Downloaded from https://iranpaper.

ir

t
1s IEEE International Conference on Power Electronics, Intelligent Control and Energy Systems (ICPEICES-2016)

A Comparative Analysis ofPID, LQR and Fuzzy

Logic Controller for Active Suspension System
using 3 Degree of Freedom Quarter Car Model
l 2
Jumi Bharali and Mrinal Buragohain
1,2Department ofElectrical Engineering,
Jorhat Engineering College, Jorhat-785007, Assam, India
E-mail: [email protected]@gmail.com

Ab stract-Suspension system plays an important role in quarter car model of active suspension system for analysis
isolating vehicle body from road shocks and vibrations. The and research purpose. 2 DOF quarter car models are
aim of suspension system is to improve ride comfort, road successfully applied using control techniques like PID,
handling and stability of vehicles. Three different active
LQR, LQG, FUZZY, sliding mode control, composite
controllers are designed using 3 Degree of Freedom (DOF)
nonlinear feedback control [1] [2] [3] [4] [ 5]. As with
quarter car model to compare the performance of three
controllers with the passive suspension system. The three
increasing numbers of degree of freedom the
controllers designed are PID controller, Linear Quadratic characteristics of system change, a 3 DOF system is used
Controller (LQR) and Fuzzy logic controller. In this work, and a fuzzy logic controller for the suspension system is
MATLAB/SIMULINK software is used for simulation designed and analyzed to compare fuzzy logic controller
purpose and simulation result demonstrate that active with that of passive suspension system [ 6].
suspension system shows better result in comparison to In this work a 3 DOF quarter car model is modeled
passive suspension system with reference to maximum and three different control methods namely PID, LQR and
amplitude, body acceleration, suspension deflection and
fuzzy are developed and the response of the control
settling time. Also the result of comparison displays that
methods are compared with that of passive suspension
fuzzy logic control exhibits better functioning and stability as
compared to other control methods and passive as weil.
system. Fuzzy logic controller is designed using
Keywords-Degree 0/ Freedom (DOF); PID Controller; MATLAB's Fuzzy Logic Toolbox and the overall system
LQR Controller; Fuz7,Y Logic Controller; Body Acceleration; is designed using MATLAB coding and SIMULINK
Suspension Dejlection; Settling Time toolbox.
Section 11 gives description of 3 degree of freedom
I. INTRODUCTION quarter car model active suspension system. Section III
A suspension system is divided into three types gives the formulation and design of the three proposed
depending upon principle of operating: passive controllers. Section IV gives description of simulation
suspension, semi-active suspension and active suspension resuIt and comparison of control schemes. Section V gives
system. Passive suspension system consists of springs and the conclusion.
dampers. Softer dampers provide a more comfortable
11. MODELLlNG OF 3 DOF QUARTER CAR MODEL
drive and stiffer damper provide more stable drive.
Therefore, compromise has to make between driving
comfort and stability.
In semi active suspension system consists of a
variable damper that can adapt to actual demands. The
active suspension system contains separate actuator that
can exert extra force on the suspension system. Different
characteristics can be considered in design of a suspension
system [ 5]. It is not possible to optimize all these
parameters all together in a suspension system. But a
better trade- off among these parameters can be achieved
in active suspension system [ 5]
In the recent years, many researchers have
investigated different types of active suspension system
using variety of models like 1/4 or quarter car model, 1/2
or half car model, full car model etc. Initially researchers
opted for linear lower order models like 2 degree of Fig. I: 3 DOF Quarter Car Model of Active Suspension System [7]

978-1-4673-8587-9/16/$31.00 ©2016 IEEE [1)

Downloaded from https://iranpaper.ir

t
1s IEEE International Conference on Power Electronics, Intelligent Control and Energy Systems (ICPEICES-2016)

In this work, a 3 DOF quarter car model is used. It is III. CONTROLLER DESIGN
simplified version of the full car model representing most
In this work, three controllers are designed i.e. PID,
of the features of the full car model. It consists of
LQR, Fuzzy logic controller.
passenger mass along with the seat which is represented
using spring and damper. The sprung mass, also known as A. PID Controller
mas of the car body is supported on springs and dampers
and the unsprung mass represents mass of wheel. Tyre is Proportional - Integral - Derivative (PID) controller
replaced by using spring. Fig. 1 represents the 3 DOF is a commonly used control scheme in many industrial
quarter car model for active suspension system [7]. control systems. In this work, PID controller is designed
such that the output parameter stabilizes quickly under
A. Equations presence of road disturbance thereby minimizing error.
Differential equations of the system are attained by
implementing Newton's second law of motion to the seat,
sprung mass and unsprung mass and the equations are as
folIows.
mpx� + kp(xp - xs) + cp(xp - xs) = 0 (1)
msx� - kp(xp - xs) - cp(xp - xs) + ks(xs -
Xus) + cs(xs - xuJ = fa (2)
musx�s - ks(xs - xuJ - cs(xs - xuJ +
kt(xus - r ) = -fa (3)
By taking Xl = Xp ,X2 = Xp,X3 = XS'X4 = Xs,Xs = Fig. 2: PlD controller [8]
Xus'X6 = Xus , the equation can be written in terms of
The PID controller is represented by:
state variable Xl'X2'X3'X4'Xs,X6as follows de
x=AX+BU+GW �) u(t) = Kp x e( t) + Ki f.ot e(T)dT + Kd dt (11)
Where X=state input variable matrix Where Kp is the proportional gain, Ki is the integral
U=control input variable matrix gain and Kd is derivative gain, e is the positional error.
W=road input matrix. There are several tuning method used for finding proper
Xl = X 2 ( 5) Kp, Ki and Kd value. To obtain the desired response, the
X2 = -l/mp[kp(xl - X3) + Cp(X2 - x4)] ( 6) following sequence can be used as a general rule:
X3 = x4 (7) • In order to improve the rise time a proportional

- x4 [kp(X3 -
=

cp(x4 - kS(x3 - s ]
+ [,a (8) •
element should be added,
In order to improve the overshoot, a derivative

Ir m
s
Xl) +
CS(x4 -
x2) +
x6)
x)+
•
element should be added,
In order to improve the steady state error, an
(9) integral element should be added [8].
X6 In this work, genetic algorithm technique is used for

[ks(xs - x3) CS(x6 - x4) kt(xs -

=

-1/
]
- [,a ( 10) proper tuning of PID parameters.
/ mus + + r)
B. LQR Controller
The passive suspension system can be designed using
same equations with fa=O. Hydraulic dynamics of force I) LQR Controller Design/or Linear Systems with
actuator is not taken into account here and it is believed Measurable Disturbances
that the required force is applied in between the sprung
Standard LQR control approach syntheses an optimal
and unsprung mass. Table I shows value of all the system
controller for systems without disturbances. But most of
parameters used in the model [ 6].
the systems are affected by disturbances that also effect
TABLE I : SYSTEM PARAMETER VALUES control optimality. In this section, the LQR control design
Symbol System Parameter for linear systems that are excited by disturbances will be
Description Value Unit presented.
mp Passenger mass 100 Kg
ms Sprung mass 2050 Kg
2) Problem Definition
mus Unsprung mass 100 Kg
kp Stiffuess of the seat 100000 N/m Considering general form of a linear time-varying
Cp Damping coefficient of seat 6000 Ns/m
ks Stiffuess of suspension system 400000 N/m
system including disturbances as folIows:
Cs Damping coeffcient of suspension system 5000 Ns/m x=Ax(t) +Bu(t) +Bww(t)
Stif fuess of the tyre 2000000 N/m Y = Cx(t) + Dy(t) + Dw(t) (12)

[21
Downloaded from https://iranpaper.ir

t
1s IEEE International Conference on Power Electronics, Intelligent Control and Energy Systems (ICPEICES-2016)

w w
Where,Bw E Rnx and Dw E Rmx .The disturbances C. FUZZY Logic Controller
w(t) E RW is supposed to be measurable in real time. The
Fuzzy logic control (FLC) algorithm gives a means of
LQR problem for the system defined by eq.(12) is to find
converting a linguistic control technique and it is widely
a control input u(t) which minimizes the objective
used in vehicle applications. In this work, the application
function J defrned by (13).
of fuzzy logic technique to design a controller for the
] = fooo[(Cx + Du + DwW)T Qy(Cx + Du + Dww) + active vehicle suspension system to improve the
uTRuu]dt suspension system performance is demonstrated.
= fooo[(Cx + DU)T Qy(CX + Du) + 2(Cx + The steps in designing fuzzy logic controller are as
Du?QyDww + wTDwT QyDww) + uTR uu]dt folIows:
• Gaussian membership functions are defined for
= fOOO[xT (cr QyC )x + 2xT(cr QyD) u +
each variable that are used to fuzzity inputs and
uT(DTQyD + R u)u] dt + fooo{2xTC T QyDww + output of suspension system. Linguistic variables
2wTDwT QyDu + wTDwT QyDww}dt assigned to these fuzzy sets are, NB, NM, NS, ZE,
== f OO[(xT Qx + 2xTNu + uTRu)}dt + (2xTNxw w +
PS, PM and PB. Total 49 rules are made.
O
• .Mamdani type inference method and centroid
2wTNuwu + wTRww)]dt
defuzzification method are used. Output of fuzzy
== fooo[F(u, x, w, t)dt (13)
controller is actuator force based on two inputs i.e.
Different from the standard LQR problem is the the sprung mass displacement error and the
second group of terms that depends on disturbances w(t). corresponding rate of change of velocity of the
Therefore, the control law has to be a function of both x(t) suspension system.
and w(t). Fuzzy control rules are made in the form of Table 11
Using Pontryagin 's maximum principle and Euler­ of the designed fuzzy logic controller.
Lagrange equation, the optimal control is defined by,
TABLE 2: Fuzzy CONTROL RULES
u* = ux * +uw* = -Kx*x - Kw*w. (14)
Optimal state feedback controller gain Kx * is given by Dis lacement (Error)
PB PM PS ZE NS NM NB
K/ = - R - 1 (NT + BTP) (1 5
.
Optimal disturbance feed forward controller garn Kw
! �

�
PB
PM
PB
PM
PM
PM
PM
PM
PB
PM
PM
PS
PS
ZE
ZE
NS
is given by
...
e
PS PM PM PM PS ZE NS NM
g(!) ZE PM PM PS ZE NS NM NM
Kw* = �
- R - 1 {NuwT + BT[AT + (K/)TBT]-l[(K/)TNuwT -
NS PM PS ZE NS NM NM NM
'(3
0 NM PS ZE NS NM NM NM NM
Q)
(Nxw + PBw)]}w (1 6) NB ZE NS NM NM NM NM NM

P matrix must satisty the reduced form of the standard NM NS ZE ps Pt.!

Riccati equation shown in eq. (17).
P + PA + AT P - (PB + N)R-1(NT + BT P) + Q 0 (17) =

disturbance-feed fOlll'{ard controller

u,,(t} w(t)

-0.1 -0.09 -0.08 -0.07 -0.06 -0.05 -0.04 -0.03 -0.02 -0.01
input variable "DISPLACEt.!ENr
.v(t)
x = A{t)x + B(t)u + B ... {t)w Fig. 4: (a) Membership Functiün für Relative Displacement Errür

x(t)

model plant

state-feedback controller
ilput variable "VELO CITY"
Fig. 3: LQR Cüntrol Scheme für System with
Fig. 4: (b) Membership Functiün für Relative Velücity Errür
Measurable Disturbances[9]

[3)
Downloaded from https://iranpaper.ir

t
1s IEEE International Conference on Power Electronics, Intelligent Control and Energy Systems (ICPEICES-2016)

D. Result
I -PID
0
PASSIVE
0 I LQR
I-ruzzy
0

0 ...
� � . .

L� L� Lt---.
0
. �.
�

'--/ '--../
V
�

0
• • • • •
0
V
0
0.5 1. 5 2,5
3 -2 -1 2 3 4
-
time{sec.)
output variable " F ORCE' ,
x10
Fig. 6: Passenger Acceleration (mJsec2) vs. Time (sec)
Fig. 4: (c) Membership Function for Output Actuator Force
� 0.05 f\
·ß
� 0
� W l f\'\7A /'.. I
."
§ -0.05 r �'1 V -PID
PASSIVE l -
.� -LQR
2 -0. I
w
U
t -FUZZY

0:: o 0.5 1.5 2.5 3.5 4 4.5

o 0 time�
u..

Fig. 7: Suspension Detlection (m) vs. Time (sec)

-2

1.5 itl ·········<· ········, ········, ········, -PIO

I -PASSIVE
········+ ···...... , .....
. , ......... +

o
�
I�
. . . . ... .
l lW.�\ .. ....;. .. .. . ; .. .. ; .... . ; .... . ;..... .,.... . ; .........; .1- LQR . ..
- FUZZY
.
VELOCITY DISPLACEMENT
� 0.5W·!H····;f I·· ···: f\ ·····c······ ··.· ··.... ..• ..... . ;.........• .....··c· ··· ···;·· ··· -I
'g
� 0·· 1'P"/---'l--,r:=""'=--'I---'<--+-
--'l----7'---"<----7'�:-::r=..-=i-'-=-
---i-�-I
Fig. 5: Fuzzy Surface of the Proposed Controller with 49 Rules "
o

.0.5 ..
Figure 5 shows the fuzzy surface for the designed
controller using the Gaussian membership function .,��=�=�="!:='c::=:,t;:=�==::t;=�=��
o 0.5 15
. 2,5 35
. 4.5
ti��)
IV. RESULT AND DISCUSSION
Fig. 8: Seat Velocity (mJsec) vs. Time (sec)
A step input signal of amplitude 0.1 m is chosen to test
the designed system. This step signal represents a pavement
1 5
.
�1 · ·····,·· ·······,· ·······,········, ·· ······,· ····· . ;
-PASSlVE

which is a sudden change of road height of 10 cm. "

�
1 �1 ·········; ··········;···············, ···········;··............ , ....... -LQR
-ruzzy

A. PID Controller �
.�
�
o .5 ·
HI I · ··
J ... .
\- .. , ." .. . , .. .. .., .... .. , .... ....
In MATLAB/SIMULINK environment using proper
coding following values of PID parameters are found:
Kp= 10000, Ki= 23470 and Kd= 5 0000. ·0.5 ..

B. LQR Controller . 10 � ----: - :"-:-----:----

'-:----:---- - -:':----!----:3"-:-----O---"=-�
5 4 5 .
time(sec.)
.

Using Matlab coding he values for Kx * and Kw *used

in the simulation are found to be as, Fig. 9: Sprung Mass Velocity (mJsec) vs. Time (sec)

K/ =1.0e+04 * � � . .+.·:-:-
0. 1 8'
.."·:-:c ..·:-:c
..::-:
..+.. _ :-:c
..::-:
..:-:-
.. ·+'..
, ::-:
..:-:- . .+.·:-:-.. ·::-!
.. ·:-:c ..?:-·. :-:c::-:
..:-:-
..+,._ ::-:
..:-:c :-:c+"
....
..:-:- . ..:-:-:-:cc;=====;;?
... . . ....

-PID
E .16
":f 0 �
' �: ": � : �' : -PASSIV[
[-0.3288 0. 685 8 1.7 517 0.102 5 -0.34640.003 5] and - - --- -------- --------- --------- -------- -------- --------- -------

E 0.14 ' : ' ,� L: � !: -LQR

1:

Kw * -S.6011e + 03
=
______ ________
-ruzzy
_______ _________ _______

'
� 0.12
<!)

C. FUZZY Controller � 0.1

'Ö
lZl 0.08
Fuzzy logic controller is designed using � 0.06
E
MALAB/SIMULINK's FUZZY logic toolbox. Instead of
� 0.04
using variable step 'ode45 'solver, 'ode 8' solver with 0.01 � 0.02
sampling time is used throughout the experiment as the {/l

response gets much slower in 'ode 45 '. Justification factor

used are Ke= 0.85and Kec=12. Defuzzification factor used
is KO=2. Fig. 10: Sprung Mass Displacement (m) vs. Time (sec)

[41
Downloaded from https://iranpaper.ir

t
1s IEEE International Conference on Power Electronics, Intelligent Control and Energy Systems (ICPEICES-2016)

Figure 6,7,8,9 and 10 show the comparison between schemes. Simulation results depict that there are
the PID, Fuzzy and LQR based active suspension system considerable differences between the resuIts of passive
and passive suspension system. Table III, IV, V, VI and and active suspension system. Another conclusion is that
VII show the comparison of the active suspension the three control schemes proposed here gave good resuIts
system equipped with proposed controllers with passive especially for reduction of settIing time and maximum
suspension system. peak to peak amplitude.
TABLE 3: PERFORMANCE CHARACTERISTlCS
In case of performance parameters like seat
OF SUSPENSION DEFLECTlON acceleration, suspension travel, sprung mass velocity, seat
Suspension Deflection velocity, sprung mass displacement Fuzzy control scheme
LQR PID FUZZY PASSIVE gives better result as compared to other control methods.
Maximum peak value (m) 0.0432 0.0285 0 0.0685 From the simulation result, it is c1ear that all the three
Senling time (sec.) 2.2 0.7 0.6 7.5
controllers are successfully designed but Fuzzy control
Steady state error (%) 0 0 0 0
scheme gives better performance than other control
TABLE 4: PERFORMANCE CHARACTERISTlCS OF SEAT ACCELERA TlON methods and passive system.
Seat Acceleration
LQR PID FUZZY PASSIVE ACKNOWLEDGMENT
Maximum peak value 23.609 46.425 1 33.3203 23.7 147
(mJsec"2) The author would like to thank Dr. Mrinal
Senling time (sec.) 2 0.9 0.7 8.5 Buragohain, Associate Professor of Jorhat Engineering
Steady state error (%) 0 0 0 0
College, India for helpful advice and encouragement
TABLE 5: PERFORMANCE CHARACTERISTlCS OF SEAT VELOCITY throughout the period of this work.
Seat Velocity
LQR PID FUZZY PASSIVE REFERENCES
Maximum peak value 1.0868 1.77 16 0.9842 1.2525
[I] Mat Hussin Ab. Talib, Intan Z. Mat Darns, "Self-Tuning PlD
(mJsec)
Controller for active suspension system with hydraulic
Senling time (sec.) 2 I 0.8 9
actuator,"IEEE Symposium on Computers & Informatics
Steady state error (%) 0 0 0 0
(ISCI),pp.86-9 1 ,April 20 13.
TABLE 6: PERFORMANCE CHARACTERISTlCS [2] Salah G. Foda, "Fuzzy control of a quarter-car suspension system,"
OF SPRUNG MASS VELOCITY 12th International Conference on M icroelectronics,2000
Sprung Mass Velocity [3] Ismail etal"A linear model of quarter car active suspension system
LQR PID FUZZY PASSIVE using composite nonlinear feedback control,"20 12 IEEE Student
Maximum peak value 0.99 19 1.623 0.9664 1.1545 Conference on Research and Development(SCOReD).
(mJsec) [4] Elnaz Akbari, Morteza Farsadi, Intan Z.Mat Darus,Ramin Ghelichi,
Senling time (sec.) 2.1 I 0.9 9 " Observer Design for Active Suspension System U sing Sliding
Steady state error (%) 0 0 0 0 Mode Control," 20I0 IEEE Student Conference on Research and
Development (SCOReD 20 10), 13 - 14 Dec 20 10, Putrajaya,
TABLE 7: PERFORMANCE CHARACTERISTlCS OF
Malaysia K. Elissa.
SPRUNG MASS DISPLACEMENT
[5] A.H. Shirdel ,E. Gatavi, Z. Hashemiyan, "Comparison of H-oo and
Sprune; Mass Displacement
optimized-LQR controller in active suspension system,"Second
LQR PID FUZZY PASSIVE
International Conference on Computational Intelligence, Modelling
Maximum peak value 0.1525 0. 1412 0.1 0.1823
and Simulation ,20 10.
(mJsec)
Senling time (sec.) 2.2 I 0.8 8 [6] Tinnavelli Ramamohan Rao and Punjala Anusha, "Active
Steady state error (%) 0 0 0 0 Suspension System of a 3 DOF Quarter Car Using Fuzzy Logic
Control for Ride Comfort," 20 13 International Conference on
V. CONCLUSION Control, Automation, Robotics and Embedded(CARE).
[7] B.Pratheepa, "Modeling and simulation of automobile suspension
The potential for improved ride comfort and better system," Frontiers in Automobile and Mechanical Engineering
road holding using Fuzzy, LQR and PID controller design (FAME),20I O.
is examined for the analysis of a three degree of freedom [8] Sallehuddin Mohamed Haris, Wajdi S. Aboud, "International
Conference onAdvanced Mechatronics Systems (ICAMechS),20 1 1.
quarter car model. Maximum peak to peak amplitude,
[9] Tuan-Anh Nguyen, "Application of Optimization Methods to
steady state error and settIing time are used to investigate
Controller Designfor ActiveSu spensions"2006.
the performance characteristics of different controller

[5)