1 s2.0 S0307904X23001452 Main

Applied Mathematical Modelling 120 (2023) 436–462
Contents lists available at ScienceDirect
Applied Mathematical Modelling

journal homepage: www.elsevier.com/locate/apm
Trajectory tracking of Stanford robot manipulator by

fractional-order sliding mode control
Samuel Chávez-Vázquez a, Jorge E. Lavín-Delgado b, José F. Gómez-Aguilar c,
José R. Razo-Hernández d, Sina Etemad e, Shahram Rezapour e,f,g,∗
a
Tecnológico Nacional de México/CENIDET, Interior Internado Palmira S/N, Col. Palmira, Cuernavaca, C.P. 62490, Morelos, México
b
Universidad Politécnica del Estado de Guerrero (UPEG), Dirección de Ingeniería en Redes y Telecomunicaciones, Puente Campuzano,
Carretera Federal Iguala-Taxco K.M. 105, Taxco de Alarcón, C.P. 40321, Guerrero, México
c
CONACyT-Tecnológico Nacional de México/CENIDET, Interior Internado Palmira S/N, Col. Palmira, Cuernavaca, C.P. 62490, Morelos,
México
d
ENAP-Research Group, CA-Fuentes Alternas y Calidad de la Energía Eléctrica, Departamento de Ingeniería Electromecánica, Tecnológico
Nacional de México, Instituto Tecnológico Superior de Irapuato (ITESI), Carr. Irapuato-Silao km 12.5, Colonia El Copal, Irapuato, C.P.
36821, Guanajuato, México
e
Department of Mathematics, Azarbaijan Shahid Madani University, Tabriz, Iran
f
Department of Mathematics, Kyung Hee University, 26 Kyungheedae-ro, Dongdaemun-gu, Seoul, Republic of Korea
g
Department of Medical Research, China Medical University Hospital, China Medical University, Taichung, Taiwan
a r t i c l e i n f o a b s t r a c t
Article history: In this work, a fractional integral sliding-mode control scheme based on the Caputo–
Received 8 August 2022 Fabrizio derivative and the Atangana–Baleanu integral of the Stanford robot for trajectory
Revised 31 March 2023
tracking tasks is developed and presented. The coupled system is composed of the robot
Accepted 3 April 2023
manipulator and the induction motors that drive its joints. The mathematical model of
Available online 12 April 2023
the system is obtained by the Euler–Lagrange method and generalized to an arbitrary or-
Keywords: der via the Caputo–Fabrizio derivative. The actuators are controlled by fractional PI con-
Fractional integral sliding mode control trollers based on the Atangana–Baleanu integral, while a fractional integral sliding mode
Fractional PI controller control law is also developed for trajectory tracking control. In this context, a fractional
Stanford robot manipulator version of the sliding surface via the Caputo–Fabrizio derivative is introduced to improve
Induction motor the performance of the control system. In addition, to attenuate the chattering effects, a
Cuckoo search algorithm fractional integral term, based on the Atangana–Baleanu integral, is introduced on the slid-
ing surface further improving system performance with less power consumption. The con-
ventional integral sliding mode control and an optimal super-twisting sliding mode con-
trol are also introduced for comparison with the proposed control strategy. The control
schemes were tuned using the Cuckoo method. External disturbances are also considered
in the system dynamics, as well as different end-effector reference trajectories, which were
designed to carry out manufacturing tasks. Simulation results confirm the superiority of
our control scheme for trajectory-tracking applications over its conventional version and
the optimal super-twisting sliding mode control, even with external disturbances and tra-
jectory changes. To show the robustness of the proposed control scheme under different
operating conditions, all the numerical simulations were performed considering the same
orders and gains. Finally, our control law introduces the fractional derivative and integral
of Caputo–Fabrizio and Atangana–Baleanu respectively, which have not been used enough
∗
Corresponding author at: Azarbaijan Shahid Madani University, Tabriz, Iran.
E-mail addresses: [email protected] (S. Chávez-Vázquez), [email protected] (J.E. Lavín-Delgado), [email protected]
(J.F. Gómez-Aguilar), [email protected] (J.R. Razo-Hernández), [email protected] (S. Rezapour).
https://doi.org/10.1016/j.apm.2023.04.001
0307-904X/© 2023 Elsevier Inc. All rights reserved.
S. Chávez-Vázquez, J.E. Lavín-Delgado, J.F. Gómez-Aguilar et al. Applied Mathematical Modelling 120 (2023) 436–462
in the modeling and control of robotic systems. Therefore, it is interesting to analyze the
contributions and advantages that arise when using them.
© 2023 Elsevier Inc. All rights reserved.
1. Introduction
Robot manipulators have been used in different industries, for example, the automotive industry, manufacturing industry,
and aerospace industry, to mention a few [1]. In these industries, it is necessary to accurately model the dynamics of the
robot and its actuators, as well as trajectory planning and generation and the design of control schemes. In addition, the
dynamics of robots are often subjected to disturbances such as joint friction, payload variation, and external forces, among
others. These disturbances affect system performance and can cause instability so the design and implementation of robust
control strategies in these robotic systems have been the subject of particular interest for years. Also, another aspect for
considering is the type of actuators used in the joints of the robot manipulators. DC motors are often used in a variety of
real industrial applications involving robot manipulators due to their easy speed or position control despite their high main-
tenance cost. To deal with this inconvenience, induction motors are an option for replacing the CD motors due to their high
output torque, ruggedness, and low-cost [2]. However, it should be noted that their dynamics are nonlinear, which makes
them difficult to control. In this context, several researches have been developed concerning the modeling, path planning,
and control robot manipulators in recent decades [3]. In this context, several researches have been developed concerning the
modeling, path planning, and control of robot manipulators and robotic systems in general in recent decades. An adaptive
impedance controller for human-robot co-transportation was described in Yu et al. [4]. Artificial vision and force sensors
were used to determine the human hand position and to calculate the interaction force between the human and the robot.
To track the motion of the human a nonlinear control for the end-effector was also developed, taking into account input
constraints on the actuators, unknown robot dynamics, and unknown initial conditions. Simulations and experiments results
showed a desirable performance in terms of a safe interaction human-robot, and smooth behavior on the co-transportation
tasks. In [5] a framework using visual and force sensing for human-robot co-transportation tasks was developed. The human
motion is calculated through visual information, while an observer was also designed for estimating the control input, which
generates the desired movement of the robot. An adaptive impedance-based control law was designed for trajectory-tracking
tasks while a neural network was used to compensate for uncertainties in the dynamic model of the robot. Experimental
and simulation results validated the proposed framework, showing that a stable and efficient interaction behavior human-
robot is obtained while decreasing human control effort. In [6] an optimal control law for actuators of robot manipulators
was described. The dynamic model is obtained by using tensors in Riemannian geometry. The closed-loop equations result
in a system of second-order ordinary differential equations, which are solved to obtain optimal joint forces. Then, these
forces are used as control inputs to track the optimal trajectories. Simulation results on a serial robot manipulator with
revolute joints confirmed that the optimal trajectory was accurately tracked, even in the presence of disturbances. A fuzzy
scheme, based on the extreme learning machine and the sliding-mode control, to track the position of robot manipulators
was proposed in Zhou et al. [7]. The learning machine was used to mimic the control law of the sliding-mode controller and
update the parameters by online cyclic training. To ensure system stability the network adaptive learning law was calculated
through the Lyapunov theory. Simulation results demonstrated that this control scheme is suitable for accurately tracking
the trajectories of the two-link robot manipulator. In addition, chattering effects are effectively reduced. In [8] an adaptive
fuzzy inherited sliding mode control system, based on the random vector unction-link network, for tracking the position of
n-link robot manipulators was proposed. This scheme also solves the instability problem caused by uncertain friction and
load changes. The self-adjustment of the output weights was done using adaptive rules so that an accurate approximation
of the unknown nonlinear dynamics in the control system can be obtained. To further attenuate the chattering phenomenon
and improve the system performance a robust control term was added. The closed-loop system stability was proved through
the Lyapunov theory. Simulation results on the two-link robot manipulator showed the effectiveness of this control scheme.
A sliding mode control strategy, based on an extreme learning machine, for a two-link robot manipulator, was presented in
Zhou et al. [9]. The grey wolf optimization method (GWO) is combined with a differential evolution algorithm to calculate
hidden layer weights and biases. In addition, the extreme learning machine is introduced into the sliding mode controller to
cope with external disturbances and uncertainties. Finally, the discontinuous control signal is serialized to effectively reduce
the chattering effects of the sliding mode controller. Simulation results showed superior performance for trajectory tracking
in terms of precision and speed over the PID controller and other sliding mode controller techniques. In [10] an adaptive
nonsingular sliding mode control strategy for trajectory tracking tasks of a robot manipulator was described. This control
law offers rapid convergence and avoids singularities and does not require a priori knowledge of the bound of the coupling
uncertainty. To minimize the chattering effects an actuator saturation compensator was designed and implemented. The sys-
tem stability is guaranteed through the Lyapunov theory. The effectiveness and superiority of the proposed controller were
demonstrated by numerical simulations. In [11] an active fault-tolerant control system for a 3-DOF robot manipulator which
combines a fast terminal sliding mode control and extended state observer is described. The state observer is designed to
estimate uncertainties, external disturbances, and faults. Then, this information is used to combine with the controller as a
437
compensator. Experimental results confirmed the effectiveness of the proposed control system was confirmed in terms of the
capacity to attenuate the effects of faults. In [12] an intelligent control strategy for the tracking control of a two-link robot
manipulator with unknown nonlinearities was described. The control system combines a neuronal network and sliding mode
control. The output weights of the neural network were updated through an adaptive law. To reduce the computational cost,
the minimum parameter learning (MPL) technique is applied. Simulation results demonstrated the robustness of the control
scheme to external disturbances and parametric uncertainties. A nested adaptive integral terminal sliding mode controller
for high-order uncertain nonlinear systems was developed in Shao [13]. The proposed control law combines a nonsingular
terminal sliding mode control and a fractional power integral terminal sliding mode control. A nested dual-layer adaptive
law, which updates controller gains, was added to eliminate the need for the upper bounds of the disturbances. Stability
analysis demonstrated the convergence of the system in finite time. Simulation results affirmed the superiority of the pro-
posed control over the conventional method. A barrier function-based adaptive sliding mode control strategy for a class of
uncertain nonlinear systems with actuator saturation is presented in Shao et al. [14]. This control scheme contains a control
input that adapts to the time-varying disturbances and it does not require the upper bound of disturbance. Also, it avoids the
adaptive gain monotonically increasing with respect to persistent disturbances. Stability analysis confirmed the asymptotic
convergence of the system even in the presence of actuator saturation, parameter uncertainties, and external disturbances.
Experimental results on an air-floating positioning system with input saturation demonstrated the effectiveness of the pro-
posed control scheme. On the other hand, in recent years fractional calculus has been successfully applied in many areas
of science and engineering such as biology, chemistry, physics, robotics, and control theory, to mention a few [15]. In this
context, the concept of fractional PID controller is presented in Vinagre et al. [16]. An analysis in the frequency domain is
performed to show the advantages of this controller, concluding that it is suitable for both fractional and integer order sys-
tems. In [17], the authors presented a fractional sliding-mode control scheme for a 2-DoF polar robot manipulator based on
the Riemann–Liouville derivative. The stability of the integer-order closed-loop system is proven by using the smooth func-
tion as the Lyapunov function. Simulation results demonstrated that the fractional-order sliding mode controller has been
quite effective for trajectory tracking control. In [18], a control scheme that combines the advantages of a fractional PID con-
trol and fractional fuzzy logic control for a two-link robot was presented. The collaborative control scheme was tuned using
the Cuckoo algorithm. Numerical simulations showed satisfactory results of the proposed scheme in comparison with some
techniques recently reported in the literature for trajectory tracking with external disturbances and payload variations. In
[19], a fractional PI controller for the velocity of a manipulator robot was developed. The performance of this controller was
analyzed and compared with its conventional counterpart in terms of the integral of the absolute magnitude of the error
(IAE), obtaining a desirable performance for the proposed control strategy. An optimal fractional fuzzy PID controller based
on Oustaloup approximation for a two-link manipulator is described in Ardeshiri et al. [20]. The multi-objective particle
swarm method was used to obtain the gains and order of the fractional controller. This controller was compared with some
approaches in real-time simulations. Based on the trajectory tracking error, its effectiveness was shown. In [21], a robust
adaptive fractional control based on Caputo derivative for electrically driven flexible-joint robots was proposed. According to
experimental results, the proposed control law has a better performance in comparison with some controllers, and the mod-
eling uncertainties have been well compensated. In [22], a fractional fuzzy PID controller based on the Grünwald–Letnikov
derivative for a three-link robot manipulator was presented. A comparative analysis showed the superiority of the proposed
control compared with other techniques reported in the literature for trajectory tracking considering external disturbances
and model uncertainty. A deep neural network-based fractional sliding mode control for position control of a robot manip-
ulator was presented in Zhou et al. [23]. The neural network is used for compensating the system dynamics uncertainties.
According to the simulation results, this control strategy exhibits robustness against external disturbances, random noises,
and parametric uncertainties. This approach was compared with other existing techniques, showing better performance for
the position tracking error. An adaptive fractional sliding mode controller for a 2-link robot is described in Zhang et al. [24].
The authors used the time delay estimation approach to estimate external disturbances and unknown dynamic parameters.
Experimental results validated the performance of the proposed technique over other laws control. A fractional adaptive
backstepping control for the PUMA 560 arm was presented in Anjum and Guo [25]. Simulation results validated the con-
troller accuracy in terms of trajectory tracking, and convergent speed considering actuator fault. A comparative analysis
validated this fractional fault-tolerant control strategy. In [26], a fractional neural network sliding-mode controller based
on Caputo derivative for a 2-link robot was described. The neural network looks to provide robustness to the controller
in the presence of disturbances and modeling errors. In this sense, the weight update method uses the Lyapunov stability
criterion. Simulation tests have led to conclude that the proposed fractional control scheme significantly reduces the posi-
tion tracking error even when the information on the dynamics of the robot manipulator is uncertain. A fractional control
law based on the pole placement method and the Mittag–Leffler function for a single flexible link robot manipulator was
described in Singh et al. [27]. The fractional dynamic model of the system was obtained through the fractional Laplace
transform approach. The proposed controller makes the system stable with minimum overshoot and settling time. Finally,
this control strategy was compared with the conventional PID controller, demonstrating better performance. One can find
more applications of fractional calculus in other fields (see for example, Mohammadi et al. [28], Rezapour et al. [29], Matar
et al. [30], Thabet et al. [31], Baleanu et al. [32–35], George et al. [36], Baleanu et al. [37], Sethi et al. [38], Etemad et al.
[39]). In the present research, the trajectory tracking problem of the Stanford robot manipulator driven by induction motors
is addressed. In this sense, DC motors generally used as actuators in robotic systems are replaced by induction motors. The
main contribution is the design and development of a hybrid control strategy consisting of PI fractional controllers and a
438
fractional integral sliding mode control generalized to an arbitrary order via the Atangana–Baleanu integral and the Caputo–
Fabrizio derivative, which has not yet been reported in the literature. In addition, the fractional dynamics of the coupled
system robot manipulator-actuators are obtained by combining the Euler–Lagrange method and the Caputo–Fabrizio deriva-
tive, which has also not been reported. The superiority of our control scheme is confirmed through numerical simulations
and from the comparison with the integer-order integral sliding-mode controller and the optimal super-twisting sliding-
mode control. Also, trajectory changes and external disturbances are introduced in the simulations while maintaining the
same gains and orders to demonstrate the robustness of our controller.
2. Mathematical preliminaries
Here are some definitions of fractional derivatives given.
2.1. Caputo–Fabrizio derivative
The Caputo–Fabrizio derivative in Caputo sense of order ℘ ∈ [0, 1] is defined as Caputo and Fabrizio [40]
M (℘)
t
dm

℘(t − τ )

CF C ℘
a Dt f (t ) = f (τ ) exp − dτ , m − 1 < ℘ < m, (2.1)
m −℘ a dτ m m −℘
where M (℘) is the normalization function, which satisfies B(0 ) = B(1 ) = 1; m is a positive integer, and t denotes the time.
If m ≥ 1, we have
CF C (℘+m ) C ℘
CF C
a Dt f (t ) =CF
a Dt a Dtm f (t ) . (2.2)
This derivative considers integer-order initial and boundary conditions, which may have a physical meaning. However, its
kernel is local, and the integral associated is not fractional [40].
2.2. Atangana–Baleanu derivative
The Atangana–Baleanu derivative in Caputo sense of order ℘ ∈ [0, 1] is given by Atangana and Baleanu [41]
M (℘)
t ℘

ABC ℘
a Dt f (t ) = f (τ )E℘ − (t − τ )℘ dτ , (2.3)
1 −℘ a 1 −℘
with E℘(z ) the Mittag–Leffler function defined as follows
∞
zk
E℘(z ) = . (2.4)
(℘k + 1)
k=0
Its non-local kernel allows the full description of the memory [42].
2.3. Riemann–Liouville integrals
The left and right Riemann–Liouville integral of order ψ > 0, represented as RL Dψ and RL ψ
a t t Db , respectively, are given by
Marichev et al. [43]
t
RL ψ 1
a It f (t ) = f (τ )(t − τ )ψ −1 dτ , (2.5)
(ψ ) a
b
RL ψ 1
t Ib f (t ) = f (τ )(τ − t )ψ −1 dτ , (2.6)
(ψ ) t
where (z ) is the Gamma function

∞
(z ) = t z−1 e−t dt , e ( z ) > 0. (2.7)
0
2.4. Atangana–Baleanu integral
The Atangana–Baleanu integral of order ψ ∈ [0, 1], associated with the derivative (2.3), is defined as Atangana and
Baleanu [41]

ABC ψ 1−ψ ψ t
a It f (t ) = f (t ) + f (τ )(t − τ )ψ −1 dτ , (2.8)
M (ψ ) M (ψ )(ψ ) a
The integral (2.8) can be rewritten by using the right Riemann–Liouville integral (2.6) as follows [44]
ABC ψ 1−ψ ψ RL ψ
a It f (t ) = f (t ) + I f (t ), (2.9)
M (ψ ) M (ψ ) a t
439
Fig. 1. The Stanford robot.
3. Robot manipulator-induction motors dynamics
The Stanford robot is shown in Fig. 1, and its dynamics can be obtained by using the Euler–Lagrange formalism, and
written in matrix form as Lewis et al. [45]

M(q )q̈ + C q, q˙ q˙ + G(q ) + F q˙ + τ d = τ , (3.1)

where q ∈ Rn denotes the generalized coordinates, M(q ) the inertia matrix, while C q, q˙ and G(q ) represent the vectors

including centrifugal and Coriolis forces, and the gravitational forces respectively; F q˙ the friction in the joints, τ d the
external disturbances, and τ the input vector of the system. Generalizing the mathematical model (3.1) to an arbitrary order
through the Caputo–Fabrizio derivative (2.1), we obtain [46]
C 2ϑ CF C ϑ
CF C ϑ
CF C
M(q )CF
0 Dt q + C q, 0 Dt q 0 Dt q + G (q ) + F 0 Dtϑ q + τ d = τ , (3.2)
Now well, the fractional dynamics can better characterize the behavior of the robot manipulator than the traditional model
[47,48], i.e., different behavior is exhibited for each fractional order so that by selecting a suitable value, more realistic
results can be obtained. The conventional dynamic model (3.1) is recovered by setting ϑ = 1. Eq. (3.2) can be rewritten as
C 2ϑ CF C ϑ
CF C ϑ

0 Dt q + N q, 0 Dt q + h q, 0 Dt q = τ ,
M(q )CF (3.3)

C Dϑ q = C q, CF C Dϑ q CF C Dϑ q + G q and h q, CF C Dϑ q = F
CF C
with N q, CF
0 t 0 t 0 t ( ) 0 t 0
Dtϑ q + τ d . According to the parameters proposed
in Table 6.5 of [49], the dynamic model of the Stanford robot, presented in Appendix C.2 of [45], has the form
⎡ ⎤ ⎡ ⎤
m11 m12 m13 m14 m15 m16 n1
⎢m12 m22 m23 m24 m25 m26 ⎥ ⎢n2 ⎥
⎢m13 m23 m33 m34 m35 m36 ⎥ ⎢n3 ⎥
M (q ) = ⎢
⎢m14
⎥, N q, CF C ϑ
0 Dt =⎢ ⎥
⎢n4 ⎥, (3.4)
m24 m34 m44 m45 m46 ⎥
⎣ ⎦ ⎣ ⎦
m15 m25 m35 m45 m55 m56 n5
m16 m26 m36 m46 m56 m66 n6
with
m11 = 1.316 − 1.056108d2 + 11.48d22 + 2.51S22 − 5.47995S22 q3 + 6.47S22 q23 + 0.23S22 q3C5 ,
m12 = −6.47C2 d2 q3 ,
m13 = −6.47S2 d2 ,
m22 = 4.721 − 5.47995q3 + 6.47q23 + 0.23q3C5 ,
m33 = 7.252,
m44 = 0.107 + 0.203S52 ,
m55 = 0.113,
m66 = 0.0203,
m14 = m15 = m16 = m23 = m24 = m25 = m26 = m34 = m35 = m36 = m45 = m46 = m56 = 0,
440
n1 = 0,
n2 = −[2.734S2 + 6.47S2 q3 + 0.115(S2C5 + C2C4 S5 )]g,
n3 = 6.47C2 g,
n4 = 0.115(S2 S4 S5 )g,
n5 = 0.115(S2C4C5 − C2 S5 )g,
n6 = 0,
where Si = sin qi and Ci = cos qi . An induction motor is an AC electric motor that is commonly used because of its robust-
ness and its mechanically strong [50]. In this type of motor, the electric current in the rotor needed to produce torque is
obtained via electromagnetic induction from the rotating magnetic field of the stator winding. In [51] a dynamic model of
an induction motor, which considers linear magnetic circuits, was presented. In this context, a nonlinear field-oriented con-
trol is also proposed that involves a change of coordinates from the fixed stator frame to the rotating frame and a nonlinear
state feedback. Thus
dω τL
= μλrd isq − , (3.5a)
dt J
dλrd
= −αλrd + α Misd , (3.5b)
dt
dρ α Misq
= n pω + , (3.5c)
dt λrd
disd
= −γ isd + vd , (3.5d)
dt
disq
= −γ isq + vq , (3.5e)
dt
with

μ = nJpLMr , α = RLrr , σ = Ls 1 − LMs Lr ,
2
(3.6)
β = σMLr , γ = Rσs + βα M ρ = arctan λλrbra ,
where J, ω, n p , τL , M, Ls , Lr , Rs , and Rr denote the moment of inertia of the rotor and the load attached to it, the rotor
speed, the pole-pair number, the load torque, the mutual inductance, the stator inductance, the rotor inductance, the stator
resistance, and the rotor resistance, respectively; while λrd represents the rotor flux linkage; whereas vd and vq are the input
voltages. Finally, the subscripts s and r refer to the stator and rotor, respectively. According to the field-oriented control
proposed in Marino et al. [52], the dynamic model for the n actuators of the robot is given in matrix form by
˙ = BA − τ L ,
J (3.7a)
˙ = −α + MIsd , (3.7b)
I˙ sd = −γ Isd + vd , (3.7c)
I˙ sq = −γ Isq + vq , (3.7d)
with
J = diag{J1 , J2 , . . . , Jn },
= [ω 1 , ω 2 , . . . , ω n ]T ,
B = diag{μ1 J1 , μ2 J2 , . . . , μn Jn },
T
A= λrd1 isq1 , λrd2 isq2 , . . . , λrdn isqn ,
τ L = [τ L 1 , τ L 2 , . . . , τ L n ]T ,
441
T
= λrd1 , λrd2 , . . . , λrdn ,
M = diag{α1 M1 , α2 M2 , . . . , αn Mn },
T
Isd = isd1 , isd2 , . . . , isdn ,
Isq = [isq1 , isq2 , . . . , isqn ]T ,
γ = diag{γ1 , γ2 , . . . , γn },
T
vd = vd1 , vd2 , . . . , vdn ,
vq = [vq1 , vq2 , . . . , vqn ]T .

The control inputs of each actuator are designed through a PI controller [52], so that

vd = Kpvd λrdre f − λrd + Kivd λrdre f − λrd dt , (3.8a)

vq = Kpvq (τd − τem ) + Kivq (τd − τem )dt , (3.8b)
where K pv , K pvq , Kiv , and Kivq are positive constants, while λrdre f denotes the desired flux linkage, τd is the desired torque,
d d
and τem = μJ λrd isq is the electromagnetic torque. Note that term BA in Eq. (3.7a) represents the vector whose entries are
the electromagnetic torques generated by the actuators.
T
BA = μ1 J1 λrd1 isq1 , μ2 J2 λrd2 isq2 , . . . , μn Jn λrdn isqn = [τ e m 1 , τ e m 2 , . . . , τ e m n ]T , (3.9)
On the other hand, generalizing to an arbitrary order via the Caputo–Fabrizio derivative (2.1), Eqs. (3.7a)–(3.7d) can be
expressed as
C ϑ
0 Dt = BA − τ L ,
JCF (3.10a)
CF C ϑ
0 Dt = −α + MIsd , (3.10b)
CF C ϑ
0 Dt Isd = −γ Isd + vd , (3.10c)
CF C ϑ
0 Dt Isq = −γ Isq + vq , (3.10d)
Similarly, the control laws (3.8a) and (3.8b) can be generalized through the Atangana–Baleanu integral (2.8), and it fol-
lows that
ζ
vd = Kpvd λrdre f − λrd + Kivd ABC
0 It λr dre f − λrd , (3.11a)
η
vq = Kpvq (τd − τem ) + Kivq ABC
0
It (τd − τem ), (3.11b)
Now well, under the assumption of a direct mechanic matching between each joint and its induction motor, it follows
that
q = , (3.12)
where = [θ1 , θ2 , . . . , θn ]
T
denotes the angular position of the n induction motors. Differentiating Eq. (3.12) twice, we
obtain
q˙ = , (3.13a)
q̈ =
˙, (3.13b)
442
T
with = [ω1 , ω2 , . . . , ωn ]T = θ˙ 1 , θ˙ 2 , . . . , θ˙ n . The fractional version of Eq. (3.13b), via the Caputo–Fabrizio derivative (2.1),
can be expressed as
CF C 2ϑ C ϑ
0 D0 q 0 D 0 ,
= CF (3.14)
CF C Dϑ
CF C T
where 0 0
= 0
Dt2ϑ θ1 , CF
0
C D2ϑ θ , . . . , CF C D2ϑ θ
t 2 0 t n . By combining Eqs. (3.10a) and (3.14), we obtain
C 2ϑ
τ L = BA − JCF
0 D0 q , (3.15)
If the load torque imposed on induction motors is the input torque of the system, i.e. τ = τ L , by substituting Eq. (3.15) into
Eq. (3.3) we obtain
C 2ϑ CF C ϑ
CF C ϑ

0 Dt q + N q, 0 Dt q + h q, 0 Dt q = τ em ,
D(q )CF (3.16)
with D(q ) = J + M(q ) and τ em = BA. The above equation represents the coupling between the Stanford robot and the induc-
tion motors.
4. Fractional integral sliding mode control

T
The robot trajectory tracking problem consists of the design of a control law such that the current trajectory q, q˙ tracks
T
the desired trajectory qd , q˙ d despite parametric uncertainties and external disturbances [53]. Now well, the sliding mode
control is a robust nonlinear control and is easy to implement, which is usually combined with other control techniques
due to the chattering phenomenon which consists of the oscillation of the control signal caused by the discontinuity of the
switching term, where the sign function is often used, as well as theC non-modeled dynamics, and the actuators, causing
instability in the system [54]. Neglecting external disturbances h q, CF D ϑ q , the system dynamics (3.16) can be expressed
0 t
as
C 2ϑ CF C ϑ

0 Dt q + N q, 0 Dt q = τ em ,
D(q )CF (4.1)
and the computed torque for the tracking control of the nominal dynamic model without parametric uncertainties and
external disturbances is given by Shi et al. [55]
C 2ς CF C ς
CF C ϑ
τ 0 = D0 (q ) CF
0 Dt qd − Kd 0 Dt e (t ) − K p e (t ) + N0 q, 0 Dt q , (4.2)

C Dϑ q = N q, CF C Dϑ q + N, with D and N the unknown part of D q and
where D0 (q ) = D(q ) + D and N0 q, CF ( )
0 t 0 t
C Dϑ q respectively, while K and K are positive definite diagonal matrices, and e t the tracking error signal de-
N q, CF
0 t p d ()
fined as
e(t ) = q(t ) − qd (t ), (4.3)

By combining Eqs. (4.1)
assuming that τ em = τ 0 as well as the system dynamics is well known, so that
and (4.2), and
C Dϑ q = N q, CF C Dϑ q , we obtain
D0 (q ) = D(q ) and N0 q, CF
0 t 0 t
CF C 2ς C ς
0 Dt e (t ) + Kd CF
0 Dt e (t ) + K p e (t ) = 0, (4.4)
it follows that tracking
error e(t ) tends
to zero asymptotically. However, for a real robot with parameter uncertainties,
D(q ) = D0 (q ) and N q, CF C Dϑ q = N q, CF C Dϑ q , therefore Eq. (4.4) becomes
0 t 0 0 t
CF C 2ς C ς
0 Dt e (t ) + K p e (t ) = D0 (q )τ p ,
−1
0 Dt e (t ) + Kd CF (4.5)
with
CF C 2ς
τ p = D 0 Dt q + N, (4.6)
Note that regardless of the choice of matrices Kp and Kd the error e(t ) will not tend to zero. To cope deal with disturbances
caused by parametric uncertainties, a fractional integral sliding-mode control strategy is designed such that Eq. (4.4) is
satisfied. In this context, considering that there exists a nominal control law τ 0 that guarantees the local asymptotic conver-
gence of the undisturbed system, a fractional control law τ l , based on the integral sliding-mode principle, is added to reject
external disturbances [55], so that
τ sw = τ 0 + τ l , (4.7)
where τ l is a fractional disturbance compensator, while τ 0 represents the nominal control related to the undisturbed system.
On the other hand, fractional operators have been successfully applied in different control schemes providing additional
degrees of freedom while improving the performance and accuracy of closed-loop systems [56–58]. The ideal robot dynamics
can be expressed in terms of the error as follows [59]

CF C ϑ 2ς 2ς
0 (q )N0 q, 0 Dt q + D0 (q )τ 0 − 0 Dt qd (t ) = 0 Dt e0 (t ),
−D−1 −1 CF C CF C
(4.8)
443
Fig. 2. Control scheme.

C Dϑ q = N q, CF C Dϑ q , and τ

with D(q ) = D0 (q ), N q, CF
0 t 0 0 t em = τ 0 . Similarly, for the real system we obtain
C ϑ 2ς 2ς
0 Dt q + D (q )τ sw − 0 Dt qd (t ) = 0 Dt e (t ),
−D−1 (q )N q, CF −1 CF C CF C
(4.9)
with τ em = τ sw . Now well the fractional switching function is defined as Utkin et al. [59]
ς
sς ,δ (t ) = s0 (e(t ) ) + zδ (t ), (4.10)
with

ς e(t )
s0 (e(t ) ) = K I CF C ς , (4.11)
0 Dt e (t )
and

CF C ς
0 Dt e (t )
CF C δ
0 Dt z (e(t )) = − K I C ϑ
CF C 2ϑ , (4.12)
−D−1 0 Dt q + D0 (q )τ 0 − 0 Dt qd
−1
0 ( ) 0
q N q, CF
where K is a positive definite diagonal matrix. To deal with the chattering effects associated with the conventional sliding
mode controller the Caputo–Fabrizio derivative (2.1) is introduced. The initial condition zδ (0 ) is determined so that sς ,δ (0 ) =
0 is satisfied [59], i.e.
ς ς
zδ (0 ) = −s0 (0 ) = −Ke(0 ) − CF
0 Dt e (0 ),
C
(4.13)
Substituting Eq. (4.2) into Eq. (4.12) results
CF C δ C ς CF C ς
0 Dt z (e(t )) = −KCF
0 Dt e (t ) − Kd 0 Dt e (t ) − K p e (t ), (4.14)
By integrating (4.14) and substituting the resulting equation together with (4.11) into (4.10), the fractional integral sliding
surface becomes [55]
ς ς
sς ,δ (t ) = CF C ABC δ
0 Dt e (t ) + Kd e (t ) + K p 0 It e (t ) − Kd e (0 ) − 0 Dt e (0 ),
CF C
(4.15)
To further improve the performance of the controller in trajectory following tasks and add robustness in presence of exter-
nal disturbances, the Atangana–Baleanu integral (2.8) is introduced in the sliding surface [60]. Now well, the integer-order
sliding surface (4.15) is obtained by setting ς = 1 and δ = 1, thus
t
s(t ) = e˙ (t ) + Kd e(t ) + K p e(ξ )dξ − Kd e(0 ) − e˙ (0 ), (4.16)
0
Similarly, the conventional version of the computed torque (4.2) is written as

τ 0 = D0 (q ) q̈d − Kd e˙ (t ) − K p e(t ) + N0 q, q˙ , (4.17)
Finally, the switching discontinuous control law is expressed as Nadda [61]
τ l = K sat(s ), (4.18)
with K a positive constant, and

s if|s| ≤ 1
sat(s ) = (4.19)
sign(s ) if|s| > 1
where |·| stands for the absolute value of each element of the vector s. The proposed control scheme for the system Stanford
robot-induction motor for tracking trajectory tasks is shown in Fig. 2.
Thus, adding the control law (4.7) into Eq. (3.16) the closed-loop equation of the system becomes
C 2ϑ CF C ϑ
CF C ϑ

0 Dt q + N q, 0 Dt q + h q, 0 Dt q = τ em + τ sw ,
D(q )CF (4.20)
444
5. Numerical scheme
Let Eq. (5.1) be a fractional ordinary differential equation

CF C ξ
0 Dt ϒ (t ) = (t , ϒ (t )). (5.1)
By the fundamental theorem of fractional calculus associated with the derivative and integral operators in the Caputo sense
(Theorem 7 in Oliveira and de Oliveira [62]), Eq. (5.1) becomes a fractional integral equation expressed as

1−ξ ξ tn+1
ϒ (tn+1 ) − ϒ (0 ) = (tn , ϒ (tn ) ) + (t , ϒ (t ))dt , (5.2)
M (ξ ) M (ξ ) 0
According to the two-step Adams–Bashforth scheme described in Atangana and Owolabi [63], the numerical solution of
(5.1) is obtained as

1−ξ 3ξ h 1−ξ ξh
ϒn+1 = ϒn + + (tn , ϒn ) − + (tn−1 , ϒn−1 ), (5.3)
M (ξ ) 2M ( ξ ) M (ξ ) 2M ( ξ )
where h is the integration step size, and M (ξ ) a normalization function usually set to one (M (ξ ) = 1). The convergence and
stability of the numerical scheme (5.3) are demonstrated through the following theorems.
Theorem 1 (Theorem 3.2 in Atangana and Owolabi [63]). Let ϒ (t ) be a solution of the differential equation
CF C ξ
0 Dt ϒ (t ) = (t , ϒ (t )),
with a continuous function bounded for the Caputo–Fabrizio derivative (2.1), we obtain [63]

ϒn+1 = ϒn + + (tn , ϒn ) − + (tn−1 , ϒn−1 ) + Rξn ,
M (ξ ) 2M ( ξ ) M (ξ ) 2M ( ξ )

where Rnξ ≤ M.
Proof. From the definition (2.1), at the points tn+1 and tn , we obtain

1−ξ ξ tn+1
ϒ (tn+1 ) − ϒ (0 ) = (tn , ϒ (tn ) ) + (t , ϒ (t ))dt ,
M (ξ ) M (ξ ) 0

1−ξ ξ tn
ϒ (tn ) − ϒ (0 ) = (tn−1 , ϒ (tn−1 ) ) + (t , ϒ (t ))dt ,
M (ξ ) M (ξ ) 0
Therefore

1−ξ ξ tn+1
ϒ (tn+1 ) − ϒ (tn ) = [(tn , ϒ (tn ) ) − (tn−1 , ϒ (tn−1 ) )] + (t , ϒ (t ))dt
M (ξ ) M (ξ ) tn

1−ξ ξ (tn , ϒ (tn ) )
= [(tn , ϒ (tn ) ) − (tn−1 , ϒ (tn−1 ) )] + (t − tn−1 )
M (ξ ) M (ξ ) h

(tn−1 , ϒ (tn−1 ) )
n
n
(t − ti )
− (t − tn ) + (ti , ϒ (ti ) ) dt,
h
i=2 i=2 (−1 )i h
then

ϒn+1 = ϒn + + (tn , ϒn ) + + (tn−1 , ϒn−1 )
M (ξ ) 2M ( ξ ) M (ξ ) M (ξ )
tn+1 n
ξ n
(t − ti )
+ (ti , ϒi )dt ,
M ( ξ ) tn
i=2 i=2 (−1 ) h
i
By denoting the error term as

n
ξ tn+1 n
(t − ti )
Rnξ = (ti , ϒi )dt ,
M (ξ ) tn (−1 )i h
i=2 i=2
It follows that

n
R ξ = ξ tn+1 n n
(t − ti )
(t , ϒ )dt
M ( ξ ) tn
i i
∞
i=2 i=2 ( −1 )i
h
∞
445

tn+1
ξ (t − ti )
n n

≤ ( t , ϒ ) dt
M ( ξ ) tn
i i
i=2 i=2 (−1 ) h
i

∞
tn+1 n

ξ n
( t − t )
≤ i
(ti , ϒi )dt ∞
M ( ξ ) tn i
i=2 i=2 (−1 ) h
ξ
tn+1 n n
|t − ti |
< sup max |(ti , ϒi )|
M ( ξ ) tn h i∈I i∈I
i=2 i=2
ξ (n + 1)!hn+1
< M
M (ξ ) 4
Thus
n ξ
R ξ < (n + 1 )!hn+1 M.
∞ M (ξ )

Theorem 2 (Theorem 3.3 in Atangana and Owolabi [63]). Let ϒ (t ) be a solution of

CF C ξ
0 Dt ϒ (t ) = (t , ϒ (t )),
for every n ∈ N
1−ξ ξ hn+1 (n + 1)!
ϒn+1 − ϒn < (tn , ϒn ) − (tn−1 , ϒn−1 ) + ,
∞
M (ξ ) ∞
4M ( ξ )
such that if (tn , ϒn ) − (tn−1 , ϒn−1 ) ∞ → 0 when n → ∞, then ϒn+1 − ϒn ∞ → 0 when n → ∞ [63].
Proof.
tn+1
1 − ξ ξ
ϒn+1 − ϒn = [(tn , ϒn ) − (tn−1 , ϒn−1 )] + (t , ϒ (t ))dt
∞
M (ξ ) M ( ξ ) tn
t ∞

1−ξ ξ
n+1
≤ (tn , ϒn ) − (tn−1 , ϒn−1 ) ∞ + (t , ϒ (t ))dt
M (ξ ) M ( ξ ) tn
∞
tn+1 n
1−ξ ξ n
t − ti
≤ (tn , ϒn ) − (tn−1 , ϒn−1 ) ∞ + dt
M (ξ ) M ( ξ ) tn (−1 )h
i=0 i=0 ∞
1−ξ ξ hn+1 (n + 1)!
< (tn , ϒn ) − (tn−1 , ϒn−1 ) ∞ + ,
M (ξ ) 4M ( ξ )
thus, if (tn , ϒn ) − (tn−1 , ϒn−1 ) ∞ → 0 when n → ∞, then ϒn+1 − ϒn ∞ → 0 when n → ∞
Now well, the fractional state-space representation of (4.20) is defined as

q = x1 , (5.4a)
CF C ϑ
0 Dt q = x2 , (5.4b)
so that,
CF C ϑ
0 Dt x1 = x2 , (5.5a)
CF C ϑ
0 Dt x2 = D−1 (x1 )[τ em + τ sw − N(x1 , x2 ) − h(x1 , x2 )], (5.5b)
By applying the numerical scheme (5.3), the numerical solution of Eqs. (5.5a) and (5.5b) is calculated as follows

1−ϑ 3ϑ h 1−ϑ ϑh
x1n+1 = x1n + + x 2n − + x2n−1 , (5.6a)
B (ϑ ) 2B ( ϑ ) B (ϑ ) 2B ( ϑ )

1−ϑ 3ϑ h
x2n+1 = x2n + + D−1 (x1n )[τ em + τ sw − N(x1n , x2n ) (5.6b)
B (ϑ ) 2B ( ϑ )

1−ϑ ϑh
−h(x1n , x2n )] − + D−1 x1n−1 [τ em + τ sw
B (ϑ ) 2B ( ϑ )
446

−N x1n−1 , x2n−1 − h x1n−1 , x2n−1 ,
Similarly, for Eqs. (3.10b)–(3.10d) we obtain

1−ϑ 3ϑ h 1−ϑ ξh
n+1 = n + + −α n + MIsdn − + −α n−1 + MIsdn−1 , (5.7a)
B (ϑ ) 2B ( ϑ ) B (ϑ ) 2B ( ϑ )

Isdn+1 = Isdn + + −γ Isdn + vdn − + −γ Isdn−1 + vdn−1 , (5.7b)
B (ϑ ) 2B ( ϑ ) B (ϑ ) 2B ( ϑ )

Isqn+1 = Isqn + + (−γ Isqn + vqn ) − + −γ Isqn−1 + vqn−1 ., (5.7c)
B (ϑ ) 2B ( ϑ ) B (ϑ ) 2B ( ϑ )
Numerical scheme (5.3) is also applied to calculate the fractional derivative of the error in Eq. (4.15). In this context, the
second derivative of the error can be expressed as

d2 e d de(t )
= = q̈(t ) − q̈d (t ), (5.8)
dt 2 dt dt
Therefore, the fractional version of the above equation via the Caputo–Fabrizio derivative (2.1) can be written as
CF C ς

0 Dt e˙ (t ) = f(t ), (5.9)
where f(t ) = q̈(t ) − q̈d (t ) = ë(t ). Then it follows that

1−ς 3ς h 1−ς ςh
e˙ n+1 = e˙ n + + ën − + ën−1 . (5.10)
B (ς ) 2B ( ς ) B (ς ) 2B ( ς )
Moreover, a numerical approximation for the Riemann–Liouville fractional integral (2.5), called the product trapezoidal
method, is presented in Diethelm et al. [64], obtaining

k
RL ψ RL ψ
0 It f (tk ) ≈ h 0 It f (tk ) = hψ bi,k f (ti ), (5.11)
Tr
i=0
with
⎧
1+ψ
(k − 1 )⎨ − (k − ψ − 1 )kψ if i = 0
1
bi,k = × (k − i + 1 )1+ψ + (k − i − 1 )1+ψ − 2(k − i )1+ψ if 1 ≤ i ≤ k − 1
(2 + ψ ) ⎩
1 if i = k
The approximation quality of this numerical scheme is described below.
Theorem 3 (Theorem 2.5 in Diethelm et al. [64]).

ψ
i. Let f (t ) ∈ C 2 [0, T ], then there is a constant CT r depending only of ψ , that satisfies

RL ψ d2 f (t ) ψ 2
RL ψ
0 It f (tk ) − h 0 It
ψ
f (tk ) ≤ CT r sup t h .
Tr t∈[0,b] dt 2 k
ii. Let f (t ) ∈ C 1 [0, T ], and suppose that f (t ) satisfies a Lipschitz condition of order a for some ∈ (0, 1 ). Then there are
ψ ,
positive constants BT r (depending on ψ and ) and M ( f, ) (depending on f and ), such that

RL ψ ψ ψ , ψ
0 It f (tk )−h RL
0 It f (tk ) ≤ BT r M ( f, )tk h1+ .
Tr
iii. Let f (t ) = t p for some p ∈ (0, 2 ) and ι = min (2, p + 1), such that

RL ψ ψ ψ ,p ψ + p−ι ι
0 It f (tk ) − h RL
0 tI f (t )
k ≤ CT r tk h.
Tr
ψ ,p
where CT r is a constant that depends on ψ and p.
The theorem and its proof are presented in Diethelm et al. [64]. By using the numerical scheme (5.11) in Eq. (2.9), the
fractional integral actions of the PIζ and PIη controllers expressed in (3.11a) and (3.11b) respectively, as well as the fractional
integral of the error defined in (4.15) are calculated as
ABC ζ 1−ζ ζ RL ζ
0 It g1 (t ) = g1 (t ) + I g1 (t ), (5.12)
B (ζ ) B (ζ ) 0 t
447
Table 1
Induction motor parameters.
J Rotor inertia 0.0586 kg m2
M Mutual inductance 0.068H

Lr Rotor inductance 0.0699H
Ls Stator inductance 0.0699H
Rr Rotor resistance 0.15
Rs Stator resistance 0.18
np number of pole pairs 1
λrd Rotor flux linkage 1.3 Wb
ω angulas speed 220 rad
s
τL load torque 70 N m
Table 2
Gains and orders for the PIζ and PIη controllers.
PIζ , PIη
qi
K pvd Kiv
d
ζ K pvq K pvq η
q1 662.1879 725.2837 0.9427 524.0772 295.3995 0.9714
q2 859.9560 806.5905 0.9828 125.5727 422.7866 0.9651
q3 574.5307 89.5136 0.9172 180.9850 304.8358 0.9412
q4 564.9090 351.1141 0.9825 838.7095 533.7098 0.9811
q5 417.6797 636.3910 0.9241 533.8394 406.1029 0.9812
q6 924.0758 292.4280 0.9797 277.0544 164.4589 0.9711
Table 3
Gains and orders for the fractional-order integral sliding-mode control.
FOISM
qi
Kp Kd K ς δ
q1 917.3419 369.0974 0.6302 0.9999 0.9999
q2 475.5492 752.4207 0.9998 0.9998
q3 444.4064 880.3755 0.9997 0.9997
q4 142.8416 404.4501 0.9996 0.9996
q5 545.1201 107.3805 0.9995 0.9995
q6 267.6068 905.6380 0.9994 0.9994
Table 4
Gains for the integral sliding-mode control law, and the PI controllers.
PI ISM
qi
K pvd Kiv K pvq Kivq Kp Kd K
d
q1 966.2299 157.2703 416.8538 923.3089 897.9105 339.2808 0.1843

q2 412.0334 610.1649 55.5708 757.7075 495.4958 537.0337
q3 713.5951 519.9420 981.2427 289.5361 55.9789 585.8173
q4 355.0996 450.4817 895.2474 612.1964 101.3476 250.4725
q5 412.8055 993.6586 19.8131 946.8145 668.5467 80.0279
q6 756.8120 277.3352 313.3580 202.9199 247.6448 479.3901
Table 5
Gains for the super-twisting sliding-mode
controller.
STSM
qi
ci ki wi
q1 150.8626 135.5645 235.9783

q2 145.6584 354.3215 229.3164
q3 167.9831 95.5498 233.1943
q4 135.3215 102.9798 230.6453
q5 147.1918 141.3154 221.9412
q6 138.3254 215.7898 237.2154
448
Table 6
Mean absolute percentage error.
qi Trajectory 1 Trajectory 2
without external disturbances with external disturbances
FOISM ISM STSM FOISM ISM STSM FOISM ISM STSM
q1 2.33 2.54 2.49 2.83 3.54 3.43 2.54 3.64 3.33

q2 2.01 2.13 2.17 2.19 3.12 3.01 2.09 3.09 2.87
q3 1.89 2.01 1.98 2.02 2.98 2.89 1.97 3.87 3.41
q4 1.52 1.62 1.71 1.71 2.64 2.58 1.54 2.53 2.31
q5 2.12 2.23 2.19 2.19 3.23 3.13 2.14 3.16 3.01
q6 2.12 2.25 2.31 2.21 3.68 3.17 2.17 3.57 3.27
Fig. 3. Joint trajectories corresponding to (6.1) without disturbances: (a) q1 , (b) q2 , (c) q3 , (d) q4 , (e) q5 , (f) q6 .
449
Fig. 4. End-effector trajectory (6.1).
ABC η 1−η η RL η
0 It g2 (t ) = g2 (t ) + I g2 (t ), (5.13)
B (η ) B (η ) 0 t
ABC δ 1−δ δ RL δ
0 It e (t ) = e(t ) + I e(t ), (5.14)
B (δ ) B (δ ) 0 t
where g1 (t ) = λrdre f (t ) − λrd (t ) and g2 (t ) = τ d (t ) − τ em (t ). Then,
RL ζ

k
0 It g1 (t ) ≈ hζ bλd g1 (ti ),
i,k
i=0
RL η
k
0 It g2 (t ) ≈ hη bτi,k g2 (ti ), (5.15)
i=0
RL δ

k
0 It e (t ) ≈ hδ bei,k e(ti ),
i=0
with
⎧
1+ζ
⎨(k − 1 ) − (k − ζ − 1 )kζ if i = 0
1
bλd = × (k − i + 1 )1+ζ + (k − i − 1 )1+ζ − 2(k − i )1+ζ if 1 ≤ i ≤ k − 1
i,k (2 + ζ ) ⎩
1 if i = k
⎧
1+η
⎨(k − 1 ) − (k − η − 1 )kη if i = 0
1
bτi,k = × (k − i + 1 )1+η + (k − i − 1 )1+η − 2(k − i )1+η if 1 ≤ i ≤ k − 1
(2 + η ) ⎩
1 if i = k
⎧
1+δ
⎨ (k − 1 ) − (k − δ − 1 )kδ if i = 0
1 1+δ 1+δ 1+δ
= ×
(2 + δ ) ⎩(k − i + 1 ) + (k − i − 1 ) − 2(k − i )
bei,k if 1 ≤ i ≤ k − 1
1 if i = k
6. Simulation results
In this section, the performance of our fractional control strategy for a Stanford robot driven by induction motors is
verified through numerical simulations with MATLAB R2017b (64-bit version) for windows with 8 GB RAM. Table 1 presents
the induction motor parameters, with the following characteristics, 15 Kw motor, angular speed 220 rad/s, and rated torque
70 N-m [52]. The fractional order for the coupled system was set at ϑ = 1, to show the contribution of the Caputo–Fabrizio
derivative (2.1) and the Atangana–Baleanu integral (2.8) in the proposed control scheme.
The reference trajectory for the end-effector position was designed for manufacturing purposes, given by
px = 7.50 cos (t )
py = 7.52 sin (t ) (6.1)
pz = 0.1t
The joint trajectories associated with (6.1) are obtained by the inverse kinematics, whose analytical solution for the Stanford
T
robot is described in Appendix A. The initial conditions q(0 ) = 0, 23π , 0, 23π , 0, −π and q˙ (0 ) = [0, 0, 0, 0, 0, 0]T , and the
450
Fig. 5. Joint trajectories corresponding to (6.1) with disturbances: (a) q1 , (b) q2 , (c) q3 , (d) q4 , (e) q5 , (f) q6 .
Table 7
Error percentage.
qi Trajectory 1 Trajectory 2
without external disturbances with external disturbances
FOISM/ ISM FOISM/ STSM FOISM/ ISM FOISM/ STSM FOISM/ ISM FOISM/ STSM
q1 96.84 90.12 97.86 91.41 96.40 88.22

q2 88.82 91.99 87.72 90.23 88.83 90.46
q3 32.24 40.74 29.23 45.46 47.95 38.58
q4 78.94 84.23 78.35 82.45 78.93 83.01
q5 17.35 30.87 16.99 35.64 16.69 32.29
q6 70.72 92.64 61.40 90.53 70.85 90.33
451
Fig. 6. End-effector trajectory (6.1) with disturbances.
Table 8
Reference trajectory parameters (6.8).
Joint
Parameter
1 2 3 4 5 6
qn i 0.4 −0.03 0.3 0.7 0.5 0.9

q fi π π
− 70 − π4 − π2 − π3 π
2 3
Table 9
Gains and orders for PIψ and PD℘.
PIψ K pvd Kiv ψ K pvq Kivq ψ PD℘ Kp Kd ℘

d
q1 231.1 295.5 0.8975 500.4 278.3 0.9452 q1 257.2 621.1 0.9855

q2 734.9 589.0 0.8963 300.6 700.7 0.8978 q2 478.8 4945.6 0.9321
q3 721.2 205.1 0.9012 799.5 600.4 0.9685 q3 408.3 568.3 0.9257
q4 130.8 312.9 0.8665 335.0 378.6 0.9358 q4 376.7 41.6 0.9633
q5 200.3 198.2 0.9543 399.1 378.5 0.9787 q5 497.4 65.4 0.9878
q6 601.7 325.8 0.9171 410.9 654.0 0.9963 q6 621.6 47.6 0.9688
Table 10
MAPE and EP calculated for trajectory (6.8).
Trajectory 3
MAPE Error percentage

qi without external disturbances with external disturbances without external disturbances with external disturbances
FOISM P Iψ − P D℘ FOISM P Iψ − P D℘ FOISM/ P Iψ − P D℘ FOISM/ P Iψ − P D℘
q1 1.47 1.54 1.48 2.01 98.24 93.52

q2 1.99 2.12 2.01 2.82 98.35 94.34
q3 1.23 1.41 1.26 2.17 97.73 92.95
q4 0.87 1.09 0.88 1.81 97.65 93.42
q5 0.99 1.07 0.99 1.64 98.63 91.85
q6 1.01 1.22 1.02 1.89 98.75 91.34
integration step size h = 0.0 0 01 were used in all numerical simulations. Table 2 shows the gains and orders of the PIζ
and PIη controllers defined in (3.11a) and (3.11b) respectively, which have been tuned by Cuckoo algorithm [65] with the
root-mean-square error (RMSE) as the cost function to be minimized. The RMSE is defined as

1
m
2
RMSE = qdi − qi , (6.2)
m
i=1
with m the total number of measurements. Meanwhile, Table 3 contains the parameters of the fractional integral sliding-
mode controller described by Eqs. (4.2), (4.15) and (4.18). Similarly, Table 4 shows the tuned gains of the conventional
version of our proposed control scheme consisting of the PI controllers (3.8a) and (3.8b); and an integral sliding-mode
controller given by the Eqs. (4.16), (4.17), and (4.18). Finally, Table 5 shows the results of the tuning of the super-twisting
452
sliding mode control proposed in Al-Dujaili et al. [66], where the constants ci in the first column correspond to the entries
of c defined in Eqs. (4) and (9) of [66], while the second and third columns contain the values of ki and wi given in Eq. (12).
6.1. Without disturbance
Figure 3 illustrates the trajectory tracking for all the joint angles when the fractional integral sliding-mode control
scheme, its integer-order counterpart, and the optimal super-twisting sliding mode control are applied. As seen the frac-
tional integral sliding mode control exhibits a significantly better tracking performance than the other two controllers,
where the most noticeable effect is seen in Fig. 3(e) in which the overshoot has been considerably attenuated. In addi-
tion, Tables 6 and 7 show the mean absolute percentage error (MAPE) [67] and the error percentage (EP) [68] for all joints
453
Fig. 8. End-effector trajectory (6.7).
respectively, calculated as

i di
m q −q
qi
i=1
MAP E = × 100, (6.3)
m
efractional
EP = × 100, (6.4)
eother
thus validating our method. FOISM, ISM, and STSM denote the fractional-order integral sliding-mode, conventional integral
sliding-mode, and super-twisting sliding-mode controllers respectively. Note that Eq. (6.4) represents an error ratio between
our controller and some other control scheme with which you want to compare. In addition, this quotient is multiplied by
100 to obtain the index representing the error percentage. In this context, EP > 100 means that the proposed fractional
order controller has a lower performance, on the contrary, for EP < 100, it is concluded that our controller has a better
performance in terms of trajectory tracking. Therefore, Table 7 demonstrates a better performance of our control scheme.
The end effector trajectory (see Fig. 4) also confirms this fact.
6.2. With disturbance
Robot manipulators may be subject to disturbances that negatively affect their performance. These disturbances could
be caused by different factors such as external forces and moments, joint frictions, and model uncertainties, to mention a
few [69]. In this sense, the friction forces and an external force are considered in the robot dynamics to demonstrate the
robustness of our control scheme to different kinds of disturbances. For the friction term in Eq. (3.2) we have [70]
CF C CF C
F 0 Dtϑ q = fC · sign 0 Dtϑ q + fv · CF C ϑ
0 Dt q, (6.5)
T
where the first and second terms denote Coulomb friction and viscous friction respectively, while fC = fC1 , fC2 , . . . , fC6
T
are their corresponding coefficients. Now if there are external forces and moments, [fext , τ ext ] ,
T
and fv = fv1 , fv2 , . . . , fv6
acting at the end-effector, the torques generated, defined in Eq. (3.2), can be expressed as Spong et al. [71]
τ d = JT (q )[fext , τ ext ]T , (6.6)

T T
with J(q ) the Jacobian for the end-effector (Appendix B), while fext = fextx , fexty , fextz and τ ext = τextx , τexty , τextz .In this
T
context, τ d components are set to τ ext = [0, 0, 0]T and fext = 100 cos π4 N, 100 sin π4 N, 0 for robustness evaluation pur-
poses of our control strategy, while the coefficients are selected as fC = [0.5, 0.5, . . . , 0.5]T (N-m) and fv = [0.8, 0.8, . . . , 0.8]T
(N-m-s). The same gains and orders used in the previous numerical simulation were kept (see Tables 2–5). Figures 5 and
6 give a comparison between the controllers considered in this work, which demonstrates that our control strategy has bet-
ter trajectory tracking performance, even under external disturbances. Tables 6 and 7 validate our control strategy through
the mean absolute percentage error and the error percentage for all the joints.
454
6.3. Reference trajectory change
The performance of our controller to trajectory changes is now analyzed, so that, an alternative reference trajectory is
defined as
px = 7.50 cos (t )
py = 7.52 sin (t ) (6.7)
pz = −0.15
Using the same controller gains and orders (Tables 2–5), this alternative trajectory is used to determine the robustness of
the controllers to cope with trajectory changes without the need to perform a new tuning. Figures 7 and 8 demonstrate the
superiority of our controller to track different reference trajectories despite variations in them. All the joint variables tracked
455
Fig. 10. Joint trajectories corresponding to (6.8) with disturbances: (a) q1 , (b) q2 , (c) q3 , (d) q4 , (e) q5 , (f) q6 .
the reference trajectories with minor errors for the fractional control scheme. This fact is confirmed by Tables 6 and 7 which
show that the proposed fractional control scheme improves the performance indices (MAPE and EP) for all the joints.
6.4. Comparison with P Iψ − P D℘ control scheme
In [72] a fractional control scheme of the PUMA 560 manipulator (see Fig. 2 in the mentioned work) consisting of PIψ
controllers for the actuators based on the Atangana–Baleanu integral (2.8), and a PD℘ control strategy for trajectory tracking
generalized via the Caputo–Fabrizio derivative (2.1), was presented. Such a strategy is now implemented in the Stanford
robot for comparison purposes with our fractional-order integral sliding-mode control scheme (FOISMC). The reference tra-
456
jectory for each articulation has the following form:

q − q πt
!
fi ni
qdi (t ) = qni + 1 − cos , (6.8)
2 tf
with the values of the parameters defined in Table 8. Simulation time was limited to t f = 3 s and the movement of joint
angles to be controlled was observed. Table 9 shows the tuned parameters for the PIψ and PD℘ fractional-order controllers
given by Eqs. (4.6) and (6.12) of Lavín-Delgado et al. [72] respectively, which were obtained with the Cuckoo algorithm
[65]. The trajectory tracking performance of the system under both controllers is shown in Fig. 9. For an easier comparison,
the root mean square error (6.2) for all the joint variables is shown in Table 10, where it is observed that the proposed
control law provides better tracking performance compared to the PIψ − PD℘ controller. In addition, due to the benefits of
the integral action through the Atangana–Baleanu operator (2.8) in controller design, it possesses lower steady state error
and faster transient response.
On comparing the real joint trajectories with the desired ones, by considering again the external disturbances modeled by
Eq. (6.6), from Fig. 10 and Table 10 we can say that the FOISMC is indeed robust with respect to the external disturbances
and, therefore, its superiority over the PIψ − PD℘ control scheme is confirmed.
7. Conclusions
In this research, a fractional control strategy for trajectory tracking tasks of the Stanford robot driven by induction mo-
tors is designed and developed. This control strategy consists of fractional PI controllers for the actuators based on the
Atangana–Baleanu integral combined with a fractional integral sliding-mode control law generalized to an arbitrary order
through the Caputo–Fabrizio derivative and the aforementioned integral. The mathematical model of the coupled system
composed of the robot and its actuators is obtained using the Euler–Lagrange formulation and generalized via the Caputo–
Fabrizio derivative, so that, the traditional behavior is recovered by setting the order to 1. In this context, this derivative
is expressed in the Caputo sense so that it allows to give a real physic meaning to the coupled system robot manipulator-
induction motors. The integer-order integral sliding mode controller and the optimal super-twisting sliding mode control
were also developed for comparison purposes. All controllers were tuned using the Cuckoo optimization approach. To val-
idate the proposed fractional-order controller against different operating scenarios, it is put to a robustness test based on
trajectory changes and external disturbances applied on the end effector. Numerical simulations affirmed that the proposed
fractional control scheme has better performance, in terms of the mean absolute percentage error and the error percent-
age, for tracking trajectory tasks than its integer-order counterpart and the optimal super-twisting sliding mode control. The
most noticeable effect can be seen at joint 5 (q5 ), in which the overshoot has been considerably attenuated. In addition,
the PIψ − PD℘ fractional order controller described in Lavín-Delgado et al. [72] was also implemented for the Stanford robot
and compared with the proposed control scheme, in this case different reference trajectories were considered. The simula-
tion results validated the superiority of the fractional integral sliding-mode controller over the other fractional control. As
future work, obtaining real data about the Stanford robot dynamics and comparing the fractional and integer order models,
even considering other fractional derivatives reported in the literature. In addition, to reduce the necessary torques for an
assigned task, different optimization techniques, cost functions, and performance indexes will be analyzed. Finally, we can
consider adding more factors to the robustness test in future works, e.g., model the parametric uncertainties as zero-mean
Gaussian distributed noise, etc.
Funding information
Not applicable.
Code or data availability
The data shown in the figures are available from the author upon request. Other datasets were not generated or analyzed
during the current study.
Ethics approval
This paper does not contain research that requires ethical approval.
Consent to participate
All participants consented to involve in the creation of this manuscript.
Consent for publication
All participants consented to publish this manuscript.
457
Declaration of Competing Interest
Authors declare that they have no conflict of interest.
CRediT authorship contribution statement
Samuel Chávez-Vázquez: Investigation, Methodology, Validation, Writing – original draft, Writing – review & editing.
Jorge E. Lavín-Delgado: Conceptualization, Investigation, Methodology, Validation, Writing – original draft, Writing – review
& editing. José F. Gómez-Aguilar: Conceptualization, Methodology, Validation, Writing – review & editing. José R. Razo-
Hernández: Formal analysis, Investigation, Validation. Sina Etemad: Methodology, Validation, Writing – review & editing.
Shahram Rezapour: Methodology, Validation, Writing – review & editing.
Data availability
No data was used for the research described in the article.
Acknowledgments
Samuel Chávez Vázquez acknowledges the support provided by CONACyT through the assignment of doctoral fellowship.
José Francisco Gómez Aguilar acknowledges the support provided by CONACyT: Cátedras CONACyT para jóvenes investi-
gadores 2014 and SNI-CONACyT. S. Etemad and S. Rezapour would like to thank Azarbaijan Shahid Madani University.
Appendix A. Inverse kinematics
According to the Fig. 1, the Denavit–Hartenberg parameters associated with the Stanford manipulator are shown in
Table A.11.
The homogeneous transformation from the frame i to the frame i − 1 is given by
⎡ ⎤
Cθi −Sθi Cαi Sθi Sαi aiCθi
⎢ S θi Cθi Cαi −Cθi Sαi a i S θi ⎥
Ai = ⎣ ⎦. (A.1)
0 Sαi Cαi di
0 0 0 1
Thus, the position and orientation of the end-effector is given by
⎡ ⎤
r11 r12 r13 tx
⎢ r21 r22 r23 ty ⎥
T60 = A1 A2 A3 A4 A5 A6 = ⎣ ⎦, (A.2)
r31 r32 r33 tz
0 0 0 1
where,
r11 = [(C1C2C4 − S1 S4 )C5 − C1 S2 S5 ]C6 − (C1C2 S4 + S1C4 )S6 ,
r21 = [(S1C2C4 + C1 S4 )C5 − S1 S2 S5 ]C6 − (S1C2 S4 − C1C4 )S6 ,
r31 = (−S2C4C5 − C2 S5 )C6 + S2 S4 S6 ,
r12 = −[(C1C2C4 − S1 S4 )C5 − C1 S2 S5 ]S6 − (C1C2 S4 + S1C4 )C6 ,
r22 = −[(S1C2C4 + C1 S4 )C5 − S1 S2 S5 ]S6 − (S1C2 S4 − C1C4 )C6 ,
r32 = (S2C4C5 + C2 S5 )S6 + S2 S4C6 ,
r13 = (C1C2C4 − S1 S4 )S5 + C1 S2C5 ,
r23 = (S1C2C4 + C1 S4 )S5 + S1 S2C5 ,
r33 = −S2C4 S5 + C2C5 ,
tx = q3C1 S2 − d2 S1 ,
ty = q3 S1 S2 + d2C1 ,
tz = q3C2 .
Table A1
Denavit–Hartenberg parameters.
i θi di ai αi
1 q1 0 0 −90◦
2 q2 d2 0 90◦
3 0 q3 0 0◦
4 q4 0 0 −90◦
5 q5 0 0 90◦
6 q6 0 0 0◦
458
On the other hand, inverse kinematics is the problem of finding the joint variables from the given position and orientation
of the end-effector [73]. In this sense, given an arbitrary position and orientation of the end-effector
⎡ ⎤
nx ox ax px
⎢ ny oy ay py ⎥
T60 = ⎣ ⎦. (A.3)
nz oz az pz
0 0 0 1
Pre-multiplying Eq. (A.3) by A−1
1
, and equating the entries in position Eq. (3.4) we obtain
d2 = −px S1 + pyC1 . (A.4)
By defining the auxiliary variables
"
r = p2x + p2y
(A.5)
φ = tan−1 ppyx ,
then
px = rCφ ,
(A.6)
py = r Sφ .
Substituting Eq. (A.6) into Eq. (A.4)
d2
Sφ −q1 = , (A.7)
r
2
and by the fact that Sφ + Cφ2 −q = 1, from Eq. (A.7), we obtain
−q1 1
!2
d2
Cφ −q1 = 1− . (A.8)
r
Combining Eqs. (A.7) and (A.8)
# $
p d2
q1 = tan−1
y
− tan−1 " . (A.9)
px ± r 2 − d22
Equating the entries (1, 4 ) and (2, 4 ), we have
pxC1 + py S1 = q3 S2 , (A.10)
−pz = −q3C2 . (A.11)

Combining Eqs. (A.10) and (A.11), it follows that
p C + p S
x 1 y 1
q2 = tan−1 . (A.12)
pz
2
A−1
1
, and equating the entries (3, 4 ), we obtain
q3 = pxC1 S2 + py S1 S2 + pzC2 , (A.13)
then, equating the entries (1, 3 ) and (2, 3 )
axC1C2 + ay S1C2 − az S2 = C4 S5 , (A.14)
−ax S1 + ayC1 = S4 S5 . (A.15)

dasdsadas Combining Eqs. (A.14) and (A.15), we have
!
−1 −ax S1 + ayC1
q4 = tan . (A.16)
axC1C2 + ay S1C2 − az S2

4
A−1
3
A−1
2
A−1
1
, and equating the entries (1, 3 ) and (2, 3 ) we obtain

C4 (axC1C2 + ay S1C2 − az S2 ) − S4 (ax S1 − ayC1 )
q5 = tan−1 . (A.17)
S2 (axC1 + ay S1 ) + azC2
Finally, Eq. (A.3) by A−1

5
A−1
4
A−1
3
A−1
2
A−1
1
, and equating the entries (1, 2 ) and (2, 2 )
%
S5 [S2 (oy S1 + oxC1 ) + ozC2 ] − C5 {C4 [C2 (oy S1 + oxC1 ) − oz S2 ] + S4 (oyC1 − ox S1 )}
q6 = tan−1 . (A.18)
C4 (oyC1 − ox S1 ) − S4 [C2 (oy S1 + oxC1 ) − oz S2 ]
459
Appendix B. Jacobian
The Jacobian of the end-effector is calculated as Tsai [74]

z0 × ( o6 − o0 ) z1 × ( o6 − o1 ) z2 z3 × ( o6 − o3 ) z4 × ( o6 − o4 ) z5 × ( o6 − o5 )
J60 (q ) = z0 z1 0 z3 z4 z5
, (B.1)
where
& & &
0 −l2 S1 d3C1 S2 − l2 S1
o0 = o1 = 0 , o2 = l2C1 , o3 = o4 = o5 = o6 = d3 S1 S2 + l2C1 ,
0 0 d3C2
& & & &
0 −S1 C1 S2 −C1C2 S4 − S1C4
z0 = 0 , z1 = C1 , z2 = z3 = S1 S2 , z4 = −S1C2 S4 + C1C4 ,
1 0 C2 S2 S4
&
(C1C2C4 − S1 S4 )S5 + C1 S2C5
z5 = z6 = (S1C2C4 + C1 S4 )S5 + S1 S2C5 ,
−S2C4 S5 + C2C5
Thus, from Eq. (B.1) we obtain
⎡ ⎤
J11 J12 ... J16
⎢ J21 J22 ... J26 ⎥
J60 = ⎢ . .. ⎥
⎣ .. ..
.
..
. .
⎦, (B.2)
J61 J62 ... J66

with
J11 = −d3 S1 S2 − l2C1 , J32 = −d3 S2 ,
J12 = d3C1C2 , J33 = J64 = C2 ,
J13 = J44 = C1 S2 , J42 = −S1 ,
J14 = J24 = J34 = 0, J43 = J53 = J63 = 0,
J15 = J25 = J35 = 0, J45 = −C1C2 S4 − S1C4 ,
J16 = J26 = J36 = 0, J46 = (C1C2C4 − S1 S4 )S5 + C1 S2C5 ,
J21 = d3C1 S2 − l2 S1 , J55 = −S1C2 S4 + C1C4 ,
J22 = d3 S1C2 , J56 = (S1C2C4 + C1 S4 )S5 + S1 S2C5 ,
J23 = J54 = S1 S2 , J65 = S2 S4 ,
J31 = J41 = J51 = J62 = 0, J66 = −S2C4 S5 + C2C5 ,
J61 = 1, J52 = C2 .
References
[1] D. Shi, J. Zhang, Z. Sun, G. Shen, Y. Xia, Composite trajectory tracking control for robot manipulator with active disturbance rejection, Control Eng.
Pract. 106 (2021) 1–8, doi:10.1016/j.conengprac.2020.104669.
[2] N. Farah, M.H. Talib, Z. Ibrahim, et al., Analysis and investigation of different advanced control strategies for high-performance induction motor drives,
TELKOMNIKA (Telecommun. Comput. Electron. Control) 18 (6) (2020) 3303–3314, doi:10.12928/telkomnika.v18i6.15342.
[3] M.W. Spong, An historical perspective on the control of robotic manipulators, Annu. Rev. Control, Robot., Auton. Syst. 5 (2022) 1–31, doi:10.1146/
annurev-control-042920-094829.
[4] X. Yu, B. Li, W. He, Y. Feng, L. Cheng, C. Silvestre, Adaptive-constrained impedance control for human-robot co-transportation, IEEE Trans. Cybern. 52
(12) (2021) 13237–13249, doi:10.1109/TCYB.2021.3107357.
[5] X. Yu, W. He, Q. Li, Y. Li, B. Li, Human-robot co-carrying using visual and force sensing, IEEE Trans. Ind. Electron. 68 (9) (2020) 8657–8666, doi:10.
1109/TIE.2020.3016271.
[6] J.A. Rojas-Quintero, J.A. Rojas-Estrada, J. Villalobos-Chin, V. Santibanez, E. Bugarin, Optimal controller applied to robotic systems using covariant control
equations, Int. J. Control 95 (6) (2022) 1576–1589, doi:10.1080/00207179.2020.1865570.
[7] Z. Zhou, H. Ji, Z. Zhu, Online sequential fuzzy dropout extreme learning machine compensate for sliding-mode control system errors of uncertain robot
manipulator, Int. J. Mach. Learn. Cybern. 13 (2022) 1–17, doi:10.1007/s13042- 022- 01513- x.
[8] Z. Zhou, B. Wu, Adaptive sliding mode control of manipulators based on fuzzy random vector function links for friction compensation, Optik 227
(2021) 1–14, doi:10.1016/j.ijleo.2020.166055.
[9] Z. Zhou, C. Wang, Z. Zhu, Y. Wang, D. Yang, Sliding mode control based on a hybrid grey-wolf-optimized extreme learning machine for robot manipu-
lators, Optik 185 (2019) 364–380, doi:10.1016/j.ijleo.2019.01.105.
[10] H. Sai, Z. Xu, S. He, E. Zhang, L. Zhu, Adaptive nonsingular fixed-time sliding mode control for uncertain robotic manipulators under actuator saturation,
ISA Trans. 123 (2022) 46–60, doi:10.1016/j.isatra.2021.05.011.
[11] Q.D. Le, H.-J. Kang, An active fault-tolerant control based on synchronous fast terminal sliding mode for a robot manipulator, Actuators 11 (7) (2022)
1–17, doi:10.3390/act11070195.
[12] I. Saidi, N. Touati, Sliding mode control of a 2DOF robot manipulator: a simulation study using artificial neural networks with minimum parameter
learning, Recent Patents Mech. Eng. 15 (2) (2022) 241–253, doi:10.2174/2212797614666210614160557.
[13] K. Shao, Nested adaptive integral terminal sliding mode control for high-order uncertain nonlinear systems, Int. J. Robust Nonlinear Control 31 (14)
(2021) 6668–6680, doi:10.1002/rnc.5631.
460
[14] K. Shao, J. Zheng, R. Tang, X. Li, Z. Man, B. Liang, Barrier function based adaptive sliding mode control for uncertain systems with input saturation,
IEEE/ASME Trans. Mechatron. 27 (6) (2022) 4258–4268, doi:10.1109/TMECH.2022.3153670.
[15] J. Sabatier, O.P. Agrawal, J.T. Machado, Advances in Fractional Calculus: Theoretical Developments and Applications in Physics and Engineering, first
ed., Springer, Dordrecht, Netherlands, 2007.
[16] B.M. Vinagre, I. Podlubny, L. Dorcak, V. Feliu, On fractional PID controllers: a frequency domain approach, IFAC Proc. Vol. 33 (4) (20 0 0) 51–56, doi:10.
1016/S1474- 6670(17)38220- 4.
[17] Y. Guo, B.-L. Ma, Global sliding mode with fractional operators and application to control robot manipulators, Int. J. Control 92 (7) (2019) 1497–1510,
doi:10.1080/00207179.2017.1398417.
[18] R. Sharma, S. Bhasin, P. Gaur, D. Joshi, A switching-based collaborative fractional order fuzzy logic controllers for robotic manipulators, Appl. Math.
Model. 73 (2019) 228–246, doi:10.1016/j.apm.2019.03.041.
[19] C. Copot, An application to robot manipulator joint control by using fractional order approach, J. Appl. Nonlinear Dyn. 8 (1) (2019) 55–66, doi:10.5890/
JAND.2019.03.005.
[20] R.R. Ardeshiri, M.H. Khooban, A. Noshadi, N. Vafamand, M. Rakhshan, Robotic manipulator control based on an optimal fractional-order fuzzy PID
approach: SiL real-time simulation, Soft Comput. 24 (5) (2020) 3849–3860, doi:10.10 07/s0 050 0- 019- 04152- 7.
[21] A. Izadbakhsh, P. Kheirkhahan, Adaptive fractional-order control of electrical flexible-joint robots: theory and experiment, Proc. Inst. Mech. Eng., Part
I 233 (9) (2019) 1136–1145, doi:10.1177/0959651818815384.
[22] J. Kumar, V. Kumar, K. Rana, Fractional-order self-tuned fuzzy PID controller for three-link robotic manipulator system, Neural Comput. Appl. 32 (2020)
7235–7257, doi:10.10 07/s0 0521- 019- 04215- 8.
[23] M. Zhou, Y. Feng, C. Xue, F. Han, Deep convolutional neural network based fractional-order terminal sliding-mode control for robotic manipulators,
Neurocomputing 416 (2020) 143–151, doi:10.1016/j.neucom.2019.04.087.
[24] Y. Zhang, X. Yang, P. Wei, P.X. Liu, Fractional-order adaptive non-singular fast terminal sliding mode control with time delay estimation for robotic
manipulators, IET Control Theory Appl. 14 (17) (2020) 2556–2565, doi:10.1049/iet-cta.2019.1302.
[25] Z. Anjum, Y. Guo, Finite time fractional-order adaptive backstepping fault tolerant control of robotic manipulator, Int. J. Control Autom. Syst. 19 (1)
(2021) 301–310, doi:10.1007/s12555- 019- 0648- 6.
[26] X. Zhang, W. Xu, W. Lu, Fractional-order iterative sliding mode control based on the neural network for manipulator, Math. Probl. Eng. 2021 (2021)
1–12, doi:10.1155/2021/9996719.
[27] A.P. Singh, D. Deb, H. Agrawal, V.E. Balas, Modeling, stability and fractional control of single flexible link robotic manipulator, in: A.P. Singh, D. Deb,
H. Agrawal, V.E. Balas (Eds.), Fractional Modeling and Controller Design of Robotic Manipulators, vol. 194, Springer, Cham, 2021, pp. 83–98.
[28] H. Mohammadi, S. Kumar, S. Rezapour, S. Etemad, A theoretical study of the Caputo–Fabrizio fractional modeling for hearing loss due to Mumps virus
with optimal control, Chaos, Solitons Fractals 144 (2021) 110668, doi:10.1016/j.chaos.2021.110668.
[29] S. Rezapour, S. Kumar, M.Q. Iqbal, H. A, S. Etemad, On two abstract Caputo multi-term sequential fractional boundary value problems under the
integral conditions, Math. Comput. Simul. 194 (2022) 365–382, doi:10.1016/j.matcom.2021.11.018.
[30] M.M. Matar, M.I. Abbas, J. Alzabut, M.K.A. Kaabar, S. Etemad, S. Rezapour, Investigation of the p-Laplacian nonperiodic nonlinear boundary value
problem via generalized Caputo fractional derivatives, Adv. Differ. Equ. 2021 (2021) 68, doi:10.1186/s13662- 021- 03228- 9.
[31] S.T.M. Thabet, S. Etemad, S. Rezapour, On a coupled Caputo conformable system of pantograph problems, Turkish J. Math. 45 (1) (2021) 496–519,
doi:10.3906/mat-2010-70.
[32] D. Baleanu, S. Etemad, S. Rezapour, A hybrid Caputo fractional modeling for thermostat with hybrid boundary value conditions, Bound. Value Probl.
2020 (2020) 64, doi:10.1186/s13661- 020- 01361- 0.
[33] D. Baleanu, S. Etemad, S. Rezapour, On a fractional hybrid integro-differential equation with mixed hybrid integral boundary value conditions by using
three operators, Alex. Eng. J. 59 (5) (2020) 3019–3027, doi:10.1016/j.aej.2020.04.053.
[34] D. Baleanu, H. Mohammadi, S. Rezapour, Analysis of the model of HIV-1 infection of CD4+ T-cell with a new approach of fractional derivative, Adv.
Differ. Equ. 2020 (2020) 71, doi:10.1186/s13662- 020- 02544- w.
[35] D. Baleanu, A. Jajarmi, H. Mohammadi, S. Rezapour, A new study on the mathematical modelling of human liver with Caputo–Fabrizio fractional
derivative, Chaos, Solitons Fractals 134 (2020) 109705, doi:10.1016/j.chaos.2020.109705.
[36] R. George, S.M. Aydogan, F.M. Sakar, M. Ghaderi, S. Rezapour, A study on the existence of numerical and analytical solutions for fractional integro-
differential equations in Hilfer type with simulation, AIMS Math. 8 (5) (2023) 10665–10684, doi:10.3934/math.2023541.
[37] D. Baleanu, S. Etemad, H. Mohammadi, S. Rezapour, A novel modeling of boundary value problems on the glucose graph, Commun. Nonlinear Sci.
Numer. Simul. 100 (2021) 105844, doi:10.1186/s13662- 019- 2407- 7.
[38] A.K. Sethi, M. Ghaderi, S. Rezapour, M.K.A. Kaabar, M. Inc, H.P. Masiha, Sufficient conditions for the existence of oscillatory solutions to nonlinear
second order differential equations, J. Appl. Math. Comput. (2021) 1–18, doi:10.1007/s12190- 021- 01629- 3.
[39] S. Etemad, I. Avci, P. Kumar, D. Baleanu, S. Rezapour, Some novel mathematical analysis on the fractal-fractional model of the AH1N1/09 virus and its
generalized Caputo-type version, Chaos, Solitons Fractals 162 (2022) 112511, doi:10.1016/j.chaos.2022.112511.
[40] M. Caputo, M. Fabrizio, A new definition of fractional derivative without singular kernel, Prog. Fract. Differ. Appl. 1 (2) (2015) 1–13, doi:10.12785/pfda/
010201.
[41] A. Atangana, D. Baleanu, New fractional derivatives with nonlocal and non-singular kernel: theory and application to heat transfer model, Therm. Sci.
20 (2) (2016) 763–769, doi:10.2298/TSCI160111018A.
[42] A. Atangana, I. Koca, Chaos in a simple nonlinear system with Atangana–Baleanu derivatives with fractional order, Chaos, Solitons Fractals 89 (2016)
447–454, doi:10.1016/j.chaos.2016.02.012.
[43] O.I. Marichev, S.G. Samko, A.A. Kilbas, Fractional Integrals and Derivatives: Theory and Applications, first ed., Gordon and Breach Science Publishers,
Switzerland, 1993.
[44] D. Baleanu, A. Fernandez, On some new properties of fractional derivatives with Mittag–Leffler kernel, Commun. Nonlinear Sci. Numer. Simul. 59
(2018) 444–462, doi:10.1016/j.cnsns.2017.12.003.
[45] F.L. Lewis, D.M. Dawson, C.T. Abdallah, Robot Manipulator Control: Theory and Practice, second ed., CRC Press, Boca Raton, 2003.
[46] G. Jumarie, Lagrangian mechanics of fractional order, Hamilton–Jacobi fractional PDE and Taylor’s series of nondifferentiable functions, Chaos, Solitons
Fractals 32 (3) (2007) 969–987, doi:10.1016/j.chaos.2006.07.053.
[47] W. Chen, H. Sun, X. Li, et al., Fractional Derivative Modeling in Mechanics and Engineering, first ed., Springer, Singapore, 2022.
[48] S. David, J.M. Balthazar, B. Julio, C. Oliveira, The fractional-nonlinear robotic manipulator: modeling and dynamic simulations, AIP Conf. Proc. 1493 (1)
(2012) 298–305, doi:10.1063/1.4765504.
[49] R.P. Paul, Robot mAnipulators: mAthematics, pRogramming, and cOntrol. The Computer Control of Robot Manipulators, first ed., MIT Press, Cambridge,
1981.
[50] D.R.K. Mayangsari, et al., Design PLC on induction motors using star-delta with manual and automatic technique, INAJEEE (Indonesian J. Electr. Electron.
Eng.) 5 (2) (2022) 38–45, doi:10.26740/inajeee.v5n2.p38-45.
[51] R. Marino, P. Tomei, C.M. Verrelli, Induction Motor Control Design, first ed., Springer Science & Business Media, London, 2010.
[52] R. Marino, S. Peresada, P. Valigi, Adaptive input-output linearizing control of induction motors, IEEE Trans. Autom. Control 38 (2) (1993) 208–221,
doi:10.1109/9.250510.
[53] A. Green, J.Z. Sasiadek, Dynamics and trajectory tracking control of a two-link robot manipulator, J. Vib. Control 10 (10) (2004) 1415–1440, doi:10.1177/
1077546304042058.
[54] L.T. Aboserre, A.A. El-Badawy, Robust integral sliding mode control of tower cranes, J. Vib. Control 27 (9–10) (2021) 1171–1183, doi:10.1177/
1077546320938183.
461
[55] J. Shi, H. Liu, N. Bajçinca, Robust control of robotic manipulators based on integral sliding mode, Int. J. Control 81 (10) (2008) 1537–1548, doi:10.1080/
00207170701749881.
[56] S.-Y. Chen, H.-H. Chiang, T.-S. Liu, C.-H. Chang, Precision motion control of permanent magnet linear synchronous motors using adaptive fuzzy
fractional-order sliding-mode control, IEEE/ASME Trans. Mechatron. 24 (2) (2019) 741–752, doi:10.1109/TMECH.2019.2892401.
[57] Y. Fang, J. Fei, D. Cao, Adaptive fuzzy-neural fractional-order current control of active power filter with finite-time sliding controller, Int. J. Fuzzy Syst.
21 (5) (2019) 1533–1543, doi:10.1007/s40815- 019- 00648- 4.
[58] J. Fei, X. Liang, Adaptive backstepping fuzzy neural network fractional-order control of microgyroscope using a nonsingular terminal sliding mode
controller, Complexity 2018 (2018) 1–12, doi:10.1155/2018/5246074.
[59] V. Utkin, J. Guldner, J. Shi, Sliding Mode Control in Electro-Mechanical Systems, second ed., CRC Press, Boca Raton, 2017.
[60] A.-J. Muñoz-Vázquez, V. Parra-Vega, A. Sánchez-Orta, Fractional integral sliding modes for robust tracking of nonlinear systems, Nonlinear Dyn. 87 (2)
(2017) 895–901, doi:10.1007/s11071- 016- 3086- 5.
[61] S. Nadda, Integral sliding mode control for position control of robotic manipulator, in: 2018 5th International Conference on Signal Processing and
Integrated Networks (SPIN), 2018, pp. 639–642, doi:10.1109/SPIN.2018.8474267.
[62] D.S. Oliveira, E.C. de Oliveira, On a Caputo-type fractional derivative, Adv. Pure Appl. Math. 10 (2) (2019) 81–91, doi:10.1515/apam- 2017- 0068.
[63] A. Atangana, K.M. Owolabi, New numerical approach for fractional differential equations, Math. Model. Nat. Phenom. 13 (1) (2018) 1–21, doi:10.1051/
mmnp/2018010.
[64] K. Diethelm, N.J. Ford, A.D. Freed, Detailed error analysis for a fractional adams method, Numer. Algorithms 36 (1) (2004) 31–52, doi:10.1023/B:
NUMA.0 0 0 0 027736.85078.be.
[65] X.-S. Yang, S. Deb, Engineering optimisation by Cuckoo search, Int. J. Math. Model. Numer. Optim. 1 (4) (2010) 330–343, doi:10.1504/IJMMNO.2010.
035430.
[66] A.Q. Al-Dujaili, A. Falah, A.J. Humaidi, D.A. Pereira, I.K. Ibraheem, Optimal super-twisting sliding mode control design of robot manipulator: design and
comparison study, Int. J. Adv. Robot. Syst. 17 (6) (2020) 1–17, doi:10.1177/1729881420981524.
[67] H. Chaudhary, V. Panwar, N. Sukavanam, B. Chahar, Imperialist competitive algorithm optimised adaptive neuro fuzzy controller for hybrid force
position control of an industrial robot manipulator: a comparative study, Fuzzy Inf. Eng. 12 (4) (2020) 435–451, doi:10.1080/16168658.2021.1921378.
[68] A. Shrivastava, V.K. Dalla, P.N. Dal, Space debris manipulation by cooperative redundant planar robots with minimized trajectory error, Arabian J. Sci.
Eng. 47 (12) (2022) 15285–15302, doi:10.1007/s13369- 022- 06573- 3.
[69] S. Torres, J. Mendez, L. Acosta, V. Becerra, On improving the performance in robust controllers for robot manipulators with parametric disturbances,
Control Eng. Pract. 15 (5) (2007) 557–566, doi:10.1016/j.conengprac.20 06.10.0 03.
[70] N. Vuong, A.H. Marcelo Jr., Y. Li, S.Y. Lim, Improved dynamic identification of robotic manipulators in the linear region of dynamic friction, IFAC Proc.
Vol. 42 (16) (2009) 167–172, doi:10.3182/20090909- 4- JP- 2010.0 0 030.
[71] M. Spong, S. Hutchinson, M. Vidyasagar, Robot Modeling and Control, Wiley New York, 2006.
[72] J. Lavín-Delgado, J. Solís-Pérez, J. Gómez-Aguilar, R. Escobar-Jiménez, Trajectory tracking control based on non-singular fractional derivatives for the
PUMA 560 robot arm, Multibody Syst. Dyn. 50 (3) (2020) 259–303, doi:10.1007/s11044- 020- 09752- y.
[73] Y. Wei, S. Jian, S. He, Z. Wang, General approach for inverse kinematics of nR robots, Mech. Mach. Theory 75 (2014) 97–106, doi:10.1016/j.
mechmachtheory.2014.01.008.
[74] L.-W. Tsai, Robot Analysis: The Mechanics of Serial and Parallel Manipulators, first ed., John Wiley & Sons, New York, 1999.
462

1 s2.0 S0307904X23001452 Main

Uploaded by

Copyright:

Available Formats

1 s2.0 S0307904X23001452 Main

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

1 s2.0 S0307904X23001452 Main

Uploaded by

Copyright:

Available Formats

Applied Mathematical Modelling 120 (2023) 436–462

Contents lists available at ScienceDirect

Applied Mathematical Modelling

Trajectory tracking of Stanford robot manipulator by

Here are some deﬁnitions of fractional derivatives given.

2.1. Caputo–Fabrizio derivative

2.2. Atangana–Baleanu derivative

2.3. Riemann–Liouville integrals

where (z ) is the Gamma function

2.4. Atangana–Baleanu integral

Fig. 1. The Stanford robot.

3. Robot manipulator-induction motors dynamics

˙ = −α + MIsd , (3.7b)

Isq = [isq1 , isq2 , . . . , isqn ]T ,

vq = [vq1 , vq2 , . . . , vqn ]T .

4. Fractional integral sliding mode control

e(t ) = q(t ) − qd (t ), (4.3)

Fig. 2. Control scheme.

Similarly, the conventional version of the computed torque (4.2) is written as

Let Eq. (5.1) be a fractional ordinary differential equation

By denoting the error term as

Theorem 2 (Theorem 3.3 in Atangana and Owolabi [63]). Let ϒ (t ) be a solution of

Now well, the fractional state-space representation of (4.20) is deﬁned as

Theorem 3 (Theorem 2.5 in Diethelm et al. [64]).

J Rotor inertia 0.0586 kg m2

M Mutual inductance 0.068H

q1 966.2299 157.2703 416.8538 923.3089 897.9105 339.2808 0.1843

q1 150.8626 135.5645 235.9783

without external disturbances with external disturbances

FOISM ISM STSM FOISM ISM STSM FOISM ISM STSM

q1 2.33 2.54 2.49 2.83 3.54 3.43 2.54 3.64 3.33

Fig. 4. End-effector trajectory (6.1).

without external disturbances with external disturbances

q1 96.84 90.12 97.86 91.41 96.40 88.22

Fig. 6. End-effector trajectory (6.1) with disturbances.

qn i 0.4 −0.03 0.3 0.7 0.5 0.9

PIψ K pvd Kiv ψ K pvq Kivq ψ PD℘ Kp Kd ℘

q1 231.1 295.5 0.8975 500.4 278.3 0.9452 q1 257.2 621.1 0.9855

MAPE Error percentage

FOISM P Iψ − P D℘ FOISM P Iψ − P D℘ FOISM/ P Iψ − P D℘ FOISM/ P Iψ − P D℘

q1 1.47 1.54 1.48 2.01 98.24 93.52

6.1. Without disturbance

Fig. 8. End-effector trajectory (6.7).

6.2. With disturbance

τ d = JT (q )[fext , τ ext ]T , (6.6)

6.3. Reference trajectory change

6.4. Comparison with P Iψ − P D℘ control scheme

jectory for each articulation has the following form:

Code or data availability

All participants consented to involve in the creation of this manuscript.

Consent for publication

All participants consented to publish this manuscript.

Declaration of Competing Interest

Authors declare that they have no conﬂict of interest.

CRediT authorship contribution statement

No data was used for the research described in the article.

Appendix A. Inverse kinematics

−pz = −q3C2 . (A.11)

−ax S1 + ayC1 = S4 S5 . (A.15)