A Fuzzy Sliding Mode Controller for Planar

4-Cable Direct Driven Robot

Xavier Aguas ∗ Andres Cuaycal ∗ Israel Paredes ∗
Marco Herrera ∗
∗
Escuela Politécnica Nacional, Quito, Ecuador, e-mail:
{xavier.aguas,andres.cuaycal, israel.paredes, marco.herrera}
@epn.edu.ec

Abstract: Cable Direct Driven Robots (CDDRs) are a special class of parallel robots, these
are formed from replacing all the supporting rigid links by cables. Unlike the traditional robots,
these are good candidates for performing a wide range of potential applications. In this paper
the rotational move and moment resistance on the end-effector are not considered on Planar
4-Cable CDDR model, all 4 cables convene in a single point and the end-effector is modeled
as a point mass. The aim of this paper is to present a robust control using a Sliding Mode
Controller (SMC) with a Fuzzy-PI as sliding surface. Step change reference test and, tracking
trajectory test were performed in order to observe the behavior of the Planar 4-Cable CDDR
during the trajectory of the end-effector. In order to prove the effectiveness of the proposed
control law, simulation from MatLab carried out on a Planar 4-Cable CDDR and the results
were compared with a PI Controller and a conventional SMC in terms of integral square error
(ISE) performance index. Only the kinematic model of Planar 4-Cable CDDR is considered in
this paper.

Keywords: Cable Direct Driven Robot, Sliding Mode Control, Fuzzy Controller

1. INTRODUCTION implementing a human’s heuristic knowledge about how to

Robots have made formidable progress into industries
for manufacturing and assembly. Traditional robots with
serial or parallel structures are unsuitable since the
workspace requirements are higher as in (Oh and Agrawal,
2005) are presented. For these motives, cable-driven mech-
anisms have received attention and have been recently
studied. The advantages shown in (Zanotto, 2011), cables
and tendon-like components in robotics have been the
interest of many research in the last years. Cable-direct-
driven robots (CDDRs) are structurally similar to parallel
robots (Jin et al., 2013)(Williams et al., 2001) wherein
the end-effector is supported in parallel by n cables by n
motors. As compared to rigid links, cables are lighter and
can handle larger loads guaranteeing considerable ranges
of motion. However, one disadvantage is cables can only
exert tension and cannot push on the end-effector. This
property makes feedback control of CDDRs more defi-
ant than conventional parallel robot as in (Babaghasabha
et al., 2015), where an adaptive controller in task space
coordinates for a planar cable-driven parallel robot with
uncertainties in dynamic and kinematic parameter is de-
signed. In (Khosravi and Taghirad, 2014) a robust PID
controller is presented for the cable-driven robot to ensure
that all cables remain in tension. On the other hand,
sliding mode control (SMC) is a nonlinear technique with
robustness against the model uncertainties and ability
of the disturbance rejection, in (Mohammad and Ehsan,
2008) SMC for a robot manipulator in order to deleting
the oscillations of the response is presented. Fuzzy control
affords a methodology for representing, manipulating, and
control a system. Fuzzy logic presents the ability to imitate
the human mind to effectively occupy modes of reasoning
that are approximate rather than exact. Designing of a
Fuzzy Logic Controller (FLC) can show a lengthy process
when performed heuristically (Nabi, 2013). FLC is identi-
cal to a conventional PID controller (Ghosh et al., 2015).
In case of FLC, control strategics are expressed in terms
of fuzzy rules and this set of well defined rules is known as
fuzzy algorithm. Planar 4-Cable CDDR mode (Williams II
and Gallina, 2003) is considered in this paper since no
rotational move and no moment resistance are required
on the end-effector, all 4 cables convene in a single point
and the end-effector is modelled as a point mass. A Sliding
Mode Control with a Fuzzy PI as sliding surface for the
kinematic model of Planar 4-Cable CDDR is proposed.
The performance of this controller is compared with a PI
Controller and a conventional SMC in terms of integral
square error (ISE) index. The simulation results show the
ability the proposed controller in comparison with the
conventional controllers for trajectory tracking and step
change reference.
The organization of this paper is as follows: Section(2)
describes the model of Planar 4 Cable-CDDR. Section(3)
presents the design of controller for Planar 4 Cable-CDDR.
Section(4) provides the simulation results. Finally, the
conclusions are presented.
2. MODEL OF PLANAR 4-CABLE CDDR
In this paper, we use only the reverse kinematic pose so-
lution (Williams II and Gallina, 2003). Given the position
 T
L˙1
  
Pr = [x y] is found the cable lengths Li (i = 1, ..., 4). cosθ1 sinθ1  

T
The end-effector position Pr = [x y] is possible to get L˙2  cosθ2 sinθ2  ẋ
 = (5)
by geometrical considerations (Euclidean Norm) with each L˙3  cosθ3 sinθ3  ẏ
fixed ground link vertex Ai (Motor Position). In Fig. 1 L˙4 cosθ4 sinθ4
shows the Planar 4-Cable CDDR scheme. It is important
to mention that each cable angle depend of quadrant.
The reverse kinematic pose solution is: The alternate forward velocity solution with L̇ as inputs,
is calculated through the left pseudoinverse:
q
2 2
Li = (x − Aix ) + (y − Aiy ) (1) Ṗr = J ∗ L̇ (6)

−1
where J ∗ = J T J J T is the left pseudoinverse.
and the cable angles are given by:
  3. DESING OF CONTROLLERS FOR PLANAR
−1 y − Aiy 4-CABLE CDDR
θi = tan (2)
x − Aix
This section designs different controllers for Planar 4-
Cable (CDDR) based on the kinematic model for trajec-
tory tracking. The designed controllers are a PI Controller,
a SMC and a Fuzzy Logic Controller.

3.1 PI Controller

In order to design a PI controller, it is necessary to know

the error. The control scheme of this controller is presented
in Fig. 2.
Pd (t ) e(t ) Pr (t )
+
-
Kp +
-
Plant

 Ki

Fig. 2. PI control scheme.

The tracking error vector is defined as follows:
Fig. 1. Planar 4-Cable CDDR Scheme (Williams II and
Gallina, 2003). e(t) = Pd (t) − Pr (t) (7)

where Pd (t) is the desired position vector and Pr (t) is the

th robot position vector.
Considering the i (i = 1, ..., 4) cable vector-loop closure
T T This controller has the following form:
equation [x y] = [Aix + Li cosθi Aiy + Li sinθi ] is
calculated the velocity kinematics equations: Z t
     P I(t) = Kp e(t) + Ki e(t)dt (8)
ẋ cosθi −Li sinθi L̇i 0
= (3)
ẏ sinθi Li cosθi θ˙i Applying this controller to the kinematic model of Planar
4-Cable CDDR is necessary to multiply by J, getting L̇i
(i = 1, ..., 4) as control actions.
Inverting ith cable Jacobian matrix: Z t
     C(t) = L̇i = J(K p e(t) + Ki e(t)dt) (9)
L̇i cosθ sinθi ẋ
˙ = sinθi i − cosθi/Li (4) 0
θi /Li ẏ
where Kp and Ki are tuning parameters. These parameters
have been selected by trial and error until achieving the
lowest ISE index.
In order to build the inverse velocity solution of Planar
4-Cable CDDR (L̇ = J Ṗr ) is necessary to eliminate the 3.2 Sliding Mode Controller
second row θ˙i as (Williams II and Gallina, 2003) to relation
the cable length rates and the end-effector velocity as: This section shows the design of a SMC with a PI sliding
surface (See Fig. 3) from the kinematic model of Planar
4-Cable CDDR as in (Villacres et al., 2017) designs a In order to design ud (t), a positive-definite Lyapunov func-
conventional SMC with a PID sliding surface. tion V is defined:

SMC
Pr (t ) 1 T
Pd (t )
+
e(t )
Kp +
uc (t)  ud (t)
Plant V = σ(t) σ(t) > 0 (18)
-
+
2
The derivative of the function V must be negative-definite:
 K i
T
V̇ = σ(t) σ̇(t) < 0 (19)

By replacing (15), V̇ is:

Fig. 3. SMC control scheme.
h i
T
In order to design a conventional SMC, this sliding surface V̇ = σ(t) (Ṗd (t) − J ∗ L̇) + λe(t) (20)
is considered as:
 n Z The control law is defined as L̇ = uc (t) + ud (t) and
d substituting:
σ(t) = +λ e(t)dt (10)
dt T
h  i
V̇ = σ(t) Ṗd (t) − J ∗ (uc (t) + ud (t)) + λe(t) (21)
where n is the order of the system and λ is a positive
constant. To satisfy (19) and replacing (16) in (21), V̇ should be:
This controller has two components: a continuous uc and T
a discontinuous ud . V̇ = σ(t) [−Jk1 sign(σ(t))] < 0 (22)
where k1 > 0.
u(t) = uc (t) + ud (t) (11)
Therefore, by analysing:
The system under study is the first order n = 1, for this if σ(t) > 0 → sign(σ(t)) > 0
reason the derivative part of the surface is eliminated. The (23)
if σ(t) < 0 → sign(σ(t)) < 0
surface is expressed as:
Z Finally, to reduce the chattering effect (Camacho and
σ(t) = e(t) + λ e(t)dt (12) Smith, 2000), ud (t) can be rewritten as a sigmoid function:
 
σ(t)
Now, the surface must be derived for the development of ud (t) = J k1 (24)
|σ(t)| + δ
the controller:
where δ is a chattering parameter reduction.
σ̇(t) = ė(t) + λe(t) (13)
3.3 Fuzzy - Sliding Mode Controller
By substituting (7) in (13):
In this section, a PI-Fuzzy as sliding surface is proposed to
σ̇(t) = (Ṗd (t) − Ṗr (t)) + λe(t) (14) SMC. The selection of this surface is based on the qualita-
tive knowledge about the process to be controlled and was
designed using Fuzzy logic toolbox in Matlab/Simulink.
For the surface design, there are two inputs and one
By replacing (6) in (14), σ̇(t) can be rewritten as: output. The surface is defined by:
σ̇(t) = (Ṗd (t) − J ∗ L̇) + λe(t) (15) Z t
σ(t) = sF (e(t), e(t)dt) (25)
The continuous part of the controller is provided with the 0
condition to keep the output on the sliding surface σ̇(t) = 0 The control scheme of Fuzzy-SMC Controller is shown in
and considering the control law as uc (t) = L̇: Fig. 4.

uc (t) = J(Ṗd (t) + λe(t)) (16) Pd (t ) e(t ) SMC Pr (t )

+
-
Kp Fuzzy uc (t)  ud (t) Plant

By completing the SMC control law, the discontinuous

part ud (t) is added:  Ki

uSM C = J[Ṗd (t) + λe(t)] + J[k1 sign(σ(t))] (17)

where ud (t) is responsible for reaching sliding surface and Fig. 4. Fuzzy-SMC control scheme.
is composed of a non-linear function sign(σ(t)) which
switches about the sliding surface and k1 > 0 is a The Member Functions (MFs)of the error and its integral
tuning parameter. These consideration were taken from are defined in Fig. 5. The range for universe of discourse
(Camacho Quintero, 2017). for the two variables are [-2 2].
 4. SIMULATION RESULTS

The controllers were implemented in Matlab/Simulink

Fig. 5. MFs for the input variables (Ghosh et al., 2015).
In this design seven membership functions namely, NB
(Negative Big), NM (Negative Medium), NS (Negative
Small), ZE (Zero), PS (Positive Small), PM (Positive
Medium), and PB (Positive Big) are defined with sym-
metric triangles having 50% overlap (Ghosh et al., 2015).
2017a using the kinematic model of Planar 4-Cable
CDDR. The test were run on a computer with an In-
tel(R)Core(TM) i7-5500U CPU @ 2.40GHz with 8,00 GB
RAM, running Windows 10. Fig. 8 shows the simulator
developed to observe the behavior of the cables during the
trajectory. Two tests were performed:
• Step Change Reference Test
• Tracking Trajectory Test
Movement CDDR
0.5
0.4

The output member function is shown in Fig. 6 and has 0.3

seven triangular functions as the input member functions 0.2
but is defined in the range [-0.6 0.6]. These rules are listed
0.1
in Table 1.

axis Y
0

-0.1

-0.2

-0.3

-0.4

-0.5
-0.5 -0.4 -0.3 -0.2 -0.1 0 0.1 0.2 0.3 0.4 0.5
axis X

Fig. 6. MF for the output variable (Ghosh et al., 2015).

Fig. 8. Planar 4-CDDR Simulator designed in Mat-
lab/Simulink.
The output surface is shown in Fig. 7. The forty-nine fuzzy
rules are based on sliding mode principle (Palm, 1992). The simulation has a duration of 110 seconds with a
sampling time of 0.1 for each test and uses ODE45 (Solve
non-stiff differential equations). The physical parameters
of Planar 4-Cable CDDR are presented in Table 2.
Table 2. Parameters Planar 4-Cable CDDR
Parameter Unit Description
A1 =(-0.35,-0.35) [m] Motor position 1
A2 =(0.35,-0.35) [m] Motor position 2
A3 =(0.35,0.35) [m] Motor position 3
A4 =(-0.35,0.35) [m] Motor position 4

Finally, the values of Kp and Ki for PI Controller and λ,

k1 and δ for SMC have been selected by trial and error
until achieving the lowest ISE index. In Table 3 are shown
these values.
Table 3. Tuning Parameter for the Controllers
Fig. 7. Output Surface. Tuning Parameter Value
Kp 10
Ki 0.08
λ 0.5
Table 1. Fuzzy Control Rules
k1 1
δ 0.1
R
e/e NB NM NS ZE PS PM PB
NB NB NB NB NM NS NS ZE
NM NB NB NM NM NS ZE PS
NS NB NM NM NS ZE PS PM
We used (26) and (27) to compare the controllers based on
ZE NB NM NS ZE PS PM PB ISE index. ∆1 corresponds to the comparison between PI
PS NM NS ZE PS PM PM PB Controller and Fuzzy-SMC Controller and ∆2 corresponds
PM NS ZE PS PM PM PB PB to the comparison between conventional SMC and Fuzzy-
PB ZE PS PM PM PB PB PB SMC Controller. The variation ∆1 and ∆2 is given by the
following equations:
 e1 − e2 Y vs. Time
∆1 % = e1 +e2 · 100 (26) -0.192
Reference
2
-0.194 PI
SMC
e1 − e3 SMC-Fuzzy
∆2 % = e1 +e3 · 100 (27) -0.196

2
-0.198

where e1 , e2 and e3 represent error values for Fuzzy-

y [m]
-0.2
SMC Controller, PI Controller and conventional SMC
respectively. -0.202

-0.204

4.1 Step Change Reference Test

-0.206

In this test, a step change is made form the reference -0.208

0 1 2 3 4 5 6
Xr = [0 0] to the point Xr = [0.2 − 0.2]. In Fig. 9 Time [s]
illustrates the end-effector path.
Fig. 11. y vs. time, Step Change Reference Test.
0
Reference
PI
SMC
-0.05 SMC-Fuzzy
Table 4. ISE Step Change Reference Test
PI SMC Fuzzy-SMC ∆1 % ∆2 %
-0.1
X 0.0080 0.0039 0.0017 78.57 129.89
Y 0.0080 0.0039 0.0017 78.57 129.89
-0.15

-0.2

4.2 Tracking Trajectory Test: Square

-0.25
0 0.05 0.1 0.15 0.2 0.25

In this test, the selected trajectory is a square whose side

Fig. 9. XY Graph, Step Change Reference. length is 0,4[m]. In Fig. 12 illustrates the efficiency of the
three designed controllers for this desired trajectory.
Fig. 10 and Fig. 11 present x and y positions during the
change reference, respectively. The percentage overshoot
of Fuzzy-SMC Controller is less than conventional SMC
and the setting time less than PI Controller.
Square Trajectory
0.25
Reference
X vs Time 0.2
0.208 PI
SMC
Reference 0.15 SMC-Fuzzy
0.206 PI
SMC 0.1
SMC-Fuzzy
0.204
0.05
y [m]

0.202 0
x [m]

-0.05
0.2

-0.1
0.198
-0.15
0.196
-0.2

0.194 -0.25
-0.25 -0.2 -0.15 -0.1 -0.05 0 0.05 0.1 0.15 0.2 0.25
x [m]
0.192
0 1 2 3 4 5 6
Time [s]
Fig. 12. XY Graph, Square trajectory.
Fig. 10. x vs. time, Step Change Reference Test.

In Table. 4 shows ISE index comparation between the

three controllers for step change reference test where In Fig. 13 presents a image zoom at beginning of trajectory
the performance of Fuzzy-SMC Controller is the best shows that Fuzzy-SMC Controller has lower overshoot
according to ISE index. than conventional SMC but its setting time is better.
 Square Trajectory
REFERENCES
-0.175
Reference Babaghasabha, R., Khosravi, M.A., and Taghirad, H.D.
PI
SMC
(2015). Adaptive control of kntu planar cable-driven
-0.18
SMC-Fuzzy parallel robot with uncertainties in dynamic and kine-
matic parameters. In Cable-Driven Parallel Robots, 145–
-0.185
159. Springer.
Camacho, O. and Smith, C.A. (2000). Sliding mode
y [m]

-0.19
control: an approach to regulate nonlinear chemical
processes. ISA transactions, 39(2), 205–218.
-0.195
Camacho Quintero, O.E. (2017). A blended sliding mode
control with linear quadratic integral control based on
-0.2
reduced order model for a vtol system.
Ghosh, A., Sen, S., and Dey, C. (2015). Design and real-
-0.205
-0.2 -0.195 -0.19 -0.185 -0.18 -0.175
time implementation of a fuzzy pi controller on a servo
x [m] speed control application. In Signal Processing and
Integrated Networks (SPIN), 2015 2nd International
Fig. 13. XY Graph Zoom, Beginning of Square trajectory. Conference on, 384–387. IEEE.
Jin, X., Jun, D.I., Pott, A., Park, S., Park, J.O., and Ko,
In Fig. 14 presents a image zoom at the corner of the S.Y. (2013). Four-cable-driven parallel robot. Interna-
square trajectory showing that Fuzzy-SMC Controller is tional Conference on Control, Automation and Systems
smoother than the others and the proposed controller is (ICCAS 2013), 879–883.
the best for reaching the trajectory. Khosravi, M.A. and Taghirad, H.D. (2014). Robust pid
control of fully-constrained cable driven parallel robots.
Square Trajectory Mechatronics, 24(2), 87–97.
-0.186
Reference
Mohammad, A. and Ehsan, S.S. (2008). Sliding mode pid-
-0.188
PI controller design for robot manipulators by using fuzzy
SMC
SMC-Fuzzy
tuning approach. In Control Conference, 2008. CCC
-0.19 2008. 27th Chinese, 170–174. IEEE.
-0.192
Nabi, A. (2013). Design of fuzzy logic pd controller for
a position control system. International Journal of
y [m]

-0.194 Engineering and Management Research (IJEMR), 3(2),

31–34.
-0.196
Oh, S.R. and Agrawal, S.K. (2005). Cable suspended
-0.198 planar robots with redundant cables: controllers with
positive tensions. IEEE Transactions on Robotics, 21(3),
-0.2
457–465. doi:10.1109/TRO.2004.838029.
-0.202 Palm, R. (1992). Sliding mode fuzzy control. In [1992
0.1985 0.199 0.1995 0.2 0.2005 0.201 Proceedings] IEEE International Conference on Fuzzy
x [m]
Systems, 519–526. doi:10.1109/FUZZY.1992.258681.
Villacres, J., Herrera, M., Sotomayor, N., and Camacho,
Fig. 14. XY Graph Zoom, Square trajectory.
O. (2017). A fuzzy sliding mode controller from a
reduced order model: A mobile robot experimental ap-
In Table. 5 shows the ISE comparation between the three plication. In Control, Decision and Information Tech-
controllers for tracking trajectory test. nologies (CoDIT), 2017 4th International Conference
Table 5. ISE Tracking Trajectory Test on, 0674–0678. IEEE.
Williams, R.L., Gallina, P., and Rossi, A. (2001). Planar
PI SMC Fuzzy-SMC ∆1 % ∆2 % cable-direct-driven robots, part i: Kinematics and stat-
X 0.0090 0.0038 0.0018 78.57 133.33 ics. In Proceedings of the 2001 ASME Design Technical
Y 0.0090 0.0038 0.0018 78.57 133.33 Conference, 27th Design Automation Conference, 26,
178–186.
5. CONCLUSION Williams II, R.L. and Gallina, P. (2003). Translational
planar cable-direct-driven robots. Journal of Intelligent
In this paper, PI, SMC and Fuzzy-SMC controllers were and Robotic systems, 37(1), 69–96.
designed for end-effector position control of Planar 4-Cable Zanotto, D. (2011). Analysis and development of cable-
CDDR based on the kinematic model. These controllers driven robotic devices.
were able to perform trajectory tracking and step change
reference, but the results indicate that SMC with Fuzzy-PI
as sliding surface with forty-nine rules presents lower set-
ting time and ISE than PI controller and, the conventional
SMC. The performance of the controller was evaluated
in terms of integral square error (ISE) index and these
results demonstrate the effectiveness of proposed controller
showing an acceptable accuracy.