Professional Documents
Culture Documents
Venkat Es An 2014
Venkat Es An 2014
Marina Bay Sands, Singapore, 10-12th December 2014 (ICARCV 2014) Fr34.5
Abstract—This paper addresses the issue of developing a sliding mode control are being used for practical control of
robust and efficient controller for a manipulator for underwater these manipulators. Most of the above mentioned control
applications by proposing a terminal sliding mode control scheme schemes require the knowledge of all the states of the system
along with a disturbance observer and incorporating the same for during control which is not feasible for low-cost applications.
3-RRR serial spatial manipulator to ensure finite time Consequently, there is a need of a robust control scheme
convergence as asymptotic convergence in undesirable in which is able to adapt to these disturbances and perform
underwater tasks like positioning and tracking of a trajectory. efficiently even when the knowledge of all the state vectors
The performance of the proposed scheme is studied using are not available. Hence, in this paper, emphasis is given to
extensive numerical simulations depicting practical
developing a control scheme for a 3-RRR serial manipulator
circumstances with external disturbances and parameter
which is able to perform flawlessly in the presence of
uncertainties within the system. The capability of the control
scheme to overcome hydrodynamic forces and moments parameter variations and external disturbances. The proposed
including added mass effects, damping effects are extensively control scheme will perform even in the presence of limited
studied and validated for the control scheme. The dynamic state measurements.
modelling is done using the Euler-Lagrangian approach involving The control scheme proposed includes a disturbance
the energy associated with the system. The results are presented observer which estimates the states, the measurements of
after analyzing the trajectory tracking capabilities of the which are unavailable and hence the cost-effectiveness of the
manipulator in the presence of external disturbances and model
system increases manifold. Only the position measurements of
uncertainties.
the actuators are assumed to be available as they are easy to
Keywords—terminal sliding mode control, underwater obtain using potentiometers. One of the shortcomings of the
manipulator control, disturbance observer, finite time aforementioned control schemes is that they ensure asymptotic
convergence. stability with infinite time for convergence. This issue is
addressed in this paper and finite time convergence of the
I. INTRODUCTION controller is attained using the proposed control scheme. In
Underwater manipulators are an integral part of underwater this paper, the dynamic modelling of the system is done using
vehicles which are used extensively for underwater the Euler-Lagrangian energy formulation method. The
exploration and investigation. Hence it is crucial to understand dynamic equations of the manipulator are then improvised
the dynamics of these manipulators when they are subject to considering the hydrodynamic effects on the manipulator by
hydrodynamic conditions when coupled with an underwater incorporating a damping vector which cumulatively considers
vehicle. More specifically, the development of a self- all the dynamic effects like skin friction, buoyancy forces and
governing manipulator has been studied in [1 and 2]. An balancing. The performance of the manipulator is assessed by
analysis of the tracking capabilities of these manipulator numerical simulations which establish the effectiveness of the
systems were performed in [2]. controller in negating the external disturbances and parameter
variations and ensuring finite time convergence.
As compared to their terrestrial counterparts, underwater
manipulators are subject to a lot of external disturbances like The paper continues on as follows: the kinematic and the
the presence of varying hydrodynamic forces, underwater dynamic modelling of the manipulator is derived in section 2.
currents, etc., their non-linear and time varying nature. Several Section 3 provides the proposed control scheme involving the
approaches have been proposed in literature to overcome these terminal sliding mode control with the disturbance observer
shortcomings and numerous control schemes have been incorporated into it followed by the stability analysis for the
proposed for the same including hybrid force and position same. The performance analysis of the controller is elucidated
control, adaptive control, etc. [3-8]. Apart from these by performing positioning and trajectory tracking in section 4.
disturbances, sensor noises and actuator limitations make the The conclusions of the paper are presented in section 5.
controller more complicated. In spite of the development of
these control schemes, traditional controllers such as PID,
Figure 2. A render of the manipulator with a gripper (the gripper has not
been included in the analysis)
θ1 = a tan 2( y , x );
θ 2 = a tan 2( ab − bc ,ac + bd );
(2)
θ3 = a tan 2(sin θ 3 ,cos θ 3 )
where,
1567
earlier, the Euler-Lagrangian energy formulation method has where ς = [θ1 θ 2 θ 3 ]T is the vector of joint space variables
been used to determine the dynamic formulations. The motive
behind choosing this method is its simplicity, the recursive for the manipulator representing the joint angles, M(ς ) is the
nature and the accuracy with which it predicts the relationship inertia matrix, C(ς , ς ) is the Coriolis matrix representing the
between the input forces and the joint parameters. The energy centripetal effects on the manipulator, D(ς , ς ) is the matrix
associated with the system is the sum of the potential energy
and the kinetic energy of the system. The effect of the comprising of the damping effects on the manipulator
hydrodynamic parameters will be added to the manipulator instigated by the hydrodynamic effects on the manipulator,
dynamic equations at a later stage. The kinetic energy of the G(ς ) is the gravity term which represents the restoring forces
system is given by in the manipulator. Here, is the vector of joint forces and
dis = idis + edis is the vector of disturbances consisting of
1 §1 1 1 ·
KE = I yy1θ12 + ¨ m2 vcT vc2 + I yy 2θ12 + I zz 2θ22 ¸ both internal and external disturbances, where idis represents
2 ©2 2 2 2 ¹
internal disturbances in the system which includes internal
§1 1 1 2·
+ ¨ m3vcT vc3 + I yy 3θ12 + I zz 3 θ2 + θ3 ¸; ( ) (4) system noises, process noises during measurement, etc.,
©2 3 2 2 ¹ and edis represents the external disturbances like payload
variations, vibrations, etc.
where, vc1 and vc2 are the velocities of the center of mass of
III. CONTROL SCHEME OF THE UNDERWATER
links 2 and 3. MANIPULATOR
The eventual objective of this paper is to develop a control
ª xc º ªθ1 º ª xc º ªθ1 º scheme for the manipulator discussed here which is robust and
« 2» « » « 3» « »
= « yc2 » = J 2 «θ2 »; vc3 = « y c3 » = J 3 «θ2 »
responsive to disturbances – both internal and external – and
vc2
uncertainties while at the same time having a finite time of
« zc » «θ3 » « zc » «θ3 » (5)
convergence. A study of the literature suggests that most of
¬ 2¼ ¬ ¼ ¬ 3¼ ¬ ¼
the controllers described till date assume that the system
parameters are accurate and the disturbances are measured by
ª − cos θ 2 sin θ1 − sin θ 2 cos θ1 0º sensors directly. But the state measurements available are
cos θ 2 cos θ1 − sin θ 2 sin θ1 0»
L1 «
J2 = generally inaccurate and it is challenging to directly measure
2 « » the disturbances like the friction effects, the buoyancy
«¬ 0 cos θ 2 0»¼ (6)
variation, and uncertainties in the model parameters. Hence, a
control scheme which is able to adapt itself to these
ª L2 L2 º disturbances and uncertainties is proposed here wherein the
«− L1 cos θ 2 + 2 cos (θ2 + θ 3 ) sin θ1 − L1 sin θ 2 cos θ1 − 2 sin (θ2 + θ 3 ) cos θ1 » disturbances are estimated explicitly by a disturbance observer
« L2 L2 »
J 3 = « L1 cos θ 2 + cos (θ2 + θ 3 ) cos θ1 − L1 sin θ 2 sin θ1 − sin (θ2 + θ 3 ) sin θ1 » (7) and the control inputs are determined by the use of a terminal
« 2 2 »
« 0 L2 cos θ 2
L2
cos (θ2 + θ 3 ) » sliding mode control – which is a modified version of the
«¬ 2 »¼
sliding mode control – to ensure finite time convergence by
the addition of a non-linear in the design of the sliding surface
The potential energy of the system is given as follows for faster convergence as compared to the asymptotic stability
guaranteed by the sliding mode control.
§d · § L ·
PE = m1 g ¨ 1 ¸ + m2 g ¨ d1 + 1 sin θ 2 ¸ Disturbance Observer: The disturbances in the system can be
© 2¹ © 2 ¹
estimated by a standard observer which was proposed for a
§ L · (8) manipulator [10]. Here, it is assumed that the joint
+ m3 g ¨ d1 + L1 sin θ 2 + 2 sin (θ 2 + θ 3 )¸; acceleration measurements are available.
© 2 ¹
where g is the acceleration due to gravity. The potential
energy has been derived considering the first link to be vertical
{
τˆdis = − Lτˆ d + L M̂ (ς )ς + Ĉ (ς ,ς )ς − τ } (11)
M (ς )ς + C (ς ,ς )ς + D (ς ,ς )ς + G (ς ) = + dis (10) Differentiating (12) with respect to time gives
1568
z = − L ⋅ z + L(− τ − ξ (ς ,ς )) (13)
The inertial parameters of the links have been found from
the solid model of the manipulator and hence emulate the
where, L, is the observer gain matrix and ξ (ς ,ς ) = ας , is the actual parameters. The efficiency of the control scheme is
disturbance observer auxiliary matrix ( α is a positive portrayed here using a few tracking and positioning tasks the
constant). The disturbance vector can hence be estimated as details of which are given below.
Description of the task: The effectiveness of the manipulator
τˆ dis = z + ξ (ς ,ς ) (13) is validated by comparing its performance with conventional z = − L ⋅ z
control schemes namely – a PID controller and a sliding mode
control without disturbance compensation – with the proposed
Terminal Sliding Mode Control: The modified sliding mode control scheme which is a terminal sliding mode control
surface for thee control scheme is given by coupled with a disturbance observer. Since the main
applications involve positioning and trajectory tracking for
s = ς~ + γς~ β (14)
pick and place operations, the manipulator is subject to a
simple positioning task where it is instructed to move to a
where ς~ = ςd − ς and ς~ = ς d − ς are the errors in joint predefined position starting from its initial position and a
velocities and joint positions respectively. The control vector trajectory tracking application where it is commanded to track
for determining the input torques is given by a circular trajectory in space the details of which Table 4.
As explained above the proposed control scheme is
(
τ = M (ς ) ςd + K s s β
)− τˆdis (15) compared with a traditional PID controller and a SMC without
disturbance compensation. The control scheme of both are
elucidated below.
where β is a positive proportional constant ( 0 < β < 1 )
τ = §¨ K Pς~ + K I ς~dt + K Dς~ ·¸ + G (τ )
PID with gravity (16)
which has been used to provide the non-linearity required in
the sliding surface which in turn will ensure a finite time © ³ ¹
convergence. Here γ and K s are the gain values which are to be
tuned to ensure faster convergence. When s → 0 , then the
actual joint velocities and positions converge to the desired §s· SMC (17)
τ = K s s + K sat sat ¨¨ ¸¸
values as ς → ςd and ς → ς d resulting in the state equilibrium ©φ ¹
of the system.
where K P , K I and K D are the proportional, integral and
IV. PERFORMANCE ANALYSIS
derivative gains of the PID controller and s is the sliding
vector used in sliding mode control. The gain values have
The performance of the system is gauged by performing been tuned in such a way that all these controllers show
extensive numerical simulations involving positioning and satisfactory performance in ideal conditions and hence can be
trajectory tracking of the end effector. The simulations have compared with the proposed controller to estimate its
been run on the manipulator using the passive parameters performance. The underwater effects that are bound to act on
presented in Table 1. The shape of the links have been the manipulator – such as the underwater currents, damping
assumed to be cylindrical though the performance of the effects, etc. – have been included in the disturbance vector
control scheme is least affected by it as it would accommodate which has been exclusively appended to the system to
these variations with ease. The control is essentially a joint simulate the underwater conditions that the manipulator will
space control wherein the desired task space parameters are be subjected to while performing any of the positioning or
transformed to joint space parameters using the inverse tracking operations.
kinematic solution which has been established in the Section
2. The gain values which have been used for the proposed A. Task 1: The positioning capabilities of the manipulator
controller are given Table 2. are assessed here. The initial and final positions of the
manipulator in terms of their joint space parameters are
TABLE II. PRE-DEFINED PARAMETERS FOR THE CONTROLLER given in Table 3. The main motive here is to see the
performance of the control scheme in positioning the end
Parameters Values assigned
effector to a pre-defined position starting from any random
Ks 10 position.
L 0.1M (ς )
α 2
β 0.75
1569
TABLE III. POSITIONING IN JOINT SPACE The proposed controller on the other hand performs
exceptionally compared to the other two by reaching a steady
Joint space
Initial position Final position state more rapidly. Moreover, the fact to be appreciated is that
parameter
it does so with no or minimal oscillations when compared to
θ1 0.5 Π
2 the PID controller. The main motive behind designing this
θ2 -0.8 Π controller is finite time convergence and the results illustrate
6 the fact this has been achieved without compromising the time
θ3 1.8 Π of settling and simultaneously ensuring the absence of
4
oscillations. The steady state error is noticeable in SMC quite
B. Task 2: The trajectory equation in task space is given in evidently because of its asymptotic convergence characteristic
Table 3. The initial velocity vector of the links is set to and the absence of a disturbance observer.
zero and the trajectory tracking time is to be to 7 seconds.
As can be observed from the equations the radius of the This shows that the disturbance observer coupled with the
circular trajectory has been set to be 0.2m and the proposed controller performs competently as well, as it is able
coordinates of the center point is defined to be (0.3m, to filter out the disturbances, noises and the uncertainties
0.3m, 0.1m). present in the system parameters thus exemplifying the
disturbance compensation capabilities of the controller.
TABLE IV. TRAJECTORY EQATIONS IN TASK SPACE
Task space
Time dependent Trajectory Equation
parameter
x 0.3 + 0.2 sin (0.1t )
1570
quite acceptable and adequate for underwater applications. Even the sliding mode control scheme – which has always
As seen in Fig. 6, the distortions in the trajectory shown by touted as an efficient scheme in annulling disturbances – fails
both the PID controller and the SMC are well beyond the to perform under the action of these extensive disturbances as
range of a reasonable performance. But the proposed can be gauged from the figures. The argument that can be
controller performs with appreciable effectiveness far discerned from these figures is that under the action of
outperforming the other controllers. disturbances – which include the external disturbances like the
ocean currents and the like – the terminal sliding mode control
has better capabilities in negating these aforementioned
disturbances.
Consequently, it is apparent from the above discussion that
the proposed controller can perform efficiently with reduction
in the torque fluctuations required by the actuators under
identical conditions. The controller is also able to recover from
disturbances, both internal and external while tracing a pre-
defined path. The parameter estimation capability of the
controller is also efficient and the adaptability of the system to
the dynamic changes in the system are remarkable. In all, the
performance of the controller is outstanding, given that it is
able to converge in finite time to the system states even in the
presence of disturbances, uncertainties and noises.
1571
compensation capabilities of the disturbance observer
incorporated with the proposed control scheme. This
guarantees the performance of the control scheme for usage in
position and trajectory tracking applications which is the sole
motive behind developing this underwater manipulator.
ACKNOWLEDGMENT
1572