications, intelligent
USA: Addison Wesley, 1986, ch. 6. control, etc.
[45] R. Kelly, V. Santibanez, and A. Loria, Control of Robot Manipulators in Dr. Patre is a Life Member of the Indian Society for Technical Educa-
Joint Space, London, U.K.: Springer-Verlag, 2005, ch. 10–12. tion (ISTE) and the Instrument Society of the Institution of Electronics and
[46] T. Yatoh, S. Sagara, and M. Tamura, “Digital type disturbance compensa- Telecommunication Engineers (IETE), and a Senior Member of the Institution
tion control of a floating underwater robot with 2 link manipulator,” Artif. of Engineering and Technology (IET). He is a reviewer for several international
Life Robot., vol. 13, no. 1, pp. 377–381, 2008. journals.
[47] A. Sabanovic and K. Ohnishi, Motion Control Systems, Singapore: Wiley,
2011, ch. 4.
[48] B. J. Choi, S. Kwak, and B. K. Kim, “Design and stability analysis of
single-input fuzzy logic controller,” IEEE Trans. Syst. Man Cybern. B,
Cybern., vol. 30, no. 2, pp. 303–309, Apr. 2000.
[49] P. S. Londhe, B. M. Patre, and A. P. Tiwari, “Design of single-input fuzzy
logic controller for spatial control of advanced heavy water reactor,” IEEE Laxman M. Waghmare (M’12) received the B.E.
Trans. Nuclear Sci., vol. 61, no. 2, pp. 901–911, Apr. 2014. and M.E. degrees from the Shri Guru Gobind Singhji
[50] K. Ishaque, S. Abdullah, S. Ayob, and Z. Salam, “A simplified approach to Institute of Engineering and Technology, Nanded, In-
design fuzzy logic controller for an underwater vehicle,” J. Ocean. Eng., dia, in 1986 and 1990, respectively, and the Ph.D. de-
vol. 38, no. 1, pp. 271–284, Jan. 2011. gree in instrumentation engineering from the Indian
[51] J. J. E. Slotine and W. Li, Applied Nonlinear Control, Englewood Cliffs, Institute of Technology (IIT) Roorkee, Roorkee, In-
NJ, USA: Prentice-Hall, 1991, ch. 4. dia, in 2001.
Currently, he is a part of a research project spon-
Pandurang S. Londhe was born in Daund, Pune, sored by Naval Research Board (DRDO) in progress.
India, in 1988. He received the B.E. degree in instru- He has one book and 130+ publications in na-
mentation and control engineering from Pune Univer- tional/international conferences and journals to his
sity, Pune, India, in 2011 and the M.Tech. degree in credit. Ten students have completed Ph.D. under his guidance and research
instrumentation engineering from Marathwada Uni- work of eight Ph.D. students is in progress. He also guided 42 M.Tech. the-
versity, Nanded, Maharashtra, India, in 2013. Cur- ses. His field of interest includes intelligent control, process control, and image
rently, he is working toward the Ph.D. degree in processing. He worked as a resource person for many continuing education
in the field of control of autonomous underwater programs (CEPs), attended many short-term courses, and participated in work-
robots at the Shri Guru Gobind Singhji Institute shop/conferences organized by different national/international agencies.
of Engineering and Technology, Nanded, Maharash- Dr. Waghmare is a recipient of the K.S. Krishnan Memorial national award
tra, India. for the best system oriented research paper (published in the IETE Research
He has published six papers in international journals and seven papers in na- Journal). He is a member of professional societies such as the Instrument Soci-
tional/international conferences. His area of interest includes intelligent control ety of India (ISI), the Institution of Engineers (IE), the Institution of Engineering
of underwater robots and nuclear reactor power control. and Technology (IET), and the Indian Society for Technical Education (ISTE).

You might also like