Study on Forward Collision Warning System

Adapted to Driver Characteristics and Road Environment

International Conference on Control , Automation and Systems 2008

Oct. 14-17, 2008 in COEX,
Seoul, Korea

Masumi Nakaoka, Pongsathorn Raksincharoensak and Masao Nagai

Department of Mechanical Systems Engineering,

Tokyo University of Agriculture and Technology

Contents

1. Motivation and objectives

2. Warning distance algorithm

3. Experimental study by driving simulator

4. Analysis on driver behavior at decelerating situation

5. Driving experiment in urban area

6. Conclusion

1
 Motivation
Rear--end collision :
Rear Lane change Others
1.5% 23.6%
Rear-end
Rear-
It covers 30% of all accidents in Japan Frontal 31.4%
collision
2.5%
Single vehicle
crash Encounter
5.2% 27.0%
Between
pedestrian
8.8% Authority：ITARDA

Operate wrongly
Cause of accident : 9%

Driver cognition miss covers 70 % Judge wrongly Frontal

20% inattention
28%

Not confirming
safe condition
43%

Authority：ITARDA
3

Motivation
Forward collision warning system
brake
on

warning

V Vp Conventional collision risk index

R
R • TTC (Time to collision) V − V
p

R
• THW (Time headway)
V
host vehicle preceding vehicle

The design of forward collision warning system which fits

the driver behavior and environment is required
4

2
 Objectives
To propose a forward collision warning algorithm which is adaptable to
the driver characteristics and the road condition
Main focuses Verify the effectiveness of algorithm by driving simulator

Conduct the parameter study in urban road driving

by using the experimental vehicle
Framework of individual adaptation of driver assistance system design
Adaptation to driver behavior and
Normal driving road condition
characteristics Activation
Collision risk signal
Collision risk
index
assessment
calculation

Current driver behavior characteristics

Warning
sensors
device HMI
Large-scale individual
driver/vehicle/
environment
database Warning
Sensing sound
Storage
5

Warning distance algorithm

Preceding vehicle starts braking
V
Rw Vp, ap

Vp2

Vτr V2 2ap Rstop

2a
V, a

Host vehicle starts braking 2

Vp V2
V ：Host vehicle speed [m/s]
● R0 + = V ⋅τ r + + Rstop
2a p 2a
Vp ：Preceding vehicle speed [m/s]
a ：Host vehicle deceleration [m/s2] ● a=ap=µg
(Full Braking with Maximum deceleration)
ap ：Preceding vehicle deceleration [m/s2]
Rw ：Relative distance [m] Warning distance : Rw [m]
Rstop: Relative distance V2
V2
when both vehicles stop [m] Rw = V τ r + − p + Rstop
2µ g 2µ g
τr: Braking reaction time of driver [s]
6

3
Experimental study using driving simulator

Driving simulator

• Experimental condition
Vehicle speed : 60 km/h
6.0 m/s2 if dry pavement, µ = 0.8)
Preceding vehicle’s deceleration :
3.0 m/s2 if wet pavement, µ = 0.3)
• Following conditions
(1) Normal Always pay attention in forward looking
(2) Without warning Inattentive driving : Looking aside behavior
(3) Warning with constant TTC: TTC = 2.7s
(4) Warning with the proposed algorithm Rw : τr = 0.75s , Rstop = 2.0m
7

Experimental study using driving simulator

4
 Experimental results (Time history)
accelerator
1

P [-] brake
0
Pedal stroke Preceding vehicle started braking
-1
94 95 96 97
Time [s]
98 99 100 101
while the driver looked aside
0
a [m/s2] -2
-4
Acceleration -6
-8
94 95 96 97 98 99 100 101
Time [s] The warning was activated

V [km/h] 50

Velocity
0
94 95 96 97 98 99 100 101
Time [s]
40
The driver reacted warning
R [m] and started braking
20
Distance
0
94 95 96 97 98 99 100 101
Time [s]

Warning 1

Activating 0.5 Rear-end collision was avoidable

Signal 0
94 95 96 97 98 99 100 101
Time [s]

Comparison of collision velocity

Collision velocity : The velocity at rear-end collision accident occurs.

Rear--end collision
Rear
Acceleration

0
-2 occurs
[m/s ]
2

-4
-6
-8
83 84 85 86 87 88 89
Time [s]
Collision velocity
Velocity
[km/h]

50

0
83 84 85 86 87 88 89
Time [s]
Relative Distance

40
[m]

20

0
83 84 85 86 87 88 89
Time [s]

R=0m
Collision velocity is used to investigate the effectiveness of the system
on the impact energy reduction if the crash is unavoidable
10

5
 Comparison of collision velocity
Collision velocity : The velocity at rear-end collision accident occurs.

In dry condition (µ = 0.8) In wet condition (µ = 0.3)

70 45
Without warning 5 5 Without warning
S.D
S.D 40 Warning Ttc
60 Warning Ttc 5 5
5 Ave. Warning Rw
velocity [km/h]

velocity [km/h]
Ave. Warning Rw 35
5
Collision Velocity [km/h]

50
5 5 5

Collision Velocity [km/h]

30 Number of forward collisions
Number of Forward Collisions
Number of experiments
Number of Experiments
5 5
40 5 25
5
5 5
30 5 20 5
5 5
5 5 5
Collision

Collision
15
20
5 3 5
4 10

0 0 5 1
10
5 5 0
5 5 5
5
0 0
1 2 3 1 2 3
Subject A Subject B Subject C Subject A Subject B Subject C
Subject A Subject B Subject C Subject A Subject B Subject C

Time to collision The drivers could not avoid the crash, if TTC was not set larger.
• Subject A could avoid the crash completely.
Proposed algorithm Rw
• 2 drivers could reduce collision velocity.
11

Driving experiment in urban road

Driving simulator Real-world situation

• Severe deceleration (6.0m/s2) • Mild deceleration (about 2.0m/s2)

• Be able to change the • Experimental condition is different
experimental condition easily
• Collision experiment is impossible
• Good repeatability for conducting
experiments
• Collision experiment is possible

Using the same parameter may not fit the real-world driving situation

Determine the parameter of reaction time from normal driving data

12

6
 Driving data collection
• Continuous sensing drive recorder

①Camera/Image Synthesizer Captured VDO image

Front Laptop-based
Front left Logger
Front right 6-image
Face synthesizer
Foot NTSC/DV
conversion
Rear

②Global Positioning System

Latitude, Longitude
RS/USB
conversion

③ CAN Signal Probe

Steering wheel angle, Accelerator pedal,
Brake pedal, Winker,
USB/HUB
Speed, Accelerations, Yaw rate, …
Headway distance, Relative speed
RS/USB
CAN-probe conversion
13

Driving data collection

• Experimental route for driving data collection

K
I A
B
J
D

F E

Road section 1 (Legal speed 30km/h) G

Road section 2 (Legal speed 40km/h)
Road section 3 (Legal speed 60km/h)
Traffic signal

14

7
 Analysis on driver brake reaction time
• Definition of driver brake reaction time τr

Driving database collected Start braking

by drive recorder

Extract the situation

Driver recorder data
t0 t1 t2 t0 ：Preceding vehicle brake light is ON.
operation
Pedal

t1 ：Driver releases the gas pedal.

[km/h]

t2 ：Driver hits the brake pedal.

V

V2 Vp 2
Rw = V τ r + − + Rstop Driver reaction time τr [s] : τ r = t 2 − t0
2µ g 2µ g
15

Analysis on driver brake reaction time

• Probability density function of τr at Preceding vehicle brake light is ON.
3 10

9
2
Driver brake reaction time τr [s]

2.5
2 8

7
2
6
2

4
6

2
4

1.5 5
8

1.25
10

4
8
2

6
6

1
3

6 2
4

0.5
4

1
2

0 0
0 0.5 1 1.5 2 2.5 3

THW at Preceding vehicle brake light is ON [s]

Tentatively determined as 1.25 seconds to test the proposed warning index

16

8
 Experimental setup
• Warning equipment
Measurement part Calculating part

Experimental car DSP-embedded

PC Buzzer
Velocity
Actuating signal
Relative ＋ (1 or 0)
distance

Laser radar Calculate Vp and Rw in real-time

Operating condition Rw > R

Arbitrarily set values

V2 Vp 2
Rw = V τ r + − + Rstop
2µ g 2µ g
Tire-road friction coefficient : µ = 0.7
Relative distance when both vehicles stop Rstop = 2.0 m
Driver brake reaction time : τr = 1.25 s
17

Experimental result
40

V [km/h] 20

Velocity
0
The preceding vehicle
0 2 4 6 8 10
Time [s] started braking
40

R [m] 20
Distance
0
0 2 4 6 8 10
Time [s] The warning was activated
1
P [-]
0.5
Pedal stroke
0
0 2 4 6 8 10
Time [s]
The driver reacted warning
5
Ttc ,Thw [s] and started braking

0
0 2 4 6 8 10
System Time [s]

activating 1
signal 0.5
( 0 or 1 )
0
0 2 4 6 8 10
Time [s]

18

9
 Conclusions

1. A forward collision warning algorithm based on stopping distance

was proposed. Based on this algorithm, the collision warning timing is
adaptable to the driver brake reaction time and the tire-road friction
coefficient.

2. From the experimental study using driving simulator, the proposed

warning algorithm has effectiveness in collision mitigation, compared
with the constant TTC algorithm.

3. From the preliminary study using test vehicle driving in urban area,
the brake reaction time was analyzed from normal driving data. And,
the proposed warning algorithm was tested in real-world driving.

19

Perspectives of the study

Adaptation to driver behavior and
Normal driving road condition
characteristics Activation
Collision risk signal
Collision risk
index
assessment
calculation

Current driver behavior characteristics

Warning
sensors
device HMI
Large-scale individual
driver/vehicle/
environment
database Warning
Sensing sound
Storage

Collect more data and put into the driver behavior analysis
Determine the parameter of the proposed warning index
to improve the acceptance of the warning system

This research has been conducted as a part of the Core Research for Evolutional
Science and Technology (CREST) research programs entitled “Mobility Sensing for
Safety and Security”, funded by Japan Science and Technology Agency (JST). The
authors would like to deeply thank for their financial support for executing this research.
20

10
Thank you for your kind attention

21

22

11
 Evaluation of the proposed algorithm
by using driving simulator

23

Comparison of brake reaction time

Driver reaction time: Time interval between time instant when the preceding vehicle
start braking and time instant when the driver starts hitting on
the brake pedal

In dry condition In wet condition

4 4
Normal
Normal
3.5 S.D 3.5
S.D Without warning
Without warning Warning Ttc
Ave.
Ave.
3 Warning Ttc 3 Warning Rw
Warning Rw
Reaction time [s]

2.5 2.5
Reaction time [s]

2 2

1.5 1.5

1 1

0.5 0.5

0 0
1 2 3 1 2 3 C
Subject A Subject B Subject C Subject A Subject B Subject

24

12
 Experimental results ( in wet condition )

Pedal Stroke
P [-]

[-]
0
Pedal stroke Preceding vehicle start braking
-1
72 73 74 75 76 77 78 79 80 81 82 83 while the driver looks aside
Time [s]
Acceleration Preceding Vehicle
[m/s ] 0
a [m/s2]
2

-2
Acceleration Host Vehicle
-4
72 73 74 75 76 77 78 79 80 81 82 83
Time [s] The warning is activated
Velocity

V [km/h]
[km/h]

50 Preceding Vehicle
Velocity Host Vehicle
0
72 73 74 75 76 77 78 79 80 81 82 83
Time [s]
The driver reacts warning
Warning Activation Relative Distance

40
Relative Distance and starts braking
R [m]
[m]

20 Rw
Distance
0
72 73 74 75 76 77 78 79 80 81 82 83
Time [s]
Headway

4
1 on
Warning
[-]

0.52
Rear-end collision is avoided
[s]
Signal

Activation Signal
Time

off0
72
72 73
73 74
74 75
75 76
76 77
77 78
78 79
79 80
80 81
81 82
82 83
83
Time [s]
Time [s]
Time to Collision

25

Conventional collision risk index

26

13
 TTC & THW
Symbols：
R: Relative distance
V Vp Rɺ : Relative velocity
R
V: Host vehicle speed
Vp: Preceding vehicle speed

R Preceding vehicle assumes

TTC (Time to collision) [s] a fixed speed.
V − Vp

THW (Time headway) [s] R

Preceding vehicle assumes stopping.
V

27

Analysis on collected driving data

28

14
 Classification of decelerating situation
The decelerating situation

Divide into three situations

Start braking

B1

Low speed vehicle

B2

Stopping vehicle

B3

29

Analysis on driver reaction time

• Histogram of τr
0.1

0.08
Probability

0.06

0.04

0.02

0
0 1 2 3 4
1.25 τr [s]
The pattern of driver reaction time did not appear

30

15
 Classification of decelerating situation
in urban road
Approaching the vehicle
B1
which starts braking
V2 V2
Rw = Vτ r + − p + Rstop
2µ g 2µ g

B2 Approaching the low speed vehicle

Rw = V ⋅ THW * + Rstop

（THW *：Aim of time headway）

B3 Approaching the stopping vehicle

V2
Rw = V ⋅ tch +
2µ g
（tch：Brake reaction time）

31

Conceptional methods of friction

estimation

32

16
Interacting Multiple Model for Friction Estimation
1
0.9
Input Object Observation 0.8

Friction coefficient
0.7
Steering angle lateral acceleration,
0.6
yaw rate 0.5
0.4
0.3
0.2
0.1

Filters 0
0 5 10 15
Time [ s ]
Estimated value
Mode A Occurrence probability

Definitive
Mode A estimated
High friction road Assembly part of value
estimated value

Mode M Estimated value

Occurrence probability

Mode M
Low friction road
33

How to capture the friction coefficient?

Telematics
Telecommunication ＋ Informatics

Car to car
Car to infrastructure

34

17
 Acknowledges

35

• いい質問が出たとき
• That’s a very good question.
• I am glad you asked this question.
• I appreciate your question.

• 質問の英語が良く分からないとき
• Sorry, I didn’t quite get your question.
• I am sorry. I didn’t understand your question. Could you please repeat your
question?
• Mr. Chairman, could you help me, please?

• 質問に的確に答えたか確認する．
• Did I answer your question?
• Does this answer your question?

36

18
ある点を指摘され，困ったとき
１．まだそれについて考えたことがないとき
I haven’t considered that point yet. Thank you for your comments.
2. 今後の課題とさせていただきます．
We are going to study about that issue in near future.
3． (しつこい外人に突っ込まれた)やっぱ個別に議論したい
I will be happy to discuss the matter with you in private.
4． 最悪状態：答えられない．誰も助けてくれない．
I am sorry that I cannot answer that question.
I am sorry that I can’t give any details at this point.

37

19