• 沒有找到結果。

The human-in-the-loop design approach to the longitudinal automation system for the intelligent vehicle, TAIWAN iTS-1

N/A
N/A
Protected

Academic year: 2021

Share "The human-in-the-loop design approach to the longitudinal automation system for the intelligent vehicle, TAIWAN iTS-1"

Copied!
6
0
0

加載中.... (立即查看全文)

全文

(1)

2006 IEEEInternational Conferenceon Systems, Man,and Cybernetics

October 8-11, 2006, Taipei, Taiwan

The

Human-in-the-loop Design

Approach to the

Longitudinal

Automation System for

the

Intelligent Vehicle, TAIWAN

iTS-i

Hsin-Han

Chiang, Jau-Woei Perng, Bing-Fei Wu, Shinq-Jen Wu, and Tsu-TianLee

Abstract-This paper presents the integration design and implementation of alongitudinal automation system with the interaction ofhuman-in-the-loop (HITL). Theproposedsystem has a hierarchical structure composed and consists of an

adaptive sensory processor, a supervisory control and a

regulation control. The adaptive sensory processor routes the information from on-board sensorstoavoidmissing detection of thevehicle ahead. Based ontherecognized measurementfrom the adaptive sensory processor, the supervisory control determines the desiredvelocityfor thevehiclesoas tomaintain safetyandsmoothoperationindifferent modes. Theregulation controlutilizessoft-computing techniqueanddrives the throttle action to execute the desired velocity commanded from the supervisorycontrol. The feasible sensorydistance iswithin40 m, and the according driving velocity can achieve 100 km/h upward. Thechallenge in low velocity operation can also be handled by the regulation control against gear changes and torque converter. Among experimental tests under various kindsoftrafficflows, the systemvalidnessisexhibitedand also the preferable comfort is achieved through theexaminationof international standard ISO 2631.

I. INTRODUCTION

Vehicle automations are currently being introduced into the assistance to relieve human drivers from undesired routines of driving task. Sincemanystudies have shown that over 90 percent ofhighway accidents are occurred due to

driver-relatederrors,themain initiative is toimprove safety ofthe automation system interacted with thehuman driver. As to the vehicle longitudinal automation control, some famous practical works [1-3] introduce the features like adaptive cruise control (ACC). Themoreunderstandings of the automation of vehicle longitudinal control task are provided in [4]. Besides, to increase more capability being with human drivers, the conception ofhuman-in-the-loop (HITL) needstobemoreconsideredintothesystemdesign.

The safety consideration into the vehicle longitudinal automation is the safety headway distance adjustment to

avoidthecollision withthe vehicle ahead. Toaccommodate the automation system to different driving conditions, the

corresponding

operation

modes andthe automatic transition frame should be developed. There is no initial demand for roadenvironment,and alsofor theaction of thehumandriver

H. H.Chiang,J. W.Perng, B. F.Wu, and T. T. LeearewithNational Chiao-TungUniversity, Hsinchu, 300, Taiwan,R.0. C.(authorstoprovide

phone:886-3-5712121-54346;fax:886-3-5729749; e-mail:hsinhan.ece90g @dnctu.edu.tw;jwperng(cn.nctu.edu.tw;bwuQcssp.cn.nctu.edu.tw;ttlee@ cn.nctu.edu.tw)

S. J. Wuis with Da-YehUniversity, Changhwa, Taiwan,R.0. C.(e-mail:

jen(dmail.dyu.edu.tw)

I 1 OMNi VOOO , I;:Svw;

Fig. 1. Overall structure ofthehuman-in-the-loop longitudinal auto-mation system.

which can be viewed as a disturbance to the automation system. It is essential to consider the adequatestability and robustness into the controlling design. Moreover, drivers' comfort is also the most important initiative counted for automationdesign.

This paper proposes the longitudinal automation

system,

which is implemented on a real vehiclenamed TAIWAN iTS-1,with the interaction of the human driver. Thespecial modificationstotheautomation system are toensure safety and smooth operation with workload reduction for the human driver. Finally, experimental results in real road environments document the performance of the overall system integration extended to the capability of HITL design.

II. SYSTEM CONFIGURATION AND OBJECTIVES AsillustratedinFig. 1,theroad environmentmainlyrefers

to the longitudinal direction in front of the subject vehicle, andtheon-boardsensorsinclude laser radar andanglesensor

of thesteering wheel (SW). The adaptive sensoryprocessor deals with measurements fromradaraccordingtothe action of humandriver andthevelocityofthe vehicle. Basedonthe recognized data, the supervisory control makes thedecision of operation modes and desired velocity command to the

regulationcontrol for the execution ofvelocity tracking. The throttlepedalisdrivenbythelongitudinalautomation system

toachieve thecontrollingobjective. Althoughthe brakepedal

(2)

TABLE I

CRITICAL THRESHOLD OF COMFORTINISO 2631 STANDARD Nodiscomfort ( -0.315 m/sA2)

Alittle discomfort (0.315 0.63m/sA2)

Somewhat discomfort (0.5 -1m/sA2) Discomfort (0.8 1.6m/sA2)

Extremely discomfort (1.25 -2.5m/sA2) Verydiscomfort (2 m/sA2 )

and reserved for the human driver totake the awareness of emergency.

As far as the longitudinal automation convinced to the human driver isconcerned, the controllingoperation for the subject vehicle comprises thevelocity tracking cruise mode and automatic vehicle following mode. In the velocity tracking cruise

mode,

theobjective is to controlthe subject vehicle to track any desired velocity commanded from the human driver.Atthetimeof thedetectionfrom the

preceding

vehicle, theautomation system will automatically switchto

the automatic vehiclefollowingmode. The objective ofthis modeistomakesureofsafety headway distancemaintaining forthesubject vehicle without the need of vehicle-to-vehicle (v-v)communication. Inaddition,the automation system will automatically switch back to the

velocity tracking

cruise mode if there isnovehicle detectioninthe front.

In these both operation modes, the ride comfort of the human driver or passengers isconsideredintotheautomation design. As the comfort analysis concern, the international standard ISO 2631-1:1997 [5], which defines the means to

evaluate vibration with respect to human responses, is employed to examine the ride

quality

of the

proposed

longitudinal automation system. This standard specifies direction and location of measurements, equipment to be used, duration ofmeasurementsand

frequency

weighting,

as

well as methods of assessment of measurement and evaluation ofweightroot-mean-square(r-m-s) accelerationin meterspersecondsquared

(m/s2).

The index ofthis standard is determinedbythe

frequency-weighted

acceleration which

canbecalculatedby usingtheexpressionas

a=

(Wa.)

2 ,(1

where

a,

denotes

frequency-weighted

acceleration,

Wi denotes theweightingfactor of

interesting

axesof

body,

and ai denotes ther-m-s acceleration for the i-thone-thirdoctave

band.

The comfortanalysisisassessedby checking thata,, over sampledhorizon doesnotreachacritical thresholdvalue,as

depictedinTableI. Astothelongitudinal automationdesign, the motion which performs in the

longitudinal

axis is the interesttobeexaminedfor ride comfort ofthesubject vehicle. Thefollowing section describes the designingapproachfor the longitudinal automation system to be assured that the safetyand comfortrequirementscanbe both achieved.

III. LONGITUDINALAUTOMATIONSYSTEMDESIGN In this section, the three components of the proposed vehicle longitudinal automation system are introduced, respectively.

Specified radar dstance

(orheadwaydistance, d)

Look-ahead

offset,Yd

Look-ahead distance d.

Adaptvescanningangle 20. 'II,

l

Fig. 2. Illustration ofthevehiclefollowingscenario on curved roads.

A. Adaptive Sensory Processor

As illustratedin Fig. 2, thepreceding vehicle can not be detectedby the laserradar in front ofthe followingvehicle if the scanning beam is too narrow. This fact might bring the emergency due todrastic acceleration action of the system. Nevertheless, if the scanning beam is too broad, some unexpected objects which areneighborwith theroad will be detected. Therefore, it is necessary to design the adaptive sensory scheme such that the preceding vehicle can be correctly detectedevenifonthecurved roads.

Byconsidering theturning behaviorof the vehicle, a linear bike model of lateral dynamics from steering angle

gf

to lateralvelocity

v,

andyaw rate r is employedas

[V!]

+LvX

Bg

(2)

where

_

_f+

)

-aCr

+

bC,I

j C

A=

Ka

+'>Ct;

(a2Cf +b2C)

jr

B Lid

aand bdenotethe distances fromthefront andrearaxlesto thecenterofgravity (CG)of thevehicle, mdenotesthe total

massofthevehicle, Vdenotes the forward velocity,

A,

denotes the yaw moment ofinertia, and

Cf

and

Cr

denote the total corneringstiffnessofthe frontandreartires, respectively.

Inaddition,todetect theexistence oftheprecedingvehicle, the look-ahead information of the subject vehicle must be considered. The dynamics of the point at a look-ahead distance of themoving vehiclecanbedescribedas

d =

Ved

-v, -rd

£ad

V/Rf

-r

(3) (4) where Ydand

6d

denote the lateral offset from the centerline and theangle between tangent to the road and the vehicle at a look-ahead distanced,and

RJ

denotesthe road curvature.

Notethat here the roadcurvatureisrestricted to be constant for theassumptionofthesteady-state analysis. The reason is that the vehicle will notreach a steady-state condition while travelinga road with varying curvatures, and then the static relation is difficulty to be investigated. As shown in Fig. 2, the subject vehicle (or called following vehicle) and the preceding vehicle both travel a curve with a constant

(3)

TABLE II

(9f,

0)WITHVARYING VELOCITIES AND RADIUS OF CURVED ROADS

Rf 200m 300m 400m 500 m v 40 km/h (19.6,6.9) (14.5,5.1) (10.9,3.8) (8.7,3.1) 60km/h (23.9,9.2) (15.5,6.1) (11.6,4.6) (9.3,3.7) 80km/h (25.1,11.9) (16.8,7.9) (12.6,5.9) (10.1,4.7) 100km/h (27.6,16.6) (18.4,11.1) (13.8, 8.3) (11.1, 6.6) Unit: degree

curvature

l/Rf.

Byconsidering asteady-state motionthat the subject vehicle tracksthe curvedroadperfectly at a constant velocity, the variationsofthevehiclelateral dynamics (2) and look-ahead dynamics (3) and (4) can be set to zero, i.e.,

rv

=r d =

ed

=0. In the following, the subscript ofss

denotesthe value atsteady-state condition. Throughdirection calculation, the steady-state steering anglecanbe obtainedas

1 mV (aCf-bCr)

Rf (a+b)CfCr

Thesteady-state look-ahead lateral offset canbe adoptedas

Y,/s =

hss

-

Rf

(6)

where h - R2

+d2+2Rfd(-v,ss

V)

Besides, at steady-state the lateral dynamics (2) can be

presented

as

ALvs -Bdf (7)

r5 andthis holds ifandonlyif

v,ss

- T (8)

where T=

b+amV2/(a+b)C.

Notethatthefixed values oftheyawrate canbe obtained by

rs2=

V/Rf

(9)

whichisassumedinthesteady-stateturning.

By substituting (8) and (9) into (6), the look-ahead steady-state look-ahead lateraloffsetcanberewrittenas

,

Rf2

+d2

2+2dT

-R

(10)

From(5)and(10), oneobtains

Ydss R, R,2+d' +2dT fR (11)

sfs

(a+b mV2(aC

bC,

)

/(a

+b)Cf

r)

Itisreasonablytoassume

d2+2dT /Rj<<1

(12)

and thefollowing

approximation

via

Taylor's

expansion

can

beobtainedas

vRf2 +

d-

+2dT= R + &+2dT (13)

2R

Through (13), the relation (

11)

becomes

Ydss,

d 2dT (14)

3,ss

2(a+ b

V2P)

where P=

m(aC1

-

bC,

)/(a+

b)CI

C,.

To obtain the relation between the adaptive scanning angle and the steering angle, the followingapproximation is applied as

tanO=

y,

/Id (15)

where

d,

denotes the specified distance in the feasible range forlaser radar.

By substituting (14) into (15) with the small angle assumption which assumes tan

0.^

,one canobtain

0

d2+2dT

Ko

(16)

6fs

2d,

(a

+b-

V2P)

d,

The real input to the vehicle is the steering wheel (SW), and the steering angle can be substituted with the SW angle through a constant ratio [6], i.e., 3=i4, (normally the value of is isbetween 18and22for passenger vehicles). In(16),the resulting feature is that the adaptive ratio between the adaptivescanning angle and the steering angle is independent ofthe road curvature. It can be observed that the adaptive ratio updates with the vehicle velocity. During the vehicle turning in a higher velocity, the look-ahead lateral offset increases such thatthe scanning beam isturnedmoredegree withthe samedirectiontothe SW. As toless road curvatures, thelook-aheadlateral offset decreasessuch that the scanning angle is turned less degree. TableIIdepicts this result.

B. SupervisoryControl

There are two stages in the supervisory control. In the first stage, thedesired acceleration is determinedaccording to the selected operation mode and available feedback signals. In the second stage, the desired acceleration is converted to the desiredvelocitywhich ispassed into theregulationcontrol.

The velocitytracking cruise mode is to design the desired acceleration such that the subject vehicle can track proper velocities commanded by the driver and avoid feeling discomfort.Define the velocity tracking erroras

eV

-V Vds

(17)

and selecttheslidingsurfaceas

Scc

=e =-

Vd,5

(18)

ToforceScc

0,

the controllaw can be chosen as

SCc

=-KCCSCC

(I19)

where Kcc >0 is chosenby thedesigner.

Equation

(19)

satisfies the global asymptotically stable requirementand also

SCcS,

< 0. Notethat thechoiceof(19) usually includes the discontinuous sign function,e.g.,sgn(S), in typical literatures of sliding mode control (SMC). However, thechattering phenomenawill be causedthatalso bringssomeunforeseen noiseofhigh frequency. Instead,the continuous function is feasible to be

implemented

for its simplicity.

By differentiating

(18),

the desired acceleration will be easily solved with (19). To achieve therequirement ofride comfort,itshouldkeeptheacceleration commandbounded to

the desiredone

aMfilax

especiallywhile the initialvalue ofevis large. Therefore, theslidingsurface in

(18)

can be modified into

af

+ee

(4)

Rulebas

P.>

Fig. 3. Theblockdiagramoftheclosed-loop velocity regulation control.

wherea1is thecurrentacceleration of thefollowing

vehicle,

and

sa(

=x,

f as x <1I

s

sign(x),

as

.xl

>I

For the case of

laj

+ev

I<afmax,

the sliding surface (20) is identicaltothe one of(18). The desired acceleration canbe solvedas

afd

=Ve,

-K(C(V

-Vt)

(21)

While

laf

+eR

I>

ajmax,

the

sliding

surface becomes

Scc

=-at af

n-xX

(22)

and thedesired accelerationis

a

=d

a a (23)

fdes KC'C' fmax

Notethatthejerkin(23)canbe

neglected

forthe intention of constraintinthedesired

acceleration,

i.e.,

afdes=±afmax

The control law in (21) israther simple. Since only the velocity ofthe subject vehicle is

required

for the

velocity

trackingmode,

implementation

of this

design

is easyandstill associatedwith theconsiderationof ride comfort.

Continuingly, the

objective

of the automatic vehicle following modeforheadway distance

tracking

isto

design

the control law of the desired acceleration. To

begin

with the development of a

sliding

surface, the vehicle

following

dynamics in terms of

using

the relative distance R are

presentedas

R

Xp

-X

(24)

Rk

=Vp-V

(25)

whereXand Varethesubject vehiclepositionandvelocity, and

Xp

and VParetheprecedingvehicle

position

andvelocity. Byemployingthe

fixed

headwaytime strategy,thedesired followingdistance lawaccordingtothevelocityofthesubject vehiclecanbeobtainedby

R,ds

aV+L

(26)

Rdes -ca

(27)

wherea isregardedasthedesiredheadwaytime, andLcanbe viewedas aminimum safety distanceortypically avehicle's length.

The error between thedesired and relative headway distance isdefinedas

eR =R-

Res

(28)

andthesliding surfacecanbedesignedas

S;F=

--af

+a1masat('+e

R (29)

f max

Forthe case of

a1

+eR

K<

afmax,

this sliding surface(29) is obviously stable since SVF OaseR O.Bychoosing thesame

control law as(19), thedesiredaccelerationcanbederivedas aff1e

I-(K1,,,

SyP +

R)

(30)

Theresultinthecaseof

laf

+eR

I>

afmax

isthe similarto(23) and the controllaw

afdes=±

afmax

isused instead. In(30), only theheadway distanceand its changerate arerequired and the stability is also

guaranteed.

Notethat there isnoknowledge of the preceding vehicle existing in (30). Although the states related to the preceding vehicle can be obtained by v-v communication device, it can notbesupposed that all other vehiclesareequipped.

In the second stage, the conversion from the desired accelerationtothevelocityisdesignedas

VcGesa=

cges-k/

(V-Vles

)

(31)

where k, >0 isadampinggain.

As to the selection logic between these two modes, this autonomy scheme can be achieved by adopting the min-operation in the secondstage,namely,

V1

=min{k7V,Vv,4

(32)

Once the preceding vehicle is detected, the final desired velocity for the subject vehicle will be determined by the automatic vehiclefollowing control, i.e.,

VJ=vVF;

otherwise, thesubject vehicle will be backtothevelocity trackingcruise control, i.e.,

Vf

=Vcc.Thisapproachisintuitivebut also easy toimplementation.

C.

RegulationControl

The objective of the regulation control is to execute the desired velocity commanded from the supervisory control. The vehicle longitudinal dynamics canbe described by a set of system composed of various linear and nonlinear subsystems, e.g., engine, automatic transmission in the gear box, brake system, and the rubber tires with respect to roads,

etc. Indeed, it is verydifficult for control designing based on this complicated model. As to the ill-conditioned and complexmodel ofvehiclelongitudinal dynamics, it motivates theemployment of fuzzy logic control (FLC) in the regulation controldesign.

(5)

TABLE III

FuZZY RULE BASE OF THESFLC

D, NB NS ZO PS PB

u NBu NSu ZOu PSu PBu

INHW JVSI ZO PS PB

-1 -0.2 0 0.2 1

A`ru NSWu ZOo PSa PBRo

-1 -0).7 0.7 1

Fig. 4. Membership function of the fuzzy input(up) and the fuzzyoutput

(below).

The block diagram of the regulation control with the vehicle longitudinal dynamics is shown in Fig. 3. The regulation control scheme is composed of a proportion

-derivative (PD) controller and a FLC. There is a single

control input defined by the error ofthe commanded and

current velocity, i.e.,e=V -V, and the control output is the

applied voltage to the throttle motor actuator. The

characteristics ofthethrottlemotor actuator canbemodeled as one saturation function with a transport delay. H(s, V)

presentsthe dynamics from the derived throttle angleto the

vehiclevelocity.

This PD-type FLC with a single-input is convincingly

representative to the single-input FLC (SFLC) proposed in

[7]. For conventional FLC's, the fuzzy rule base is constructed in a two-dimension (2-D) space for using the error and error change phase-plane, i.e., (e, e?). It can be

inspected thatmost 2-D fuzzyrule bases have the so-called

skew-symmetricproperty.Onenewfuzzyinputcomposedof

theerroranderrorchangecanbepresentedas

Ds

=kde+kle

(33)

Therefore, the 2-D fuzzy rule base of the error and error

changephase-planecanbe reducedinto 1-D spaceofDsfor

SFLCaslisted in TableIII.Both therangesof thefuzzyinput

and output arethesamefrom-1 to 1,and thecorresponding

membershipfunctionsareplottedinFig. 4, respectively.

Inthedefuzzication operation, thecenter ofmass(COM)

method isappliedtocalculatethecontrol output

5 5

U = ,Z ,(Ds)xu / u,(Ds) (34)

whereArepresentstheweightingvalue of each rulei,andui

is thecrispvalueof each ruleconsequence.

There are many advantages for applying this PD-type

SFLC. Regardless of the controlled plant dynamics, it

requires onlyone fuzzy input anda 1-D space offuzzyrule

base. Therefore,thenumberoftuningparametersinFLC can

be greatly reduced. The computation load can also be alleviated for that the numberoffuzzyrules is considerably

decreased.

Fig. 5. Theexperimentalscenarioonthe real road environment.

IV. Experimental Results

The proposed longitudinal automation system is implemented on a commercial vehicle named TAIWAN

iTS-I,Savrian, manufactured by Mitsubishimotorcompany.

The feedback velocity of the vehicle is measured from the wheel-velocity instrument of the front left-tire. For the headway distance measurement, one laser radar (LMS291,

manufactured by SICK), is employed to the automation

systembythe connection through RS-232. The permissible

distanceofradar inforwarddirection issetto80m,however,

the feasible distance inpractice is 40m.Astothe angle ofthe scanning beam, the maximum value for radar is 50 degree in bothleft andright planes. Itisassumed that theturning angle for the scanning beam is adaptive to be according to the algorithm derivedin Section III. However,notethat thereis

no auto-turning schemeinthe function ofLMS291, thus in

this work, the acquired sensory distance is determined

accordingtothemeasuredonewiththecalculated expanding

angle of radar.Inaddition, the anglesensorisinstalledaround

the axis of the steering wheel, and through CAN bus tothe automation system. The whole automation is built in the real-time Microautobox (MABX), which is a compact

stand-aloneunit withrapidprototype ofcontroldesign. The

more detailed information of MABX is available at

http://www.dspace.com/ww/en/pub/home/products/hw/mica

utob.cfm. The throttlepedalisadjusted byaDC-motor with

thefeedbacksignal ofthethrottlepositionsensor,toyield the velocity trackingforvehiclecontrol.

To exhibit the validness of the proposed longitudinal automation system, manyexperimentshave bee taken inthe expressway (Chutung-Nanliao segment, Taiwan), in which the legal highest velocity is 90 km/h. The scenario of

experiments attherealroad is illustrated in Fig. 5. Initially

the subject vehicle is in the automatic vehicle following mode, and maintains the safety headwaydistanceaccording

tothecurrentvelocity(about75 km/h). Thesampled history

ofexperimentswiththe scheme ofadaptivesensoryprocessor

isdepictedinFig. 6. As shown in the secondgraph,here the fixedheadwaytime issetas 1 second thesame,suchthatthe

desiredheadwaydistance isabout21 m.There isnomissing

detection of the preceding vehicle during the vehicle

(6)

50 60 70 80 90

..I

0 10 20 30 40 00 60 70 00 90 90 ~80~ >700 10 20 30 40 50 60 70 80 90 1 commard 4S ...--measured 0 0 10 20 30 40 50 60 70 80 90 time(sec)

Fig. 6.Theexperimental scenarioonthe real road environment.

followingcontrol mode, and thethrottle action also

performs

smooth, asin the bottomgraph. Once thepreceding vehicle changes the original lane, the subject vehicle automatically switchestothevelocity cruise

tracking

mode and accelerates tothe

original

desired

velocity

90

km/h,

asshown in thethird graph.Itcanbeseenthat thesteady-statethrottlevoltage after the acceleration is larger values for the

higher

velocity than the slower velocity. If one vehicle cuts in the forward direction,orradar detects the slowervehicle ahead within the feasible range (40 m), then the throttle pedal of the

subject

vehicle will beadjustedtofollow the

preceding

vehiclewith thesafetydistanceaccording thecurrentvelocity. Besides the high velocityoperation, althoughnot

shown,

the

operation

of low velocity (20 km/h

upward)

can also be handled

by

the regulation control against gearchanges and torque converter of the vehicleengine.

During these experimental tests of the

longitudinal

automation system, the acceleration of the

subject

vehicleis recorded forcomfort analysis.Oneaccelerometer is locatedat the center of gravity of the

subject

vehicle for

only

the longitudinal motion is ofinterest.InTable IV, it is clearto see thatthecomfortindexa, withindifferent

sampling

intervals areallextremely satisfied with the constraint ofnodiscomfort in ISO2631 standard. Therecorded acceleration data of the subject vehiclearenotrestrictedtoonlyone operation mode but also include the transition process. Even if the human driveractsbraketoomuch such that the vehicleisdecelerated tomuchlowervelocity, yet theautomation systemperforms smooth accelerationtothedesired velocityunder therequest for comfort. Both theselections ofthesaturated acceleration

a1

max andheadwaytime acan beadjusted from thehuman

driveraccording topersonal driving behavior and favor. The braking action is reserved to the human driver to take the awareness of emergency, while the concentration on maintaining a safety headway distance is delivered to the longitudinal automation system. Therefore, inaddition tothe comfortrequirement,theworkload reductionfromthehuman driver can also be achieved.

TABLEIV

FREQUENCY-WEIGHTEDACCELERATION DURINGDIFFERENTSAMPLING INTERVALS

Sampling interval(sec.) a,(m/s2)

0o 30 0.0522

0-40 0.0688

0 50 0.0585

0 60 0.0755

V. CONCLUSIONANDFUTURE WORKS

Integration ofhuman-in-the-loop designintoalongitudinal

automation designis presentedinthis paper, andtheoverall

system is successfully implemented on apassenger vehicle

testedinreal road environments.Thelongitudinal automation

system is composed of the adaptive sensoryprocessor, the

supervisory control, and theregulation control. The system

safetyisimproved byinclusion ofadaptivesensoryschemeto prevent the missing detection of the preceding vehicle on

curved roads. The supervisory control isdesignedto switch between different modesautomaticallyandoperatewithinthe bound acceleration constraint without therequirementofv-v

communication.Basedonthereference velocitycommanded

from the supervisory control, the regulation control is to

executethevehiclevelocitytrackingthroughthethrottle. The proposed automation system istoassist the humandriver in

the velocityand inter-vehicle space control such asto yield

theworkload reduction ofdriving. Finally, the experimental resultsat real roads verify the validness of the longitudinal automation system. Furthermore, the ride comfort is also achievedthroughtheexaminationofthestandardISO 2631.

ACKNOWLEDGMENT

This work is supported by the program of Promoting

Academic Excellence of Universities under Grant. No. 91X104 EX-91-E-FA06-4-4. We also thank Tseng-Wei Chang and Tien-Yu Liao for the cooperation of hardware work.

REFERENCES

[1] H. Raza and P. Ioannou, "Vehicle following control design for

automatedhighwaysystems,"IEEETrans.Contr. Syst.,pp.43-60, Dec.

1996.

[2] R. Rajamani et al., "Design and experimental implementation of longitudinal control foraplatoon ofautomated vehicles," Trans. ASME

J. Dyn. Syst. Meas. Contr., vol. 122,pp.470476, Sep. 2000.

[3] J. E. Naranjo et al., "Adaptive fuzzy control for inter-vehicle gap

keeping," IEEE Trans. Intell. Transport. Syst., vol. 4, no. 3, pp.

132-142,Sep. 2003.

[4] A. Vahidi and A. Eskandarian, "Research advances in intelligent

collisionavoidance andadaptive cruise control," IEEE Trans. Intell. Transport.Syst., vol. 4,no.3,pp. 143-153, Sep. 2003.

[5] "ISO 2631/1: Mechanical Vibration and Shock-Evaluation ofHuman

ExposuretoWhole-body Vibration, Part 1. General Requirements,"

1997.

[6] S.J. Wuetal.,"Theautomatedlane-keeping design foranintelligent

vehicle," in Proc. Intell. Veh. Symp. 2005, Las Vegas, 2005, pp. 508-5 13.

[7] B.J. Choi, S. W. Kwak, and B. K. Kim, "Design stability analysisof

single-input fuzzy logic controller,"IEEETrans.Syst Man,and Cyber. B, vol. 30,no.2,pp.303-309,April 2000.

a 10lo ~0 10 I,.-f 10 20 30 40

I

HOl l , iool- ---i

數據

Fig. 1. Overall structure of the human-in-the-loop longitudinal auto- auto-mation system.
Fig. 2. Illustration of the vehicle following scenario on curved roads.
TABLE II
Fig. 3. The block diagram of the closed-loop velocity regulation control.
+3

參考文獻

相關文件

Salas, Hille, Etgen Calculus: One and Several Variables Copyright 2007 © John Wiley &amp; Sons, Inc.. All

1 As an aside, I don’t know if this is the best way of motivating the definition of the Fourier transform, but I don’t know a better way and most sources you’re likely to check

了⼀一個方案,用以尋找滿足 Calabi 方程的空 間,這些空間現在通稱為 Calabi-Yau 空間。.

Understanding and inferring information, ideas, feelings and opinions in a range of texts with some degree of complexity, using and integrating a small range of reading

Writing texts to convey information, ideas, personal experiences and opinions on familiar topics with elaboration. Writing texts to convey information, ideas, personal

• ‘ content teachers need to support support the learning of those parts of language knowledge that students are missing and that may be preventing them mastering the

 Promote project learning, mathematical modeling, and problem-based learning to strengthen the ability to integrate and apply knowledge and skills, and make. calculated

Now, nearly all of the current flows through wire S since it has a much lower resistance than the light bulb. The light bulb does not glow because the current flowing through it