JarrahaMAShaoutbA-FuzzyModularAutonomousIntelligentCruiseControl(AICC)System.pdf

121

Journal of Intelligent & Fuzzy Systems 11 (2001) 121–134

IOS Press

Fuzzy modular autonomous intelligent cruise

control (AICC) system

M.A. Jarrah a , ∗ and A. Shaout b

a School of Engineering American University of Sharjah P.O. Box 26666, Sharjah, United Arab Emirates

b University of Michigan-Dearborn, The Electrical and Computer Engineering Department, Dearborn, MI 48128,

USA

Abstract . An autonomous intelligent cruise control system was designed and simulated based on measured relative distance,

speed, and acceleration. These constitute the fuzzy inputs. The results have shown that the system can be robust to noisy distance

measurements and modular in structure to ease its implementation. A simple way of estimating the road condition was developed,

implemented and tested in simulation. The results show that based on the measured relationship between the deceleration and

brake pressure, the brake pressure output gain can be adjusted to prevent lock up of the tires and the consequent loss of stability.

Two controller implementations were tested. The ﬁrst one is based on heuristic knowledge of the system using Mamdani inference

system. The second model was based on ofﬂine adaptive neuro-fuzzy controller model.

1. Introduction

erations such as cruise, ABS, air-conditioning, and

steering. Winkelman [19] patented a self-tuning speed

control system for a vehicle with throttle control us-

ing speed measurements and integrating the automatic

transmission system into his model. Kato [7] presented

self-tuning pole-assignment based cruise control sys-

tem using least-square method. Fuzzy cruise control

applications were being implemented by the Japanese

and European automotive companies [11]. Hosaka [3]

presented a fuzzy-logic based cruise control system us-

ing two inputs: velocity difference and acceleration.

Takahashi [18] patented a fuzzy control system for au-

tomotive vehicle by controlling vehicle driving behav-

ior by harmonization of driving condition with bet-

ter speed tracking, less overshoot and undershoot, and

quicker convergence than conventional PID controllers.

Kayano and Itah [9] have reached similar conclusions.

Murai [13] added means of detecting uphill climb, de-

termining current gear position, and inﬂuence of the

automatic transmission to down-shift and up-shift, and

means of determining the difference between the max-

imum engine output after a down-shift. Kawano [8]

patented a cruise control device for motor vehicles that

increase driver comfort by minimizing throttle angle

swing.

Investigation of autonomous and intelligent cruise

control system is important in modern automobiles.

The cruise-control is becoming a standard feature of

modern automobile models. New features are being

added continuously. Some of the advanced cruise con-

trol systems are generically termed as intelligent cruise

control (ICC). The primary goal of ICC is to maintain a

safe distance between vehicles on the road. Mercedes

Benz uses a distance controller as a synonym for its

intelligent cruise control system, which stresses the im-

portance of maintaining the safe distance goal. The sec-

ondary goal is maintaining the cruising speed. Main-

taining a safe distance will improve the driver’s safety

by giving him time to react, brake, or change lane. It

will also improve the driver’s comfort by reducing hard

braking, transmission shift jerks, and tailgating if the

vehicle behind is also using intelligent cruise control.

The focus of the present investigation is to propose new

ICC based on neuro-fuzzy methods.

Fuzzy logic as well as conventional control has been

used for controlling various aspects of automobile op-

Muller and Nocker [12] presented intelligent

∗ Corresponding author. E-mail: mjarrah@aus.ac.ae.

1064-1246/01/$8.00

(

)

122

M.A. Jarrah and A. Shaout / Fuzzy modular autonomous intelligent cruise control

aicc

system

cruise control with the goal of maintaining a safe dis-

tance between the vehicle and the vehicle in front if

one exists. Brien et al. [14] addressed the modeling

and control of lateral motion of a highway vehicle to

track the center of the present lane, on both curved

and straight highway sections. Raza and Ioannon [5]

studied vehicle following control design for automated

highway systems. Their on-board vehicle control sys-

tem’s objective was to be able to accept and process

inputs from the driver, the infrastructure, and other ve-

hicles, perform diagnostics and provide the appropriate

commands to actuators so that the resulting motion is

safe.

Many authors have addressed the platoon problem.

Godbole and Lygeros [4] presented longitudinal con-

trol laws for vehicles moving in an intelligent high-

way system environment. Their controller maintains a

safe spacing while maneuvering in the platoon. Sheik-

holeslam and Dosoer [17] proposed longitudinal con-

trol of a platoon with no communication. Autonomous

intelligent cruise control was studied by Ioannon and

Chein [5] and compared with human driver models.

Their computer simulation of an AICC vehicle follow-

ing a single lane, without passing demonstrated the ef-

fectiveness of the AICC system. Fujioka and Suzuki [2]

addressed the control of both longitudinal and lateral

platoon using sliding mode controller design for their

autonomous intelligent vehicles. Yang and Tongue [21]

developed platoon control laws during vehicle entry

and exit using information from the lead, preceding,

and following vehicles in addition to the vehicle itself.

Kim et al. [10] presented fuzzy throttle and brake

control for platoons of smart cars using an adaptive

fuzzy system. Their simulation results show that a car

equipped with throttle and brake controllers will be

able to maintain a constant gap while driving downhill.

Neural network system was used to learn the fuzzy

rules for different vehicle types.

Most recent study by Yi et al. [23], a throttle/brake

control law for the intelligent cruise control (ICC) sys-

tems has been proposed in this paper. The ICC system

consists of a vehicle detection sensor, a controller and

throttle/brake actuators. The results indicate the pro-

posed ICC throttle/brake control law can provide satis-

factory vehicle-to-vehicle distance and velocity control

performance.

Autonomous Intelligent Cruise Control (AICC) have

been under investigation for at least 20 years, and Mer-

cedes Benz have begun development into AICC since

1980. Three major obstacles, cost, distance sensor

problems, and the ability to interpret the distance mea-

surements have prevented AICC systems from becom-

ing standard equipment in today’s automobiles. The

customer perceived cost versus utility for an AICC sys-

tem is much lower than that of the actual variable cost,

i.e., the true cost to build an AICC system is still much

higher than what the customers expect to pay. The de-

celeration control is fairly straightforward and easy to

implement, but distance sensing and correct interpreta-

tion of the noisy distance measurements are very difﬁ-

cult problems to overcome. State of the art cruise con-

trol system are characterized by added features such as

the distance sensor and deceleration control [12]. Dis-

tance measurements are made using distance sensors.

The challenges associated with these technologies can

be summarized as follows:

Distance Sensor: Distance measurements between

vehicles are obtained using laser or radar based dis-

tance. Regardless of the senor used, the same type of

strategy is used in detecting the distance between two

vehicles. The distance sensor usually scans through a

predetermined angle range, 6–12 degrees, per fraction

of a second, and the distance can be determined from

the results of the scan. Ideally, a sweep instead of a

scan should be performed, but this is impractical due

to the huge amount of calculations involved with a true

sweep. Each type of distance sensor has its own advan-

tages and disadvantages. Laser has the advantage of

being cheaper, fewer parts, less complex, and smaller to

package. Some of the disadvantages are narrow beam

width, and difﬁculty to determine exact distance if the

beam strikes a glassy or transparent surface such as re-

ﬂector and tail light, which can lead to inaccurate dis-

tance measurements. The narrow beam width means

that more samples must be taken each time to cover the

same angle range. Leica, NEC, and Mercedes Benz

have demonstrated that laser can be effectively used as

a mean for distance sensing. Doppler radar can also be

used. Its advantages are broader beam width, and not

as sensitive to change in material as the laser. The beam

width depends on the radar frequency used, which usu-

ally is in the microwave spectrum; thus the beam width

is in the cm range. The larger beam width implies

that fewer samples are needed to cover the same angle

range. The disadvantages are higher cost, more parts,

higher complexity, and larger package size. Thomp-

son/CFS, Phillips, Lucas, Hella, and TRW have em-

ployed some type of Doppler radar in their intelligent

cruise control.

Distance Sensing Problem: It is very easy to mea-

sure the safe distance between moving objects, with the

present laser or radar technology if the moving objects

(

)

M.A. Jarrah and A. Shaout / Fuzzy modular autonomous intelligent cruise control

aicc

system

123

1/s

distance

Radar Signal

Front Object

distance

Relative distance

velocity

ABS FLC

Vehicle Dynamics

Acceleration

ABS Fuzzy Model

Radar sensor

(Front Object Speed)

Velocity

Command

Safe Distance

Throttle

Cruise FLC

Cruise Fuzzy Model

Fig. 1. AICC vehicle block diagram.

remain in the same straight path. But this becomes a

very difﬁcult problem when interaction between vehi-

cles, road curvature, drivers’ intention, trafﬁc, weather

conditions, and driving situations are involved. Besides

the problems discussed in the ‘distance sensor’ section

above, other problems associated with normal driving

increase distance noise. For example, what happens if

the vehicle in front goes around a sharp bend, changes

lane to enter or exit a highway, come to a complete stop,

etc.? What happen if there is oncoming trafﬁc, another

vehicle veer into the lane, etc.? The problem is that how

to measure the distance between the AICC equipped

vehicle and the vehicle moving at approximately the

same cruising speed in the same lane. One way to min-

imize the noisy distance measure is to narrow the angle

scan, but then sharp bends, oncoming trafﬁc, trees or

telephone poles can also fool the distance sensor. An

optimal angle scan range can be determined through

testing and experiment, but a foolproof distance sen-

sor cannot be developed because these will be always

unexpected factors or noise in a driving environment.

One solution to most of these problems is to have the

driver provide information to the controller. The driver

can tell the controller how many objects are around.

He can instruct the controller in the case of a crowded

scene. In the worst case, the driver can use his fuzzy

logic to estimate the distance.

Difﬁculty in Interpreting the Noisy Distance Mea-

sure: Given that the distance sensor will always be

noisy, the question is: can the electronic module handle

the noise properly and gracefully, i.e. can the electronic

module be smart enough to remove noise from the real

data?

with PID tuning and fuzzy logic control based on PID

classical concepts prove to be inadequate [12]. They

cannot handle the noisy distance data and the result is a

very jerky ride. A ﬂexible and more intelligent type of

control, one that can mimic driver behavior, is needed

to handle all the driving situations and conditions.

Deceleration Control: ABS braking, transmission

down shifting, or a combination of both is used as the

mean to slow down the vehicle in order to maintain the

safe distance between the two vehicles. As can be seen

above, there are numerous intelligent cruise control

systems under development for the future automobile.

Below is a discussion of the proposed intelligent

cruise control system, which is based upon distance,

speed, and acceleration measurements. The outputs

of the present system is the brake pressure and the

throttle demands. It is assumed that the distance is free

of noise for the ﬁrst part of the simulation. Distance

measurements can be compensated for the noise by

several means as follows:

1. Use of a low pass ﬁlter with low corner frequency,

which limits the performance of the system when

the rate of change of events is high. This will

result in good performance when the driver view

is slowly varying.

2. Sensor noise can be reduced by having a sen-

sor array, using signal processing techniques, and

differential structure to eliminate noise while en-

hancing the true signals. This technique proved

to be useful in treatment of noisy measurements.

The second case is assumed to be possible and driver

monitoring will be required to interfere if any problems

Experiment with classical linear PID control

(

)

124

M.A. Jarrah and A. Shaout / Fuzzy modular autonomous intelligent cruise control

aicc

system

stop

veryslow

slow

Medium

Fast

Veryfast

0.5

-10

Speed

near

far

veryfar

0.5

100

150

position

none

light

medium

heavy

veryheavy

0.5

-2

break

Fig. 2. Mamdani Fuzzy inputs (relative distance and relative speed) membership functions, and output brake pressure or throttle (driving force)

membership functions.

Table 1

Fuzzy rules for AICC System

Distance

F res = F b + f r W cos( θ s )

(2)

+ R a ± W sin( θ s )+ R t

Where f r is the rolling resistance coefﬁcient, W

is the vehicle weight, θ is the angle of the slope of

the road with respect to the horizontal axis, R a is the

aerodynamic drag, R t is the transmission resistance,

g is the gravitational acceleration, and F thrust is the

thrust force developed by the engine.

The braking force developed on the tire-road inter-

face [20] is given by:

M M L

M H

arise. The second method will insure better ride qual-

ity. A noise of 1 Hz frequency and 50 m amplitude is

what one can expect in extreme cases having a relative

speed of 350 km/hr. Higher frequency noise will be

associated with much lower distance amplitude.

T b − Σ Iα an

F b =

(3)

Where T b is the applied brake torque, I is tire mo-

ment of inertia, α an is the angular deceleration, and r

is rolling radius of the tire.

The acceleration of the vehicle during braking is

given by:

2. Vehicle model

W F res

This acceleration should be limited to prevent tire

lock-up. Tire lock-up is a major safety factor in any

automatic braking system. It will cause directional

stability loss when the lock-up occurs ﬁrst in the rear

tires. Tire lock-up for a given vehicle will depend on

The vehicle longitudinal acceleration can be esti-

mated by the following equation [20]:

g =

a =

F res + F thrust

(1)

The resultant retarding force can be expressed as

follows:

(

)

M.A. Jarrah and A. Shaout / Fuzzy modular autonomous intelligent cruise control

aicc

system

125

1.5

Negative large

Negative small

Positive large

Positive small

Zero

0.5

-0.015

-0.01

-0.005

0.005

0.01

Distance Error 1000

1.5

Neg_small

Zero

Neg_large

Pos_large

Pos_small

0.5

-0.06

-0.04

-0.02

0.02

0.04

0.06

Relative speed 1000

m/sec

1.5

Deceleration

Acceleration

0.5

-0.5

-0.4

-0.3

-0.2

-0.1

0.1

0.2

0.3

0.4

0.5

m/sec 2

Realtive Acceleration 10

Fig. 3a. Sugeno fuzzy inputs (relative distance, relative speed, and relative acceleration) membership functions.

Negative small

Neg_small

(

Distance

error)

(

Relative

speed

)

Zero

(

Distance

error)

(

Relative

speed

)

Relative speed

Distance Error

Driving

force

Fuzzy membership

functions

Inputs

w 1

w 1 *z 1

Negative small

Distance

Error

Zero

w i *z i

Driving

force

Relative

speed

w 2 *z 2

Neg_small

w i

w 2

Zero

Fig. 3b. Sugeno fuzzy Inference system structure using two inputs (relative distance and relative speed).

r = µl 1 /L +(1 − K bf ) f r

the condition of the road and on the load of the vehicle.

The design of the controller should enforce that the

deceleration should be less than that corresponding to

either front or rear-tire lock-up. This is given by the

following simple relation for the front tire [20]:

;

(4)

1 − K bf + µh/L

Where µ is the friction coefﬁcient, l 1 and l 2 are the

distances from front and rear tire centers to the center

of gravity of the vehicle respectively, L is the distance

between the tire centers, h is the height of the vehicle

center of mass, K bf is the proportion of the total braking

force on the front axle, and the subscript f stands for

µl 2 /L + K bf f r

K bf − µh/L

f =

;

Plik z chomika:

Inne pliki z tego folderu:

Inne foldery tego chomika: