Research Article

International Journal of Advanced

Robotic Systems
January-February 2019: 1–16
A stable proportional–proportional integral ª The Author(s) 2019
DOI: 10.1177/1729881418819956
tracking controller with self-organizing journals.sagepub.com/home/arx

fuzzy-tuned gains for parallel robots

Francisco G Salas1 , Jorge Orrante-Sakanassi2,

Raymundo Juarez-del-Toro1 and Ricardo P Parada3

Abstract
Parallel robots are nowadays used in many high-precision tasks. The dynamics of parallel robots is naturally more complex
than the dynamics of serial robots, due to their kinematic structure composed by closed chains. In addition, their current
high-precision applications demand the innovation of more effective and robust motion controllers. This has motivated
researchers to propose novel and more robust controllers that can perform the motion control tasks of these manipulators.
In this article, a two-loop proportional–proportional integral controller for trajectory tracking control of parallel robots is
proposed. In the proposed scheme, the gains of the proportional integral control loop are constant, while the gains of the
proportional control loop are online tuned by a novel self-organizing fuzzy algorithm. This algorithm generates a perfor-
mance index of the overall controller based on the past and the current tracking error. Such a performance index is then
used to modify some parameters of fuzzy membership functions, which are part of a fuzzy inference engine. This fuzzy engine
receives, in turn, the tracking error as input and produces an increment (positive or negative) to the current gain. The
stability analysis of the closed-loop system of the proposed controller applied to the model of a parallel manipulator is carried
on, which results in the uniform ultimate boundedness of the solutions of the closed-loop system. Moreover, the stability
analysis developed for proportional–proportional integral variable gains schemes is valid not only when using a self-organizing
fuzzy algorithm for gain-tuning but also with other gain-tuning algorithms, only providing that the produced gains meet the
criterion for boundedness of the solutions. Furthermore, the superior performance of the proposed controller is validated
by numerical simulations of its application to the model of a planar three-degree-of-freedom parallel robot. The results of
numerical simulations of a proportional integral derivative controller and a fuzzy-tuned proportional derivative controller
applied to the model of the robot are also obtained for comparison purposes.

Keywords
Parallel robot, tracking control, fuzzy control, gain-scheduling, stability analysis

Date received: 13 November 2017; accepted: 25 November 2018

Topic: Robot Manipulation and Control

Topic Editor: Andrey V Savkin
Associate Editor: Steven Su

1
Facultad de Contaduria y Administracion, Universidad Autonoma de Coahuila, Torreon, Mexico
2
Division de Estudios de Posgrado e Investigacion, CONACYT-Tecnologico Nacional de Mexico/Instituto Tecnologico de Hermosillo, Hermosillo,
Mexico
3
Programa de Ingenieria en Tecnologias de Manufactura, Universidad Politecnica de Gomez Palacio, Gomez Palacio, Mexico

Corresponding author:
Francisco G Salas, Facultad de Contaduria y Administracion, Universidad Autonoma de Coahuila, Blvd Revolucion 151 Oriente, Col. Centro, 27000,
Torreon, Coahuila, Mexico.
Email: francisco.salas@uadec.edu.mx

Creative Commons CC BY: This article is distributed under the terms of the Creative Commons Attribution 4.0 License
(http://www.creativecommons.org/licenses/by/4.0/) which permits any use, reproduction and distribution of the work without
further permission provided the original work is attributed as specified on the SAGE and Open Access pages (https://us.sagepub.com/en-us/nam/
open-access-at-sage).
2 International Journal of Advanced Robotic Systems
Introduction
Parallel robots are nowadays used in many high-precision
industrial tasks, such as machining,1,2 welding,3,4 packa-
ging,5,6 as well as flight simulators7 and telescopes.8,9
These high-precision applications demand the implemen-
tation of position regulators, for point-to-point motion
tasks, and tracking controllers, when the end-effector
motion has to follow a prescribed trajectory of motion.10
Moreover, parallel robots consist of closed kinematic
chains that can be modeled, roughly speaking, as a collec-
tion of open kinematic chains in which their extremes (at
the end-effectors) are joined together.11 This parallel struc-
ture can be mathematically represented as a dynamic model
of a serial robot plus a set of holonomic constraints. There-
fore, as it will be explained later, the dynamic model of
parallel robots has inherently a more complex structure
than the dynamic model of serial robots. This has led
researchers in robot control to conceive more robust con-
trollers to cope with inaccuracies and parameter uncertain-
ties that may exist in the dynamic models of parallel robots.
Hence, we can see that the development of trajectory track-
ing controllers for parallel robots is nowadays an important
and interesting problem.
Recent approaches for control of parallel manipulators
found in the literature include model-based controllers12,13
and not model-based controllers.12,14,15 In the study by
Ren et al.,12 a comparison of several control approaches
for robot tracking of a three-degree-of-freedom (3-DOF)
parallel robot is presented. They compare the performance
of an adaptive controller, a proportional integral (PI)-type
synchronized controller (model-based), a conventional
proportional integral derivative (PID) controller, and an
adaptive synchronized controller (not model-based control-
ler). In Diaz-Rodriguez et al.’s16 work, a reduced model-
based control of a 3-DOF parallel robot is proposed. The
reduced model is obtained by considering a simplified
model with a set of relevant parameters. In the study by
Borbonnais et al.,13 a computer torque controller and a
conventional PID controller are implemented for a novel
five-bar parallel robot.
An interesting two-loop control scheme is proposed by
Kelly and Moreno17 for motion control of manipulators in
operational space. The proposed scheme is composed of a
joint velocity inner loop and an operational space kinematic
control outer loop. The motivation of the cited work is the
fact that in many industrial robots the motion control of
each joint is performed using inner joint velocity control
loops and outer position control loops.17–19 In addition, it is
shown that the overall closed-loop dynamics of the two-
loop control scheme can be seen as a cascade system, and it
presents sufficient conditions for the Global Exponential
Stability of the closed-loop systems. Inspired by Kelly and
Moreno’s work,17 Soto and Campa20 proposed a two-loop
control scheme for redundant manipulators. In such a
scheme, an outer kinematic controller generates the joint
velocity commands for an inner velocity controller for each
joint of the manipulator. A recent application of a two-loop
motion control scheme for a parallel robot is reported by
Campa et al.,21 where a kinematic model-based operational
space tracking controller with a two-loop structure for a
hexapod-type parallel robot is proposed. The two-loop
structure contains a Resolved Motion Rate Controller
(RMRC) in the outer loop and a joint velocity PI in the
inner loop.
On the other hand, not model-based controllers are not
sensitive to model uncertainties or inaccuracies. This can
be an important advantage when dealing with the dynamic
model of parallel robots, which have inherently a more
complex structure than the dynamic model of serial
robots, as it was pointed above. Regarding classic not
model-based controllers, such as PD or PID, these can
improve their robustness when used in variable gains
schemes or when a nonlinear function of the error is
added. In these schemes, either the controller gains or
some nonlinear control action are set-up online according
to a measure of the performance of the controller. In the
study by Salinas et al.,15 a family of nonlinear PID-like
controllers is proposed in which an integral action of a
nonlinear function of the position error is added to the
control signal. Recent approaches of variable gains con-
trol schemes for parallel robots include 14 in which a
discrete-time motion controller is proposed. In such work,
the output of an Adaptive Neuro-Fuzzy Inference System
is used in the control law in addition to a conventional PID
controller. In Villareal-Cervantes and Alvarez-Gallegos’
work,22 a PID control off-line tuning scheme using Evo-
lutionary Algorithms is proposed. The scheme is tested on
a planar parallel robot with a five-bar mechanism. In
Ouyang et al.’s work,23 a nonlinear PD control for a par-
allel manipulator system is proposed, in which the P and D
gains are certain class of nonlinear bounded functions of
the position error. In the study by Zhu and Zhang,24 a
fuzzy-tuned PID control strategy is proposed for regula-
tion control of a parallel robot. In Stan et al.’s work,25 a
fuzzy controller performs the fine tuning of the gains of a
PID controller. A coarse tuning of these gains is previ-
ously done. The fuzzy controller is applied to a Cartesian
parallel robot of 3-DOF known as ISOGLIDE. In another
related recent work,26 a fuzzy-PID controller is designed
for control of a 6-DOF hydraulically driven parallel robot.
The PID gains are tuned by a conventional fuzzy algo-
rithm which uses the joint position errors and its temporal
derivatives as input variables. However, stability analysis
of these variable gains schemes for not model-based con-
trollers for parallel robots have not been reported in the
literature. The stability of the resultant closed-loop system
is an important concern in the design process of control-
lers for manipulators.
Given the larger complexity of the dynamic models of
parallel robots, and the need of improved robustness of the
controllers before parameter variations or uncertainties, in
Salas et al. 3

this work, we propose a novel design of a two-loop P-PI
controller for tracking control of parallel manipulators. The
outer loop is a joint position P control loop, and the inner
loop is a joint velocity PI control loop. The gains of the
inner PI control loop are constant, while the gains of the
outer P control loop are tuned by a self-organizing fuzzy
(SOF) algorithm, which is a novel algorithm that generates
the gain depending on the current tracking errors. This
algorithm employs a set of fuzzy rules that can be updated
online according to a performance index of the controller
response. This gain tuning scheme provides the needed
increased robustness in presence of perturbations and para-
meter uncertainty of the overall controller, as well as
improved results when it is tracking fast changing trajec-
tories. The usefulness and performance of the proposed
controller is tested by numerical simulations, in compari-
son with two controllers, a PID controller, and a fuzzy-
tuned PD (FPD) controller. Moreover, the stability analysis
of the closed-loop system is carried out, giving as a result
the proof of the Uniform Ultimate Boundedness of the
solutions of the closed-loop system.
The main contributions of this work are the design and
the stability analysis of a tracking control structure based
on a two-loop scheme composed of two not model-based
controllers, that is, P and PI. In addition, the gains of the P
control loop can be tuned online. Moreover, the P-gain tun-
ing is performed by a novel SOF algorithm. Furthermore, the
stability analysis of the resulting closed-loop system is car-
ried out. It is worth to mention that the result of the stability
analysis holds for any tuning algorithm for tuning the P
gains, by only ensuring that some conditions on the bound-
edness of the produced gains are satisfied. Therefore, this is a
general result for a family of two-loop P-PI tracking control-
lers in which the P gain of the outer control loop is variable.
According to the best knowledge of the authors, this is an
original result not published before elsewhere.
The remaining of this article is organized as follows. In the
“Preliminaries” section, the dynamic model of the parallel
robot as well as some of their properties useful for the stability
analysis are described. In the section “Proposed controller,”
the full description of the proposed controller is given. In
section “Stability analysis,” the development of the stability
analysis in the sense of Lyapunov of the closed-loop system is
carried out, and the result of this analysis is presented. In
section “Simulations,” the elements of the dynamic and kine-
matic model of a 3-DOF parallel robot are described as well as
the desired trajectory for simulations. Also, the numerical
simulations conducted are presented and discussed. Finally,
some concluding remarks are given.
pffiffiffiffiffiffiffiffiffi
capital letters, for example, A. k x k¼ x T x represents the
Euclidean norm of vector x. l max fAg and l min fAg repre-
sent the largest and smallest eigenvalues of matrix A,
respectively.
Dynamic model of the robot
The joints of a parallel manipulator can be classified as n
actuated joints, represented by vector q 2 Rn , and m non-
actuated joints, represented by vector b 2 Rm . Moreover,
the l operational or Cartesian variables of the end-effector
can be represented with vector x 2 Rl . Let us define the
vector of constrained variables r as
r ¼ ½qT bT xT  T 2 Rs
with s ¼ n þ m þ l. By applying the Euler–Lagrange
formulation, the dynamic model of a parallel robot27
in absence of friction is in general formulated as
M 0 ðrÞ€
r þ C 0 ðr; rÞ
_ r_ þ g0 ðrÞ ¼ tr þ D T ðrÞl ð1Þ
where M 0 ðrÞ 2 Rss is the symmetric positive definite
_ 2 Rss is the Coriolis and centripe-
inertia matrix, C 0 ðr; rÞ
tal forces matrix, g ðrÞ 2 Rs is the vector of gravitational
0
forces, tr 2 Rs is the vector of generalized forces associated
to scalar variables of r, DðrÞ ¼ @gðrÞ=@r 2 RðkÞðnÞðsÞ is
the Jacobian matrix of the system holonomic constraints
gðrÞ 2 RðkÞðnÞ , with k as the minimum number of coordinates
in the space that describes the position of the end-effector,
and l 2 RðkÞðnÞ is the vector of Lagrange multipliers.
Kinematic transformation
Let us define a kinematic transformation between vectors r_
and q_ as
r_ ¼ RðrÞq_ ð2Þ
to obtain a so-called reduced dynamic model. Notice that
the matrix RðrÞ 2 RðsÞðnÞ can be constructed with the
Jacobian matrices J b ðrÞ and J x ðrÞ from the differential
kinematic models b_ ¼ J b ðrÞq_ and x_ ¼ J x ðrÞq,
_ as
2 3
I
6 7
RðrÞ ¼ 4 J b ðrÞ 5; I 2 Rnn
J x ðrÞ
The derivative with respect to time of equation (2) is
r _
€ ¼ RðrÞ q_ þ RðrÞ€q ð3Þ

Preliminaries By substituting r_ from equation (2) and r

€ from equation
(3) in equation (1), we get
Notation
M 0 ðr ÞRðrÞ
_ q_ þ M 0 ðrÞRðrÞ €q þ C 0 ðr; rÞRðrÞ
_ q_ ð4Þ
In this work, vectors are denoted with italic-bold lowercase
0 T
letters, for example, x or v. Matrices are denoted with þ g ðrÞ ¼ tr þ D ðrÞl
4 International Journal of Advanced Robotic Systems

Figure 1. Proposed controller. SOF: self-organizing fuzzy.

After multiplying equation (4) by RT ðrÞ

_ and rearranging MðqÞ €q þ Cðq; qÞ
_ q_ þ gðqÞ ¼ t ð7Þ
terms we obtain
_ and gðqÞ are the inertia matrix, the
where MðqÞ; Cðq; qÞ,
R T ðr ÞM 0 ðr ÞRðrÞ €q ð5Þ Coriolis and centripetal forces matrix, and the gravitational
þ ½R T ðrÞM 0 ðrÞRðrÞ
_ þ R T ðrÞC 0 ðr; rÞRðrÞ
_ q_ forces vector of the reduced model, respectively, and t is
the vector of forces associated with the actuated joints q.
þ R T ðrÞ g0 ðrÞ ¼ R T ðrÞtr The model in equation (7) exhibits the following properties.
where the term with the product of the Jacobian matrix of Property 1. M(q) is symmetric and positive definite.29
the constraints by the Lagrange multipliers vanishes
because it belongs to the null space of D T ðrÞ. By defining _
Property 2. MðqÞ _ is skew-symmetric.11,29
 2Cðq; qÞ
Mðr Þ ¼ R T ðr ÞM 0 ðrÞRðrÞ Property 3. There exists a constant k C > 0 such that
_ ¼ R T ðrÞM 0 ðrÞRðrÞ
Cðr; rÞ _ þ R T ðrÞC 0 ðr; rÞRðrÞ
_ _ k  k C k q_ k, for all q 2 Rn .30
k C 0 ðq; qÞ
gðrÞ ¼ R T ðrÞg0 ðrÞ Property 4. For robots with only revolute joints, there exists
t ¼ R T ðrÞtr a constant k 0 such that k gðqÞ k k 0 , for all q 2 O  Rn .

equation (5) can be rewritten as
q þ Cðr; rÞ
MðrÞ € _ q_ þ gðrÞ ¼ t
11
In Ghorbel et al.’s work, it is proven, by means of the
implicit function theorem, that there exists a unique para-
metrization of r 2 Nr , inside a neighborhood Nr
r ¼ hðqÞ
for any q 2 Nq inside a neighborhood Nq , whenever the
system is not in a singular configuration.
In addition, Muller28 established that, for a parallel
machine, a subset q of d joint variables determines its con-
figuration, in virtue of that there exists a smooth mapping
that assigns to each r, the parallel machine configuration as
r ¼ ðqÞ, where the map 1 is a local parametrization of
the d dimensional manifold V
V ¼ fr 2 Vn ; gðrÞ ¼ 0g
where V represents the set of all admissible configurations
of the parallel machine, and gðrÞ represents the holonomic
constraints.
In consequence, without loss of generality, we can write
down the matrices of the dynamic model MðrÞ; Cðr; rÞ, _
_ and gðqÞ, respectively. Thus, the
and gðrÞ as MðqÞ, Cðq; qÞ,
dynamic model in equation (6) takes the form
Proof. For robots with only revolute joints, the entries of the
vector of gravitational forces g0 ðrÞ (see equation (1)) are
ð6Þ
sinusoidal or trigonometric functions with finite coeffi-
cients. Thus the entries are upper bounded. Now, lets con-
sider the vector gðrÞ ¼ R T ðrÞ g0 ðrÞ. Note that the entries
of this vector are upper bounded for all q 2 O  Rn , where
O denotes a space of configurations of the robot in which
the matrix RðrÞ is well posed.
Proposed controller
In this work, we propose a controller for trajectory tracking
of parallel manipulators. The structure of the proposed con-
troller is a two-loop cascade structure, with a P position
controller in the outer loop and a PI controller in the inner
loop. The overall structure of the controller is shown in
Figure 1. Note that the desired joint position and velocity
vectors qd and q_ d , respectively, are represented in the Fig-
ure 3, as well as the position error and velocity error vectors
q~ ¼ qd  q and q~_ ¼ q_ d  q,
_ respectively. We define the
velocity control law of the outer loop as
q; q~_Þ~
vd ¼ Fð~ q þ q_ d ð8Þ
q; q~_Þ is a diagonal matrix of variable gains
Notice that Fð~
of n  n dimension. The values of each diagonal element of
Salas et al.
Figure 2. Simplified structure of the SOF controller. SOF: self-
organizing fuzzy.
the matrix are tuned by an algorithm, represented in Figure
1 as the block labeled SOF controller, that evaluates the
performance of the current gain values by measuring
the position and velocity errors. Such an algorithm will
be described in a later section in this article. Regarding the
Figure 1, the velocity error v ~ in the inner loop can be
written as
q; q~_Þ~
~ ¼ Fð~
v q; q~_Þ q~ þ q~_
q þ q_ d  q_ ¼ Fð~ ð9Þ
where q_ d  q_ has been substituted by q~_. Notice that, in
Figure 1, the diagonal arrow shown over the block labeled
as Fð~q; q~_Þ represents the tuning of the gains by the SOF
algorithm.
In the inner loop, the control law can be written as
t ¼ K vp v
~ þ K vi x ð10Þ
where
ðt
x_ ¼ v;
~ x¼ ~
vðsÞ ds ð11Þ
0
K vp and K vi are n  n diagonal matrices of constant
gains. These matrices can be defined for a later analysis as
K vp :¼ k p I ð12Þ
K vi :¼ aK vp ð13Þ
where k p and a are positive constants and I 2 Rn is the
identity matrix.
SOF controller
The proposed algorithm for tuning the gains of the outer P
loop is a novel SOF controller. This controller generates the
gains based on the current position and velocity errors. The
performance of the overall control strategy is evaluated to
produce a performance index. This parameter is then used
to adjust the parameters of the membership functions of the
system to obtain an improved control signal. In a fuzzy
controller, the membership functions are functions that
map a crisp input variable to a membership grade that can
take continuous values in the interval ½0; . . . ; 1.31 This
process is known as fuzzification. In summary, by continu-
ously evaluating the performance of the controller and gen-
erating the respective changes in the fuzzy membership
parameters, the performance of the tracking is improved.
A simplified block diagram of the structure of the SOF
controller is shown in Figure 2. The input signals are the
5
Figure 3. Structure of the input stage.
ith position and velocity errors q~i and ~q_i . The output is i ,
which is the i th diagonal element of variable gains matrix
F. In the following subsections, the stages of the SOF con-
troller are described.
Input stage. A block diagram of the structure of the input
stage is shown in Figure 2. In this stage, two processes are
performed: scaling and quantization, to normalize the lev-
els of input signals before passing them to the following
stage. The scaling is done by multiplying by a scaling
factor. Next, the scaled signal is subject to a quantization
process by mapping it into two sets of discrete values
S R ¼ f1; 2; 3; 4g and S F ¼ f6; 5:5; 5; . . . ; 5; 5:5; 6g.
These two discrete sets correspond to two types of discre-
tization which are conveniently labeled regular and fine,
respectively. Moreover, the values of the discrete set S R
corresponds to the linguistic variables zero, small, medium,
and large, labeled with letters Z, S, M, and L. Position and
velocity errors ~qi and ~q_i are scaled and quantized. In addi-
tion, the output i is also fed back into the scaling and
quantization processes. The output signals from regular
quantization of ~qi and ~q_i , labeled as epI and evI , respec-
tively, are used to address a performance index table. This
table, which is described in a subsequent section, produces
an appropriate value of the performance index.
On the other hand, the output signals from fine quanti-
zation of ~qi and ~q_i , labeled as epF and evF , respectively, are
submitted to a fuzzification process. The output signal
from fine quantization of i is stored in a First Input–First
Output (FIFO) serial-shifting memory. In this memory,
each introduced value is stored during four time periods
(see Figure 3). Then, the sum of the stored value and the
current performance index is stored into a four-element
serial-shifting memory. The output of these four memory
elements are used as the centers of four fuzzy membership
functions.
Fuzzification. The fuzzy membership functions selected are
triangular shaped functions, with constant width and adjus-
table central values. These functions are shown in Figure 4.
6 International Journal of Advanced Robotic Systems

performance is considered good and the performance index

is set to zero. If epI is L (Large) and evI is S (Small), we can
assume, similarly, that the errors are decreasing and the
performance index is set to zero. If epI and evI are S, M
(Medium) or L, it is considered that the performance is
poor, then the performance index is set to an appropriate
positive or negative value to modify the rule, to adjust the
produced gain value accordingly.

Figure 4. Fuzzification of discretized errors. Stability analysis

In this work, the approach that we address to carry out the
Table 1. Performance index table. stability analysis of the closed-loop system is considering
the SOF-P-PI controller simply as a variable gains P-PI
epI =evI Z S M L controller, without taking into account the dynamics of the
Z 0 0 0 0 tuning algorithm used for tuning the variable gains. We
S 0 2 2 2 only assume that the values of the tuned gains of the pro-
M 0 4 4 4 portional outer loop can be upper and lower bounded, as
L 0 0 2 2 well as the change ratio of these gains.
Z: zero; S: small; M: medium; L: large.
Closed-loop system
These central values are considered as the inference rules of By substituting equation (10) in equation (7) we obtain
the fuzzy mechanism. When crisp values epf and evf are MðqÞ €q þ Cðq; qÞ ~ þ K vi x
_ q_ þ gðqÞ ¼ K vp v
fuzzified by the four membership functions, the resulting
memberships grades are the sets f1p ;  2p ;  3p ; 4p g and Since the velocity error in the inner loop and its tem-
f1v ; 2v ; 3v ;  4v g, respectively. Next, these membership poral derivative are
grades are fed into a fuzzy inference process that uses the v ¼ vd  q_
min function for each pair of fuzzy membership grades of
position and velocity errors. The output I K from this infer- v~_ ¼ v
_ d  €q
ence algorithm is the vector we can write
~_ þ Cðq; qÞ
MðqÞv _ v ~ þ K vi x
~ þ K vp v ¼ MðqÞv
_ d þ Cðq; qÞv
_ d þ gðqÞ
I K ¼ ½ minð1p ; 1v Þ; minð 2p ; 2v Þ; minð3p ; 3v Þ; minð4p ; 4v Þ
ð14Þ
After adding aMðqÞv ~ þ aCðq; qÞx
_ to both sides of
Defuzzification. Defuzzification is carried out by the Mean equation (14), we obtain
of Maxima algorithm. In this algorithm, the two greatest
values of vector I K are multiplied by the central value of ~_ þ av
MðqÞ½v ~ þ Cðq; qÞ½
_ v ~ þ ax þ K vp v
~ þ K vi x
the fuzzy membership function from which the member- ¼ MðqÞ½v
_ d þ av
~ þ Cðq; qÞ½v
_ d þ ax þ gðqÞ
ship grade is obtained. The output U i is the arithmetic
average of both products. Finally, this output is descaled ð15Þ
by multiplying by a reverse scaling gain k r , and the product ~ T 2 R 3n . Let
qT xT vT
Let define the state vector z ¼ ½~
is added to the past period gain the term dðt; zÞ
i ðtÞ ¼ i ðt  T Þ þ k r U i _ d þ av
dðt; zÞ ¼ MðqÞ½v ~ þ Cðq; qÞ½v
_ d þ ax þ gðqÞ
where i ðt  T Þ is the new value of the gain and i ðt  T Þ ð16Þ
is the past value.
be considered as a disturbance. From equations (9), (11),
(15), and (16) the closed-loop system is
Performance index table. Performance index values are gen- 2 3 2 3
erated according to the current position and velocity error q~ ~  Fð~
v q; q~_Þ~
q
d6 7 6 7
values. In Table 1, the selected values for the performance 4x5¼4 v~ 5 ð17Þ
index are shown. These values were selected according to dt
~
v av~  M 1 ðqÞDðq; q; _ zÞ
the following criterion: if epI or evI are Z (Zero), we can
assume that the errors are decreasing. Then, the where
Salas et al. 7

21 3
_ zÞ ¼ Cðq; qÞ½
Dðq; q; _ v ~ þ ax þ K vp v
~ þ K vi x  dðt; zÞ 0 0
2
6 7
Note that the origin of the closed-loop system in 6 7
6 7
equation (17) is not an equilibrium. In the current case, P1 ¼ 6 0 ak p þ 12 a 2 l min fMg 12 al max fMg 7;
6 7
the stability of a nominal system will be investigated 6 7
4 5
first, to investigate the uniform and uniform ultimate 0 12 al max fMg 1
2
l min fMg
boundedness of the solutions of the closed-loop system 21 3
in equation (17). 2
0 0
6 7
6 7
6 7
P2 ¼ 6 0 ak p þ 12 a 2 l max fMg 1al
max fMg 7
6 2 7
Stability of the nominal system 6 7
4 1 1
5
0 al max fMg l max fMg
When the disturbance dðt; zÞ is equal to zero, the closed- 2 2

loop system is a nominal system

2 3 2 3 Then, the Lyapunov candidate function in equation (19)
q~ ~  Fð~
v q; q~_Þ~
q satisfies
d6 7 6 7
4x5¼4 ~
v 5 ð18Þ    
dt  k q~ k  2  k q~ k 2
v~ av 1
~  M ðqÞEðq; q; _ zÞ    
   
l 1  k x k   V ð~ ~  l2  k x k 
q; x; vÞ ð20Þ
   
where k v~ k   kv~ k 
_ zÞ ¼ Cðq; qÞ½
Eðq; q; _ v ~ þ ax þ K vp v
~ þ K vi x
l 1 ¼ l min fP 1 g; l2 ¼ l max fP2 g ð21Þ
Notice that the origin z ¼ 0 2 R is the only equili-
3n
To ensure that the Lyapunov candidate function is pos-
brium point of equation (18). We propose the positive itive definite, it is enough to prove that matrix P1 is positive
definite Lyapunov candidate function definite, which is satisfied if
2 3
2
V ð~
1 1
~ ¼ q~ T q~ þ ax T 4k p I þ aMðqÞ5x
q; x; vÞ aðl 2max fMg  l min fMgÞ
2 2 kp < ð22Þ
2l min fMg
ð19Þ
T 1 T The temporal derivative of the Lyapunov candidate
þ ax MðqÞv ~ þ v ~
~ MðqÞv
2 function in equation (19) is
V_ ð~ ~ ¼ q~ T q~_ þ 2ak p x T x_ þ a 2 x T MðqÞx_
q; x; vÞ
The first term of equation (19) can be written as
1 ~T ~
q q ¼ 12 k q~k2 . The second term of V ð~ ~ can be
q; x; vÞ 1 T
2 _
þ a 2 x T MðqÞx þ ax_ MðqÞv
~
lower bounded as follows 2
   
1 1 2 _
þ ax T MðqÞ ~_
~ þ ax T MðqÞv
v
ax k p I þ aMðqÞ x  a l min fMg þ ak p k xk2
T
2 2
1 T_
þv ~_ þ v
~ T MðqÞv ~
~ MðqÞv
and it can be upper bounded as 2
   
1 1
ax T k p I þ aMðqÞ x  a 2 l max fMg þ ak p k xk2 which along the trajectories of the nominal system in equa-
2 2
tion (18) yields
The third and fourth terms can be lower and upper
V_ ð~
q; x; vÞ ¼ q~T v q; q~_Þ~
~  q~T Fð~ q  a 2 xT K vp x
bounded as ð23Þ
T
T ~ T K vp v
v ~  2ag x v~
 al max k x kk v
~ k ~  al max k x kk v
ax MðqÞv ~ k;
1 1 T 1 where the Property 2 has been used, as well as equations
~ 2
l min fMg k vk ~ MðqÞv
v ~ 2
~  l max fMg k vk (12) and (13). The upper bounds of each term of equation
2 2 2
(23) can be written as
Thus V ð~ ~ in equation (19) can be bounded as
q; x; vÞ q~T v~  k q~ kk v
~ k;
2 3T 2 3 2 3T 2 3
k q~ k k q~ k k q~ k k q~ k
T
q; q~_Þ q~
 q~ Fð~   lFl k q~k2 ;
6 7 6 7 6 7 6 7
4 k x k 5 P 1 4 k x k 5  V ðq~; x; v
~ Þ  4 k x k 5 P2 4 k x k 5  a 2 xT K vp x   a 2 ðk p þ gÞ k xk2 ;
kv~ k kvk kv~ k kv~ k v ~ T K vp v~  ~ 2;
 ðk p þ gÞ k vk
where  2agx v T
~  2ag k x kk v ~ k
8 International Journal of Advanced Robotic Systems

where lFl is defined as ~ T þ ax T Þ d ðt; zÞ  ð1 þ aÞ& 0 k z k þð1 þ aÞ& 1 k z k2

ðv
 
lFl ¼ inf q; q~_Þg
l min fFð~ þ ð1 þ aÞ& 2 k z k3
q~;q~_2Rn
ð31Þ
Thus, equation (23) can be upper bounded as After substituting equations (26) and (31) as the upper
2 3T 2 3 bounds of equation (28) and making some arrangements,
k q~ k k q~ k
6 7 6 7 we obtain
V_ ð~ ~  4 k x k 5 Q4 k x k 5
q; x; vÞ ð24Þ
kv~ k kv~ k V_  ð1 þ aÞ& 0 k z k ðl 3  ð1 þ aÞ& 1 Þk z k2
ð32Þ
þ ð1 þ aÞ& 2 k z k3
where
2 3 Now we define a ball Br  Rn , centered in the origin,
lFl 0 12 with radio r, such that
6 7
6 7
Q¼6 a 2 ðk p þ gÞ ag 7 Br :¼ fz 2 Rn :k z k< rg ð33Þ
6 0 7
4 1 5
2 ag kp þ g inside which there exist conditions for the temporal deri-
vative of the Lyapunov candidate function V_ to be negative
To ensure that V_ ðt; zÞ is negative definite, matrix Q has definite, as it will be addressed later. Equation (32) can now
to be positive definite, which is satisfied if be written as
kp þ g V_  ð1 þ aÞ& 0 k z k ðl 3  ð1 þ aÞð& 1 þ r& 2 ÞÞk z k 2
lFl > ð25Þ
4k p ðk p þ 2gÞ ð34Þ

From equation (24), V_ ðt; zÞ can also be bounded as Let introduce a positive constant e < 1, such that equa-
tion (34) now can be written as
V_ ðt; zÞ  l3 k zk2 ð26Þ
V_  ð1 þ aÞ& 0 k z k
where
 ðl 3  ð1 þ aÞð& 1 þ r& 2 ÞÞð1  eÞk z k2 ð35Þ
l 3 ¼ l min fQg ð27Þ  ðl 3  ð1 þ aÞð& 1 þ r& 2 ÞÞek z k 2

Thus, it can be concluded the exponential stability of the Investigating the positiveness of the sum of the first and
origin of the nominal system by satisfying equations (22) the third terms of equation (35) we have that
and (25).
ðl 3  ð1 þ aÞð& 1 þ r& 2 ÞÞe k z k2 þ ð1 þ aÞ& 0 k z k< 0
when
Stability of the overall closed-loop system ðl3  ð1 þ aÞð& 1 þ r& 2 ÞÞe k z k 2 > ð1 þ aÞ& 0 k z k;
The temporal derivative of the Lyapunov candidate func- ð1 þ aÞ& 0
tion along the trajectories of the overall closed-loop system kzk >
ðl3  ð1 þ aÞð& 1 þ r& 2 ÞÞe
in equation (17), after simplifying terms can be written as
V_ ð~q ; x; vÞ q Tv
~ ¼~ ~ ~ q T Fð~ q_ Þ ~
q; ~ q  a 2 x T K vp x Thus, by denoting
v ~  2agx T v
~ T K vp v ~ þ ðv ~ T þ ax T Þ dðt; zÞ ð1 þ aÞ& 0
 :¼   ð36Þ
ð28Þ l3  ð1 þ aÞð& 1 þ r& 2 Þ e
The last term of equation (28) can be upper bounded as from equation (35) we can write
ðv T T
~ þ ax Þ d ðt; zÞ  ð1 þ aÞ k z kk dðt; zÞ k ð29Þ  
V_   l 3  ð1 þ aÞð& 1 þ r& 2 Þ ð1  eÞ k z k2 ; 8 k z k> 
where it has been taken into account that k v ~ kk z k,
ð37Þ
k x kk z k. The perturbation can be upper bounded as (see
equation (1U) in Appendix 1) Notice that the coefficient of k z k 2 in equation (37) is
negative, and the temporal derivative of the Lyapunov can-
k dðt; zÞ k & 0 þ & 1 k z k þ& 2 k z k 2 ð30Þ
didate function V_ is negative definite whenever
where & 0 , & 1 , and & 2 are positive constants (see equations
l 3 > ð1 þ aÞð& 1 þ r& 2 Þð1  eÞ ð38Þ
(1V) to (1X) in Appendix 1). By substituting equation (30)
in equation (29), we have is satisfied.
Salas et al. 9

Now we are ready to establish the main stability result in

the following.

Proposition 1. Let consider the dynamic model of a parallel

robot in equation (7), together with the control law given by
equations (9) to (11), where Fð~ q; q~_Þ is a diagonal matrix of
variable gains, and K vp and K vi are diagonal matrices of
constant gains defined in equations (12) to (13). Also, the
parameter k p satisfies equation (22). In addition, the para-
meter  in equation (36), where l3 was defined in equation
(27), & 0 , & 1 , and & 2 are defined in equations (1V) to (1X)
(see Appendix 1), satisfies
sffiffiffiffiffiffi
l1
< u
l2

where l 1 and l 2 were defined in equation (21), and u is the

radius of a ball D :¼ z 2 R 3n : k z k< u, inside which the
temporal derivative of the Lyapunov function in equation
(37) is negative definite. Moreover, l 3 satisfies equation Figure 5. Structure of the parallel robot 3-RRR.
(38). Then, the solutions zðtÞ of the closed-loop system in
equation (17) with initial state Simulations
sffiffiffiffiffiffi
l1 To test the effectiveness of the proposed control approach,
zðt 0 Þ < u ð39Þ we conducted numerical simulations of the proposed con-
l2
troller applied to the model of a planar parallel 3-revolute-
are uniformly ultimately bounded with ultimate bound revolute-revolute (RRR) joints robot. In the following
given by subsection, the matrices of the dynamic model of the par-
sffiffiffiffiffiffi allel 3-RRR robot are presented. Next, the desired trajec-
l2 tory for the end-effector of the robot is described. Finally,
k z k ; 8 t > t 0 þ T
l1 the simulations conducted are described.

with T  0. A Lyapunov function to demonstrate this is

equation (19).
Proof. First, note in equation (20) that the Lyapunov candi-
date function V is upper and lower bounded by positive
definite functions of k z k. Moreover, the time derivative
of V along the trajectories of the closed-loop system is
upper bounded by a negative definite function of k z k (see
equation (37)), for all k z k  where  is a positive para-
meter. By invoking the Theorem 4.18 in Khalil’s study32
and recalling the described functions and parameters, the
Proposition 1 is proven.
pffiffiffiffiffiffiffiffiffiffiffiffi
Remark 1. For any initial state zðt 0 Þ < l 1 =l 2 u (inside the
ball D) in equation (39), the trajectories of the solution zðtÞ
of the system will stay below the uniform ultimate bound
pffiffiffiffiffiffiffiffiffiffiffiffi
l 2 =l1 , after a finite time t 0 þ T . This bound can be
reduced by adjusting the bounds of the perturbation, which
depends on the bounds of the desired trajectory and on the
bounds of the variable gains (see equations (30) and equa-
tions (1V) to (1X) in Appendix 1).
Remark 2. Notice that the radius u of the ball D is not necessarily
equal to the radius r of the ball Br in equation (33). A further
analysis of this can be consulted in the study by Salas et al.33
Kinematic structure of the parallel 3-RRR robot
The kinematic structure of the robot is shown in Figure 5.
This structure is designated as 3-RRR, which represents a
robot with three legs or kinematic chains, each having
three rotational joints, from which the joint to the fixed
platform is actuated.33 The joints to the fixed platform are
located in the dots labeled A1 , A 2 , and A 3 . The mobile
platform, which is considered the end-effector of the
robot, is a triangular equilateral platform with
P ¼ ðxp ; yp Þ as its geometric center.
The equation of the dynamic model of the robot used for
simulations is equation (1). The nonzero elements of the
matrix MðrÞ 2 R 99 are29
M 11 ¼ M 22 ¼ M 33 ¼ s 1 ;
M 44 ¼ M 55 ¼ M 66 ¼ s 2 ;
M 77 ¼ M 88 ¼ M 99 ¼ mp ;
M 14 ¼ M 41 ¼ s 3 cosðq1  b 1 Þ;
M 25 ¼ M 52 ¼ s 3 cosðq2  b 2 Þ;
M 36 ¼ M 63 ¼ s 3 cosðq3  b 3 Þ
2
where s 1 ¼ m 1 lc1 þ m 2 l 21 þ I 1 , s 2 ¼ m 2 lc2
2
, s 3 ¼ m2 l 1 lc2 .
10 International Journal of Advanced Robotic Systems

_ 2 R 99 are
The nonzero elements of the matrix Cðr; rÞ Table 2. Dynamic parameters of the model.

C 14 ¼ s 4 sinðq 1  b 1 Þq_ 1 ; Parameter Value Units

C 41 ¼ s 4 sinðq  b 1 Þb_ 1 ;
1 m1 2.1992 kg
C 25 ¼ s 4 sinðq 2  b 2 Þq_ 2 ; m2 2.0485 kg
mp 5.8579 kg
C 52 ¼ s 4 sinðq  b 2 Þb_ 2 ;
2 I1 0.0264 kg m2
C 36 ¼ s 4 sinðq 3  b 3 Þq_ 3 ; I2 0.0228 kg m2
Ip 0.2504 kg m2
C 63 ¼ s 4 sinðq3  b 3 Þb_ 3 l 0.265 m
where s 4 ¼ m 2 l 1 lc2 . lc1 0.1365 m
lc2 0.1463 m
Since the robot workspace is limited to the horizontal r 0.2887 m
plane, it can be considered that the robot is not affected by
gravity forces. The nonzero elements of the matrix
DðrÞ ¼ @gðrÞ=@r 2 Rð6Þð9Þ are  pffiffiffi
L cos 3L sin
E 21 ¼ 2l xp  þ ð41Þ
D 11 ¼ l 1 sinðq1 Þ; 2 6
D 14 ¼ l 2 sinðb 1 Þ; 0 pffiffiffi 12
D 19 ¼ r sinð þ 1 Þ; L cos 3 L sin
E 31 ¼ @xp  þ A
2 6
D21 ¼ l 1 cosðq1 Þ;
0 12 ð42Þ
D 24 ¼ l 2 cosðb 1 Þ; pffiffiffi
L sin 3L cosA
D 29 ¼ r cosð þ  1 Þ; þ @y p  
2 6
D32 ¼ l 1 sinðq 2 Þ;
 pffiffiffi
D35 ¼ l 2 sinðb 2 Þ; L sin 3L cos
E 12 ¼ 2l yp þ  ð43Þ
D 39 ¼ r sinð þ 2 Þ; 2 6
D 42 ¼ l 1 cosðq 2 Þ;  pffiffiffi
L cos 3L sin
D 45 ¼ l 2 cosðb 2 Þ; E 22 ¼ 2l xp þ þ R ð44Þ
2 6
D 49 ¼ r cosð þ  2 Þ;
 pffiffiffi 2
D 53 ¼ l 1 sinðq3 Þ; L cos 3L sin
E 32 ¼ xp þ þ R
D56 ¼ l 2 sinðb 3 Þ; 2 6
 pffiffiffi 2 ð45Þ
D 59 ¼ r sinð þ 3 Þ; L sin 3L cos
þ yp þ 
D63 ¼ l 1 cosðq3 Þ; 2 6
D 66 ¼ l 2 cosðb 3 Þ;  pffiffiffi pffiffiffi
3L cos 3
D69 ¼ r cosð þ 3 Þ E 13 ¼ 2l yp þ  R ð46Þ
3 2
where  1 ¼ p=6,  2 ¼ 5p=6, and 1 ¼ 3p=2 [radians].  pffiffiffi
3L sin 1
The vector of Lagrange multipliers l 2 R 6 can be com- E 23 ¼ 2l xp   R ð47Þ
puted by the expression 3 2
 pffiffiffi 2
l ¼ ½DðrÞM 1 ðrÞD T ðrÞ1 3L sin 1
E 33 ¼ xp   R
 ½DðrÞM 1 ðrÞðCðr; rÞ _
_ r_  tÞ  Dðr; _ r
rÞ _ 3 2
 pffiffiffi pffiffiffi 2 ð48Þ
(see Garcia-Gamez et al.’s work29). 3L sin 3
þ yp þ  R
The parameters of the dynamic model are shown in 3 2
Table 2. The inverse kinematic model34 is with L ¼ 0:5 m as the length of the side of the mobile
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi platform, and R ¼ 0:62 m as the length of the fixed plat-
E 1i + E 21i þ E 22i  E 23i form, from A1 to A2 (see Figure 5).
qi ¼ 2 tan1 ; i ¼ 1; 2; 3
E 3i  E 2i
where
Desired trajectory
 pffiffiffi The desired trajectory for the end-effector of the manipu-
L sin 3L cos lator is an elliptical trajectory in operational space is
E 11 ¼ 2l yp   ð40Þ
2 6 given by
Salas et al. 11

xd ¼ xh þ U x cos ðwtÞ Table 3. Gains of the PI control loop.

yd ¼ yh þ U y sin ðwtÞ Parameter Value Units

where xh and yh are the initial position coordinates of the Kvp diagf5; 10; 8g Nm s=rad
end-effector. The desired trajectory in joint space is Kvi diagf50; 20; 70g Nm=rad
obtained by computing the inverse kinematic model given PI: proportional integral.
in equations (40) to (48). The values of the initial position
used in simulations are: xh ¼ 0:31 m and yh ¼ 0:1789 m.
Table 4. Gains of the PID controller.
The initial value of the orientation of the end-effector is
h ¼ 1:4811 rad, which is a constant value for the whole Parameter Value Units
trajectory. This initial position yields the initial joint
Kp diag f9, 50, 450g Nm/rad
position vectors qh ¼ ½0:4033 2:4984  1:6903 T and Ki diag f10, 10, 10g Nm/rad s
bh ¼ ½0:7835 1:3101  2:8781 T rad. The amplitudes Kd diag f40, 70, 10g Nm s/rad
used in simulations are U x ¼ 0:08 m and U y ¼ 0:06 m.
PID: proportional integral derivative.
The trajectory must be completed in a period of time
of 8 s. The coordinates of the joints A 1 , A 2 , and A 3
(see Figure 5) are A1 ¼ ½0 0, A2 ¼ ½0:62 0, and A3 ¼ Table 5. Partitions of the fuzzy membership functions of the FPD
controller.
½0:31 0:5369 m.
An important issue regarding the motion of parallel Joint k1p k 2p k3p k1d k 2d k3d
manipulators is the avoidance of singular configurations.
1 9.5 90 200 3.7 16.1 60
These configurations exist in the workspace of the manip-
2 9.5 90 200 3.7 16.1 60
ulators and are represented by the mathematical singularity 3 9.5 90 200 0.61 2.83 5.29
of the jacobian matrices of the differential kinematic Units [Nm/rad] [Nm s/rad]
model. It is well known that passing or approaching a sin-
gular configuration in the trajectory of the robot can result FPD: fuzzy-tuned proportional derivative.
in high joint velocities and important uprising of the
mechanical load, including an eventual loss of the control The selected gains of the inner loop PI control loop are
of the mechanism.35 There exists, in the literature of motion shown in Table 3. These gains were selected by a trial and
of parallel robot, several approaches for singularity avoid- error process to avoid exceeding the upper bound of the
ance, from trajectory planning to redundancy schemes.36,37 actuators, particularly at the beginning of the test. For com-
In this work, we address this problem by designing a parison purposes, simulations of two more controllers: a
desired trajectory that is free of singular configurations. classic PID controller and a FPD controller applied to the
With that objective in mind, we compute the reciprocal 3-RRR parallel robot model were performed. The gains of
of the condition number G 1 ðLÞ ¼ sl =sL , as is proposed the PID controller were chosen by means of a trial and error
by Alba,37 where sl and sL denote the minimum and max- procedure, to obtain the best possible performance from the
imum singular values of a jacobian matrix L, respectively. robot. Moreover, we were careful in avoiding to exceed the
Notice that a G 1 ðLÞ ¼ 0 corresponds to a singular con- upper bound of the nominal torque of the actuators at the
figuration. The jacobian matrix computed is the matrix beginning of the test. These gains are shown in Table 4.
J x ðrÞ of the differential kinematic model x_ ¼ J x ðrÞq.
_ As The FPD controller is the proposed by Salas et al.,33
1
a result of the computing of G ðJ x ðrÞÞ along the desired which is inspired on the work by Llama et al.38 Both gains
trajectory, the minimum computed value is 0.1117, which P and D are tuned by a fuzzy inference engine. Three input
indicates that it never reaches a zero value. fuzzy membership functions and three output fuzzy mem-
bership functions are used. The partitions of these functions
are shown in Table 5. Similar to the work reported by Salas
Simulations et al.,33 the inference mechanism is supported on three
A practical issue concerning the actuators of a robot is to fuzzy inference rules
avoid saturating the actuators with torque control refer-
ences that exceed the rated values, particularly at the begin-  If S ðj~qi jÞ then B ðyÞ
ning of the test. In this work, to avoid actuator saturation in  If M ðj~qi jÞ then M ðyÞ
a similar way to real-time experiments, we established an  If B ðj~qi jÞ then S ðyÞ
upper bound for the demanded torques from each actuator.
This upper bound prevent the controllers to produce a tor- where S ðj~qi jÞ, M ðj~qi jÞ, and M ðj~qi jÞ represent the mem-
que control signal that a real actuator cannot accomplish in bership grades of the input variable ~qi to the fuzzy sets
real-time experiments. The upper bound for the torque con- represented by linguistic variables Small (S), Medium
trol signals was set to 4 Nm. (M), and Big (B). On the other hand, B ðyÞ, M ðyÞ, and
12 International Journal of Advanced Robotic Systems

0.6 0.02
SOF-PPI SOF-PPI
FPD FPD
0.4 PID 0.01 PID

0.2 0

q̃3 [rad]
q̃1 [rad]

-0.01
0

-0.02
-0.2

-0.03
-0.4 0 1 2 3 4 5 6 7 8
0 1 2 3 4 5 6 7 8
t [s] t [s]

Figure 6. Position errors in joint 1. SOF: self-organizing fuzzy; Figure 8. Position errors in joint 3. SOF: self-organizing fuzzy;
FPD: fuzzy-tuned proportional derivative; PID: proportional FPD: fuzzy-tuned proportional derivative; PID: proportional
integral derivative; P-PI: proportional–proportional integral. integral derivative; P-PI: proportional–proportional integral.

0.2 5
τ 1 [Nm]

SOF-PPI
FPD 0
0.1 PID
−5
0 1 2 3 4 5 6 7 8
0 5
q̃2 [rad]

τ 2 [Nm]

0
-0.1

−5
0 1 2 3 4 5 6 7 8
-0.2
5
τ 3 [Nm]

-0.3 0
0 1 2 3 4 5 6 7 8

t [s] −5
0 1 2 3 4 5 6 7 8
t [s]
Figure 7. Position errors in joint 2. SOF: self-organizing
fuzzy; FPD: fuzzy-tuned proportional derivative; PID: pro-
portional integral derivative; P-PI: proportional–proportional Figure 9. Commanded torques by SOF-P-PI. SOF: self-organizing
integral. fuzzy; P-PI: proportional–proportional integral.

respectively, while the largest tracking error peaks obtained

S ðyÞ represent the fuzzy membership grades to obtain the
crisp variable from the output membership functions.
The defuzzification is carried out by a simplified version
of the Center-of-Gravity method,38 which can be written as
k T ðj~
qi jÞ
y¼
k ðj~
qi jÞ k
where k ¼ ½k 1 k 2 k 3  T is the output partition.
Simulation results are shown in Figures 6 to 12. In
Figures 6 to 8, the errors in joints 1 to 3, respectively,
obtained by the three controllers, are depicted. The largest
tracking error peaks obtained by the PID controller were
0.2, 0.18, and 0.0055 rad for the joints 1, 2, and 3,
by the SOF-P-PI controller were 0.01, 0.01, and 0.005 rad,
for the joints 1, 2, and 3, respectively. The largest tracking
error peaks obtained by the FPD were 0.06, 0.02, and
0.0053 rad, for the joints 1, 2, and 3, respectively. It can
be noticed in Figures 6 to 7 that both errors produced by the
PID controller in joints 1 and 2 have much larger over-
shooting and settling times that the errors produced by the
SOF-P-PI controller. On the other hand, the overshooting
and settling times produced by the FPD controller are only
a little larger than in the SOF-P-PI controller. In addition, it
can be noticed in Figure 8 that although the overshooting
response in the three controllers for the joint 3 are similar,
the steady-state error of the SOF-P-PI controller is smaller
Salas et al. 13

20 100

Φ11 [s− 1 ]
τ 1 [Nm]

0 50

−20 0
0 1 2 3 4 5 6 7 8 0 1 2 3 4 5 6 7 8
1 100

Φ22 [s− 1 ]
τ 2 [Nm]

0 50

−1 0
0 1 2 3 4 5 6 7 8 0 1 2 3 4 5 6 7 8
5 100

Φ33 [s− 1 ]
τ 3 [Nm]

0 50

−5 0
0 1 2 3 4 5 6 7 8 0 1 2 3 4 5 6 7 8
t [s] t [s]

Figure 10. Commanded torques by PID. PID: proportional Figure 12. Evolution of P gains by SOF algorithm. P: propor-
integral derivative. tional; SOF: self-organizing fuzzy.

k p3 [Nm/rad] k p2 [Nm/rad] k p1 [Nm/rad]

5
200
τ 1 [Nm]

0 100

-5 0
0 1 2 3 4 5 6 7 8 0 1 2 3 4 5 6 7 8

5 200
τ 2 [Nm]

0 100
0
-5 0 1 2 3 4 5 6 7 8
0 1 2 3 4 5 6 7 8
5 200
τ 3 [Nm]

0 100
0
-5 0 1 2 3 4 5 6 7 8
0 1 2 3 4 5 6 7 8
t [s] t [s]

Figure 11. Commanded torques by FPD. FPD: fuzzy-tuned Figure 13. Evolution of P gains by FPD. P: proportional;
proportional derivative. FPD: fuzzy-tuned proportional derivative.

than the steady-state error of the PID and the FPD control-
lers. In Figures 9 to 11 the demanded torques by the SOF-P-
PI, the PID, and the FPD controllers, respectively, are
shown. Observe in these figures that, in the entire simula-
tion time, the torques are below the predefined value for
maximum torques. In Figure 12, the temporal evolution of
the variable P gains in the outer loop are shown. This tem-
poral evolution exhibits the increase of the gain values
according to the adopted criteria in the SOF algorithm, to
obtain small tracking errors. Also, in Figures 13 and 14, the
temporal evolution of the P and D gains, respectively, of
the FPD controller are shown.
A useful quantitative measurement of the performance
of a controller can be done by computing the L2 norm of the
position or tracking errors. The L 2 norm can be obtained
with the formula
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ðT
1
k q~ kL 2 ¼ q~ T q~ dt
T  t0 t0
Here, T is the computed time period and t 0 is the initial
computing time. The total computing time considered
must be the period of steady state behavior, this is, it must
start after the transient behavior has passed. The result of
computing the L 2 norm of the position errors from the
three controllers, SOF-P-PI, PID, and FPD is shown in
Table 6. The period of steady state behavior is from
2.5 s to 8 s.
14 International Journal of Advanced Robotic Systems

measurement result was obtained that shows the better

k v 2 [Nms/rad] k v 1 [Nms/rad]

100 results of the SOF-P-PI controller.

50
Declaration of conflicting interests
0
0 1 2 3 4 5 6 7 8 The author(s) declared no potential conflicts of interest with respect
100 to the research, authorship, and/or publication of this article.

50 Funding
0 The author(s) disclosed receipt of the following financial support
0 1 2 3 4 5 6 7 8 for the research, authorship, and/or publication of this article: This
k v 3 [Nms/rad]

6 work was partially supported by PRODEP/SEP Mexico.

4
2 Francisco G Salas
0 1 2 3 4 5
t [s]
Figure 14. Evolution of D gains by FPD. D: derivative; FPD: fuzzy-
tuned proportional derivative.
Table 6. L 2 norm of q~ from the three controllers.
SOF-P-PI PID
0.0022 0.1107
SOF: self-organizing fuzzy; PID: proportional integral derivative; FPD:
fuzzy-tuned proportional derivative; P-PI: proportional–proportional
integral.
Conclusion
In this article, we have presented a novel two-loop P-PI
tracking controller for parallel manipulators, in which the
gains of the outer loop are online tuned by an original SOF
algorithm. This controller is a not model-based scheme
which improves the robustness in the presence of perturba-
tions and parameter uncertainty of the constant gains two-
loop approach for tracking control, by implementing a
variable gains loop.
The stability analysis of the closed-loop system has been
developed, resulting in the uniform ultimate boundedness
of the solutions of the closed-loop system. It is worth to
mention that this result can be generalized not only to P-PI
tracking controllers with variable gains tuned by an SOF
algorithm but also to other variable gains P-PI tracking
control schemes for parallel robots, only providing that the
gains produced by the tuning algorithm satisfy the bound-
edness criterion. This is an original result not published
elsewhere.
The usefulness of the proposed controller was validated
by performing numerical simulations of the application of
the controller to a parallel manipulator. In addition, simula-
tions of a PID controller and a FPD controller applied to the
parallel manipulator have been conducted for comparison
purposes. The simulation results of the three controllers
were contrasted, from which a significant quantitative
ORCID iD
http://orcid.org/0000-0001-9870-9760
6 7 8
References
1. Barnfather J, Goodfellow M, and Abram T. Positional capa-
bility of a Hexapod robot for machining applications. Int J
Adv Manuf Tech 2017; 89: 1101–1111.
2. Kelaiaia R. Improving the pose accuracy of the delta robot in
machining operations. Int J Adv Manuf Tech 2017; 91:
2205–2215.
FPD
3. Huapeng W, Heikki H, and Pekka P. Mobile parallel robot for
0.0245 assembly and repair of ITER vacuum vessel. Ind Robot 2008;
35(2): 160–168.
4. Qinchuan L, Weifeng W, Ji’nan X, et al. A hybrid robot for
friction stir welding. Proc Instit Mech Eng C J Mec Eng Sci
2015; 229(14): 2639–2650.
5. Fugui X and Xin-Jun L. Analysis of the kinematic character-
istics of a high-speed parallel robot with Schönflies motion:
Mobility, kinematics and singularity. Appl Mech Mater 2016;
11(2): 135–143.
6. Pierrot F, Reynaud C, and Fournier A. Delta: a simple and
efficient parallel robot. Robotica 1990; 8(02): 105–109.
7. Huang Z and Cao Y. Property identification of the singularity
loci of a class of Gough-Stewart manipulators. Int J Robot
Res 2005; 24(8): 675–685.
8. Nan R, Li D, Chengjin J, et al. The five-hundred-meter aper-
ture spherical radio telescope (FAST) project. Int J Mod Phys
D 2011; 20: 989–1024.
9. Enferadi J and Shahi A. On the position analysis of a new
spherical parallel robot with orientation applications. Robot
Cim Int Manuf 2016; 37: 151–161.
10. Siciliano B, Sciavicco L, Villani L, et al. Robotics. London:
Springer-Verlag, 2009.
11. Ghorbel FH, Chételat O, Gunawardana R, et al. Modeling and
set-point control of closed chain mechanisms: Theory and
experiment. IEEE Trans Cont Sys Technol 2000; 8: 801–815.
12. Ren L, Mills JK, and Sun D. Experimental comparison
of control approaches on trajectory tracking control of a
3-DOF parallel robot. IEEE Trans Cont Sys 2007; 15(5):
982–988.
13. Borbonnais F, Bigras P, and Bonev IA. Minimum-time tra-
jectory, planning and control of a pick-and-place five-bar
parallel robot. IEEE/ASME Trans Mechatron 2015; 20(2):
740–749.
Salas et al.
14. Ren Q and Bigras P. Discrete-time parallel robot motion
control using adaptive neuro-fuzzy inference system based
on improved substractive clustering. In: IEEE international
conference on fuzzy systems, Vancouver, Canada, 24–29 July
2016, pp. 1000–1006. IEEE.
15. Salinas A, Moreno-Valenzuela J, and Kelly R. A family of
nonlinear PID-like regulators for a class of torque-driven
robot manipulators equipped with torque-constrained actua-
tors. Adv Mech Eng 2016; 8(2): 1–14.
16. Diaz-Rodriguez M, Valera A, Mata V, et al. Model-based
control of a 3-DOF parallel robot based on identified relevant
parameters. IEEE/ASME Trans Mechatron 2013; 18(6):
1737–1744.
17. Kelly R and Moreno J. Manipulator motion control in opera-
tional space using joint velocity inner loops. Automatica
2005; 41: 1423–1432.
18. Corke P. The Unimation Puma servo system. Technical
Report Report MTM-226, CSIRO Division of Manufacturing
Technology, Preston, Australia, 1994.
19. Nilsson K. Industrial Robot Programming. PhD Thesis, Lund
Institute of Technology, Department of Automatic Control,
Sweden, 1996.
20. Soto I and Campa R. Two-loop control of redundant manip-
ulators: analysis and experiments on a 3-DOF planar arm. Int
J Adv Robot Syst 2013; 10: 1–8.
21. Campa R, Bernal J, and Soto I. Kinematic modeling and
control of the Hexapod parallel robot. In: American control
conference, Boston, MA, USA, 6–8 July 2016, pp.
1203–1208. American Automatic Control Council.
22. Villareal-Cervantes MG and Alvarez-Gallegos J. Off-line
PID control tuning for a planar parallel robot using DE var-
iants. Expert Syst Appl 2016; 64: 444–454.
23. Ouyang P, Zhang W, and Wu F. Nonlinear PD control for
trajectory tracking with consideration on the design for con-
trol methodology. In: Proceedings of the IEEE international
conference on robotics and automation, Washington DC,
11–15 May 2002, pp. 4126–4131. IEEE.
24. Zhu C and Zhang H. Parallel robotics control strategy study
based on Fuzzy-PID. In: Huang DS, Gan Y, Bevilacqua V,
et al. (eds) Lecture notes in computer science, advanced intel-
ligent computing ICIC, Vol 6838, Berlin: Springer, 2011,
p. 213.
25. Stan S, Balan R, Matie V, et al. Kinematics and fuzzy control
of ISOGLIDE3 medical parallel robot. Mechanika 2009; 75:
62–65.
26. Sang D and Han C. Fuzzy PID control of six degrees of
freedom parallel manipulators in electro hydraulic servo sys-
tem. In: Xing S, Chen S, Wei Z, et al. (eds) Lecture notes in
electrical engineering, unifying electrical engineering and
electronics engineering, vol 238. New York: Springer,
2014, pp. 1971–1981.
27. Merlet J. Parallel robots, 2nd ed. The Netherlands: Springer,
2006.
28. Muller A. Internal preload control of redundantly actuated
parallel manipulators—its application to backlash avoiding
control. IEEE Trans Robot 2005; 21: 668–677.
15
29. Garcia-Gamez C, Campa R, and Santibanez V. Stability anal-
ysis of conventional controllers in parallel manipulators: the
3-RRR as case of study. In: XI Congreso Mexicano de Robót-
ica 2009 (in spanish). Celaya, Gto, México: Asociación Mex-
icana de Robótica.
30. Khalil W and Dombre E. Modeling, identification and control
of robots. London: Kogan Page Science, 2002.
31. Wang LX. A course in fuzzy systems and control. Upper
Saddle River: Prentice-Hall, 1997.
32. Khalil H. Nonlinear systems, 3rd ed. New Jersey: Prentice-
Hall, 2002. p. 172.
33. Salas F, Santibanez V, and Llama M. Fuzzy-tuned PD track-
ing control of a 3-RRR parallel manipulator: stability analysis
and simulations. Intell Autom Soft Co 2014; 20: 159–182.
34. Jun W, Jinsung W, and Zheng Y. A comparison study on the
dynamics of planar 3-DOF 4-RRR, 3-RRR and 2-RRR par-
allel manipulators. Robot Cim Int Manuf 2011; 27: 150–156.
35. Tsai L. Robot analysis: the mechanics of serial and parallel
manipulators. New York: John Wiley and Sons, 1999.
36. Alba O, Wenger P, and Pamanes J. Performance indices for
kinematically redundant parallel planar manipulators. Prob
Mech 2006; 1: 22–40.
37. Alba O. Optimization of trajectories for redundant parallel
robots. PhD Thesis, Instituto Tecnologico de La Laguna,
Torreon, Mexico, 2007. In Spanish.
38. Llama M, Kelly R, and Santibanez V. Stable computed torque
control of robot manipulators via fuzzy self-tuning. IEEE
Trans Syst Man Cybern B 2000; 30: 143–150.
Appendix 1
In this appendix, we obtain the upper bound of the pertur-
bation of the system.
Bounds over the perturbation
Next we will find the upper bounds of each term of the
perturbation in equation (16)
_ d þ av
dðt; zÞ ¼ MðqÞ½v ~ þ Cðq; qÞ½v
_ d þ ax þ gðqÞ
_
Bounds over Cðq; qÞ½vd þ ax
_
Cðq; qÞ½v d þ ax  k C k q
_ kk vd þ ax k ð1AÞ
where Property 3 has been used. By bounding the velocity q_
in equation (9) we have
q_ k vd k þ k v
~ k ð1BÞ
Substituting equation (1B) in equation (1A) we have
k Cðq; qÞ½v
_ d þ ax k
 k C ½k vd k þ k v
~ k ½k vd k þa k x k
 k C ½k vd k þ k z k ½k vd k þa k z k
 k C ½k vd k2 þ ða þ 1Þ k vd kk z k þa k zk2 
ð1CÞ
16 International Journal of Advanced Robotic Systems

where it has been taken into account that k q~ kk z k and which can be upper bounded as
k x k k z k. vd can be upper bounded as 
vd  k qd kM þ lFu k q ~k k q~_ k ðlFu þ 1Þ k z k ð1L
ð1DÞ
 k q_ d kM þ lFu k z k
where it has been taken account that k v
~ kk z k. Substi-
where lFu has been defined as tuting equation (1L) in equation (1K) yields
 
_ d k €qd kM þ lFru k z k þlFu ðlFu þ 1Þ k z k
v ð1MÞ
lFu ¼ sup l max fFð~q; q~_Þg
q~;q~_2Rn
Substituting equation (1M) in equation (1J) yields
Substituting equation (1D) in equation (1C) yields _ d þ av
MðqÞ½v ~
k Cðq; qÞ½v
_ d þ ax k  l max fMg½k €q d kM þ lFru k z k þlFu ðlFu þ 1Þ k z k
 k C ½½k q_ d kM þ lFu k z k 2 þakv ~ k
þ ða þ 1Þ½k q_ d kM þ lFu k z k k z k þa k zk2 
   l max fMg k q€ d kM
2
 k C lFu þ k C ða þ 1ÞlFu þ a k zk2
  þ l max fMg½lFru þ lFu ðlFu þ 1Þ þ a k z k
2
þ 2k C lFu k q_ d kM þ k C ða þ 1Þ k q_ d kM k z k þ k C k q_ d kM ð1NÞ
ð1EÞ By defining
By defining z 4 :¼ l max fMg k €qd kM ð1OÞ
2
z 1 :¼ k C k q_ d kM ð1FÞ z 5 :¼ l max fMg½lFru þ lFu ðlFu þ 1Þ þ a ð1PÞ
z 2 :¼ 2k C lFu k q_ d kM þ k C ða þ 1Þ k q_ d kM ð1GÞ equation (1N) can be written as
2
z 3 :¼ k C lFu þ k C ða þ 1ÞlFu þ a ð1HÞ _ d þ av
MðqÞ½v ~  z4 þ z5 k z k ð1QÞ
equation (1E) can be written as Bound over gðqÞ gðqÞ can be upper bounded as
2
k Cðq; qÞ½v
_ d þ ax k z 1 þ z 2 k z k þz 3 k zk ð1IÞ k gðqÞ k k 0 ð1RÞ
_ d þ av
Bound over MðqÞ½v ~ where we have used the Property 4. By defining
_ d þ av
MðqÞ½v ~  l max fMg k v_ d þ av
~ k z 6 :¼ k 0 ð1SÞ
ð1JÞ
 l max fMg½k v
_ d k þa k v
~ k
equation (1R) can be written as:
By recalling that the temporal derivative of vd in equa-
k gðqÞ k z 6 ð1TÞ
tion (8) is
_ q; q~_Þ q~ þ Fð~ Bounds over the perturbation. By combining equations
_d ¼€
v qd þ Fð~ q; q~_Þ q~_
(1I), (1Q), and (1T), and substituting in equation (16), we
_ d can be upper bounded as
v obtain the upper bound of the perturbation

qd kM þ lFru k q~ k þlFu k q~_ k

_d kk€
kv ð1KÞ k dðt; zÞ k  & 0 þ & 1 k z k þ & 2 k z k2 ð1UÞ

where lFru has been defined as where the coefficients have been defined as (see equations
  (1F) to (1H), (1O), (1P), (1S))
lFru ¼ sup l max fFð~_ q; q~_Þg
q~;q~_2Rn
& 0 :¼ z 1 þ z 4 þ z 6 ð1VÞ

From equation (9), we can write & 1 :¼ z 2 þ z 5 ð1WÞ

q~_ ¼ v q; q~_Þ q~
~  Fð~ & 2 :¼ z 3 ð1XÞ