Adaptive cruise control for an intelligent vehicle

Full text

Adaptive Cruise Control for an Intelligent Vehicle
Worrawut Pananurak, Somphong Thanok, Manukid Parnichkun
School of Engineering and Technology
Asian Institute of Technology
P.O. Box 4,Klongluang,Prathumthani 12120, Thailand
worrawpa@scg.co.th, somphongt@kmutnb.ac.th, manukid@ait.ac.th
Abstract—In this research, an adaptive cruise control system
is developed and implemented on an AIT intelligent vehicle. To
develop the adaptive cruise control system, the original throttle
system and braking system of the vehicle have to be modified.
The original throttle valve which is controlled by a cable from
the accelerator pedal is modified to the drive-by-wire system
by using a dc motor with a position control algorithm. The
braking system is modified by using a dc servo motor to directly
control the brake pedal. A proportional and derivative control
with error compensation algorithm is proposed to perform the
velocity control mode. In the distance control mode, a fuzzy
logic algorithm is applied. Inputs of the fuzzy controller are
distance error and relative velocity read from a laser range
finder. The experiments on a racing circuit show that the vehicle
can perform adaptive cruise control efficiently.
Index Terms—adaptive cruise control, cruise control, lidar,
radar, automotive technology
I. INTRODUCTION
Cruise control system is developed for highway driving.
This system is useful for driving in the roads which are
big, straight, and the destination is farther apart. When
traffic congestion is increasing, the conventional cruise
control becomes less useful. The adaptive cruise control
(ACC) system is developed to cope up with this situation.
The conventional cruise control provides a vehicle with
one mode of control, velocity control. On the other hand,
ACC provides with two modes of control, velocity and
distance control. ACC reduces the stress of driving in dense
traffic by acting as a longitudinal control pilot. ACC can
work like the conventional cruise control that it is used for
maintaining the vehicle’s preset velocity. Unlike the cruise
control, however, ACC can automatically adjust velocity in
order to maintain a proper distance between obstacle and
the vehicle equipped with ACC. This is achieved by using
laser or radar to measure the relative distance between the
host vehicle and a vehicle in front.
Low-speed ACC is one of the systems, which operates
under congested traffic to maintain the distance behind the
obstacle vehicle. This type of ACC system is sometimes
called ”stop-and-go ACC.” Early versions may only perform
a ”stop and wait” function which requires drivers to initiate
a resumption of forward movement when appropriate. The
reason is that manufacturers are hesitant to offer such a
system to automatically operate in complex low-speed traffic
environments, which may have bicycles and pedestrians.
The general low-speed ACC system is operated at very
low speed (approximately 5 km/hr) and requires the driver
to interfere to stop and restart vehicle motion. Low-speed
ACC was introduced to the Japanese market in 2004.
High-speed ACC system is the evolution of the cruise
control. The system provides velocity control as in
conventional cruise control when there is no vehicle in front
of the host vehicle. If a vehicle runs in front of the host
vehicle at a slower speed, the throttle and braking system
are controlled to maintain the inter-vehicle gap which is set
by the driver. The host vehicle will run at the preset velocity
again when the way ahead is no obstructed, resulting from
either the slower vehicle ahead changes the lane or the
driver of the host vehicle moves to the other lane. The
first ACC systems were designed to operate at moderate to
high velocity, 40 km/hr and above. Most European systems
operate from 30 km/hr and higher because this is a typical
speed limit in city areas. The upper speed range goes as
high as 200 km/hr. Bishop, R.H. [2] mentioned that ACC
systems should be designed to have a limited braking
authority, on the order of 0.25g (full braking in a typical car
is 1.0g). In cases where the distance to the vehicle ahead
is near and the braking authority of the host vehicle is
inadequate to maintain the inter-vehicle gap, audible alerts
are sounded to force the driver to take control of the vehicle.
In this research, the ACC system is developed on an AIT
intelligent vehicle, Mitsubishi Galant, 1993. The authors
propose a fuzzy control algorithm to perform the ACC
function. Inputs of the fuzzy controller are distance error
and relative velocity. These inputs are read from the laser
range finder from SICK, LMS 291. Outputs of the controller
are braking command and velocity command.
II. HARDWARE
To develop the ACC system for the AIT intelligent vehi-
cle, hardware and sensors are designed and installed on the
platform.
A. AIT Intelligent Vehicle
The intelligent vehicle is developed on Mitsubishi Galant
GLSi, 1993 as shown in Fig. 1. This vehicle has a 2.0 liter,
gasoline powered engine, automatic transmission.
Proceedings of the 2008 IEEE
International Conference on Robotics and Biomimetics
Bangkok, Thailand, February 21 - 26, 2009
978-1-4244-2679-9/08/$25.00 ©2008 IEEE 1794

Fig. 1. The AIT intelligent vehicle, Mitsubishi Galant GLXi, 1993
B. Throttle Valve Control System
The original throttle valve control system is changed to
a drive-by-wire system so as to be able to control by the
ACC system. A 12v dc servo motor is installed to control
the throttle valve position. A potentiometer is installed at
the accelerator pedal to measure the pedal position. The
drive-by-wire controller is developed on an ARM7 micro-
controller which reads the required throttle position from
output voltage from the potentiometer. Fig. 2 shows the
modified throttle valve control system.
C. Automatic Braking Control System
To automatically control the braking system, a Cool Mus-
cle dc servo motor is used to control the braking system. The
motor, CM1C23L20 with 25:1 gear box ratio, is chosen. This
motor is a closed loop vector drive servo system utilizing
an H-infinity controller. A motor driver with a 32-bit RISC
CPU, a magnetic encoder, and a power management unit
are built into the motor. The motor is controlled by an
Arm7 microcontroller via serial communication. The motor
is installed in the vehicle as shown in Fig. 3. Rotational
movement of the motor is transferred to linear movement
by a pulley and a steel cable. If the motor rotates, the brake
pedal will be pull down by the cable.
D. Distance Sensor
The sensor used to measure the distance to the vehicle in
front is the laser range finder, SICK LMS 291. The sensor
is installed at the front bumper of the vehicle. This sensor
measures the distance based on a time-of-flight principle
(lidar). A single laser pulse is sent out and reflected by an
Fig. 2. The modified throttle valve control system
(a) (b)
Fig. 3. Automatic braking system: (a) Cool Muscle motor installed in the
vehicle, (b) break pedal connected with accelerator cable
object surface within the range of the sensor. The elapsed
time between transmitting and receiving of the laser pulse
is used to calculate the distance between the object and the
sensor.
III. SOF TWAR E
A. Drive-by-wire throttle valve position control
Block diagram of the position control algorithm of the
throttle valve is shown in Fig. 4. The inner loop is the
velocity control the outer loop is the position control. The
sampling time of the control loop is 2 ms. Digital low-pass
filter with 20 Hz cut-off frequency is used to filter noise in
the velocity signal.
With servo position control, the actuator is applied with a
control signal which is proportional to the amount of position
error and velocity error, as expressed in Equations (1) and
(2). The results of the Bode plot and tracking performance
are shown in Fig. 5 and 6, respectively.
Outputp=Kp(θ−θr)(1)
Outputv=Kv(ω−ωr)(2)
B. Adaptive Cruise Controller
Basically, the ACC has two operating modes. The first
mode is the velocity control mode which is operated when
there is no obstacle ahead of the host vehicle. The other is
the distance control mode which is operated when the host
vehicle finds the obstacle vehicle in front.
In the velocity control operation, the vehicle controls its
velocity by the ARM7 microcontroller, using proportional
and derivative control algorithm with command compen-
sator. The block diagram is shown in Fig. 7. The proportional
controller is used to remove the speed error. The derivative
Fig. 4. Throttle position control block diagram
1795

Fig. 5. Bode diagram of first-order low pass filter for velocity sensor
Fig. 6. Tracking result of motor position control
controller is used to reduce the overshoot and oscillation of
the velocity response. The control signal of the proportional
and derivative control can be described by Equation (3).
OutputP D =KP+KD
de
dt (3)
For distance control, the authors propose a fuzzy logic
algorithm to control the host vehicle. The distance and the
relative velocity between the host vehicle and the obstacle
are the inputs of the fuzzy controller which is implemented
in a PC. Fuzzy controller is suitable for multi-parameters and
nonlinear control problems, and the system transfer function
is not required. Human experience and experimental results
are used to design the controller. Block diagram of the
system is shown in Fig. 8.
In this research, Mamdani’s fuzzy inference method (FIS)
is applied. The singleton membership function of the outputs
is used. This type of output membership function makes it
convenient for the defuzzification process because it simpli-
fies the computational efforts compared with other types of
membership function. The entire fuzzy system is developed
and implemented by using MATLAB and Microsoft Visual
Basic compiler. Fig. 9 shows the block diagram of the fuzzy
controller. Two input variables are ”relative velocity” and
”distance error.” The block at the middle represents all the
fuzzy inference rules which are the control strategy of the
system. The output variables are commands to control brake
and velocity set point ratio. The inputs are defined as shown
in Fig. 10 and Equations (4) and (5).
Fig. 7. Cruise control block diagram
Fig. 8. ACC block diagram
disterror =distsensor −distset (4)
Relative Speed =Vobs −Vhost (5)
By using MATLAB, it is easy to select and adjust the
shapes of the membership functions. In this research, the
authors select trapezoid and triangle shape because it is easy
for programming. The ranges of membership function are
shown in Fig. 11 and 12. Each linguistic variable contains
seven terms. The meanings of each input variable are as
follows
NL : negative large PS : positive small
NM : negative medium PM : positive medium
NS : negative small PL : positive large
Z : zero
Fig. 9. Fuzzy control block diagram
Fig. 10. Fuzzy input definition
1796

Fig. 11. Membership functions of distance error
Fig. 12. Membership functions of relative velocity
The membership functions of the output are shown in Fig.
13. The outputs are divided into two sides. The negative
side represents the braking command. The positive side
represents the velocity ratio command.
NVL : negative very large PS : positive small
NL : negative large PM : positive medium
NM : negative medium PL : positive large
NS : negative small PVL : positive very large
Z : zero
The fuzzy rules for the ACC are the collection of
linguistic statements. These statements describe how the
fuzzy inference system should make a decision according to
the inputs. The example of fuzzy rules is shown as follows.
If distant error is negative large and relative speed is
negative large then command is negative very large.
Two input variables each with seven membership func-
tions are associated with nine singleton outputs. Totally there
are 49 rules. The rules are designed from experience and
adjusted by experiments. Table I shows the final fuzzy rules.
There are 5 steps to compute the output of the fuzzy
interference system.
1) Determining a set of fuzzy rules
2) Fuzzifying the inputs using the input membership
functions
3) Combining the fuzzified inputs according to the fuzzy
rules to establish rule strength
4) Finding the consequence of the rule by combining the
rule strength and the output membership function
5) Defuzzifying the output
6) Defuzzifying method is the weighted average of all
rule outputs, computed from Equation (6).
Fig. 13. Membership function of output command
TABLE I
THE FU ZZY RU LE O F OUTP UT COMMAND
Output Distance Error
NL NM NS Z PS PM PL
Relative
Velocity
NM NVL NL NM NM NS NS Z
NS NL NM NS NS NS Z Z
Z NM NS Z NS Z PS PS
PS NM Z Z Z Z PM PL
PM NS Z Z PS PM PL PVL
PL NS Z Z PS PL PVL PVL
Final Output =PN
i=1 wiZi
PN
i=1 wi
(6)
Where
Final Output = Average of all outputs
wi= Weight of membership function of Zi
Zi= output i
The control surface of this fuzzy algorithm is shown in
Fig.14.
IV. EXP ER IM EN T RES ULT
Two types of experiments are conducted. In the first
experiment, the velocity control mode experiment is set
up to evaluate the transient and steady state responses of
the system. Step input is provided to evaluate the transient
response parameters; slope, time constant, and settling time.
Steady state response is tested at various speed set points to
determine steady state error and so as to design an appro-
priate compensator. In the second experiment, the distance
Fig. 14. Fuzzy Control Surface
1797

control mode is set up. The fuzzy parameters; fuzzy rules
and range of membership functions are adjusted according to
the responses from the experiment so as to make the vehicle
follow the obstacle vehicle in front.
A. Velocity control experiment: Transient response test
For velocity control experiment, transient response is
tested by using step input at 20, 30 and 40 km/hr. The pro-
portional and derivative gains are 2.2 and 0.02 respectively.
Velocity set point and vehicle velocity data are recorded and
then plotted. Fig. 15 shows the results of 30 km/hr step
response.
B. Velocity control experiment: Steady state response test
The main objective of steady state response experiment is
to determine steady state error in order to design the com-
mand compensator to minimize this error. The experiment is
tested at various speed set points. The speed set points and
vehicle speed data are recorded. Fig. 16 shows steady state
response test at the speed varied in between 35-60 km/hr.
Without integral term, there exists steady state error. This
problem is solved by using command compensation method.
From the experiment result, the steady state error depends on
the vehicle velocity. The compensation is determined from
the ratio of velocity set point with the real vehicle velocity
as shown in Equation (7).
Compensator =Set Point
Velocity (7)
After adding the command compensator, the velocity
control experiment is tested again. The results show that the
compensator can decrease the steady state error efficiently
as shown in Fig. 17.
C. Distance Control Experiment
The distance control experiment is tested by using a
passenger car as an obstacle. The experiment is set up on
the straight part of the Bangkok Racing Circuit, Bangkok.
Firstly, the experiments are conducted for tuning of fuzzy
parameters. The target of the fuzzy tuning is that the
controller can control the vehicle like a human control and
Fig. 15. Transient response (30 km/hr step input)
also can maintain the distance between the vehicles. After
tuning the fuzzy parameters, the experiment is conducted to
evaluate the controller performance. The distance between
two cars is set at 15 meters and the velocity set points are
set at range between 0 to 45 km/hr. The data such as distance,
relative velocity, velocity set point and host vehicle velocity
are recorded. The experiment results are shown in Fig. 18.
The distance between the two vehicles at the beginning is
around 18 m. and the relative velocity is 0 m/s. The output
command is braking command that causes the host vehicle to
stop. The obstacle vehicle is then accelerating at time 20 sec.
At this point, the output of fuzzy rule is speed command that
causes the host vehicle to move forward. At time 32 sec, the
obstacle vehicle starts decelerate that causes the host vehicle
decelerates and stops at time 40 sec. The distance between
two vehicles is around 15 m. After time 40 sec, the obstacle
vehicle starts accelerating again and the host vehicle also
speeds up.
V. CONCLUSIONS
In this research, the adaptive cruise controller is designed
and developed on an AIT intelligent vehicle. The mechanical
throttle valve control is replaced by the electronic throttle
control, drive-by-wire system. The drive-by-wire system
uses a dc servo motor to control the throttle valve position.
The control algorithm of the throttle valve is proportional
and derivative control. The braking system of the vehicle is
modified by adding the Cool Muscle dc servo motor. The
velocity controller on ARM7 microcontroller is implemented
with proportional and derivative control algorithm. The
distance controller on a PC platform uses fuzzy algorithm.
The inputs of the fuzzy are the distance and relative velocity
which come from SICK LMS291 laser range finder. The
outputs from the controller are separated into 2 groups.
The first output is the command to accelerate the vehicle.
The other output is the command to decelerate the vehicle.
When the output is the braking command, the Cool Muscle
motor is actuated and the speed command is cleared. In
contrary, when the output is the speed command, the braking
Fig. 16. Velocity control result at 35-60 km/hr without command
compensator
1798

Fig. 17. Velocity control result at 0-50 km/hr with command compensator
Fig. 18. Distance control experiment at 45 km/hr velocity set point and
15 m distance
command is cleared and the speed set point is sent to the
ARM7 microcontroller.
The ACC system which is developed for the AIT in-
telligent vehicle is able to control the vehicle to run at
desired velocity when operated in velocity control mode and
efficiently maintain the distance between the host vehicle and
the obstacle vehicle.
ACKNOWLEDGMENT
This research project is financially supported by National
Electronics and Computer Technology Center
REFERENCES
[1] Bageshwar, V. L. Garrard, W. L. and Rajamani, R. ”Model Predictive
Control of Transitional Maneuvers for Adaptive Cruise Control Vehi-
cles” IEEE Trans.Tech. vol 53., Sep 2004, pp.1573-1585
[2] Bishop, R.H. ”Intelligent Vehicle Technology and Trends” Artech
House Publishers 2005.
[3] Bosch,R. ”Bosch Acc Adaptive Cruise Control”, Robert GmbH. 2003.
[4] Daniel, C & Johan, S. ”Adaptive Cruise Control for Heavy Vehicles
Hybrid Control and MPC”, (Master research Linkping University, De-
partment of Electrical Engineering, 2003) Sweden Linkping University.
[5] Kilian, C. T. ”Modern Control Technology: Component and System
(2nded)”. Delmar Thomson Learning 2000.
[6] W. D. Jones. ”Keeping cars from crashing” IEEE Spectrum. Sep 2001,
40-45
1799