Professional Documents
Culture Documents
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
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
pffiffiffiffiffiffiffiffiffi
this work, we propose a novel design of a two-loop P-PI capital letters, for example, A. k x k¼ x T x represents the
controller for tracking control of parallel manipulators. The Euclidean norm of vector x. l max fAg and l min fAg repre-
outer loop is a joint position P control loop, and the inner sent the largest and smallest eigenvalues of matrix A,
loop is a joint velocity PI control loop. The gains of the respectively.
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 Dynamic model of the robot
the gain depending on the current tracking errors. This
algorithm employs a set of fuzzy rules that can be updated The joints of a parallel manipulator can be classified as n
online according to a performance index of the controller actuated joints, represented by vector q 2 Rn , and m non-
response. This gain tuning scheme provides the needed actuated joints, represented by vector b 2 Rm . Moreover,
increased robustness in presence of perturbations and para- the l operational or Cartesian variables of the end-effector
meter uncertainty of the overall controller, as well as can be represented with vector x 2 Rl . Let us define the
improved results when it is tracking fast changing trajec- vector of constrained variables r as
tories. The usefulness and performance of the proposed
r ¼ ½qT bT xT T 2 Rs
controller is tested by numerical simulations, in compari-
son with two controllers, a PID controller, and a fuzzy- with s ¼ n þ m þ l. By applying the Euler–Lagrange
tuned PD (FPD) controller. Moreover, the stability analysis formulation, the dynamic model of a parallel robot27
of the closed-loop system is carried out, giving as a result in absence of friction is in general formulated as
the proof of the Uniform Ultimate Boundedness of the
solutions of the closed-loop system. M 0 ðrÞ€
r þ C 0 ðr; rÞ
_ r_ þ g0 ðrÞ ¼ tr þ D T ðrÞl ð1Þ
The main contributions of this work are the design and where M 0 ðrÞ 2 Rss is the symmetric positive definite
the stability analysis of a tracking control structure based _ 2 Rss is the Coriolis and centripe-
inertia matrix, C 0 ðr; rÞ
on a two-loop scheme composed of two not model-based tal forces matrix, g ðrÞ 2 Rs is the vector of gravitational
0
controllers, that is, P and PI. In addition, the gains of the P forces, tr 2 Rs is the vector of generalized forces associated
control loop can be tuned online. Moreover, the P-gain tun-
to scalar variables of r, DðrÞ ¼ @gðrÞ=@r 2 RðkÞðnÞðsÞ is
ing is performed by a novel SOF algorithm. Furthermore, the
the Jacobian matrix of the system holonomic constraints
stability analysis of the resulting closed-loop system is car-
ried out. It is worth to mention that the result of the stability gðrÞ 2 RðkÞðnÞ , with k as the minimum number of coordinates
analysis holds for any tuning algorithm for tuning the P in the space that describes the position of the end-effector,
gains, by only ensuring that some conditions on the bound- and l 2 RðkÞðnÞ is the vector of Lagrange multipliers.
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. Kinematic transformation
According to the best knowledge of the authors, this is an
Let us define a kinematic transformation between vectors r_
original result not published before elsewhere.
and q_ as
The remaining of this article is organized as follows. In the
“Preliminaries” section, the dynamic model of the parallel r_ ¼ RðrÞq_ ð2Þ
robot as well as some of their properties useful for the stability
to obtain a so-called reduced dynamic model. Notice that
analysis are described. In the section “Proposed controller,”
the full description of the proposed controller is given. In the matrix RðrÞ 2 RðsÞðnÞ can be constructed with the
section “Stability analysis,” the development of the stability Jacobian matrices J b ðrÞ and J x ðrÞ from the differential
analysis in the sense of Lyapunov of the closed-loop system is kinematic models b_ ¼ J b ðrÞq_ and x_ ¼ J x ðrÞq,
_ as
carried out, and the result of this analysis is presented. In 2 3
I
section “Simulations,” the elements of the dynamic and kine- 6 7
matic model of a 3-DOF parallel robot are described as well as RðrÞ ¼ 4 J b ðrÞ 5; I 2 Rnn
the desired trajectory for simulations. Also, the numerical J x ðrÞ
simulations conducted are presented and discussed. Finally,
The derivative with respect to time of equation (2) is
some concluding remarks are given.
r _
€ ¼ RðrÞ q_ þ RðrÞ€q ð3Þ
equation (5) can be rewritten as Proof. For robots with only revolute joints, the entries of the
vector of gravitational forces g0 ðrÞ (see equation (1)) are
q þ Cðr; rÞ
MðrÞ € _ q_ þ gðrÞ ¼ t ð6Þ
sinusoidal or trigonometric functions with finite coeffi-
11 cients. Thus the entries are upper bounded. Now, lets con-
In Ghorbel et al.’s work, it is proven, by means of the
implicit function theorem, that there exists a unique para- sider the vector gðrÞ ¼ R T ðrÞ g0 ðrÞ. Note that the entries
metrization of r 2 Nr , inside a neighborhood Nr of this vector are upper bounded for all q 2 O Rn , where
O denotes a space of configurations of the robot in which
r ¼ hðqÞ
the matrix RðrÞ is well posed.
for any q 2 Nq inside a neighborhood Nq , whenever the
system is not in a singular configuration. Proposed controller
In addition, Muller28 established that, for a parallel
machine, a subset q of d joint variables determines its con- In this work, we propose a controller for trajectory tracking
figuration, in virtue of that there exists a smooth mapping of parallel manipulators. The structure of the proposed con-
that assigns to each r, the parallel machine configuration as troller is a two-loop cascade structure, with a P position
controller in the outer loop and a PI controller in the inner
r ¼ ðqÞ, where the map 1 is a local parametrization of
loop. The overall structure of the controller is shown in
the d dimensional manifold V
Figure 1. Note that the desired joint position and velocity
V ¼ fr 2 Vn ; gðrÞ ¼ 0g vectors qd and q_ d , respectively, are represented in the Fig-
ure 3, as well as the position error and velocity error vectors
where V represents the set of all admissible configurations
of the parallel machine, and gðrÞ represents the holonomic q~ ¼ qd q and q~_ ¼ q_ d q,
_ respectively. We define the
constraints. velocity control law of the outer loop as
In consequence, without loss of generality, we can write q; q~_Þ~
vd ¼ Fð~ q þ q_ d ð8Þ
down the matrices of the dynamic model MðrÞ; Cðr; rÞ, _
_ and gðqÞ, respectively. Thus, the
and gðrÞ as MðqÞ, Cðq; qÞ, q; q~_Þ is a diagonal matrix of variable gains
Notice that Fð~
dynamic model in equation (6) takes the form of n n dimension. The values of each diagonal element of
Salas et al. 5
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
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
_ 2 R 99 are
The nonzero elements of the matrix Cðr; rÞ Table 2. Dynamic parameters of the model.
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.
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.
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- position or tracking errors. The L 2 norm can be obtained
lers. In Figures 9 to 11 the demanded torques by the SOF-P- with the formula
PI, the PID, and the FPD controllers, respectively, are sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ðT
shown. Observe in these figures that, in the entire simula- 1
tion time, the torques are below the predefined value for k q~ kL 2 ¼ q~ T q~ dt
T t0 t0
maximum torques. In Figure 12, the temporal evolution of
the variable P gains in the outer loop are shown. This tem- Here, T is the computed time period and t 0 is the initial
poral evolution exhibits the increase of the gain values computing time. The total computing time considered
according to the adopted criteria in the SOF algorithm, to must be the period of steady state behavior, this is, it must
obtain small tracking errors. Also, in Figures 13 and 14, the start after the transient behavior has passed. The result of
temporal evolution of the P and D gains, respectively, of computing the L 2 norm of the position errors from the
the FPD controller are shown. three controllers, SOF-P-PI, PID, and FPD is shown in
A useful quantitative measurement of the performance Table 6. The period of steady state behavior is from
of a controller can be done by computing the L2 norm of the 2.5 s to 8 s.
14 International Journal of Advanced Robotic Systems
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]
14. Ren Q and Bigras P. Discrete-time parallel robot motion 29. Garcia-Gamez C, Campa R, and Santibanez V. Stability anal-
control using adaptive neuro-fuzzy inference system based ysis of conventional controllers in parallel manipulators: the
on improved substractive clustering. In: IEEE international 3-RRR as case of study. In: XI Congreso Mexicano de Robót-
conference on fuzzy systems, Vancouver, Canada, 24–29 July ica 2009 (in spanish). Celaya, Gto, México: Asociación Mex-
2016, pp. 1000–1006. IEEE. icana de Robótica.
15. Salinas A, Moreno-Valenzuela J, and Kelly R. A family of 30. Khalil W and Dombre E. Modeling, identification and control
nonlinear PID-like regulators for a class of torque-driven of robots. London: Kogan Page Science, 2002.
robot manipulators equipped with torque-constrained actua- 31. Wang LX. A course in fuzzy systems and control. Upper
tors. Adv Mech Eng 2016; 8(2): 1–14. Saddle River: Prentice-Hall, 1997.
16. Diaz-Rodriguez M, Valera A, Mata V, et al. Model-based 32. Khalil H. Nonlinear systems, 3rd ed. New Jersey: Prentice-
control of a 3-DOF parallel robot based on identified relevant Hall, 2002. p. 172.
parameters. IEEE/ASME Trans Mechatron 2013; 18(6): 33. Salas F, Santibanez V, and Llama M. Fuzzy-tuned PD track-
1737–1744. ing control of a 3-RRR parallel manipulator: stability analysis
17. Kelly R and Moreno J. Manipulator motion control in opera- and simulations. Intell Autom Soft Co 2014; 20: 159–182.
tional space using joint velocity inner loops. Automatica 34. Jun W, Jinsung W, and Zheng Y. A comparison study on the
2005; 41: 1423–1432. dynamics of planar 3-DOF 4-RRR, 3-RRR and 2-RRR par-
18. Corke P. The Unimation Puma servo system. Technical allel manipulators. Robot Cim Int Manuf 2011; 27: 150–156.
Report Report MTM-226, CSIRO Division of Manufacturing 35. Tsai L. Robot analysis: the mechanics of serial and parallel
Technology, Preston, Australia, 1994. manipulators. New York: John Wiley and Sons, 1999.
19. Nilsson K. Industrial Robot Programming. PhD Thesis, Lund 36. Alba O, Wenger P, and Pamanes J. Performance indices for
Institute of Technology, Department of Automatic Control, kinematically redundant parallel planar manipulators. Prob
Sweden, 1996. Mech 2006; 1: 22–40.
20. Soto I and Campa R. Two-loop control of redundant manip- 37. Alba O. Optimization of trajectories for redundant parallel
ulators: analysis and experiments on a 3-DOF planar arm. Int
robots. PhD Thesis, Instituto Tecnologico de La Laguna,
J Adv Robot Syst 2013; 10: 1–8.
Torreon, Mexico, 2007. In Spanish.
21. Campa R, Bernal J, and Soto I. Kinematic modeling and
38. Llama M, Kelly R, and Santibanez V. Stable computed torque
control of the Hexapod parallel robot. In: American control
control of robot manipulators via fuzzy self-tuning. IEEE
conference, Boston, MA, USA, 6–8 July 2016, pp.
Trans Syst Man Cybern B 2000; 30: 143–150.
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. Appendix 1
23. Ouyang P, Zhang W, and Wu F. Nonlinear PD control for
In this appendix, we obtain the upper bound of the pertur-
trajectory tracking with consideration on the design for con-
bation of the system.
trol methodology. In: Proceedings of the IEEE international
conference on robotics and automation, Washington DC, Bounds over the perturbation
11–15 May 2002, pp. 4126–4131. IEEE.
24. Zhu C and Zhang H. Parallel robotics control strategy study Next we will find the upper bounds of each term of the
based on Fuzzy-PID. In: Huang DS, Gan Y, Bevilacqua V, perturbation in equation (16)
et al. (eds) Lecture notes in computer science, advanced intel- _ d þ av
dðt; zÞ ¼ MðqÞ½v ~ þ Cðq; qÞ½v
_ d þ ax þ gðqÞ
ligent computing ICIC, Vol 6838, Berlin: Springer, 2011,
p. 213. _
Bounds over Cðq; qÞ½vd þ ax
25. Stan S, Balan R, Matie V, et al. Kinematics and fuzzy control _
Cðq; qÞ½v d þ ax k C k q
_ kk vd þ ax k ð1AÞ
of ISOGLIDE3 medical parallel robot. Mechanika 2009; 75:
62–65. where Property 3 has been used. By bounding the velocity q_
26. Sang D and Han C. Fuzzy PID control of six degrees of in equation (9) we have
freedom parallel manipulators in electro hydraulic servo sys-
q_ k vd k þ k v
~ k ð1BÞ
tem. In: Xing S, Chen S, Wei Z, et al. (eds) Lecture notes in
electrical engineering, unifying electrical engineering and Substituting equation (1B) in equation (1A) we have
electronics engineering, vol 238. New York: Springer,
2014, pp. 1971–1981.
k Cðq; qÞ½v
_ d þ ax k
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
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
q~_ ¼ v q; q~_Þ q~
~ Fð~ & 2 :¼ z 3 ð1XÞ