You are on page 1of 14

Accepted Manuscript

Original article

Design of Direct MRAC Augmented With 2 DoF PIDD Controller: An Appli-


cation to Speed Control of a Servo Plant

R. Rajesh, S.N. Deepa

PII: S1018-3639(18)31218-2
DOI: https://doi.org/10.1016/j.jksues.2019.02.005
Reference: JKSUES 328

To appear in: Journal of King Saud University - Engineering Sci-


ences

Received Date: 15 December 2018


Accepted Date: 28 February 2019

Please cite this article as: Rajesh, R., Deepa, S.N., Design of Direct MRAC Augmented With 2 DoF PIDD Controller:
An Application to Speed Control of a Servo Plant, Journal of King Saud University - Engineering Sciences (2019),
doi: https://doi.org/10.1016/j.jksues.2019.02.005

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.
Design of Direct MRAC Augmented With 2 DoF PIDD Controller: An Application to
Speed Control of a Servo Plant
R. Rajesh*, S.N. Deepa
Department of Electrical and Electronics Engineering
Anna University, Regional Campus: Coimbatore, India.

*Corresponding author (R. Rajesh)

Ph: +91-8754971124

Email: ra.rajeshkrishna@gmail.com

orcid ID:

R. Rajesh : https://orcid.org/0000-0002-8945-5196

S.N. Deepa : https://orcid.org/0000-0003-2150-5051

Authors biography

Mr. R. Rajesh was born in Coimbatore District, Tamil Nadu, India, in March, 1994. He received B.E.
in Instrumentation and Control Engineering from PSG College of Technology, Coimbatore, India in
2015, and the M.E. in Control and Instrumentation Engineering from Anna University, Regional
Campus, Coimbatore, India in 2018. His research areas includes naturally inspired algorithms, linear
control theory, nonlinear control system, adaptive control system, optimal control theory, advanced
process control, high voltage testing and high voltage measurements. He published few papers in
national and international conferences and reputed journals.

Dr. S.N. Deepa is an associate professor of Electrical and Electronics Engineering at Anna University,
Regional Campus: Coimbatore, India. She got a under graduate in Electrical and Electronics Engineering
from Government College of Technology, Coimbatore, India in 1999 and post graduate in Control
Systems from PSG College of Technology, Coimbatore, India in 2004 and doctor of philosophy in
Electrical Engineering from Anna university, Chennai, India in 2008. She has published 9 books in
national/international Publishers, and many articles in national/international journals and conferences. Her
Research areas includes linear control system, non-linear control system design and analysis, soft computing techniques, adaptive
control system, robust control systems and medical image processing.
Design of Direct MRAC Augmented With 2 DoF PIDD Controller: An Application to Speed Control of a Servo Plant

Abstract 2017), MRAC with intelligent controller (Mohideen et al., 2013; Prakash et
al., 2015), modifying adaptive control (Datta et al., 1994; Kreisselmeier et al.,
An adaptive control with reference model provides the solution for 1982), multiple model MRAC (Narendra et al., 1997; Chen et al., 2001; Fu et
uncertainty in dynamics, transient performance characteristics of autonomous al., 2007), Variable structure MRAC (Ambrosino et al., 1984; Yu et al., 1996;
systems. The high frequency oscillation and poor settling time are the important Hsu et al., 1990), MRAC with back stepping control (Liu et al., 2007), MRAC
challenges in adaptive control methodology. This article proposes the Model
with H controller (Vargas-Martínez et al., 2010; Zhang et al., 2017),
Reference Adaptive Control (MRAC) augmented with Proportional Integral
MRAC with baseline LQR (Wang et al., 2017; Dydek et al., 2010; Lavretsky
Double Derivative (PIDD) controller for a solution of such kind of problems.
2009).
This proposed multi-loop structure consists of two loops: a primary outer loop
executes MRAC with two adjustment mechanism and a secondary inner loop The updating of real-time parameters in control loop, produces poor
act as a normal closed loop 2 DoF PIDD controller with plant dynamics. The settling time, transient instability, and high frequency oscillation due to high
function of 2 DoF PIDD controller is act as a filter for primary adaptive loop to adaptation gain which are avoided by cascaded or multi-loop controller.
control the high frequency oscillations and pulsating signals. This cascade or Tepljakov et al., 2018 presented the multi-loop MRAC for magnetic levitation
multi-loop structure will makes the secondary loop is more stable. The tuning system. In that, inner loop is conventional closed loop control system that
of 2 DoF PIDD parameters again a challenging task due to the presence of provides stability, and an outer loop employs MIT rule to perform MRAC.
extra two parameters and it was done by particle swarm optimization (PSO) They showed that multi-loop structure is very useful to make conventional
algorithm. The effectiveness of the proposed methodology has been tested in closed loop control system as an adaptive system. And also fault tolerance and
servo plant model. disturbance rejection is much better in multi-loop adaptive control structure.
Kavuran et al. 2016 used MIT rule to improve the adaptation skills of a closed
Keywords
loop PID control system for coaxial rotor system. Similarly, Rajesh et al.,
Model reference adaptive control (MRAC); PIDD controller; Particle 2018 developed enhanced MRAC with PID controller for conical tank system
swarm optimization (PSO); Adaptation rate; Servo plant; High frequency and found smooth control response, better trajectory tracking in the proposed
oscillation. controller. Parkash et al., 2012 designed MRAC with neural network
controller for nonlinear system and better performance achieved even the
1. Introduction presence of uncertainties and nonlinearities. Kavuran et al. 2017 studied
Proportional Integral Derivative (PID) control is one of the traditional MRAC of twin rotor multi-input multi-output system with error modified
linear control methodologies to control the real time applications almost a fractional order MIT rule algorithm. Hsu et al., 2003 developed model
century. On considering the PID control structure, it is simple, robust, high reference output feedback sliding mode controller; for a class of multivariable
reliability, and easy tuning mechanisms for wide a range of application. The nonlinear system by using Lyapunov theory.
tracking error due to plant nonlinearity, sensitivity to measurement noise, and
In this proposed work, MRAC with closed loop 2 DoF PIDD controllers
amplification of measurement noise are some considerable problems in fixed
are combined in multi-loop manner for servo plant speed control. This
gain controllers. Time domain and frequency domain analysis criterion cannot
structure, a plant with 2 DoF PIDD controller is in the inner loop and outer loop
alone ensure the long term robustness to meet the set point tracking. To
having an adaptive mechanism by using MIT (Massachusetts Institute of
achieve such criteria for the real world applications adaptive control has been
technology) rule which is connected in arrangement of multi-loop or cascaded
proposed.
manner. The constant control gains for the plant model is obtained by 2 DoF
Adaptive control is an efficient control approach to deal the uncertainties PIDD controller based on PSO algorithm in the inner loop and MIT rule
in modeling, time-varying parameters, fault tolerance and load or set point changes the control signal to achieve adaptation in the external loop to track the
variations in real world applications from 1960s. An adaptive control method reference model response. This configuration avoids the online updating of the
is to change the gain of the controller to compensate the corresponding controller parameters directly to the plant but adaptation in the outer loop
deviation among the plant and reference model to obtain the required which avoids the computational difficulty and produce better performance.
performance of the plant. In case performance degradation situation, adaptive By independently adjusting the two closed loop transfer functions has been
control themselves reconfigure for a new form to sustain their performance. inspired the proposal of 2 DoF PI/PID controller. 2 DoF PID controller is one
For capable of handling uncertainties, the adaptive control used in an of the famous improved versions of the conventional PID controller. Usually,
extensive variety of applications such as satellite control, autopilot, chemical conventional 1 DoF controllers tuning is achieved by load disturbance
reactors, robotics, automation, etc. The most widely used adaptive control conditions and is not on set point tracking, robustness level (Vilanova et al.,
methodologies are self-tuning regulator, gain scheduling, and MRAC. Among 2012). The 2 DoF PID controller efficiently meets both servo and regulatory
these, MRAC is one of the finest methods due to its guarantee asymptotic regulations and set point tracking capabilities which provides extra flexibility to
tracking performance and closed loop boundedness in the adaptive control the control system design. Different meta-heuristic algorithm is suggested for
mechanism. Even though, it produces some problems such as poor tuning of 2 DoF PID controller in literature such as, differential evolution
performance in an existence time delay, lack of reliability due to real-time (Sahu et al., 2013), genetic algorithm (Oh et al., 2006; Wang et al., 2001),
updating of controller parameters, oscillatory transient behavior, and poor invasive weed optimization (Manoharan et al., 2017), firefly algorithm
settling time. A lot of researchers have developed hybrid adaptive laws to (Debbarma et al., 2014), teaching learning based algorithm (Sahu et al., 2016),
improve MRAC performance which includes: MRAC with PI/PID controller cuckoo search algorithm (Dash et al., 2014), bacterial foraging algorithm
(Dey et al., 2016; Zhou et al., 2016), MRAC based on linear matrix inequality (Latha et al., 2012). 2 DoF PIDD controller used instead of 2 DoF PID
(Shi et al., 2017), fractional order MRAC (Vinagre et al., 2002; Aguila- controller in this article to improve performance and reduced oscillations.
Camacho et al., 2013), MRAC with anti-windup compensator (Sarhadi et al.,
Even though, the development of many natural inspired algorithms,
particle swarm optimization (PSO) has some unique place among researchers
due to its simplicity, simple calculations, lots of freedom given to modify the
algorithm and a substitute solution to the non-linear complex optimization
problem. The base idea was stimulated from the social activities of animals
such as fish schooling, birds clustering, etc. PSO is depended on the natural
activity of communication between the teams and to share individual
information when a team of birds or insects searching food or throughout
migration. All the birds or particle cannot find the best position, however from
the social behavior if anyone of the members within the group can find the
desired path to travel and therefore the remainder of the birds or particle track
quickly. Considering, the design of tuning PID controller, performance index (c) Connector (d) Inertial Loads
such as Integral Square Error (ISE), Integral Absolute Error (IAE), and Integral
Time Absolute Error (ITAE) was proposed in the objective function so far. IAE Fig. 1. SRV02 components
and ISE produce comparatively small overshoot but a long settling time.
However, ITAE overcomes these disadvantages, but produces stability margin
problems. This article includes time domains performance criteria such as The angular speed of the SRV02 load shaft with respect to the input motor
settling time, peak overshoot and rise time with weighting values to improve voltage can be defined by the following first order transfer function
the optimization performance. l( s ) K
 (1)
The major contributions of this article are Vm ( s )  s  1

1. The problem of oscillatory behavior due to directly updating parameter to Where  l ( s ) is the Laplace transform for the load shaft speed ω l (s) , Vm (s)
the plant is avoided by multi-loop or cascaded MRAC – 2 DoF PIDD
controller for servo plant speed control applications. is the Laplace transform of motor input voltage vm (s) , K is the steady state
gain,  is the time constant and s is the Laplace operator.
2. A PSO algorithm is used for tuning of 2 DoF PIDD controller parameters
with multi-objective function. Table 1. SRV02 components description
3. Optimum adaptation rate value for adaptive controller is analyzed and high
over shoot problem due to high adaptation rate was rectified using Number Component
proposed control structure. 1 Potentiometer
This article proposed as follows: Section 2 contains an explanation and 2 Bottom plate
mathematical modeling of servo plant. The detailed review of MRAC, 2 DoF 3 Posts
PIDD controller, and PSO algorithm are given in section 3. A proposed
4 Motor pinion gear: 72-teeth (low gear)
methodology for servo plant is given in section 4. Simulation study of
proposed direct MRAC augmented with 2 DoF PIDD controller for servo 5 Load gear: 72-teeth (low gear)
plant is given in section 5. Lastly, section 6 concludes this article. 6 Potentiometer anti-backlash gear

2. Servo plant description and mathematical modeling 7 Anti-backlash springs


8 Load shaft
The Quancer SRV02 rotary servo plant, visualized in Fig.1 contains of a
DC motor that is enclosed in a solid aluminium furnished with a planetary 9 Motor
gearbox. The motor has its individual internal gearbox which drives external 10 Gear box
gears. The SRV02 is equipped with three sensors such as a potentiometer,
11 Potentiometer
encoder, and tachometer. The potentiometer and encoder measures the angular
position of the load gear and the tachometer can be used to measure its 12 Encoder
velocity. The description of Fig.1 is tabulated in Table 1, and specifications of 13 Tachometer
the main parameters associated with the SRV02 is listed in Table 2.
14 Ball-bearing block
15 Motor connector
16 Tachometer connector
17 Encoder connector
18 S1 & S2 Connector
19 Motor pinion gear: 72-teeth (high gear)
20 Load gear: 120-teeth (high ger)
21 Bar inertial load
22 Disc inertial load
23 Thumb screws
(a) Top view (b) Front view

2.1 Electrical equations using first priciples

The DC motor armature circuit diagram and gear train is showed in Fig.2.
The back emf voltage eb  t  depends on the speed of the motor shaft m

Symbol Description Value and back emf constant of the motor k m . It opposes the current flow. The back
Vnom Motor nominal input voltage 6.0 V emf voltage is given by:
2.6 Ω
eb  t   kmm  t 
Rm Motor armature resistance
(2)
Lm Motor armature inductance 0.18 mH
kt Motor current –torque constant 7.68 × 10-3 N m/A Using Kirchoff’s law, the following equation can be write,

Km Motor back-emf constant 7.68 × 10-3 V/(rad/s) dI m (t)


Vm (t)  Rm I m (t)  Lm  kmm (t)  0 (3)
Kg High gear total gear ration 70 dt
m Motor efficiency 0.69 Since the motor inductance is much lower than the resistance, it can be ignored.
g Gearbox efficiency 0.90 Thus the (3) becomes,
J m,rotor Rotor moment of inertia 3.90 × 10-7 kg · m2 Vm (t)  Rm I m (t)  kmm (t)  0 (4)
Jtach Tachometer moment of inertia 7.06 × 10-8 kg · m2
High gear equivalent moment of Then the motor current I m (t) can be found as:
J eq 9.76 × 10-5 kg · m2
inertia without external load
Vm (t)  km m (t)
High gear equivalent viscous I m (t)  (5)
Beq 0.015 N · m/ (rad/s) Rm
damping coefficient
mb Mass of bar load 0.038 kg
Lb Length of bar load 0.1525 m 2.2 Mechanical equations using first priciples
md Mass of disc load 0.04 kg
rd Radius of disc load 0.05 m In this segment the equation of motion describing the speed of the load
shaft, l , with respect to the applied motor torque,  m , is established
mmax Maximum load mass 5 kg
f max Maximum input voltage frequency 50 Hz Since the SRV02 is one degree of freedom rotary system, Newton’s
I max Maximum input current 1A Second Law of Motion can be written as:

max Maximum motor speed 628.3 rad/s J    (6)


K gi Internal gearbox ratio 14
Where J is the moment of inertia of the body (about its center of mass),  is
K ge,low Internal gearbox ratio (low-gear) 1 the angular acceleration of the system, and  is the sum of the torques being
K ge,high Internal gearbox ratio (high-gear) 5 applied to the body. As showed in Fig.1, the SRV02 gear train along with the
m24 Mass of 24-tooth gear 0.005 kg viscous friction acting on the motor shaft, Bm , and the load shaft Bl are
m72 Mass of 72-tooth gear 0.030 kg considered. The load equation of motion is,
m120 Mass of 120-tooth gear 0.083 kg dl (t)
Jl  Bll (t)   l (t) (7)
r24 Radius of 24-tooth gear 6.35 × 10-3 m dt
r72 Radius of 72-tooth gear 0.019 m
Where J l is the moment of inertia of the load and  l is the total torque applied
r120 Radius of 120-tooth gear 0.032 m
on the load. The load inertia contains the inertia from the gear train and from
K pot Potentiometer sensitivity 35.2 deg/v any external loads attached, e.g. disc or bar. The motor shaft equation is stated
Kenc SRV02-E encoder sensitivity 4096 counts/rev as:
Kenc SRV02-EHR encoder sensitivity 8192 counts/rev dm (t)
Jm  Bmm (t)   ml (t)   m (t) (8)
Ktach Tachometer sensitivity 1.50 V/kRPM dt

Where J m is the motor shaft moment of inertia and  ml is the resulting torque
acting on the motor shaft from the load torque. The torque at the load shaft
from an applied motor torque can be written as:

 l (t)   g K g ml (t) (9)

Where K g is the gear ratio and  g is the gearbox efficiency. The planetary
gearbox that is directly attached on the SRV02 motor is represented by the N1
and N 2 gears in Fig.1 and has a gear ratio of

N2
K gi  (10)
N1
Fig. 2. SRV02 DC motor armature circuit and gear train
This is the internal gear box ratio. The motor gear N 3 and the load gear N4 are
straightly interconnected together and are observable from the outside. These
gears comprise the external gear box which has an associated gear ratio of
Table 2. SRV02 servo plant specification
N4
K ge  (11)
N3
The gear ratio of the SRV02 gear train is then given by:
 m (t) 

m kt Vm (t)  km K g l (t)  (23)
K g  K ge K gi (12) Rm

Thus, the torque seen at the motor shaft through the gears can be expressed as: If substitute (23) into (20),

(t) d 
 ml (t)  (13) J eq  l (t)   Beql (t) 
g K g  dt 
(24)
 g K gm kt (Vm (t)  km k g l (t))
Intuitively, the motor shaft must rotate K g times for the output shaft to rotate Rm
one revolution. After assembling the terms, the (24) becomes
m (t)  K g l (t) (14)
d   km g K g2m kt 
 l (t)  J eq    Beq  l (t)

It can able to find the relationship between the angular speed of the motor  dt  
Rm
 (25)
shaft, m and the angular speed of the load shaft l , by taking the time  g K gm ktVm (t)

derivative: Rm
m (t)  K g l (t) (15) This (25) can be written as

To find the differential equation that defines the motion of the load shaft d 
respect to an applied motor torque substitute (13), (15) and (7) into (8) to get  l (t)  J eq  Beq,vl (t)  AmVm (t) (26)
 dt 
the following:
Where the equivalent damping term is given by
dl (t)
JmKg  Bm K g l (t) 
dt  g K g2m kt km  Beq Rm
 d (t)  (16) Beq,v  (27)
J l  l   Bll (t) Rm
 dt    m (t)
g K g and the actuator gain equals

Assembling the coefficients in terms of the load shaft velocity and acceleration  g K gm kt
Am  (28)
gives Rm

 K J
g
2
g m  Jl  ddt(t)   K B
l
g
2
g m 
 Bl l (t)   g K g m (t) (17) 3. Methodology

Defining the following terms: 3.1 Model reference adaptive control

J eq   g K g2 J m  Jl (18) The MRAC is one of the important adaptive control algorithms which
provides feedback and control law to control of the system with parameters
Beq   g K g2 Bm  Bl (19) uncertainties and good tracking capability (Shi et al., 2017). The block diagram
of MRAC as shown in Fig.3 and it was initially introduced to progress the
Simplifies the equation as: performance of the autopilot. In the MRAC controller gains are updated based
on the variation between the uncertain plant output and reference model output
dl (t)
J eq  Beql (t)   g K g m (t) so that the response from the plant closely follows the reference model
dt response. The reference model is the model of expected or wished response
(20) from the closed loop plant output. The mechanism for parameter adjustment is
done by two ways such as gradient descent method (MIT rule) and applying
stability method. MIT rule is the original approach for adjusting the parameters
2.3 Joining the electrical and mechanical equations
of MRAC and it was developed at instrument laboratory (now Draper
laboratory) at Massachusetts Institute of Technology.
The combined expression of electrical equation and mechanical equation
embodies the load shaft speed in terms of the applied motor voltage. The motor
torque is proportional to the voltage applied and is described as

 m (t)  mkt lm (t) (21)

Where kt is the current torque constant (N .m/ A), m is the motor efficiency,
and lm is the armature current.

It can able to define the motor torque with respect to the input voltage
Vm (t) and load shaft speed l (t) , by replacing the motor armature current
given by (5) into the current- torque relationship given in (21)

m kt Vm (t)  km m (t)


 m (t)  (22)
Rm

To define this in terms of Vm and l , insert the motor load shaft speed
(15) into (21) to get: Fig. 3. General block diagram of MRAC
Consider a plant dynamics (29) 3.2 2 DoF PIDD controller

dy
 cy  du
dt (29)

Where u is the control variable and y is the measured output. Assume


that we want to obtain a closed loop system described by,

dy
 cm ym  dm rm
dt (30)
Fig. 4. 2 DoF control system structure
A natural choice of control law is,
The PID controller is commonly used because of simple design, easy
u  t   1r  t   2 y  t  implementation, and good reliability. As like PID controller, 2 DoF PIDD
(31)
controller is obtain wide range of attraction due to its good regulation, tracking
dm c c capability, without oscillations, extreme sensitivity lesser than a stated value
Certainly, if we select gain value 1  ; 2  m , then (30) and extra flexibility to the control system design due to added DoF. The PIDD
d d
becomes (29) and this is known as perfect model following. controller is used instead of PID controller in this article to improve stability,
fast response, reduced settling time, and reduced peak overshoot (Mohanty et
To apply MIT rule, introducing the error, al., 2015). Usually, derivative factor amplifies the noise signal harshly,
particularly in the real-time application which avoided by adding the derivative
e  y  ym
(32) filter N with a derivative term. Fig.4. shows the 2 DoF PIDD controller
structure. The 2 DoF PIDD control scheme generates a control signal depends
Substitute (30) in (29)
on variation between the input signal R  S  and controller output Y  S  . This
d1 calculates a weighted variance signal for each control activities according to the
y r
p  c  d 2
(33) defined set point weights   and   . The controller output is the addition of

Where p is the differential operator. the P, I and DD actions on the individual variance signals, where every exploit
is weighted according to the selected gain parameters (Zhao et al., 2012).
One possibility is to adjust the parameters in such a way that the cost
KI K DD s 2
function  J  (34) is minimized. C(s)  K P   (40)
s K DD s 2
1
KP N
 
J 1,2 
1 2
2
e
(34) KI K DD s 2
F(s)  K P     (41)
s K DD s 2
To make J small, it is reasonable to change the parameters in the 1
KP N
direction of the negative gradient of J, that is,
Where K P is the proportional gain, K I is the integral gain, K DD is the
d  
   J   e (35)
dt   double derivative gain,  and  are the proportional and derivative set point
weights correspondingly. The output of the proposed 2 DoF PIDD controller is,
Where Γ is the adaptation gain, and is the sensitive derivative
KI
which described as how error is subjective by the adjustable parameters and U(s)  K P  R(s)  Y(s)  R(s)  Y(s)
can be written as, s
K DD s 2  R(s)  Y(s) (42)
e d 
 r K s2
1 p  c  d 2 1  DD
(36) KP N
Likewise,
In the 2 DoF PIDD controller   0 , for attenuate large change in
e d 1 2
d controller output when change in input signal and chose 0    1 to avoid
 r y (37)
2  p  c  d p  c  d2
2
2 extreme change in proportional action when a set point changes (Vilanova et
al., 2012). Tuning of 2 DoF PIDD controller is similar like a 2 DoF PID
controller tuning and it is done by analytical (Vilanova et al., 2011; Taguchi et
The equations (36) and (37) cannot be used directly because the process
al., 2000; Taguchi et al., 2002) and meta-heuristic based approach (Sahu et al.,
parameters are still unknown. Hence approximations are required. By
2013; Wang et al., 2001; Sahu et al., 2016). In this article PSO algorithm is
choosing the p  c  d2  p  cm , the plant model dynamics perfectly match utilized for tuning the controller parameter.
the reference model dynamics and the tracking error becomes disappear. Then
we get following equations (38) and (39) for updating the controller
3.3 Particle swarm optimization algorithm
parameters.

 c  PSO is a unique population based stochastic search algorithm proposed by


d1
   m r  e (38) Dr. Kennedy and Dr. Eberhart in 1995 (Eberhart et al., 1995). The original idea
dt  p  cm  of PSO algorithm is inspired from a model of the communal performance of
animals such as fish schooling and bird gathering. It is based on
d2  c  communication of individual knowledge and natural learning when birds or
   m y  e (39)
 p  cm 
dt insects hunt for food or migration in the search space. Popularity of PSO in the
last decades due to its simple structure and few parameters only needed to In velocity updating rule (43) has three components, in the first term vi  t 
adjust the algorithm.
is termed as an inertia component that gives a memory of the immediate past
In PSO, every potential solution to the problem is treated as a bird or

direction. The second term c1  rand1  t   pbest i ( t )  xi t   is termed as a
particle represented as i  1,2, ,n  where n  1 . The set of particle in the
cognitive component that measures the performance of the particle relative to
colony (n) is flown in the D dimensional search space. This position of every the past performance or distance at which a particle is situated from the best
particle is represented as a X i ( t )   xi1( t ),xi2 ( t ), ,xiD ( t ) . The D solution Pbest i found by its own. The third term

dimensional space is created by X i ( t )  ld ,ud  where ( ld ) is a lower bound 


c2  rand2  t   gbest ( t )  xi t   is termed as a social component that
and ( ud ) is an upper bound. Velocity vector also represents like a position measures the performance of the particle relative to a bunch of particle or
vector Vi ( t )   vi1( t ),vi2 ( t ), ,viD ( t ) . Each particle tracks the personal best neighbors and the best solution determined by its neighbor is termed as Gbest .

particle position from its own experience and particle fitness value calculated There is no control mechanism for velocity in the first version of PSO and this
from objective function, which is represented as makes inefficient behavior of the algorithm. Later Shi et al. 1998 address this
problem and produce an inertial weight w component in velocity updating

Pbest i ( t )  pbest i 1( t ), pbest i 2 ( t ), 
, pbest i D ( t ) . The optimal value among the
rule. Clerc et al. 1999 proposed a constriction factor    for ensuring
Pbest is called Gbest and it is represented as
i convergence of optimal solution and control of the magnitude of particle
Gbest ( t )   gbest1 ( t ),gbest2 ( t ), ,gbest D ( t )  . The PSO algorithm is worked velocity. This constriction model is helpful to value selection of w, c1 and c2 .
based on (43-46). This proposed constriction model with inertia weight is (47).


vi  t  1  vi  t   c1  rand1  t   pbest i ( t )  xi  t   (43)
 w  v  t   c  rand  t   p
vi  t  1   
i 1 1 
best i ( t )  xi  t    (47)


 c2  rand2  t   gbest ( t )  xi  t   
 c2  rand2  t   gbest ( t )  xi  t 
  

xi  t  1  xi  t   vi  t  1 (44) where

2 (48)
The best local position of each particle is updated as follows: 
2 2     2  4

 i 
 pbest ( t ) if f xi  t  1  pbest ( t )
i 
pbesti ( t  1)   (45)   c1  c2 ,  4 (49)

 
xi  t  1 if f xi  t  1  pbesti ( t ) 
where f : m   is the fitness function. The global best position Gbest is
4. Proposed MRAC augmented with 2 DoF PIDD Controllerfor a
calculated by taking the conditions of all particles in the preceding three steps is
SRV02 servo plant
defined as:
The main purpose of the proposed block diagram is to avoid real-time
gbest ( t  1)  arg min f
xi
 pbesti ( t  1) (46) updating of controller parameters, reduced high frequency oscillation due to
high adaptation rate, fast settling time, and enhances the disturbance rejection.
The proposed control scheme consists of adaption in the external loop and 2
where, i  1, 2, , n and n  1 DoF PIDD controller with plant in the internal loop connected in the multi-loop
or cascade manner. The working of two loops as follows:
vi is the velocity of particle i ; xi denotes the current position of
[i]. In the outer loop, MRAC is developed using MIT rule with two
particle i with objective value fitness; Pbesti is the best past position of particle
adjustment mechanism such as 1 , and  2 which acts as a feed forward
i itself; gbesti is the global best position among the group. adaptation mechanism and feedback adaptation mechanism. The adaptive
mechanism adjusts the controller parameter depends on model error generation
rand1  t  ,rand1  t   0,1 are the uniformly distributed random numbers. c1
between servo plant and reference model e'  y  ym . When error inclines to
and c2 are the positive accelerated constraints. The constants c1 and c2
zero, the servo plant response tracks the reference model response perfectly.
express the confidence of the particle has in itself and in its neighbour and these The reference model is the expected response from the plant output. For servo
parameters improve fast convergence of the optimal solution. The selection of plant response, a maximum overshoot is less than 5% and settling time is 3 min
c2 are depends on the problem, in the first version of PSO considered for designing reference model. The transfer function of the
values of c1 and
reference model is given in equation (50).
c1  c2  2 , for better performance on real time problems recent work
17.9
propose that c1  c2  4. GR  (50)
s 2  2.67s  17.9
Fig. 5. Proposed scheme of MRAC augmented with 2 DoF PIDD controller for a SRV02 servo plant

Selection of adaptation gain (  i ) is very important in the design of In the proposed controller, an adaptive mechanism is implemented in the
outer loop which measure and update the parameters depends upon the
MRAC. Several analytical methods are proposed in the literature (Haykin
deviation from the servo plant response and reference model response
2018; Ioannou et al., 1996; Åström et al., 2013; Kuo et al., 1995). In practice,
due to non-linearity and uncertainty in modeling adaptation gain tuned from (e'  ym  y) . The updated control signal response from the adaptive
low value to high value (Slotine et al., 1991). Usually,  i value is a trade-off mechanism in the outer loop (u') is given to the input of inner loop 2 DoF
between the speed of convergence and implementation difficulty. The large PIDD controller and the parameters of 2 DoF PIDD controller is tuned by using
value of  i leads to fast convergence but too large value makes saturation in PSO algorithm.
the control states and deviation of adaption law. Adaptation gain (  i ) is selected based on the tuning values from low
[ii]. The output of MRAC mechanism u' is given to the input of the 2 range to higher range. In the conventional MRAC, the high value of adaptation
DoF PIDD controller. In the 2 DoF PIDD controller F(s) act as a filter for high gain reduces settling time, peak time but increases the frequency of oscillation
frequency oscillation and pulsating signal from the MRAC mechanism which which leads to higher overshoot as shown in Table 3. The ISE value is also
avoids plant from a jerk. For tuning the 2 DoF PIDD controller PSO algorithm reduced while increasing adaption gain (Dey et al., 2016; Shi et al., 2017)
is utilized. Specifying the multi-objective function in the meta-heuristic which improves the fault tolerance capability. The Table 3 indicates the output
algorithm is one of the way to improving the optimal performance of algorithm gets saturated after certain value of adaption rate. From this observation,
(Zamani et al., 2009) and the following objective function (51) is proposed for adaptation gain is taken as 15 for implementation of proposed technique.
tuning 2 DoF PIDD controller.

 T 
 
Table 3. Performance comparison of varying adaptation gain in
   
J K P ,K I,K DD ,    1  e( t )   2  M p
 (51) conventional MRAC
 0 
 3  tr   4  ts  Peak over Peak time Settling time
Γ ISE
shoot (%) (sec) (sec)
Where e(t ), M p , tr , and ts are error, maximum peak overshoot, rise time and
0.5 36.2 7.63 18.32 13.65
settling time. 1, , 5 are the weighting values of objective function and
1 41.3 7.03 16.53 11.25
ranges vary from 0 to 20 in this case. This proposed objective function (51) is
given to the PSO algorithm for tuning of 2 DoF PIDD controller parameters 5 49.3 6.95 12.56 8.52
( K P ,K I,K DD , ) for the SRV02 servo plant model. 10 56.3 6.23 9.52 6.25
20 59.3 5.98 8.53 5.24
5. Simulation result and discussion
40 62.3 5.73 7.82 5.08
This section contains the simulation of proposed MRAC augmented 2 DoF
PIDD controller for speed control of Quancer SRV02 servo plant. The
conventional MRAC control strategy is not capable to produce a smooth For optimal tuning of 2 DoF PIDD controller the parameters ranges are
response. High frequency control signals due to the high value of adaptation selected as K P  0,40 , K I  0,10 , K DD  0,10 ,   0,1 .
rate which spoils the sensitive components in the plant like motor, gear
mechanism, etc. To avoid such drawbacks a novel adaption mechanism is The parameters of PSO algorithm is listed in Table 4. Based on the objective
proposed for SRV02 servo plant and the simulation results shows the improved function, the optimized algorithm is executed for a specified number of
performance in the proposed controller better than conventional MRAC. The iterations, and the global best particles are the obtained gains for 2 DoF PIDD
entire simulation work is carried using MATLAB R2013a (Version 8.1.0.604) controller as shown in Table 5. Fig. 6 shows the 2 DoF PIDD controller based
platform and implemented in a PC with Intel Core2 Duo Processor with PSO algorithm is converged in the 75-80th iteration.
2.27GHz speed and 2.00 GB RAM. The system consists of SRV02 servo plant, Table 4. Values of PSO
Q8 USB data acquisition card, and power amplifier. The input ± 15 V, 3 A
peak, Faulhaber coreless DC motor is used. To detect the angular position
Vishay Spectrol model 132 potentiometer and US digital single ended encoder
is used. Tachometer is directly attached to the DC motor for speed
measurement. To generate square speed response, amplitude and offset
constant gain are taken as respectively 2.5 rad/s and 5 rad/s.
in the inner loop and
Parameters PSO
model reference
adaptive Dimension 5 control in
the outer loop Number of particles 50 connected
in a cascaded or multi-
loop manner. Number of iteration 100 The
parameters of c1 0.8 the 2 DoF
PIDD c2 1.3 controller is
tuned by using PSO
algorithm and Inertia weight ( w ) 0.9 MRAC is
developed using MIT rule. The selection of adaptation gain has been carried
out with different random values from lower range to higher range. The higher
range values of adaptation gain is produced better peak time, settling time and
Table 5. Optimized controller parameters of 2 DoF PIDD controller
ISE better than the lower range of values but increases the peak overshoot

Controller Kp Ki Kd 
Table 6. Performance comparison for speed control of servo plant
2 DoF PIDD - PSO 31.51 3.81 7.81 0.61
Peak over Peak time Settling
Fig. 7 shows the adjustment mechanism 1 and  2 . The model error Controller ISE
shoot (%) (sec) time (sec)
between the servo plant and reference model is given in Fig. 8. Fig. 9 is depicts MRAC 53.5 6.12 8.95 5.9
the control input which shows the proposed MRAC control signal produces a
Proposed MRAC 10.32 2.23 3.02 1.03
smooth response when compared to the conventional MRAC methodology.
The speed response of SRV02 servo plant is given in Fig. 10 which is clearly while increasing adaption rate.
shows the proposed MRAC augmented 2 DoF PIDD controller restricts the
The main purpose of the proposed technique is to avoid high frequency
oscillatory behavior and produce a smooth performance when compared to
oscillation and pulsating signal due to real time updating of controller
conventional MRAC. The proposed controller tracks the speed step response
parameters. It is seen that the proposed methodology for servo plant has the
accurately when a deviation between the servo plant model and reference
ability to control high oscillation, improved settling time, and better disturbance
model goes to zero. Table 6 indicates the effectiveness and better performance
rejection compared to conventional MRAC methodology. The problem of peak
of proposed MRAC for servo plant speed control is better than conventional
overshoot in the high value of adaptation rate has been controlled in the
MRAC. The disadvantage of increased peak overshoot due to high adaptation
proposed technique. This control has the ability to protect the sensitive
gain is rectified and improved settling time was obtained in proposed
components in the servo plant.
methodology. The strong disturbance is applied at 21st seconds of simulation
time but proposed MRAC methodology strongly rejects the disturbance and
tracking the reference faster than conventional MRAC mechanism. The
zoomed view of disturbance rejection is given in Fig. 11. Fig.12 depicts the
motor voltage Vm (V).

6. Conclusion
In this article, an adaptive technique has been proposed to speed control of
SRV02 servo plant. The proposed technique consists of 2 DoF PIDD controller

Fig. 6. Convergence of 2 DoF PIDD controller based PSO algorithm


Fig. 7. Adjustment mechanism ( 1 ,2 )
Error (y-ym)

Time (seconds)

Fig. 8. Convergence of an error between the servo plant and reference model in the proposed controller

Fig. 9. Control input to the servo plant


ω (rad/sec)
Fig. 10. Speed response of a servo plant

ω (rad/dec)

Fig. 11. Zoomed view of disturbance rejection of proposed controller

Fig. 12. Motor voltage Vm (V)

7. References
[1] Aguila-Camacho, N. and Duarte-Mermoud, M.A., 2013. Fractional adaptive control for an automatic voltage regulator. ISA transactions, 52(6), pp.807-815.
[2] Ambrosino, G., Celektano, G. and Garofalo, F., 1984. Variable structure model reference adaptive control systems. International Journal of Control, 39(6), pp.1339-1349.
[3] Åström, K.J. and Wittenmark, B., 2013. Adaptive control. Courier Corporation.
[4] Chen, L. and Narendra, K.S., 2001. Nonlinear adaptive control using neural networks and multiple models. Automatica, 37(8), pp.1245-1255.
[5] Clerc, M., 1999. The swarm and the queen: towards a deterministic and adaptive particle swarm optimization. In Evolutionary Computation, 1999. CEC 99. Proceedings of the 1999 Congress
on (Vol. 3, pp. 1951-1957). IEEE.
[6] Dash, P., Saikia, L.C. and Sinha, N., 2014. Comparison of performances of several Cuckoo search algorithm based 2DOF controllers in AGC of multi-area thermal system. International
Journal of Electrical Power & Energy Systems, 55, pp.429-436.
[7] Datta, A. and Ho, M.T., 1994. On modifying model reference adaptive control schemes for performance improvement. IEEE transactions on automatic control, 39(9), pp.1977-1980.
[8] Debbarma, S., Saikia, L.C. and Sinha, N., 2014. Robust two-degree-of-freedom controller for automatic generation control of multi-area system. International Journal of Electrical Power &
Energy Systems, 63, pp.878-886.
[9] Dey, R., Jain, S.K. and Padhy, P.K., 2016. Robust closed loop reference MRAC with PI compensator. IET Control Theory & Applications, 10(18), pp.2378-2386.
[10] Dydek, Z., Annaswamy, A. and Lavretsky, E., 2010, August. Combined/composite adaptive control of a quadrotor UAV in the presence of actuator uncertainty. In AIAA Guidance, Navigation,
and Control Conference (p. 7575).
[11] Eberhart, R. and Kennedy, J., 1995, October. A new optimizer using particle swarm theory. In Micro Machine and Human Science, 1995. MHS'95., Proceedings of the Sixth International
Symposium on (pp. 39-43). IEEE.
[12] Fu, Y. and Chai, T., 2007. Nonlinear multivariable adaptive control using multiple models and neural networks. Automatica, 43(6), pp.1101-1110.
[13] Haykin, S.S., 2008. Adaptive filter theory. Pearson Education India.
[14] Hsu, L., 1990. Variable structure model-reference adaptive control (VS-MRAC) using only input and output measurements: the general case. IEEE Transactions on Automatic Control, 35(11),
pp.1238-1243.
[15] Hsu, L., Costa, R.R. and da Cunha, J.P.V.S., 2003. Model‐reference output‐feedback sliding mode controller for a class of multivariable nonlinear systems. Asian Journal of Control, 5(4),
pp.543-556.
[16] Ioannou, P.A. and Sun, J., 1996. Robust adaptive control (Vol. 1). Upper Saddle River, NJ: PTR Prentice-Hall.
[17] Kavuran, G., Ates, A., Alagoz, B.B. and Yeroglu, C., 2017. An Experimental Study on Model Reference Adaptive Control of TRMS by Error-Modified Fractional Order MIT Rule. Control
Enginnering and Applied Informatics, 19(4), pp.101-111.
[18] Kavuran, G.U.R.K.A.N., Alagoz, B.B., Ates, A.B.D.U.L.L.A.H. and Yeroglu, C.E.L.A.L.E.D.D.I.N., 2016. Implementation of model reference adaptive controller with fractional order
adjustment rules for coaxial rotor control test system. Balkan Journal Of Electrical & Computer Engineering, 4(2), pp.84-88.
[19] Kreisselmeier, G. and Narendra, K., 1982. Stable model reference adaptive control in the presence of bounded disturbances. IEEE Transactions on Automatic Control, 27(6), pp.1169-1175.
[20] Kuo, S.M. and Morgan, D., 1995. Active noise control systems: algorithms and DSP implementations. John Wiley & Sons, Inc.
[21] Latha, K. and Rajinikanth, V., 2012, December. 2DOF PID controller tuning for unstable systems using bacterial foraging algorithm. In International Conference on Swarm, Evolutionary, and
Memetic Computing (pp. 519-527). Springer, Berlin, Heidelberg.
[22] Lavretsky, E., 2009. Combined/composite model reference adaptive control. IEEE Transactions on Automatic Control, 54(11), pp.2692-2697.
[23] Liu, T.H. and Hsu, H.H., 2007. Adaptive controller design for a synchronous reluctance motor drive system with direct torque control. IET Electric Power Applications, 1(5), pp.815-824.
[24] Manoharan, N., Dash, S.S., Rajesh, K.S. and Panda, S., 2017. Automatic Generation Control by Hybrid Invasive Weed Optimization and Pattern Search Tuned 2-DOF PID
Controller. International Journal of Computers, Communications & Control, 12(4).
[25] Mohanty, B. and Hota, P.K., 2015. Comparative performance analysis of fruit fly optimisation algorithm for multi-area multi-source automatic generation control under deregulated
environment. IET Generation, Transmission & Distribution, 9(14), pp.1845-1855.
[26] Mohideen, K.A., Saravanakumar, G., Valarmathi, K., Devaraj, D. and Radhakrishnan, T.K., 2013. Real-coded Genetic Algorithm for system identification and tuning of a modified Model
Reference Adaptive Controller for a hybrid tank system. Applied Mathematical Modelling, 37(6), pp.3829-3847.
[27] Narendra, K.S. and Balakrishnan, J., 1997. Adaptive control using multiple models. IEEE transactions on automatic control, 42(2), pp.171-187.
[28] Oh, T.S., Lee, W.H. and Kim, I.H., 2006. Parameter optimization of 2-DOF-PID controller using genetic algorithm. International Journal of Applied Electromagnetics and Mechanics, 24(3-4),
pp.131-145.
[29] Prakash, R. and Anita, R., 2012. Robust model reference adaptive intelligent control. International Journal of Control, Automation and Systems, 10(2), pp.396-406.
[30] Prakash, R. and Vasanthi, R., 2015. Speed control of DC-DC converter fed DC motor using robust adaptive intelligent controller. Journal of Vibration and Control, 21(15), pp.3107-3120.
[31] Rajesh, R. and Krishnan, P.H., 2018. Effects of Adaption Gain in Direct Model Reference Adaptive Control for a Single Conical Tank System. Digital Signal Processing, 10(5), pp.77-82.
[32] Sahu, R.K., Panda, S. and Rout, U.K., 2013. DE optimized parallel 2-DOF PID controller for load frequency control of power system with governor dead-band nonlinearity. International
Journal of Electrical Power & Energy Systems, 49, pp.19-33.
[33] Sahu, R.K., Panda, S., Rout, U.K. and Sahoo, D.K., 2016. Teaching learning based optimization algorithm for automatic generation control of power system using 2-DOF PID
controller. International Journal of Electrical Power & Energy Systems, 77, pp.287-301.
[34] Sarhadi, P., Noei, A.R. and Khosravi, A., 2017. Model reference adaptive autopilot with anti-windup compensator for an autonomous underwater vehicle: Design and hardware in the loop
implementation results. Applied Ocean Research, 62, pp.27-36.
[35] Shi, Y. and Eberhart, R., 1998, May. A modified particle swarm optimizer. In Evolutionary Computation Proceedings, 1998. IEEE World Congress on Computational Intelligence., The 1998
IEEE International Conference on (pp. 69-73). IEEE.
[36] Shi, Z. and Zhao, L., 2017. Robust model reference adaptive control based on linear matrix inequality. Aerospace Science and Technology, 66, pp.152-159.
[37] Slotine, J.J.E. and Li, W., 1991. Applied nonlinear control (Vol. 199, No. 1). Englewood Cliffs, NJ: Prentice hall.
[38] Taguchi, H. and Araki, M., 2000. Two-degree-of-freedom PID controllers—their functions and optimal tuning. IFAC Proceedings Volumes, 33(4), pp.91-96.
[39] Taguchi, H., 2002. Optimal tuning of two-degree-of-freedom PD controllers. In The 4th Asian control conference (Vol. 268).
[40] Tepljakov, A., Alagoz, B.B., Gonzalez, E., Petlenkov, E. and Yeroglu, C., 2018. Model Reference Adaptive Control Scheme for Retuning Method-Based Fractional-Order PID Control with
Disturbance Rejection Applied to Closed-Loop Control of a Magnetic Levitation System. Journal of Circuits, Systems and Computers, p.1850176.
[41] Vargas-Martínez, A., Puig, V., Luis, E.G.C. and Morales-Menendez, R., 2010, October. MRAC+ H∞ fault tolerant control for linear parameter varying systems. In Control and Fault-Tolerant
Systems (SysTol), 2010 Conference on (pp. 94-99). IEEE.
[42] Vilanova, R., Alfaro, V.M. and Arrieta, O., 2011. Analytical Robust Tuning Approach for Two-Degree-of-Freedom PI/PID Controllers. Engineering Letters, 19(3).
[43] Vilanova, R., Alfaro, V.M. and Arrieta, O., 2012. Simple robust autotuning rules for 2-DoF PI controllers. ISA transactions, 51(1), pp.30-41.
[44] Vinagre, B.M., Petráš, I., Podlubny, I. and Chen, Y.Q., 2002. Using fractional order adjustment rules and fractional order reference models in model-reference adaptive control. Nonlinear
Dynamics, 29(1-4), pp.269-279.
[45] Wang, Q., Ma, L., Qiang, W.Y. and Fu, P.C., 2001. Design for 2-DOF PID controller based on hybrid genetic algorithm and its application. Control and Decision, 16(2), pp.195-198.
[46] Wang, X., Chen, X. and Wen, L., 2017. The LQR baseline with adaptive augmentation rejection of unmatched input disturbance. International Journal of Control, Automation and
Systems, 15(3), pp.1302-1313.
[47] Yu, X. and Man, Z., 1996. Model reference adaptive control systems with terminal sliding modes. International Journal of Control, 64(6), pp.1165-1176.
[48] Zamani, M., Sadati, N. and Ghartemani, M.K., 2009. Design of an H∞ PID controller using particle swarm optimization. International Journal of Control, Automation and Systems, 7(2),
pp.273-280.
[49] Zhang, Q., Xu, D.Z. and Zhan, K.K., 2017. Model reference robust adaptive H∞ controller design. International Journal of Control, Automation and Systems, 15(4), pp.1507-1514.
[50] Zhao, Y.M., Xie, W.F. and Tu, X.W., 2012. Performance-based parameter tuning method of model-driven PID control systems. ISA transactions, 51(3), pp.393-399.
[51] Zhou, X., Yang, C. and Cai, T., 2016. A model reference adaptive control/PID compound scheme on disturbance rejection for an aerial inertially stabilized platform. Journal of Sensors, 2016.

Funding

This study is not received any financial support any organisation.

Conflict of Interest

The authors declare that they have no conflict of interest.

You might also like