Chiung-Yao Fang Hsiu-Lin Hsueh Sei-Wang Chen National Taiwan Normal University Department of Computer Science and Information Engineering Dangerous Driving.

Chiung-Yao FangHsiu-Lin HsuehSei-Wang Chen

National Taiwan Normal UniversityDepartment of Computer Science and Information Engineering

Dangerous Driving Event Analysis Systemby a Cascaded Fuzzy Reasoning Petri Net

Outline

IntroductionSystem flowchartCascaded fuzzy reasoning Petri netDangerous driving event analysis

systemExperimental resultsConclusions and future work

Introduction

Driver assistance systemPassive methodsActive methods

Active driver assistance systemDetection componentAnalysis component

Block Diagram of an Active DAS

Analysiscomponent

Detectioncomponent

Sensor 1

Sensor 2

Sensor n

Detection system 1

Detection system 2

Detection system n

Dangerous driving event analysis system

Warning output

…

…

Flowchart of the Analysis System

Start

Data acquisition

Driving event integration

Degree of danger computation

Degree of danger output

Behavior analysis of our vehicle

Interaction analysis between nearby vehicles

and our vehicle

CFRPN

Cascaded Fuzzy Reasoning Petri Net (CFRPN)

operation) production-(maxfunction inference :

vectorparameter :C

then ,1 If or.input vect :

output final theis then , If tor.output vec :

,C ,

stage,th in the FRPNth For the

f

XUjU

YMjY

UfY

ji

ji

ij

ij

i

ji

ji

ji

ji

ji

The Production Rules of FRPN

Rule 1:

where : confidence vector of rule associated with the only corresponding transition

Rule 2:

where : confidence vector

of the rule related to the

corresponding m transitions

)( jj CR

)()(...)()( 112211 mmmm PPPP

)( jj CR

)()(...)()( 112211 mmmm PPPP T

21 ],...,,[ mj cccC

][cC j

cmm },...,,min{ 211

},...,,max{ 22111 mmm ccc

)( 22 P

)( 11 mmP

)( 11 P

)( mmP

c

)( 11 P

)( 22 P

)( mmP

2c

mc

1c

)( 11 mmP

Fuzzy Reasoning Algorithm

over. is reasoning theotherwise,

3; Step return to ,1 then ,or If 4) Step

from Compute (3)

from Compute (2)

1 if ,1

1 if

Compute (1)

3) Step

.0Let 2) Step

. and , , , , :inputs initial FRPN Read 1) Step

11

1

1

1

1

,

00

kkγγθθ

γHIOγγ

γγ

CO

θθ

, rpHx

, rpIxx

ρ

k

CHOI

kkkk

kTkk

kk

kkk

kk

rp jiii

jiii

ik

k

ji

Some Terminology

Lateral distanceForward distance

Our vehicle

Right-front

vehicle

Forward distance

Lateral distance

Using FRPNs for Reasoning

An example Get the membership

values of Lateral distance Position Speed

Input to the fuzzy reasoning Petri net Speed change reasoning Lane change reasoning Integration

Our vehicle

Preceding vehicle

Following vehicle

Left-rear

vehicle

Right vehicle

Left-front

vehicle

Right-front

vehicle

Left vehicle

Right-rear

vehicle

Membership Functions

Lateral distance

Position

Speed

Lateral distance (m)

Position (m)

Speed (km/hr)

Example of Lane Change Reasoning

Vo : our vehicle

Vlf : left-front vehicle

(1) : Vo moves without lane change.

(2) : Vo changes to the left lane.

(3) : Vo changes to the right lane.

(4) : Vlf moves without changing lane.

(5) : Vlf changes to the left lane.

(6) : Vlf changes to the right lane.

(7) : Vlf and Vo are moving in the same lane.

(8) : Vlf and Vo are moving in different lanes.

)( 11 p

)( 22 p)( 33 p

)( 44 p

)( 55 p)( 66 p

)( 77 p

)( 88 p

Example of Lane Change Reasoning

Two production rules:

: occurrence possibility that “Vlf and Vo are moving in the sa

me lane.”

: occurrence possibility that “Vlf and Vo are moving in differe

nt lanes.”

7

)())()(())()(( 7744226611 ppppp

)())()((

))()(())()(())()((

))()(())()(())()((

886633

553344336622

552255114411

ppp

pppppp

pppppp

8

Corresponding FRPN

: left-front vehicle and

our vehicle are moving in

the same lane

: left-front vehicle

and our vehicle are moving

in different lanes

)( 77 p

)( 88 p

Example of Speed Change Reasoning

Vo : our vehicle

Vlf : left-front vehicle

(1) : forward distance between Vlf and Vo is close

(2) : speed of Vo is slower than that of Vlf

(3) : speeds of Vlf and Vo are equal

(4) : speed of Vo is faster than that of Vlf

(5) : forward distance between Vlf and Vo increases

(6) : forward distance between Vlf and Vo remains the

same

(7) : forward distance between Vlf and Vo decreases

)'(' 77 p

)'(' 66 p

)'(' 55 p

)'(' 44 p

)'(' 33 p

)'(' 22 p

)'(' 11 p

Example of Speed Change Reasoning

Three production rules:

To decide the occurrence possibilities of the following driving ev

ents:

Forward distance between Vlf and Vo increases

Forward distance between Vlf and Vo decreases

Forward distance between Vlf and Vo remains unchanged

)'(')'(')'(' 552211 ppp

)'(')'(')'(' 663311 ppp

)'(')'(')'(' 774411 ppp

Corresponding FRPN

: forward distance between

Vlf and Vo increases

: forward distance between

Vlf and Vo remains the same

: forward distance between

Vlf and Vo decreases

)'(' 55 p

)'(' 66 p

)'(' 77 p

Integration

Integration rule:

: degree of danger of the interaction be

tween Vlf and Vo

)"("))'(')(())'(')((

))'(')(())'(')((

))'(')(())'(')((

77886688

55887777

66775577

ppppp

pppp

pppp

"

Part of CFRPNRoad change

reasoning

Speed change reasoning

Integration

Freeway Driving Event SimulationObjective – provide experimental

data

Two major stages:Simulation of freeway environments Simulation of vehicle behaviors

Simulation of Freeway Environments

Given: Total length of the freeway Number of toll stations, tunnels, freeway entries

and exits

Produce: Total lengths of tunnels Locations of toll stations, tunnel entries and

exits, freeway entries and exits Locations of road signs

Simulation of Vehicle BehaviorsGiven :

Vehicle positions and moving directions (of our vehicle and nearby vehicles)

Produce : Vehicle speed Lateral distance Directional signal Braking signal

Our vehicle

Preceding vehicle

Following vehicle

Left-rear

vehicle

Right vehicle

Left-front

vehicle

Right-front

vehicle

Left vehicle

Right-rear

vehicle

Experimental Results

Conditions:Our vehicle moves without lane change.The left-front vehicle changes its lane to the

right.The speed of our vehicle is faster than that of

the left-front one.

Experimental Results

0.724

0.953

0.971

0.002

0.000

0.131

0.731

0.731

0.022

0.131

0.015

0.802

0.227

0.047

0.002

0.000

Left-front vehicle and our vehicle are moving in the same lane

Distance between left-front vehicle and our vehicle decreases

Our vehicle moves without changing lane

Our vehicle is faster than the left- front

one

Left-front vehicle

changes lane to the right

Dangerous

Degrees of Danger of Interactions Between Left-fr

ont Vehicle Vlf and Our Vehicle Vo

No. Speed of Vo– speed of Vlf (Km

/hr)

Forward distance between Vo and V

lf (meter)/(second)

Horizontal shift of Vo (meters)

Horizontal shift of Vlf (meters)

Degree of danger

1 -15 50/2.25 0 2 0.073

2 -15 50/2.25 -2 0 0.073

3 0 50/2.25 0 2 0.181

4 0 50/2.25 -2 0 0.181

5 15 50/2.25 -2 0 0.266

6 15 30/1.35 0 0 0.439

7 15 30/1.35 0 2 0.724

8 15 10/0.45 0 2 0.922

9 15 10/0.45 0 0 0.572

Degrees of Danger of Interactions Between Prece

ding Vehicle Va and Our Vehicle Vo No Speed of Vo–

speed of Va

(Km/hr)

Distance

between Vo and Va

(meter)/(second)

Horizontal

shift of Vo

(meters)

Horizontalshift of Va

(meters)

Degree ofdanger

1 -15 50/2.25 0 2 0.037

2 -15 50/2.25 -2 0 0.037

3 0 50/2.25 0 2 0.181

4 0 50/2.25 -2 0 0.181

5 15 50/2.25 -2 0 0.261

6 15 30/1.35 0 0 0.724

7 15 30/1.35 0 2 0.709

8 15 10/0.45 0 2 0.903

9 15 10/0.45 0 0 0.943

Conclusions and Future Work Dangerous driving event analysis system

Reasoning by a cascaded FRPN module Detection subsystem integration Warning drivers Future work: integrate into the driver assistance

system

Freeway driving event simulation system Provide experimental data Support other driver assistance subsystems Future work: include more road conditions