Adaptive Sliding Mode Fuzzy PID Control:

Supervisory Control Systems

Mobin Saeedi Jafar Zarei Roozbeh Razavi-Far Mehrdad Saif
Department of Electrical Department of Electrical Department of Electrical Department of Electrical
and Electronics and Electronics and Computer Engineering and Computer Engineering
Engineering, Shiraz Engineering, Shiraz University of Windsor University of Windsor
University of Technology University of Technology Windsor, Canada Windsor, Canada
Shiraz, Iran Shiraz, Iran [email protected] [email protected]
[email protected] [email protected]

Abstract—This paper proposes a novel adaptive sliding performance in practice in the presence of modeling
mode controller (ASMC) combined with a fuzzy PID algorithm uncertainties, and external disturbances [9]–[11]. However,
which leads to global asymptotic stability. The sliding mode PID control has exposed its limitations and is not able to
approach effectively solves the problem of uncertainties and satisfy the control requirements in processes with uncertainty
unmolded dynamics of the system; however, the chattering in the mathematical model of the system and nonlinearity
phenomenon is the main drawback of this method. Proposed [11]. Furthermore, difficulty in the online tuning of
solutions in literature for overcoming the chattering problem parameters, large dead zone and its weakness in dealing with
cause ultimately globally bounded stability or suffer from nonlinearity and time-varying systems are some of the main
complicated calculations. In this paper, an improved adaptive
drawbacks of it [12].
robust fuzzy PID controller is presented as the three-layer
controller which can be considered for tracking and regulation With the development of control theory, the application
tasks. The simulation results illustrate the effectiveness of the of knowledge-based systems in control, such as fuzzy control
controller scheme. Applicability to the wide range of uncertain is growing [10]. In the fuzzy control theory as an intelligent
dynamic systems without the need of designing a fuzzy system control approach, linguistic explanations of human
for each of them, having chattering-free control signals and knowledge in the control process are represented as fuzzy
ultimately asymptotically stability are the main advantages of rules or relations. Besides, due to its robustness, fuzzy
the proposed method. control can solve the problem of uncertainties and unmolded
dynamics in complex control systems, such as position
Keywords- Adaptive Sliding, Fuzzy PID, Supervisory Control
control of a magnet ball suspension system (MBSS) [13].
I. INTRODUCTION A more advanced control strategy, fuzzy PID, combines
Sliding mode control (SMC) is well-known as an effective the advantages of fuzzy logic control and traditional PID
nonlinear control scheme in the presence of unmolded control, not only result in appropriate dynamical response
dynamics, uncertainties and disturbances [1]. and better robust performance for uncertain factors but also
Fast response and acceptable transient behavior are some of has the better adaptive behavior in dealing with nonlinearity
the main advantages of this method. Because of and has higher precision than PID control [6], [14].
improvement in robustness and performance in control of Ding Hai-Shan in [14] improved the control performance
uncertain dynamic systems, integration of SMC with benefited from an offline identification system, so he uses an
adaptive approaches are widely used [2]. unknown nonlinear function of the fuzzy tree identification
method. In some studies like in [12], [15] fuzzy PID
As extensively discussed in the literature, the adaptive
approach is presented and a combination of robust and fuzzy
sliding mode controller (ASMC) leads states trajectories
approach is introduced in [16].
toward sliding surface in a finite time, meanwhile, it can
estimate some parameters of parametric and un-parametric In several studies, the combination of the fuzzy method
uncertainties of the nonlinear dynamic systems [3]. and traditional PID are presented and the effectiveness of
The main drawback of the SMC approach is the chattering controller behavior is shown [17]. However, the most
phenomenon. Chattering causes extremely high-frequency important issue is designing fuzzy sets intervals and their
control effort which may damage actuators. Wide studies robustness against disturbances and initial error conditions. It
are done to deal with this problem and some methods such is obvious that the desired behavior of fuzzy sets is
as high order sliding mode [4], replacing Sign function with maintained whenever the error and its deviation rate remain
smooth functions are studied in the literature. Most of these in the predefined intervals.
methods suppress asymptotic stability into bounded stability In this paper, the adaptive robust fuzzy PID approach is
or suffers from complicated calculations. In this paper, a utilized for stabilizing a series of dynamic systems. Several
novel ASMC presented by utilizing of fuzzy PID approach. studies have been conducted using fuzzy methods and
The traditional control usually pursues PID control due getting advantages of combination fuzzy approach and
to its simple structure, good robustness and high reliability conventional PID method, however, implementing ASMC
[5]–[7]. It is well-established and can satisfy the performance with fuzzy PID improves robustness and response behavior
requirements of most industry plants [8]. The success of PID besides of being free of designing fuzzy sets inputs and
controller design is attributed to, ease of design,
good

978-1-7281-5365-0/20/$31.00 ©2020 IEEE

Authorized licensed use limited to: UNIVERSITY OF CONNECTICUT. Downloaded on May 16,2021 at 11:38:36 UTC from IEEE Xplore. Restrictions apply.
 membership functions in a different situation and initial where C is the positive constant, [ˆ is adaptation law which
errors.
is presented as follows,
It is important to note that the simple structure of the  (5)
proposed method, robustness and its ability to generalize for [ˆ S
the most dynamic systems are some of the main features of
an adaptive robust fuzzy PID algorithm. The proposed ­°0 e  D
controller implemented on a magnet ball suspension system I* ® (6)
to illustrate the superiority of the proposed approach. °̄1 e ! D
This reminder part of this paper is structured as follows:
In Section II, the adaptive sliding mode controller design is where D is the maximum value of fuzzy MF intervals and
discussed. In section III, the combination of a fuzzy approach u fuzz is the output of the fuzzy PID controller which is
with the designed controller in section II is presented, In introduced in section V.
section IV dynamic equations of well-known Magnet Ball
Suspension system(MBSS) are explained and simulation
results are given in Section V. Finally, Section VI concludes A. Stability Analysis:
the paper.
Stability analysis is divided into two separate parts, inside
II. CONTROLLER DESIGN & SYSTEM DESCRIPTION the boundary of e  D where the fuzzy PID controller
In this paper a class of uncertain high- order system guarantees the asymptotic stability of the system, outside the
considered as below, boundary where Adaptive Sliding Mode Controller
guarantees Ultimately Globally Bounded behavior and drive
x i x i 1 , 1 d i d n  1 the system’s states into the boundary [18]. The Lyapunov
(1) function is considered as follows,
x n F (X , t )  B (X , t ) u (t )  [ (X , t )
1 T 1 (7)
V S S  [T [
where X (t ) [ x 1 , x 2 ,..., x n ]T  R n and is the system’s 2 2
states, F (X , t )  R is a nonlinear function of X and t , 
where [ [  [ , then,
ˆ
[ ( X , t ) represents an unknown model and uncertainty, B is 
V S T S  [T [
(8)
the known and non-singular gain for control input and
With calculating (S ) and replacing it in (8):
u (t )  R is the control signal.

V S T (e(x )  K * e (x ))  [T [
(9)
Definition 1. Let F is R n o R
Given an open set B  R n , F is Lipschitz-continues on the Finally:
open subset B if there exists a constant ' , (10)
where F (x )  F ( y ) d ' x  y x , y  B 
V S T (F (X , t )  Bu (t )  [ (X , t )  X d  ke )  [T [
Assumption 1. All states, i.e. X, are measurable in real- By considering assumption 2,
time.   (11)
[ [ˆ
Assumption 2. Uncertainty dynamic [ ( X , t ) is bounded
and a slow varying-parameter that satisfies the following By inserting (4) and (5) into equation (10),
 (12)
assumption. V S T ( C sign (S )  [ (X , t ))  [T [ˆ
[ (X , t ) d [M Finally,
(2)
V C S  0 (13)
Let e x n  x nd represents the error vector with respect to
the vector of desired state variables. Sliding surface In the next sections designing u fuzz and its combination with
dynamic is presented as follows, the adaptive sliding mode approach designed in Section II is
t illustrated and implemented on the MBSS system.
S (X ) e (t )  k ³ e (t )dt
0 (3) The overall view of the controller scheme is shown in Fig.
1. Simulation results are presented in V and show the
effectiveness of the proposed method in different initial
where K  R 1*n is a positive definite constant matrix.
conditions and its robustness in the presence of
Proposed control law which guarantees ultimately
uncertainties.
asymptotically stability is presented as follows,

u B 1 (F ( X , t ) C sign (S )
(4)
[ˆ  ke  u  X )I *  u
fuzz d fuzz

Authorized licensed use limited to: UNIVERSITY OF CONNECTICUT. Downloaded on May 16,2021 at 11:38:36 UTC from IEEE Xplore. Restrictions apply.
 adaptation
law


Fuzzy [ˆ S Uncertain System’s Dynamic
System

System’s Output
u B 1 (F (X , t ) C sign (S )  [ˆ  ke  u fuzz  X d ) I *  u fuzz Dynamic Equation

X
Fig. 1. The overall structure of the ASM fuzzy PID controller

kd  [ kd min , k d max ] for simplifying, k 'p and kd' can be

III. FUZZY PID CONTROLLER DESIGN
defined as follows:
In this section, the traditional fuzzy PID controller
approach is presented. In subsection IV it has been combined
with an adaptive sliding mode approach as a supervisory k p  k p min (17)
k p'
controller for improving its efficiency and overcoming any k p  k p max
uncertainties according to section II. The results are 
compared with each other in the IV. k d  k d min
k d' (18)
The control input u fuzz (t ) is determined by the following k d  k d max
equation in the time domain:
Ti DT d (19)

(14) These three variables, k 'p , k d' and D are tuned by a fuzzy
1
u fuzz (t ) k p (e (t ) T d e(t ) 
Ti ³ e (t ) dt )  '
system. The membership functions of k d' , k p and D have
been shown in Fig. 2 and Fig. 3.
and it is represented in the frequency domain as follows:
kd s k
H (s ) kp   i (15)
1 s / E s

where k p is the proportional coefficient, k d and k i are

D
differential an integration constants respectively and E is
the positive constant which satisfies the following criterion:
ki Fig. 2. Membership functions defined for D
E !! (16)
kp

Changing in PID constants coefficients affects stability

and results in a faster dynamic response with less overshoot.
The fuzzy PID improves the robustness and speediness of
response by changing these coefficients during the process.
Fuzzy PID controller used in this paper is an interface made
of two inputs, error and its rate, and three outputs, three e (t ) e (t )
coefficients.
The rule definitions are shown in Table I and they have Fig. 3. Membership functions defined for e (t ) and e(t )
been implemented as follows:
If e is (Membership Function) and e is (Membership Theorem 1: Under Assumption 2, the control law shown
Function) Then K P , K d or K i is (Membership Function). in Equation (14) guarantees that the output tracking errors
and all signals in the closed-loop system remain Ultimately
Based on what is introduced by Wang [19] without loss Globally Bounded (UGB) inner domain e d D 0 [19].
of generality, with considering k p  [k p min , k p max ] and

Authorized licensed use limited to: UNIVERSITY OF CONNECTICUT. Downloaded on May 16,2021 at 11:38:36 UTC from IEEE Xplore. Restrictions apply.
 TABLE I. RULES DEFINITION OF THE FUZZY LOGIC SYSTEM [19]

K cp K dc D e
e

IV. MBSS CHARACTERISTIC AND SIMULATION where the inductance of winding is L , R defines the
RESULTS resistance of winding and u is the applied voltage to the
winding. The state equations of the system are as follows:
In this section proposed controller in sections II and III is
x1 x 2
implemented on Magnet Ball Suspension System (MBSS).
c x 32
The schematic of the MBSS is illustrated in Fig. 4. It is a x 2 g
nonlinear system for which achieving acceptable results mx 1 (23)
using a linear PID controller with constant coefficients is R x
difficult. Therefore, it would be beneficial to use nonlinear x 3 
x 1x 3  1 u
controllers. The considered problem here is the stabilizing of L0 L0
the ball in the desired position with different initial As it is illustrated in section I, choosing an efficient
conditions in the presence of uncertainties and disturbances. interval for membership functions (error and its rate) playing
an important rule in improving the response’s features. In
this section with utilizing the supervisory approach a new
method for deducing a robust response in the presence of
u R uncertainty and different initial error is presented, with
defining the new states as follows:

L
x z1 x1 (24)
M z2 x2 
c x
z3 g  ( 3 )2
m x1
Fig. 4. Schematic diagram of MBSS

The equation (23) can be rewritten like as follows:

MBSS is a well-known dynamic system in which its
behavior is described in the literature frequently [16]. The
dynamic equations of MBSS are as follows: z1 z 2
z 2 z 3 
d 2x (25)
m mg  f (20) C 2R 2 2 m (g  z 3)
2 z3 ( )( g  z 3 ) 2 ( z 1  z 2 )  (  )( z1 ) u
dt m L L C

where x as shown in Fig. 4 is the ball position, m is the ball These new dynamic equations can be rewritten as:
mass and g defines the gravitational acceleration Magnetic
force equation is modeled as follows: 
z1 F ( Z )  B ( Z )u (26)
ci 2
f (21)
x where:
where i is the winding current and c is the positive constant. C 2R 2 2 m (g  z 3)
F ( )( g  z 3 ) 2 ( z 1  z 2 ) and B ( )( z1 )
Electric circuit equation is as follows: m L0 L0 C
The control signal is considered as (4) and rewritten as
di follows,
L Ri  u (22)
dt
u B 1 (F ( Z ) C sign (S )  [ˆ  ke  u fuzz  X d ) I *
L0 (27)
L u fuzz
x

Authorized licensed use limited to: UNIVERSITY OF CONNECTICUT. Downloaded on May 16,2021 at 11:38:36 UTC from IEEE Xplore. Restrictions apply.
 where k is a positive constant in the sliding surface S ,
and u1 is the output of fuzzy PID controller, and I * is
defined as follows:

­°0 z 1  z 1d  D
I* ® (28)
°̄1 z 1  z 1d ! D

where D is the maximum interval’s value of membership

functions intervals.
V. SIMULATION RESULTS
In this section, simulation results are presented to
illustrate the effectiveness of the proposed controller in
stabilizing of MBSS. According to the notation introduced in
sections II and III, the following parameters have been
chosen for simulation:
Fig. 7. The response of x 1 with initial condition 0.5 (m) with input’s range
m e [ 0.3, 0.3], e [ 0.5, 0.5] with adaptive robust supervisory controller
g 9.8 ,m 1Kg , c 8, l 0.1mh , k p min 15 ,
sec 2 system, x (0) [0.5, 0, 0.5]
k p max 17 , k d min 30 , k d max 40 , k 1 100 , k 2 2, The responses of the x 1 under two conditions are shown
k3 1 , x 1d 1(m ) , [ (X , t ) 0.1x 32 in Fig. 5; these responses are deduced in the absence of the
supervisor controller and with the two ranges for the fuzzy
control system’s inputs.
As it is illustrated, features of response related to the
input’s range of fuzzy control system. Fig.6 and Fig.7 show
that under the same conditions, the fuzzy PID system which
has the input range e [3,3], e [5,5] remains stable with
the initial conditions 80 (cm) for the state x 1 while the other
system becomes unstable. Fig.7 shows the responses of
x 1 with supervisory control. In this scheme, the fuzzy PID
system ‘s inputs’ ranges are e [ 0.3, 0.3], e [ 0.5, 0.5] for
their fuzzy control system’s input. As it is shown, the system
Fig. 5. The response of x 1 under two conditions for fuzzy system inputs, remains stable while the system under the same conditions in
range1: e [  0.3, 0.3], e [ 0.5, 0.5] , range2= e [ 3,3], e [ 5,5] with the absence of the supervisory controller becomes unstable.
initial condition 0.85(m)

VI. CONCLUSION:
Novel integration of the adaptive sliding mode controller
with a fuzzy PID approach is presented in this paper.
Robustness as well as chattering free control signal and
applicability to a wide range of uncertain dynamic systems,
are the main advantages of the proposed controller. It is
worth mentioning that fixed fuzzy’s sets intervals for
different initial conditions are another achievement of the
proposed method. When PID control theory is combined
with fuzzy, it can improve control performance. Choosing
the appropriate domain for the fuzzy system’s inputs is one
of the main concerns in designing a fuzzy PID controller by
considering stability and getting the best response feature. To
Fig. 6. The response of x 1 with initial condition 0.5m with input’s range overcome this problem and improve the robustness of the
controller, the adaptive sliding mode approach is employed
e [ 0.3, 0.3], e [ 0.5, 0.5] for fuzzy controller system
along with the fuzzy PID approach as a supervisory
controller. By utilizing this method, the designer can choose
fuzzy sets parameters regardless of the initial error and its
changing rate.

Authorized licensed use limited to: UNIVERSITY OF CONNECTICUT. Downloaded on May 16,2021 at 11:38:36 UTC from IEEE Xplore. Restrictions apply.
 REFERENCE
[1] S. Baek, J. Baek, and S. Han, “An adaptive sliding mode control
with effective switching gain tuning near the sliding surface,” IEEE
Access, vol. 7, 2019, pp. 15563–15572.
[2] D. Xu, Q. Liu, W. Yan, and W. Yang, “Adaptive terminal sliding
mode control for hybrid energy storage systems of fuel cell, battery
and supercapacitor,” IEEE Access, vol. 7, 2019, pp. 29295–29303.
[3] P. Mani, R. Rajan, L. Shanmugam, and Y. H. Joo, “Adaptive
fractional fuzzy integral sliding mode control for PMSM model,”
IEEE Trans. Fuzzy Syst., vol. 27, no. 8, 2019, pp. 1674–1686.
[4] P. Li, X. Yu, and B. Xiao, “Adaptive quasi-optimal higher order
sliding-mode control without gain overestimation,” IEEE Trans.
Ind. Inform., vol. 14, no. 9, 2017, pp. 3881–3891.
[5] A. S. O. Al-Mashakbeh, “Proportional integral and derivative
control of brushless DC motor,” Eur. J. Sci. Res., vol. 35, no. 2,
2009, pp. 198–203.
[6] X. Wang, “Fuzzy-PID control in electro-hydraulic proportional
valve system,” in 2010 IEEE 11th International Conference on
Computer-Aided Industrial Design & Conceptual Design 1, vol. 2,
2010, pp. 1454–1457.
[7] K. H. Ang, G. Chong, and Y. Li, “PID control system analysis,
design, and technology,” IEEE Trans. Control Syst. Technol., vol.
13, no. 4, 2005, pp. 559–576.
[8] S. Chiu, “Using fuzzy logic in control applications: beyond fuzzy
PID control,” IEEE Control Syst. Mag., vol. 18, no. 5, 1998, pp.
100–104.
[9] A. N. Ravari and H. D. Taghirad, “A novel hybrid Fuzzy-PID
controller for tracking control of robot manipulators,” presented at
the 2008 IEEE International Conference on Robotics and
Biomimetics, 2009, pp. 1625–1630.
[10] M. Van, “An enhanced robust fault tolerant control based on an
adaptive fuzzy PID-nonsingular fast terminal sliding mode control
for uncertain nonlinear systems,” IEEEASME Trans. Mechatron.,
vol. 23, no. 3, 2018, pp. 1362–1371.
[11] K. Łapa and K. Cpałka, “Flexible fuzzy PID controller (FFPIDC)
and a nature-inspired method for its construction,” IEEE Trans.
Ind. Inform., vol. 14, no. 3, 2017, pp. 1078–1088.
[12] J. Xu, Y. Du, Y.-H. Chen, and H. Guo, “Optimal Robust Control
Design for Constrained Uncertain Systems: A Fuzzy-Set Theoretic
Approach,” IEEE Trans. Fuzzy Syst., vol. 26, no. 6, 2018, pp.
3494–3505.
[13] M. Al-Greer, M. Armstrong, and V. Pickert, “Selecting appropriate
fuzzy PID control structure for power electronic applications,” J.
Eng., vol. 2019, no. 17, 2019, pp. 4457–4460.
[14] H. Ding, J. Mao, and Y. Lin, “Indirect adaptive fuzzy control based
on fuzzy tree model,” Acta Autom. Sin., vol. 34, no. 6, 2008, pp.
666–683.
[15] J. Shen, B. Xin, H. Cui, and W. Gao, “Control of single-axis
rotation INS by tracking differentiator based fuzzy PID,” IEEE
Trans. Aerosp. Electron. Syst., vol. 53, no. 6, 2017, pp. 2976–2986.
[16] S. Mohammad-Hosseini, A.-R. A. M. Farrokhi, and M.-R. Jahed-
Motlagh, “Fuzzy-sliding state-feedback control of nonlinear ball
suspension system,” IFAC Proc., vol. 35, no. 1, 2002, pp. 229–234.
[17] Y. Wang, Y. Shi, M. Cai, W. Xu, and Q. Yu, “Optimization of air–
fuel ratio control of fuel-powered UAV engine using adaptive
fuzzy-PID,” J. Frankl. Inst., vol. 355, no. 17, 2018, pp. 8554–8575.
[18] M. Saeedi, A. M. Shari, A. Khatamian; “Design and implementation
of an adaptive sliding mode controller for stabilizing
nonholonomnic wheeled mobile robot”, IEEE International
Conference on Robotics and Biomimetics (ROBIO), 2012, pp.
2248-2253.
[19] Li- Xin. Wang, A Course in Fuzzy Systems and Control, Prentice-
Hall International, Inc, 1997.

Authorized licensed use limited to: UNIVERSITY OF CONNECTICUT. Downloaded on May 16,2021 at 11:38:36 UTC from IEEE Xplore. Restrictions apply.