Escolar Documentos
Profissional Documentos
Cultura Documentos
PII: S0957-4174(16)30097-5
DOI: 10.1016/j.eswa.2016.03.006
Reference: ESWA 10575
Please cite this article as: Mehran Rahmani , Ahmad Ghanbari , Mir Mohammad Ettefagh , Robust
adaptive control of a bio-inspired robot manipulator using bat algorithm, Expert Systems With Applica-
tions (2016), doi: 10.1016/j.eswa.2016.03.006
This is a PDF file of an unedited manuscript that has been accepted for publication. As a service
to our customers we are providing this early version of the manuscript. The manuscript will undergo
copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please
note that during the production process errors may be discovered which could affect the content, and
all legal disclaimers that apply to the journal pertain.
ACCEPTED MANUSCRIPT
Highlights
T
The simulation results show the effectiveness of the proposed control.
IP
CR
US
AN
M
ED
PT
CE
AC
ACCEPTED MANUSCRIPT
Mehran Rahmani
Email: rahmani_mehran62@yahoo.com
Ahmad Ghanbari
T
Professor in Center of Excellence for Mechatronics, University of Tabriz, Iran
IP
Email: a-ghanbari@tabrizu.ac.ir
CR
Associate professor in Department of Mechanical Engineering, University of Tabriz, Iran
Email: Ettefagh@tabrizu.ac.ir
Abstract US
AN
This paper proposes a novel adaptive fractional order PID sliding mode controller (AFOPIDSMC) using a
Bat algorithm to control of a Caterpillar robot manipulator. A fractional order PID (FOPID) control is
M
applied to improve both trajectory tracking and robustness. Sliding mode controller (SMC) is one of the
control methods which provides high robustness and low tracking error. Using hybridization, a new
ED
combined control law is proposed for chattering reduction by means of FOPID controller and high
trajectory tracking through using SMC. Then, an adaptive controller design motivated from the SMC is
PT
applied for updating FOPID parameters. A metaheuristic approach, the Bat search algorithm based on the
echolocation behavior of bats is applied for optimal design of the Caterpillar robot in order to tune the
CE
parameter AFOPIDSMC controllers (BA-AFOPIDSMC). To study the effectiveness of Bat algorithm, its
performance is compared with five other controllers such as PID, FOPID, SMC, AFOPIDSMC and PSO-
AC
AFOPIDSMC. The stability of the AFOPIDSMC controller is proved by Lyapunov theory. Numerical
simulation results completely indicate the advantage of BA-AFOPIDSMC for trajectory tracking and
chattering reduction.
Keywords
Caterpillar robot, Fractional PID controller, SMC, Robust control, adaptive control, Bat algorithm.
1. Introduction
ACCEPTED MANUSCRIPT
In the field of engineering, bio-inspiration (Hsu & Juang, 2013) can be investigated in mechanisms, designing
and control significantly is modifying state of the art in robotics (Juang, Chen & Jhan, 2015). Researchers
investigate the measuring effects of animal locomotion in order to define the optimal gait for given-size robot.
Limbless animals such as Caterpillars and Snakes are among the most accessible animals in each part of the Earth
due to their structures and locomotion mechanisms. Their locomotion mechanisms are based on the flexible
movements of their bodies. Investigating the Caterpillar locomotion in order to change it into a robot will be
T
effective as an engineering application. According to Fig. 1, the movement of the Caterpillar could be categorized in
IP
different steps. First of all, the extending situation of the Caterpillar is considered as a primary status, then the
CR
endpoints contact with the ground. In the next step, the Caterpillar releases a wave, hence, its rear endpoint moves to
grasp the ground. In the last step, the frontal section moves, and finally the Caterpillar in a further situation return to
US
its initial position. This feature can be used in a robot in various fields such as inspecting gas pipes (Yamashita,
Matsui, Kawanishi, Kaneko, Murakami, Omori, & Asama, 2011; Wang, Song, Wang, Guo, & Tan, 2011; Bodnicki
AN
& Kamiński, 2014; Zhou, Tao, Cheng, Liu, & Fu, 2013), using in medical issues (Zarrouk, Sharf, & Shoham, 2010,
2011, 2012; Zarouk & Shoham, 2013). Ghanbari, Rostami, Noorani and Fakhrabadi (2008), considering the
inspiration of Caterpillar locomotion of the nature, proposed the model of the movements of the robot. They
M
classified the locomotion mechanism into two sub-mechanisms. As a result, they demonstrate the snapshot of
Caterpillar gait locomotion using Matlab program. As follows; the dynamic equations are obtained using Euler-
ED
Lagrange equations. And then, the trajectory tracking of joints has been optimized by Genetic Algorithm in order to
minimize the consumed effort. A reduction of 5 to 37 percent in torque consumption has been obtained (Ghanbari &
PT
Noorani, 2011). Hopkins and Gupta (2014) designed a robot that was inspired from the snake. Regarding its hyper-
redundant body, it can pass through tight spaces. This robot is faster than its previous structures that rectilinear gaits
CE
were used in them. The frictional force mainly occurs by high speed linear motion. The Worm robot has been built
in different structures such as Nickel Titanium Coil Actuators (Kim, Hawkes, Cho, Joldaz, Foleyz, & Wood, 2009;
AC
Seok, Onal, Cho, Wood, Rus, & Kim, 2013) and Origami (Onal, Wood, & Rus, 2011, 2013).
PID controller is a convenient method in trajectory tracking that is acceptable for accurately implementing and
tracking performances. The PID controller includes three independent parameters. By tuning the three parameters in
the PID controller, the controller can provide control operation designed for particular developing demands. Ayala
and Coelho (2012) presented a multi-objective genetic algorithm for PID parameters` tuning applied to a robotic
ACCEPTED MANUSCRIPT
manipulator. The proposed NSGA- optimization method provides an application process to implement a robust
solution for trajectory tracking in closed loop. Different methods such as Neural Networks (Yu & Rosen, 2013),
Fuzzy logic (Pan et al., 2015), Fuzzy Neural Network (Wai & Muthusamy, 2014) and etc., efficiently suggested
tuning the parameters of PID controllers. In recent years, the fractional controller has been developed in (Lan, Gu,
Chen, Zhou, & Luo, 2014; Tang, Zhang, Zhang, Zhao, & Guan, 2013; Muresan, Dulf, Copot, De Keyser, & Ionescu,
2015; Vinagre, Monje, Calderón, & Suárez, 2007). Podlubny (1999) proposed the fractional order PID controller as
T
an extension of PID controller. It utilizes fractional order Integrals and derivatives and also it can supply robustness
IP
and obtains higher efficiency than the common integer controllers (Li, Luo, & Chen, 2010). The suggested method
CR
in (Li, Luo, & Chen, 2010) is based on frequency analysis and is considered for SIS systems. Dumlu and Erenturk
(2014) observed that transient and steady-state error values have been reduced with the PI D controller for the
US
Maryland manipulator tracking control in comparison to the conventional PID controller.
In most cases that PID controller has been used, the lack of high robustness and the average of trajectory tracking
AN
are mentioned. SMC gives us the best tracking performance. This method is very useful and popular because of its
robustness in comparison to decompensation certainty in dynamic model. Different approaches have been carried
M
out for improving trajectory tracking robot arms by using SMC, such as an adaptive sliding mode control
(Soltanpour, Khooban, & Khalghani, 2014) and an expression of the sliding mode controller (SMC) for robot arms
ED
in generalized velocity components (GVC(Herman, 2005). Capisani and Ferrara (2012), while using SMC, have
minimized the value of chattering. In some cases, the robot arms are in contact with the surface. The force of the
PT
contact surface that can be obtained by using a sensor is at the risk of damaging. Zeinali and Notash (2010), using
SMC had reached to good trajectory tracking without using the sensor.
CE
A Metaheuristic optimization algorithm can be utilized in order to tune the controller parameters in different
engineering problems. Maldondo, Castillo and Malin (2013) proposed the optimization of the type-2 membership
AC
functions for the average approximation of type-2 fuzzy controller using PSO. They compared the simulation results
of the optimization using the PSO approach with genetic algorithm. Pedro, Danger, Dahunsi and Ali (2014)
proposed a nonlinear control method utilizing dynamic neural network-based input-output feedback linearization for
a quarter- car active vehicle suspension systems. They optimized the gains of the proposed controllers and the
weights of the dynamic neural network using PSO. Hashim, El-Ferik and Abido (2015) proposed a novel practical
ACCEPTED MANUSCRIPT
method based on fuzzy rules for online tuning L1 adaptive controller parameters. They reduced both the tracking
error and the controller input signal range with fuzzy controller, which is optimally tuned using PSO. Therefore, we
can apply PSO in different engineering problems as an effective optimization algorithm (Tavakkoli- Moghaddam,
Azarkish & Sadeghnejad- Barkousaraie, 2011; Zhu, Wang, Wang & Chen, 2011; Li & Wu, 2011). BA is completely
stronger than PSO, genetic algorithm and Harmony Search (Yang, 2010). The mainly reason is that BA uses an
appropriate combination of significant advantages of mentioned algorithms. Abd- Ekazim and Ali (2016) proposed
T
for optimal tuning of PI controllers for load frequency controller design. The numerical simulation results
IP
demonstrated the superiority of BA in comparing with Simulated Annealing in PI controller optimization. A PID
CR
cascade controller applied to the control of an interconnected, multi area thermal system. Controller parameters are
tuned at the same time using powerful evolutionary computational approach BA (Dash, Sakia & Sinha, 2015).
US
Osaba, Yang, Oiaz, Lopez-Garcia and Carballedo (2016) proposed a discrete version of the BA to solve the famous
Travelling Salesman problems. Furthermore, they presented a development in the fundamental structure of the
AN
classic BA. The experimental results have demonstrated that the proposed developed BA outperforms completely all
the other alternatives in most of the problems. So, BA can be considered as a powerful optimization tool (Svecko &
SMC has some good advantages such as acceptable tracking performance, resistance to disturbances and its
ED
disadvantages are poor stability and chattering. Because of the vast benefits of the FOPID controller, the main
problem of this controller can be stated as its low tracking performances. In this paper, a new robust adaptive control
PT
has been proposed, in which it is suggested that control system is able to eliminate their disadvantages, of FOPID
controller and a new robust adaptive control, to control a Caterpillar robot manipulator. An adaptive controller
CE
design is motivated from the sliding mode control and is applied for updating FOPID parameters. In the following,
Bat algorithm optimization was used to design and select the control parameter using an optimal approach. The
AC
stability of the proposed control system can be guaranteed considering the Lyapunov stability theorem. The
performance of the suggested control system has been compared with classical PID, SMC, FOPID, AFOPIDSMC
and PSO-AFOPIDSMC. The simulation results have been shown that the performance of proposed controller can be
The rest of this paper was organized as follows. In Section 2, the summary was explaining the motion of the
robot. In Section3, obtaining dynamic equations was defined through utilizing the Euler - Lagrange equation. The
ACCEPTED MANUSCRIPT
Section4 included the sliding mode controller. In Section5, the fractional-order PID controller was described.
Section6 defined adaptive robust fractional-order PID sliding mode controller. The implementation of fractional-
order operator was delineated In Section 7. In section 8, the bat algorithm was applied for tuning of proposed
controller parameters. Section9 was about simulation results. At the end, the conclusion and contributions of the
T
2. Caterpillar kinematic description
IP
In this section the main motion pattern of Caterpillar is defined to develop dynamic relations. A series of the
CR
connected joints is shown in that figure. They created robot forward locomotion. The distance movement of a
x 2l (1 cos )
US (1)
The main motion pattern could be categorized into four sub- mechanisms M1, M2, M2, M1. The joint angles at
AN
the end of each step are determined in Table 1. At least one link is unmoving in each sub-mechanism because the
adequate friction isprovided. The unmoving links are demonstrated with dark colors in Fig. 3 and Fig. 4. A
M
symmetry property in sequence motion of the robot can be discovered in Fig. 2. Therefore, the modeling of the two
sub-mechanisms, M1, M2, according to Fig.3 and Fig.4, is sufficient. At the Next step, coordinate systemis reserved
ED
for the last two sub-mechanisms and then contrary trajectory orientation is used. The angles in these models are
formulated in local coordinate system. So they will be converted into a global coordinate system since each sub-
mechanismis solved.The sub-mechanism M1 is known as a 3-R Planar manipulator. The geometries constrain of
PT
CE
n (2)
sin i 0
i 1
n, i can be described as the number of the joints, the angle of ith link with due to positive direction of horizontal
AC
axes respectively.
0 0, i i 1 i (3)
3. Dynamic modeling
In this section the dynamic equations of sub-mechanism M1 will be obtained. According to Fig.3 and Fig.4, the
ACCEPTED MANUSCRIPT
manipulator is assumed to move in vertical plane x-y. Pi and Vi are position and velocity of origin of ith link. So, the
velocity centroid of pth link and its square can be denoted as (Ghanbari & Noorani, 2011):
lp (4)
v p v p p sin p i cos p jˆ
2
p p 1 (5)
lp2
l l C
2
vp p 2 i j i j ij
4 i 1 j 1
T
Li denotes the length, mi represents the mass and Ii defines the moment of inertia of ith link.
IP
the abbreviationis used as follows:
CR
C ij cos i j , S ij sin i j
Also, the height of the centroid point of pth link is defined by the following equation (Ghanbari & Noorani,
2011):
1
h p Pp 1. jˆ l p sin p
p
l j sin j 2p sin p
l
US (6)
AN
2 j 1
1
3
2 1 (7)
m p p p 2p ml 2 (2812 1622 432 3612c12 1223c 23 1231c 31 )
2 p 1
24
ED
m
1
V p g hp mgl 5sin 1 3sin 2 sin 1
PT
p 1
2 (8)
Where Qk describes nonconservative forces, represents joint torques exerted to the links, FR and Ff denote the
CE
normal and frictional forces apply on the tip of the last link. Virtual works, with application of all the non-
conservative forces in the system and with respect to Coulomb friction model, is defined as:
AC
3 3 3
w . Ff i N jˆ . p 1 p p p 1 N l p sin p p l p cos p p
p 1 p 1 p 1
3 3 3
p 1
p p 1 l p cos p sin p N p p p 1 p N p Q p p
p 1 p 1
(9)
Where
p l p (cos p sin p ) , and sgn( tip . et ) , It should be described that the i=1, …,n,do not
i
ACCEPTED MANUSCRIPT
recognize an autonomous coordinate system. FR which is Lagrange’s coefficient is emerged due to the geometric
The motion equations can be obtained using Euler-Lagrange relations, so it can be represented as:
d
Q
dt (10)
If Eqs. (7), (8), and (9) are substituted in Eq. (10), the motion equations for sub-mechanism M1 will be obtained
T
(Ghanbari & Noorani, 2011):
IP
2
14 9c12 3c13 1 2
0 9 s12 3 s13 1 5cos 1 1 1 0 1 1
CR
2
ml m l m gl
6c12 8 3c 23 2 9 s12 0 3 s 23 22 3cos 2 0 1 1 2 2 FR
2 6 2 cos 0 0 1
3c13 3c 23 2 3 3 s
13 3 s 23 0 2 3 3 3
3
(11)
15 s12
2
9 s13 3 s14 1
2
9 s 23 3 s 24 2
7 cos 1
5cos 2
AN
ml 12 14 9c 23 3c 24 2 ml 15 s12
2 0 m gl
6 9c13 9c 23 8 3c 34 3 6 9 s13 9 s 23 0 3 s 34 2 2 3cos
3 3
3c14 3c 24 3c 34 2 4 3 s14 3 s 24 3 s 34 0 2 cos
4
4
1 1 0 0 1 1
M
0 1 1 0 2 2
F
0 0 1 1 3 3 R
0 0 0 1 4 4
ED
(12)
M C 2 G D FR (13)
Where , , are the joint positions, velocity and acceleration vectors.M () denotes the mass matrix of the
CE
manipulator, C() expresses the centrifugal coefficient matrix, G() is the gravity vector, and according to
Kronecker Delta , D is the subtraction matrix with D mn= mn- m(n-1) resolution. represents the applied torque on
AC
joints, d , is the vector of un-modeled dynamic disturbances. Obviously, Eq.(13) can be written as:
M 1 C 2 M 1 G d t M 1 D M 1 FR (14)
Where d t M 1
d , is external disturbance,from Eq.(14), the dynamic equations for a worm robot
defined as:
ACCEPTED MANUSCRIPT
Y 2 PG Q R FR (15)
Where Y M
1
C , P M 1 , and Q M 1 D , R M 1 and u t are the control
vector. Y , P , Q , and F present some uncertainties of parameter variations; So, Eq(15) can be arranged as:
Y Y 2 P P G Q Q u t R R FR (16)
T
By definition of l,u as lower and upper uncertainty values, the uncertainties can be bounded as:
IP
CR
4. Sliding mode controller
US
Sliding mode control is well known because of its robustness with a high tracking trajectory and its supplying of a
robust solution in order to estimate parameters, fault diagnosis and control. Sliding mode control (SMC) methods
AN
are developed to design a control system with complex, high-order nonlinear dynamic plant operating under
uncertainconditions and external disturbances, low sensitivity to plant parameter variations and modeling errors.
Determination of sliding surface s(t) is the most important and valuable factor in SMC design, which it can have a
M
good performance because of sliding surfaces. Therefore in order to define sliding surface as (Amer, Sallam, &
Elawady, 2011): s e e , where diag [1 , 2 , 3 ] has known as a vector of slopes, which is named bandwidth
ED
To obtain desired performance, the control endeavor being derived in the solution of s t 0 , without uncertainty
CE
s r (18)
AC
s Y 2 PG Qu t RFR r (19)
Therefore, if external disturbance occured, the equivalent control endeavor cannot compensate the favorable control
performance;Threfore, a secondary control endeavor can be designed to remove the effect of unpredictable
disturbances.
ACCEPTED MANUSCRIPT
The secondary control endeavor deals with reaching control endeavor byus(t). In this issue, the Lyapunov function
1 2
V t s t (21)
2
With V 0 0 and V t 0 for s t 0 .The reaching condition which can guarantee the trajectory tracking of
T
position error can be written as:
IP
V t s t s t 0 , s t 0 (22)
CR
The equivalent control term ueq(t) in Eq.(20) is completed by a control term us(t). On the other hand, the SMC law
u t ueq t u s t US (23)
AN
According to Eq. (22), the reaching control u s t can be obtained as:
ss s Y Y 2 P P G Q Q u t R R FR r
M
s Y Y 2 P P G Q Q Q 1 r Y 2 PG RFR u s t
(24)
R R FR r
ED
Y Q Y Q P PQ
1 1
Q G Q 1Q r Q 1R Q R FR s Q Q u s
PT
ss s 2
(25)
CE
Eq. (25) will be less than zero, of course certainty of this equation should be met, the reaching control can be
written as:
AC
u s t sign s Q Q
1
Y
Q 1Y Q 2 P PQ 1Q G Q 1Q r (26)
Visibly, if Eq. (26) is substituted with Eq. (25), V t 0 is obtained. Therefore, the reaching control can be
determined as:
Where r d e is the reference acceleration which permits the desired acceleration to be shifted according to
the velocity error. K is the bounds of uncertainty vector which has to be known as a prior sign(s) is a Signum
function.
First of all, one must choose Ks for joint angles, which represents a switching gain of SMC, in order to compensate
T
IP
Chattering is one of the disadvantages of SMC; So, the common method to reduce the chattering problem is to use
saturation function sat(s,). sgn(s) should be replaced with sat(s,), Where saturation functions can be defined as:
CR
sgn(s i ) , si 0
sat (s , ) si i 1, 2,3, 4
, si
PID controllers have been used widely as a conventional system in the most sections of industrial structures. It
ED
calculates the difference error between processing variables, which they are joint angles and desired setpoint in this
system. It uses Kp, Ki, Kd parameters in order to minimize the error which each of them causes a change of system
PT
properties.
de (t )
u PID (t ) K p e (t ) K i e (t )dt K d
dt
AC
e (t ) (t ) d (t )
(28)
Where Kp, Ki, Kdare the proportional, integral, derivative gains in Eq. (28), respectively.
d e (t ) d e (t )
U FOPID (t ) K P e (t ) K i Kd (29)
dt dt
Where and are fractional order operators; The FOPID in S domain can be expressed as:
1
U FOPID (S ) (K P K i K d S )E (S ) (30)
S
T
IP
6. Adaptive robust fractional-order PID sliding mode control
CR
FOPID and SMC have advantages and disadvantages. The low tracking trajectory is one of the FOPID
drawbacks. On the contrary, SMC is robust and stable against any perturbation and the main disadvantage of SMC is
US
chattering. Thus, according to Fig.5, by combining both of FOPID and SMC with regard to their effects, it can be
obtained a suitable control approach, which has the best performances. It can be written as:
AN
U (t ) U FOPID U SMC (31)
M
The control endeavor is derived in the solution of s (t ) 0 , which can be written as:
e e 0 (32)
ED
The notice that tracking error will tend to zero ( e (t ) 0 ), when time goes to infinity ( t ).
PT
The Lyapunov function is being selected to prove the stability of proposed controller, which it can be denoted as:
CE
1 2
v s (33)
2
AC
v ss (34)
ACCEPTED MANUSCRIPT
ss s Y Y 2 P P G Q Q u t R R FR r
s Y Y 2 P P G Q Q Q 1 r Y 2 PG RFR u s t u FOPID
(35)
R R FR r
Y Q
T
ss s Y Q 2 P PQ 1Q G Q 1Q r Q 1R Q R FR
1
(36)
Q Q U FOPID s Q Q u s s Q Q u FOPID
IP
CR
Noticeably, if Eq. (26) is substituted with Eq. (36), it guarantees V t 0 .
According to the gradient approach and the chain rule (Chang & Yan, 2005), and using Eq.(29) and Eq.(36), the
K i S (38)
K i U FOPID K i dt
ED
It is considered that the approach is utilized for implementing the fractional order integrators and differentiators.
Different authors have considered specific definitions such as Grunwald-Letinikov (G-L), Reimann and Lioville (R-
L), Oustaloup’s approximation, Cauto definition, MittagLefller (M-L) and etc. for the implementation of fractional
order operators; the fundamental fractional order differentiator and integrator can be denoted as follows:
ACCEPTED MANUSCRIPT
d 0
0 R (0 ) 0
dt
a0 Dt0 1 R (0 ) 0 (40)
t
(d ) 0
0
R (0 ) 0
Where R stands for Real, a0 is related to initial conditions and 0 is desired fractional order operator that can be
T
complex number. In some recent researches, the fractional order is carried out with the Oustaloup’s recursive
IP
approximation method. The approximation transfer which is available is similar to fractional operator S when is
CR
N ou
S W f zo
S K ou S W f po
(41)
K ou N ou
US
Where Kou is gain, W f zo indicates zeros and W f po indicates poles of the filter and can be defined as follows
AN
(Sathya & Ansari, 2015):
K ou N ou (1/2) ( /2)
M
W h 2 N ou 1
W f po W bou ou (42)
W b
ou
ED
K ou N ou (1/2) ( /2)
W h 2 N ou 1
W f zo W bou ou (43)
W b
PT
ou
K ou W h (44)
CE
ou
Therefore, is the order of fractional deferential or integrator; 2Nou+1 is the order of approximation; W bou ,W hou
AC
is the frequency range (Sathya & Ansari, 2015). This approximation is selected over other approaches because of its
feasibilities of implementing it in real hardware utilizing higher-order infinite impulse response type digital or
analog filters for the non-integer order of differential-Integrator. In this research, the value of N and frequency range
BAT bio-inspired algorithm is an optimization algorithm, inspired by the echolocation behavior of common bats
in locating their foods. It is proposed in (Xine-she, 2010; Yang, 2011) and is utilized for solving many optimization
problems.The bat algorithm utilizes the echolocation behavior of bats. These bats emit a very loud sound pulse and
listen for the echo that reflected from the surrounding things. Their signal bandwidthranges are contingent on the
species, and increases using harmonics. The ith bat flies randomly with velocity i at positionxiwith a fixed frequency
T
fmin. The bat changes its wavelength and loudness A to find food.Such strong echolocation capabilities of microdots
IP
can be associated with the objective function in order to be optimized. The bat behavior in searching the optimal
CR
solution can be formulated to optimizealgorithm. By idealizing some of the echolocation features of microbats, the
bat-inspired algorithms can be developed. The bat’s echolocation properties are to solve an optimization problem
US
which are created by the following assumptions (Sathya & Ansari, 2015).
3) They regulate their wavelength/frequency and can adjust pulse emission rate ri[0-1] according to the
M
4) When they come close to the prey, their loudness changes from large A0 to small Amin values.
Frequency occures in a range of [fmin, fmax] in practical implementations and is selected in such a way that it is
PT
analogous to the size of the domain of interest. Rules need to be determined to set their positions and velocities in
the d-dimensional search space for a virtual bat to solve an optimization problem. The new position xi+ and velocity
CE
i+ at time step t are defined as follows (Mitić & Miljković, 2015):
x it x it 1 it (47)
Where ξ[0-1] is the random vector taken from a uniform distribution, x* is the current global best solution
among all N bats. For local search, once a solution is chosen among the recent best solution, the new solution is
ACCEPTED MANUSCRIPT
obtained on the basis of recent loudness Ai of the bat and the maximum allowed variance max(var) during a time
stop as follows:
When a bat is found its prey, the loudness decreases and the rate of pulse emission increases. The bat is moving
T
IP
Ait 1 Ait , rit 1 rr [1 e t ] (49)
CR
Where and are constant. The initial emission rate is r0[0-1] and the initial boundness is Ai[0.1-0.9] and
==0.9. Bat algorithm is utilized for tuning the [ , , , , K] parameters of proposed controller for a Caterpillar
US
robot manipulator. The objective function of this problem is defined as (Zeng, Lu, Dai, Zhang, Chen, Zheng, &
Peng, 2014):
AN
J
0
(w 1 e (t ) w 2u 2 (t ))dt w 3t u (50)
M
In that objective function 0
w 2u 2 (t )dt occurs because of avoiding the exporting a large control value. Secondly,
ED
for evaluating the rapidity the response of a control system and the rising time w3tu is used. A flow chart of bat bio-
9. Simulation results
The proposed BA-AFOPIDSMC controller is applied for control of a Caterpillar robot manipulator. Also,
AC
different comprative approaches are applied to demonstrate the effectiveness of the proposed Bat algorithm for
tuning AFOPIDSMC parameters. The equations of movement and dynamic relations of the arm areobtained. All
simulations have been done using MATLAB program. In order to demonstrate the performance of the proposed
controller, simulation for each of sub-mechanisms are carried out. The characteristics of links for both sub-
mechanisms are tabulated in Table 2. Design of bat bio-inspired algorithm is carried out for tuning proposed
controller parameters of a Caterpillar robot manipulator. The block diagram of system is illustrated in Fig. 5. The
ACCEPTED MANUSCRIPT
parameters which are used for proposed controller in this paper are: total population= 40; number of iteration= 22;
loudness A= 0.5; wavelength r = 0.5; frequency fmin=0.2, fmax=0.8. In AFOPIDSMC controller optimum the gain
values are chosen by changes the mentioned parameters. The PSO parameters are defined as: population size =40,
number of iteration= 22, velocity constant c1=2, velocity constant c2=2 (GirirajKumar, Jayaraj & R.Kishan, 2010).
The tuned values of the proposed controller parameters are listed in Table 3. The variables of objective function for
T
9.1. Sub-mechanism M1
IP
, ,
CR
Acoording to Table 1, the desired reference joint angles have been selected as 4 4 4 . The parameters of
the PID controller areKp=diag850, 850, 850, Ki=diag250, 245, 247, Kd=diag300, 295, 298. For the
US
Fig. 7 shows the joint angle control since the classical PID, conventional SMC, FOPID, AFOPIDSMC, PSO-
AN
AFOPIDSMC and BA-AFOPIDSMC controllers are applied. Therefore, as you can see in Fig. 7, the maximum
overshoot and setting time in FOPID controller is lower than conventional PID controller, but it doesn’t have a
conventional trajectory tracking in comparison to SMC. In addition, The conventional sliding mode controller has
M
too much chattering. The incident of the unfavorable chattering phenomenon in the robotic system motivates high
order nonlinear dynamics and finally the closed-loop system will be unstable. Also in Fig. 7, with application of BA-
ED
AFOPIDSMC, the chattering will be removed and maximum overshoot, setting time converge to zero in a limited
time for joints 1, 2, 3, respectively. The total population and number of iteration in BA are the same in PSO. So, it is
PT
clear that with same characteristics in both BA and PSo, the performance of BA is much more stronger than PSO.
Fig. 8 shows the tracking error of joints under PID, conventional SMC, FOPID, AFOPIDSMC, PSO-AFOPIDSMC
CE
and BA-AFOPIDSMC controllers. The velocity of joints under mentioned controllers demonstrates in Fig. 9.
AC
9.2. Sub-mechanism M2
The desired reference joint angles are , , , in respect to Table 1. The parameters of the PID controller
4 4 4 4
are chosen as: Kp=diag850, 850, 850, 850, Ki=diag250, 245, 247, 246, Kd=diag300, 295, 298, 300.For the
SMC the controller gain is selected as: =diag150, 150, 150, 150. Fig. 10 demonstrates that chattering will be
ACCEPTED MANUSCRIPT
eliminated under proposed controller in comparison to others. The tracking error under BA-AFOPIDSMC controller
converges to zero, which depicted in Fig. 11. However, the proposed controller exhibits the best performance with
the reduction of maximum overshoot and setting time. The velocity of joint angles has demonstrated in Fig.12.
The motion of a Caterpillar robot has been simulated through the MATLAB program under Hybrid PID sliding
T
IP
10. Conclusion
CR
This paper proposes a novel AFOPIDSMC using BA for control of a Caterpillar bio-inspired robot manipulator.
All of the AFOPIDSMC optimized to improve accuracy. The BA-AFOPIDSMC controller method demonstrates
US
several advantages, including the priority supplying a nonlinear feedback, provides a model-free control since
standard SMC needs information of the dynamic system, excellent trajectory tracking performance, and chattering
AN
reduction by SMC. Also, it represents an effective robustness. Additionally, the performance of BA-AFOPIDSMC
is compared with five other controllers such as PID, FOPID, SMC, AFOPIDSMC and PSO-AFOPIDSMC. The
M
simulated results confirm the advantages of the BA-AFOPIDSMC approach to the Caterpillar robot motion problem.
In future, the proposed controller approach will be applied experimentally to control of a Caterpillar robot
ED
manipulator. As well as, BA method will be compared with multi-objective Bat algorithm in order to tune
References
Abd-Elazim, S.M., & Ali, E.S. (2016). Load frequency controller design via BAT algorithm for nonlinear interconnected power
AC
Dash, P., Saikia, L. C., & Sinha, N. (2015). Automatic generation control of multi area thermal system using Bat algorithm
optimized PD–PID cascade controller. International Journal of Electrical Power & Energy Systems, 68, 364-372.
Dumlu, A., & Erenturk, K. (2014). Trajectory Tracking Control for a 3-DOF Parallel Manipulator Using Fractional-Order
Control. Industrial Electronics, IEEE Transactions on, 61(7), 3417-3426.
Ghanbari, A., & Noorani, S. M. R. S. (2011). Optimal trajectory planning for design of a crawling gait in a robot using genetic
algorithm. International Journal of Advanced Robotic Systems, 8(1), 29-36.
Ghanbari, A., Rostami, A., Noorani, S. M. R. S., & Fakhrabadi, M. M. S. (2008). Modeling and simulation of inchworm mode
locomotion. In Intelligent Robotics and Applications (pp. 617-624). Springer Berlin Heidelberg.
GirirajKumar, S. M., Jayaraj, D., & Kishan, A. R. (2010). PSO based tuning of a PID controller for a high performance drilling
machine. International Journal of Computer Applications, 1(19), 12-18.
Hashim, H. A., El-Ferik, S., & Abido, M. A. (2015). A fuzzy logic feedback filter design tuned with PSO for L1 adaptive
controller. Expert Systems with Applications, 42(23), 9077-9085.
T
Herman, P. (2005). Sliding mode control of manipulators using first-order equations of motion with diagonal mass matrix.
Journal of the Franklin Institute , 11.
IP
Hopkins, J. K., & Gupta, S. K. (2014). Design and modeling of a new drive system and exaggerated rectilinear-gait for a snake-
inspired robot. Journal of Mechanisms and Robotics, 6(2), 021001.
Hsu, C. H., & Juang, C. F. (2013). Multi-objective continuous-ant-colony-optimized FC for robot wall-following control.
CR
Computational Intelligence Magazine, IEEE, 8(3), 28-40.
I. Podlubny, Fractional-order systems and PI_D_controller, IEEE Transactionson Automatic Control 44 (1999) 208–214.
Juang, C. F., Chen, Y. H., & Jhan, Y. H. (2015). Wall-following control of a hexapod robot using a data-driven fuzzy controller
learned through differential evolution. Industrial Electronics, IEEE Transactions on, 62(1), 611-619.
Juhász, Z., & Zelei, A. (2013). Analysis of worm-like locomotion. Mechanical Engineering, 57(2), 59-64.
2234). IEEE. US
Kim, S., Hawkes, E., Cho, K., Joldaz, M., Foleyz, J., & Wood, R. (2009, October). Micro artificial muscle fiber using NiTi spring
for soft robotics. In Intelligent Robots and Systems, 2009. IROS 2009. IEEE/RSJ International Conference on (pp. 2228-
Lan, Y. H., Gu, H. B., Chen, C. X., Zhou, Y., & Luo, Y. P. (2014). An indirect Lyapunov approach to the observer-based robust
control for fractional-order complex dynamic networks. Neurocomputing, 136, 235-242.
AN
Li, C., & Wu, T. (2011). Adaptive fuzzy approach to function approximation with PSO and RLSE. Expert Systems with
Applications, 38(10), 13266-13273.
Li, H., Luo, Y., & Chen, Y. Q. (2010). A fractional order proportional and derivative (FOPD) motion controller: tuning rule and
experiments. Control Systems Technology, IEEE Transactions on, 18(2), 516-520.
Maldonado, Y., Castillo, O., & Melin, P. (2013). Particle swarm optimization of interval type-2 fuzzy systems for FPGA
M
IEEE.
Onal, C. D., Wood, R. J., & Rus, D. (2013). An origami-inspired approach to worm robots. Mechatronics, IEEE/ASME
Transactions on, 18(2), 430-438.
Pan, D., Gao, F., Miao, Y., & Cao, R. (2015). Co-simulation research of a novel exoskeleton-human robot system on humanoid
CE
374.
Seok, S., Onal, C. D., Cho, K. J., Wood, R. J., Rus, D., & Kim, S. (2013). Meshworm: a peristaltic soft robot with antagonistic
nickel titanium coil actuators. Mechatronics, IEEE/ASME Transactions on, 18(5), 1485-1497.
Soltanpour, M. R., Khooban, M. H., & Khalghani, M. R. (2014). An optimal and intelligent control strategy for a class of
nonlinear systems: adaptive fuzzy sliding mode. Journal of Vibration and Control , 18.
Svečko, R., & Kusić, D. (2015). Feedforward neural network position control of a piezoelectric actuator based on a BAT search
algorithm. Expert Systems with Applications, 42(13), 5416-5423.
Tang, Y., Zhang, X., Zhang, D., Zhao, G., & Guan, X. (2013). Fractional order sliding mode controller design for antilock
braking systems. Neurocomputing, 111, 122-130.
Tavakkoli-Moghaddam, R., Azarkish, M., & Sadeghnejad-Barkousaraie, A. (2011). A new hybrid multi-objective Pareto archive
PSO algorithm for a bi-objective job shop scheduling problem. Expert Systems with Applications, 38(9), 10812-10821.
ACCEPTED MANUSCRIPT
Vinagre, B. M., Monje, C. A., Calderón, A. J., & Suárez, J. I. (2007). Fractional PID controllers for industry application. A brief
introduction. Journal of Vibration and Control, 13(9-10), 1419-1429.
Wai, R. J., & Muthusamy, R. (2014). Design of Fuzzy-Neural-Network-Inherited Backstepping Control for Robot Manipulator
Including Actuator Dynamics. Fuzzy Systems, IEEE Transactions on, 22(4), 709-722.
Wang, Y. K., Song, C. N., Wang, Z. L., Guo, C., & Tan, Q. Y. (2011). A SMA Actuated Earthworm-Like Robot. In Intelligent
Computing and Information Science (pp. 619-624). Springer Berlin Heidelberg.
Xin-She, Y. A New Metaheuristic Bat-Inspired Algorithm, Nature Inspired Cooperative Strategies for Optimization (NISCO
2010). Studies, 284, 65-74.
Yamashita, A., Matsui, K., Kawanishi, R., Kaneko, T., Murakami, T., Omori, H., ... & Asama, H. (2011, December). Self-
localization and 3-D model construction of pipe by earthworm robot equipped with omni-directional rangefinder. In
Robotics and Biomimetics (ROBIO), 2011 IEEE International Conference on (pp. 1017-1023). IEEE.
Yang, X. S. (2011). Bat algorithm for multi-objective optimisation. International Journal of Bio-Inspired Computation, 3(5),
T
267-274.
Yu, W., & Rosen, J. (2013). Neural PID control of robot manipulators with application to an upper limb exoskeleton.
IP
Cybernetics, IEEE Transactions on, 43(2), 673-684.
Zarrouk, D., Sharf, I., & Shoham, M. (2010, May). Analysis of earthworm-like robotic locomotion on compliant surfaces. In
Robotics and Automation (ICRA), 2010 IEEE International Conference on (pp. 1574-1579). IEEE.
CR
Zarrouk, D., Sharf, I., & Shoham, M. (2011). Analysis of wormlike robotic locomotion on compliant surfaces. IEEE
Transactions on Biomedical Engineering, 58(2), 301-309.
Zarrouk, D., Sharf, I., & Shoham, M. (2012, May). Experimental validation of locomotion efficiency of worm-like robots and
contact compliance. In Robotics and Automation (ICRA), 2012 IEEE International Conference on (pp. 5080-5085). IEEE.
Zarrouk, D., & Shoham, M. (2013, May). Energy requirements of inchworm crawling on a flexible surface and comparison to
Zeng, G. Q., Lu, K. D., Dai, Y. X., Zhang, Z. J., Chen, M. R., Zheng, C. W., ... & Peng, W. W. (2014). Binary-coded extremal
optimization for the design of PID controllers. Neurocomputing, 138, 180-188.
AN
Zhou, M., Tao, Y., Cheng, L., Liu, W. T., & Fu, X. (2013). A Biomimetic Earthworm-Like Micro Robot Using Nut-Type
Piezoelectric Motor. In Intelligent Robotics and Applications (pp. 129-135). Springer Berlin Heidelberg.
Zhu, H., Wang, Y., Wang, K., & Chen, Y. (2011). Particle Swarm Optimization (PSO) for the constrained portfolio optimization
problem. Expert Systems with Applications, 38(8), 10161-10169.
M
ED
PT
CE
AC
ACCEPTED MANUSCRIPT
Table 1. Joint angles at the end of each step (Ghanbari & Noorani, 2011).
1* 2* 3* 4*
Start 0 0 0 0
M1 0
T
M2
IP
M 2 0
CR
M1 0 0 0 0
US
AN
Table 2. Parameters of Caterpillar robot model.
Mass (m) Link length (l) Friction Contact force Time step (t)
M
T
IP
CR
Fig. 1. Caterpillars locomotion (Juhász & Zelei, 2013).
US
AN
M
ED
T
IP
CR
US
AN
M
ED
Fig.5. Block diagram of the adaptive robust fractional order PID sliding mode controller.
PT
CE
AC
ACCEPTED MANUSCRIPT
Start
Initialize parameters
T
Determine pulse frequency for each bat
IP
Simulate the system and evaluate performance
CR
index J using Eq. 50 and evaluate fitness
US
Generate new positions for bats by adjusting frequency,
and updating velocities and positions using Eq. 49
AN
Select the bat with the best position, and generate a local
position around the selected best position of the bat
M
Yes
If iteration <n
AC
No
Output the best bats
End
Fig.6. Flow chart for a bat algorithm for tuning of adaptive robust fractional PID sliding mode controller
parameters.
ACCEPTED MANUSCRIPT
a 3 b 0.5
1.4 Reference Reference
2.5 1.2 PID PID
1
FOPID 0 FOPID
SMC SMC
2 0.8
AFOPIDSMC AFOPIDSMC
0.6 PSO-AFOPIDSMC PSO-AFOPIDSMC
-0.5
theta1 (rad)
theta2 (rad)
-0.1 0 0.1 0.2 BA-AFOPIDSMC BA-AFOPIDSMC
1.5
1
-1
-0.7
-0.8
0.5
-0.9
T
-1.5 -1
0 -1.1
-1.2
IP
0 0.1 0.2
-0.5 -2
0 5 10 15 0 5 10 15
time (sec) time (sec)
CR
c 0
Reference
-0.75 PID
-0.2
-0.8 FOPID
-0.85 SMC
-0.4 AFOPIDSMC
-0.9
US
-0.95 PSO-AFOPIDSMC
theta3 (rad)
-0.8
AN
-1
-1.2
-1.4
0 5 10 15
M
time (sec)
Fig. 7. Responses of joint angles (a, b, c) under BA-AFOPIDSMC, PSO-AFOPIDSMC, AFOPIDSMC, SMC,
FOPID and PID controllers in Sub-mechanism M1
ED
PT
CE
AC
ACCEPTED MANUSCRIPT
a 1.5 b 1.2
PID 0.4 PID
FOPID 1 FOPID
1 0.3
SMC 0.8 0.2 SMC
AFOPIDSMC 0.1 AFOPIDSMC
0.5 PSO-AFOPIDSMC 0.6 PSO-AFOPIDSMC
0
theta1 error (rad)
-1 -0.2 -0.2
T
-0.4 -0.4
-1.5 -0.6 -0.6
0 0.2 0.4
IP
-2 -0.8
0 5 10 15 0 5 10 15
time (sec) time (sec)
CR
0.6
PID
0.4 FOPID
SMC
AFOPIDSMC
0.2 PSO-AFOPIDSMC
theta3 error (rad)
US
BA-AFOPIDSMC
0
0.2
-0.2
0.15
0.1
AN
-0.4
0.05
0
-0.6
0 0.1 0.2 0.3 0.4
-0.8
0 5 10 15
M
time (sec)
Fig. 8. Responses of tracking error of joints (a, b, c) under BA-AFOPIDSMC, PSO-AFOPIDSMC, AFOPIDSMC,
SMC, FOPID and PID controllers in Sub-mechanism M1
ED
PT
CE
AC
ACCEPTED MANUSCRIPT
a 80 b 20
10
PID
60 0
FOPID 10
SMC
-10 AFOPIDSMC 0
40 PSO-AFOPIDSMC
thetadot1 (rad/sec)
thetadot2 (rad/sec)
-20 BA-AFOPIDSMC -10
20 0 0.1 0.2 0.3 0.4
10
-20
0 5
-30 PID
0 FOPID
T
-20
-40 SMC
0 0.1 0.2
AFOPIDSMC
-40
IP
-50 PSO-AFOPIDSMC
BA-AFOPIDSMC
-60 -60
0 5 10 15 0 5 10 15
time (sec) time (sec)
CR
c 20
10
US
thetadot3 (rad/sec)
-10
6
-20
4
AN
-30 2
PID
0 FOPID
-40 SMC
0 0.1 0.2 0.3 AFOPIDSMC
-50 PSO-AFOPIDSMC
BA-AFOPIDSMC
-60
M
0 5 10 15
time (sec)
Fig. 9. Responses of angular velocity joints (a, b, c) ) under BA-AFOPIDSMC, PSO-AFOPIDSMC, AFOPIDSMC,
ED
a 3 b 1.5 -0.5
Reference
2 1
2.5 PID -1
FOPID 0.5
1.5
SMC -1.5
2 AFOPIDSMC 0
1
PSO-AFOPIDSMC -2
theta2 (rad)
-0.5
theta1 (rad)
T
AFOPIDSMC
0 PSO-AFOPIDSMC
-3
BA-AFOPIDSMC
IP
-0.5 -3.5
0 5 10 15 0 5 10 15
time (sec) time (sec)
c d
CR
0.5 2
0
1.5
-0.5
1
US
theta3 (rad)
theta4 (rad)
-1
0.5
-1.5 Reference Reference
-0.6
PID 1 PID
-0.8 0
AN
-2 FOPID FOPID
-1 0.9
SMC SMC
-1.2
AFOPIDSMC -0.5 0.8
AFOPIDSMC
-2.5 -1.4
PSO-AFOPIDSMC PSO-AFOPIDSMC
-1.6
0 0.1 0.2 0.3
BA-AFOPIDSMC 0 0.1 0.2
BA-AFOPIDSMC
-3 -1
0 5 10 15 0 5 10 15
M
Fig. 10. Responses of joint angles (a, b, c, d) under BA-AFOPIDSMC, PSO-AFOPIDSMC, AFOPIDSMC, SMC,
FOPID and PID controllers in Sub-mechanism M2
ED
PT
CE
AC
ACCEPTED MANUSCRIPT
a 1.5 b 3
PID 1 PID
1 FOPID FOPID
SMC 2 SMC
0.5
AFOPIDSMC AFOPIDSMC
0.5 PSO-AFOPIDSMC PSO-AFOPIDSMC
1 0
theta1 error (rad)
T
-1 -2
-1.5
0 0.1 0.2 0.3
IP
-2 -3
0 5 10 15 0 5 10 15
time (sec) time (sec)
c d
CR
2.5 2
0.8 PID 0.1 PID
2 FOPID FOPID
0.6 1.5 0
SMC SMC
0.4 AFOPIDSMC
1.5 -0.1 AFOPIDSMC
0.2 PSO-AFOPIDSMC 1 PSO-AFOPIDSMC
theta3 error (rad)
US
theta4 error (rad)
-1 -1
-1.5 -1.5
0 5 10 15 0 5 10 15
time (sec)
M
time (sec)
Fig. 11. Responses of tracking error of joints (a, b, c, d) under BA-AFOPIDSMC, PSO-AFOPIDSMC,
AFOPIDSMC, SMC, FOPID and PID controllers in Sub-mechanism M2
ED
PT
CE
AC
ACCEPTED MANUSCRIPT
a 80 b 50
PID
15
60 FOPID
10 SMC
AFOPIDSMC
40 5
PSO-AFOPIDSMC
thetadot1 (rad/sec)
thetadot2 (rad/sec)
0
0 BA-AFOPIDSMC
20 0.2 0.4 0.6
30
0 20
-50 PID
10 FOPID
-20
SMC
T
0
AFOPIDSMC
-40 0 0.2 0.4
PSO-AFOPIDSMC
BA-AFOPIDSMC
IP
-60 -100
0 5 10 15 0 5 10 15
time (sec) time (sec)
c d
CR
60 120
PID PID
FOPID 100 FOPID
40 SMC SMC
AFOPIDSMC 80 AFOPIDSMC
PSO-AFOPIDSMC 5 PSO-AFOPIDSMC
thetadot3 (rad/sec)
thetadot4 (rad/sec)
20
US
BA-AFOPIDSMC 60 BA-AFOPIDSMC
0
0 40
25 -5
20 20
-20 0 0.1 0.2 0.3
AN
15
10
0
-40 5
-20
0
0 0.1 0.2 0.3
-60 -40
0 5 10 15 0 5 10 15
M
Fig. 12. Responses of angular velocity joints (a, b, c, d) under BA-AFOPIDSMC, PSO-AFOPIDSMC,
AFOPIDSMC, SMC, FOPID controllers in Sub-mechanism M2
ED
PT
CE
AC
ACCEPTED MANUSCRIPT
y
y
0 0 0
0 0.1 0.2 0.3 0.4 0 0.1 0.2 0.3 0.4 0 0.1 0.2 0.3 0.4
x x x
T
0.4 0.4 0.4
IP
0.3 0.3 0.3
y
y
0.2 0.2 0.2
CR
0.1 0.1 0.1
0 0 0
0 0.1 0.2 0.3 0.4 0 0.1 0.2 0.3 0.4 0 0.1 0.2 0.3 0.4
US
x x x
y
0.2 0.2 0.2
0 0 0
0 0.1 0.2 0.3 0.4 0 0.1 0.2 0.3 0.4 0 0.1 0.2 0.3 0.4
x x x
y
y
0 0 0
T
IP
0.4 0.4 0.4
y
0.2
y
y
0.2 0.2
CR
0 0 0
US
x x
y
0.2 0.2 0.2
0 0 0
M