2009 5th International Colloquium on Signal Processing & Its Applications (CSPA)

Modelling and Controller Design for a Cruise Control

System
Khairuddin Osman, Mohd. Fuaad Rahmat, Mohd Ashraf Ahmad

Abstract- This paper presents the performance of different dampers, springs and masses can be reduced to much simpler
control approaches that consist of conventional, modern and form of model such as moving cart. By the reduction, the
intelligent controller that employed in a cruise control system. system seems much simpler and great amount of complex and
The cruise control system is one of the most enduringly popular tedious calculations has been avoided. For example,
and important laboratory models for teaching control system
calculations that involve the masses, springs, dampers and
engineering. The system is widely used because it is very simple
to understand and yet the control techniques cover many other related products. In modelling a system cruise control, it
important classical and modern design methods. In this paper, is important that the model should represent the most general
the mathematical modelling for linear and nonlinear dynamic situation or case so that the model will take into accounts all
model of the cruise control system is obtained. PID, state space of the important parameters, including those that are due to
and artificial intelligence controller (fuzzy logic) are designed for disturbances which directly or indirectly affect the overall
linear model. Meanwhile, PID with feedforward controller is performance of the system.
proposed for nonlinear model with disturbance effect. After modelling the cruise control system, the design of the
Feedforward-type PD controller is proposed in this study in controller such as PID control can be applied and the stability
order to eliminate the gravitational and wind disturbance effect.
analysis based on linear state-space model or transfer function
Simulation will be carried out using. Finally, a comparative
assessment of the impact of each controller on the basis of results is analyzed. Furthermore, some modern controller such as
acquired. state space controller [1] and intelligent controllers for cruise
control such as fuzzy control [2], neural control [3], can also
I. INTRODUCTION be found. Although a simple and quick simulation of the
feedback system may be handy for a quick check of the
Automobile cruise control system is functional as an design, a more accurate simulation should be done by
automatic speed control for a car. Thus, it maintains the speed applying the controller to the original nonlinear model.
of the car throughout a journey. The output of the system This paper present several controllers design such as PID
which is speed is controlled by the controller in order to controller, state space controller and artificial intelligence
provide the desired speed at which the car is to be maintained. controller (fuzzy logic) for linear model. Meanwhile, the
Normally, the drivers have to press step the acceleration pedal nonlinear model focuses on PID with feedforward controller.
consistently, to maintain the car’s speed. The controller Feedforward-type PD controller is proposed in this study in
provides comfortability and easiness to drivers when driving order to eliminate the gravitational and wind disturbance
the car. Comfortability means driving without having to effect. Finally, a comparative assessment of the impact of each
control the pedals frequently and less tiring. Easiness means controller on the system performance is presented and
controlling the speed of the car by pressing buttons instead of discussed.
pedals.
The problem of cruise control system is to maintain the II. MODELLING THE CRUISE CONTROL SYSTEM
output speed of the system as set by input signal. The
example, if driver desires to cruise or maintain the car speed at The purpose of the cruise control system is regulating the
110 meter per second, then the system should be able to vehicle speed so that it follows the driver’s command and
produce the desired output. This can be achieved by various maintains the speed at the commanded level. Base on the
methods of controller such as using proportional-integral- command signal vR from the driver and the feedback signal
derivatives (PID) controller, state-space controller, fuzzy logic from the speed sensor, the cruise controller regulates vehicle
controller and many more. The cruise control system also speed v by adjusting the engine throttle angle u to increase or
experienced problem when there are disturbances from decrease the engine drive force Fd. The longitudinal dynamics
external system such as gravitational force and wind gust of the vehicle as governed by Newton’s low (or d’Alembert’s
speed. However these short-falls can be overcome by adding principle) is
cascade controller and feedforward controller to cruise control
d (1)
system. Fd  M v  Fa  Fg
Modelling is a task that requires creatively and problem- dt
solving skills. A complex model of a car with where M(dv/dt) is the inertia force, Fa is the aerodynamic drag
and Fg is the climbing resistance or downgrade force. The
forces Fd, Fa, and Fg are produced as shown in the model of
Fig. 1 [4], where vw is the wind gust speed, M is the mass of

978-1-4244-4152-5/09/$25.00 ©2009 IEEE 254

Authorized licensed use limited to: ULAKBIM UASL - YILDIZ TEKNIK UNIVERSITESI. Downloaded on November 26,2024 at 11:26:51 UTC from IEEE Xplore. Restrictions apply.
 the vehicle and passenger(s), θ is the road grade, and Ca is the
aerodynamic drag coefficient. The throttle actuator and In the equation, δv means that the output is discrete and δFd
vehicle propulsion system are modeled as a time delay in also means that drive force is discrete. The symbol v means
cascade with a first order lag and a force saturation the desired and δu(t-τ) is the time delay of the engine. Up to
characteristic. The following parameter values are adopted this point, both the state and output equations are written in
from [5]. However, some values need to be modified so that time domain. The linearized model provides a transfer
the block diagram could represent the same model with function can be obtained by solving the state-equations for the
slightly different values just to provide computing and ratio of V ( s ) / U ( s ) .
calculation challenges rather than reusing the identical values: C1e s
C1=743, T=1s, τ=0.2s, M=1500kg, Ca=1.19N/(m/s)2, V ( s )
(8)
 MT
Fdmax=3500N, Fdmin=-3500N, and gravity constant g=9.8m/s2. U ( s )  2C a v  1
s   s  
 M  T
vw

 Using the power series expansion approximation for the

Fg   Fa
 Mg sin(  ) Ca ( v  v w ) 2 time delays, the time delay of the transfer function of the
 system is approximated to be

C1e s
Fd max Fd  F 1 v 1 1
u
Ts  1 M s
v
s 1  (9)
Fd min
e  
1  s 1
s
Fig. 1. Vehicle longitudinal by dynamic model 

The controller design for this system begins by simplifying Substitute this expression into the plant transfer function of
the model. Consider to sell all the initial conditions to zero.
The same applies to the disturbance parameters. Hence, it is C1
(10)
assumed is no wind gust and no grading exists during the V ( s ) MT
G p ( s)  
movement of the car. Applying this zero initial condition to U ( s )  2C a v  1  1
s   s   s  
the block diagram, the model is left with the forward path and  M  T   
the unity feedback loop of the output speed. Since the state-
variables have been chosen to be the output speed and the It is obvious that the system described by this transfer
drive force, the corresponding state and output equations are function is third order system, as a result of the time delay
found to be: approximation. Despite of that, the transfer function has been
successfully linearized. The upcoming calculation shall be as
v 
1
( Fd  C a v 2 ) (2) difficult as if the linearization has not been done. At least the
M complexity of the calculations should be reduced.
Hence, after substituting the values of the constants into
1
Fd  C1u (t  T )  Fd  (3) equation (10), the final form of the linearized transfer function
T derived from the block diagram through state equation is
shown below.
yv (4)
V ( s) 2.4767 (11)
G p (s)  
However, a problem of non linearity arises. There is a U ( s ) s  0.0476s  1s  5
squared term in the equation (2). One way to overcome this
problem is to linearize all of the state-equations by The linearized system equations can also be represented in
differentiating both left and right hand sides of the equations state-space form. This can be done by Linear Ordinary
with M, Ca, C1, T and v remain constant. After differentiating, Difference Equation [7]. The state-space representation is
the state-equations become shown in equation (12-13). The mathematical model in state
space form is used to design the pole placement controller in
d 1 (5) the next section.
v  (2C a vv  Fd )
dt M
 v1   0 1 0   v1   0  (12)
d  1 (6) v     v    0  u (t ) 
Fd  C1u (t  T )  Fd   2  0 0 1  2   
dt T  v3    6.0476  5.2856  0.238   v3   2.4767 

and the output equation becomes

 v1 
(13)
y (t )  1 0 0v 2 
y  v (7)
v3 

255
Authorized licensed use limited to: ULAKBIM UASL - YILDIZ TEKNIK UNIVERSITESI. Downloaded on November 26,2024 at 11:26:51 UTC from IEEE Xplore. Restrictions apply.
 which is design for a linearized small-signal model. When it is
III. CONTROLLER DESIGN used to control the original plant, the nominal input u0 must be
reinserted to introduce the nominal throttle angle and velocity.
The problem of cruise control system is to maintain the With controller as described, the injection of u0 can be
output speed v of the system as set by input signal base on the accomplished by setting u0 as the initial condition of the
command signal vR from the driver. Cruise controller design is integrator when cruised control is initiated. In practice, the
applied assuming a single-loop system configuration with a required nominal values are determined by the state of the
linear model and nonlinear model, as shown in Fig. 2. The vehicle at the moment when the cruise control is activated.
controller function Gc(s) is designed to augment or modify the The basic concept of feedforward (FF) control is to measure
open-loop function in a manner that produces the desired important disturbance variables and take corrective action
closed-loop performance characteristics. The plant functions before they upset the process to improve the performance
Gp(s) represent the actuators and the controller part of the result. A feedforward control system is shown in Fig. 4, where
system, and the plant parameters are determined primarily by disturbances are measured and compensating control actions
functional aspects of the control task. Before make any are taken through the feedforward controller. Deviations in the
decision of controller design, a few of design specification controlled variables can be calculated as
have been set. In this design, we take two considerations to be
met which are settling time Ts less than 5s and percentage of (15)
v  G p G ff D  Gd D
overshot %OS is less than 10%. Controller designs are
dividing into two sections such as design for linear model and
design for nonlinear model. Linear model controllers are where Gp is the process transfer function model, Gd is the
focused to three controller such as proportional-integral- disturbance model, Gff is the feedforward controller, u is a
derivatives (PID) controller, state-space controller and fuzzy vector of the manipulated variables, and D is a vector of
logic controller. For nonlinear model, proportional-integral- disturbances where D1 and D2 are disturbances from external
derivatives (PID) with feedforward and feedforward-type PD system such as gravitational force and wind gust speed.
controller are proposed.
D1 D2
D


v R (s ) G c (s) G p (s ) v(s )
e(s) u (s)


 
 
u v
Fig. 2. The cruise control system configuration

Nonlinear model using PID controller is design of the Fig. 4. A feedforward control system
controller has been based on use of the linearized model.
Although a simple and quick check of the design, a more In order to make Δv zero, a feedforward controller of the
accurate simulation should be done by applying the controller following form is designed. Supposed that,
to the original nonlinear model as shown in Fig. 3.
 kd  (16)
G d   
 d s 1
u0

 kp  (17)
 Gp   

K p e(t )  K i  e( ) d  K d
de(t ) u  u  ( s  1)( s  1)( s  1) 
vR
dt
v  p1 p2 p3 


Then from equation (16-17), the ideal feedforward controller,

1  ( p1 s  1)( p 2 s  1)( p 3 s  1)  (18)

Fig. 3. PID control applied to the nonlinear plant
G ff  G p Gd  k ff  
 ( d s  1) 
The practical issues need to be considered where controller
where feedforward controller gain is k ff   k d / k p . In process
is described by
plant applications, Gff is typically chosen to be of static form,
de(t ) (14) and it is this approach that will be adopted in this project. The
u (t )  K p e(t )  K i  e( )d  K d solutions of Gp refer process reaction curve method [4] and
dt
solutions of Gd refer equation (11). After simulate, if the
performance of the system is not really good, the values of kff

256
Authorized licensed use limited to: ULAKBIM UASL - YILDIZ TEKNIK UNIVERSITESI. Downloaded on November 26,2024 at 11:26:51 UTC from IEEE Xplore. Restrictions apply.
 will be tuned manually to get the best response where the TABLE 1
value of feedforward controllers gain kff1 and kff2 are 0.5 and 1.
Controller
Add Feedforward-type PD Control (FFPD) control is a Specifications
new technique to improve feedforward control is to measure PID State Space Fuzzy Logic
important disturbance variables and take corrective action Percent Overshoot
10.2% 10% 1.91%
before they upset the process. Based on the model, the FFPD (%OS)
controller combined a feedforward path with the proportional- Peak Time (TP) 3.54s 2.97s 3.16s
derivative (PD) control inside feedforward controller Gff. To
cooperate with the feedforward action, a PD mode was applied Settling Time (TS) 5.5s 5s 3.37s
for small error conditions to eliminate the steady state offset.
Rise Time (TR) 1.7s 1.38s 2.21s
To design the FFPD, the process and to get the value of kff in
FF are applied to KP1 and KP2. The values of KD will be tuned Percent Steady State
0.01 0.01 0.01
error (%ess)
manually to get the best response where the both KD1 and KD2
are 1.
B. Comparison Results for Nonlinear Model
IV. RESULT AND ANALYSIS The original non-linear model provided the realistic
scenario. By comparing the results presented in Fig. 6 and
A. Comparison Results for Linear Model Table 2 where compare the percent overshoot %OS, peak time
To perform comparison between controller design for linear TP and settling time TS to know the stability of the system. It is
model cruise control system, one of the first things that must be noted that the performances to original nonlinear model (PID),
done during controller design is deciding upon a criterion for modified by adding the feedforward (FF) and adding the
measuring how good a response is. However, in dynamic feedforward-type PD (FFPD) respect to any possible
systems where the transient behaviour is also important, it disturbance occurred. Without adding feedforward, the
becomes important to introduce several other criterions. Fig. 5 percentage overshoot for jerking are big and settling time also
shows the overall step responses of the velocity (m/s) versus increase compared adding the feedforward controls. Clearly,
the time (s) are compared. The most common are compare the the advantages of feedforward-type PD are it can decrease the
percent overshoot %OS, peak time TP, settling time TS, rise time jerking and provide better result in terms of performance.
TR and the percent steady state error (%ess) to know the
stability of the system. All the specifications values of the 160
Desired
responses are summarized in Table 1. By comparing the 140 PID
characteristics of proportional-integral-derivatives (PID) PID-FF
PID-FFPD
controller, state space controller and fuzzy logic controller, 120

fuzzy logic controller is more stable than proportional-integral- 100

derivatives (PID) controller and state-space controller.
Velocity (m/s)

80

220 60
Desired
200
Fuzzy 40
180 PID
State space
20
160

140 0
0 50 100 150 200 250 300 350 400 450 500
Velocity (m/s)

120 Time (s)

100
Fig. 6. Comparison of without feedforward controller and adding feedforward
80
controllers responses
60

40 V. CONCLUSION
20

0 The mathematical model for cruise control system has

0 2 4 6 8 10 12 14 16 18 20
Time (s) been derived successfully. The plant is consists of two plant
which are linear model and nonlinear model. Several
Fig. 5. Comparison of the overall controller responses controller of conventional, modern and intelligent scheme
have been successfully designed to control the cruise control
system. However, the analysis results had shown that to
achieve better simulation result, a controller has to be applied
to the nonlinear model. This method will lead to real
performance of the system in actual condition. On the other
hand, if just compared on the controllers design only it will

257
Authorized licensed use limited to: ULAKBIM UASL - YILDIZ TEKNIK UNIVERSITESI. Downloaded on November 26,2024 at 11:26:51 UTC from IEEE Xplore. Restrictions apply.
 not give a good result and valid because each controller have REFERENCES
their own tuning method. And there is a possibility where by
using only one suitable controller, we can achieve the desired [1] Carneige Mellon. 1997. Control Tutorial for Matlab, Website of the
University of Michigan.
output and response. That why, the analysis is focused on a
http://www.engin.umich.edu/group/ctm/examples/cruise/ccSS.html
nonlinear system with one PID controller. The feedforward [2] Muller, R. Nocker, G. Daimler-Benz AG and Stuttgart. 1992.
(FF) and feedforward-type PD (FFPD) also included for Intelligent Cruise Control with Fuzzy Logic, IEEE Intelligent Vehicles
further study and to compare the operation of the system when '92 Symposium.
[3] St. Germann and R. Isermann. Jun 1995. Nonlinear Distance and
disturbance occurred.
Cruise Control for Passenger Cars, American Control Conference,
IEEE American Control Conference.
ACKNOWLEDGMENT [4] Lewis, Paul H. and Yang Chang. 1997. Basic Control System
Engineering. Prentice-Hall, Inc. New Jersey, USA.
The authors gratefully acknowledge the contribution and [5] Frank, Andrew A., S. J. Liu, S. C. Liang. 1989. Longitudinal Control
Concepts for Automated Automobiles and Trucks Operating on a
supervision of Assoc. Prof. Dr Mohd. Fuaad Rahmat for his
Cooperative Highway. Society of Automotive Engineers Technical
willingness to spare his time for knowledge and expertise Paper Series (891708).
sharing. Also indebted to Universiti Teknikal Malaysia [6] Thomas E.Marlin. 2002. Process Control: Designing Processes and
Melaka (UTeM) for their encouragement and financial Control Systems for Dynamic Performance. 2nd Ed . Mc. Graw Hill.
[7] Ogata, Katsuhiko. 2002. Modern Control Engineering. 4th Ed. Prentice-
support.
Hall, Inc. New Jersey, USA.

Table 2
Disturbance

Specifications Wind gust speed Gravitational force

Increase Decrease Increase Decrease

Percent Overshoot (%OS) -5% 4.55% -5.64% 5.64%

PID Peak Time (TP) 102.20s 202s 301.90s 402.10s

Settling Time (TS) 132.50s 232s 336.50s 436.40s

Percent Overshoot (%OS) -2.27% 2.64% -2.73% 2.82%

PID-FF Peak Time (TP) 100.80s 201s 300.80s 401.10s

Settling Time (TS) 125.60s 224.60s 331.70s 432.30s

Percent Overshoot (%OS) -1% 1.09% -1.18% 1.18%

PID-FFPD Peak Time (TP) 100.30s 201.20s 301.90s 402.40s

Settling Time (TS) 121.90s 222.40s 325s 425.10s

258
Authorized licensed use limited to: ULAKBIM UASL - YILDIZ TEKNIK UNIVERSITESI. Downloaded on November 26,2024 at 11:26:51 UTC from IEEE Xplore. Restrictions apply.