Current Accident Analysis and AEB Evaluation Method for

Final AsPeCSS Workshop
Current Accident Analysis and
AEB Evaluation Method for
Pedestrians in Japan
July 1st, 2014
National Traffic Safety and Environment Laboratory
Kenichi Ando
Outline
Pedestrian accident in Japan
Regulation and JNCAP (passive)
Pedestrian accident analysis
Evaluation method of AEB
2
18.000
1.400.000
16.000
1.200.000
Injured
781,494
14.000
Fatalities
12.000
Vehicle Fleet
74,434,936
1.000.000
800.000
10.000
8.000
Accidents
629,021
600.000
6.000
Fatalities
400.000
4.000
4,373
Government Targets
Targets
of fatality reduction
2.000
200.000
0
Accidents and Injured
Traffic Accident in Japan
0
1990
1995
2000
2005
2010
2015
2020
Target of 2015 ⇒3000
Target of 2018 ⇒2500
3
Traffic Accident in Japan
■Comparison
Vehicle occupants
Motorcyclists
Motorized bicycle
Bicyclists
Pedestrians
Other
of Fatalities in 2002 and 2013
28.8%
2,416
41.2%
3,463
11.9%
997
8.6% 9.3%
726 777
2002
N = 8,396
36.2%
1,584
32.4%
1,415
10.6%
13.7% 6.7% 465
600 295
2013
N = 4,373
2013 National Police Agency
4
Injured Body Regions in Pedestrian
Head,Face
Arm
Leg
Neck
Abdomen
other
8.0%
20.7%
20.2%
Thorax,Back
Hip
Pelvis
Whole body
33.5%
11.6%
N = 61,001
Injured
5.1%
1.0%
3.7%
56.3%
16.8%
10.1%
7.4%
Fatal
N = 1,584
4.2%
0%
20%
40%
60%
80%
1.5%
100%
2013 ITARDA
5
Age in Pedestrian
<6
7 - 15
5.5%
16 - 39
24.1%
40 - 64
29.1%
65 - 74
13.6%
75 <
15.7%
Injured
N = 64,128
11.9%
7.8%
21.7%
20.3%
47.6%
N = 1,634
Fatal
< 1.5%
0%
20%
40%
60%
80%
100%
2012 ITARDA
6
History of Regulation and JNCAP
Regulation 2003
2004
2005
2006 –
2010
2011
2012
2013
2014
2013
2014
Head
Leg
JNCAP
EEC 2003/102
Base
2003
2004
2005
2006 –
2010
UN-R127 Base
2011
2012
Head
Leg
EEC 2003/102
Base
UN-R127 Base
7
Trend of J-NCAP Pedestrian Protection Test Results
Head Protection Performance
Leg Protection Performance
Level 1 Level 2 Level 3 Level 4 Level 5
20
18
16
14
12
10
8
6
4
2
0
1
1
5
6
1
4
6
6
3
2
13
10
1
6
11
7
8
7
1
3
7
3
7
9
8
4
4
3
1
1
1
1
1
1
2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013
3
4
4
1
2
Introduction
of regulation
Number of 台数
tested vehicles
Number of tested
台数 vehicles
Level 1 Level 2 Level 3 Level 4 Level 5
20
18
16
14
12
10
8
6
4
2
0
12
12
12
1
1
2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013
Introduction
of regulation
8
Main Body Regions Injured among Fatal Pedestrians
1999 and 2009 in Real-world Accidents
< Year 1999 >
n=940
n=165
n=68
n=149
< Year 2009 >
n=246
80
63
60
68
64
68
60
40
20
n=459
n=139
n=42
n=279
n=170
56
43
62
51
58
100
8
7
12
7
12
8
6
4
22
21
16
15
3
21
Mini
van
80
60
Light
Light
Box
van passenger cargo
van
car
Head
Chest
19
40
20
Hip
19
13
10
14
0
Sedan
Distribution (%)
Distribution (%)
100
19
27
Sedan
Mini
van
16
12
12
Others
12
14
9
14
15
16
0
Light
Light
Box
van passenger cargo
van
car
9
Technical Countermeasure for Pedestrian Protection
Head Protection Test
Fatal Accident in Japan
To reduce the fatality, a reduction of
impact velocity is a key.
Pedestrian
1,584
36%
4,373
Bicycle
Motor
600
cycle
11% Moped 14%
7%
Year 2013
Detection
AEB for Pedestrian Detection is effective.
(1) Near miss incidents
Car
1,415
32%
Auto Alert & Braking
(2) The effect of impact velocity reduction
(3) The current performance of AEB
10
Analysis of Near-miss Incidents Data
11
Near-miss Incident Data from J-SAE
• Data: Driving recorders are installed in taxis
2005 to 2009: 38,000 incidents as Near-miss data
105 taxis in Tokyo
20 taxis in Shizuoka
Recorder installed
When a taxi driver brakes with high
deceleration, the information are
recorded for 15 seconds.
Sudden brake
10 seconds
camera
5 seconds
Possibly recorded for 15 seconds
• Forward motion picture
• Car traveling velocity
• Brake signal
• Acceleration
• NTSEL: 163 car-to-pedestrian near-miss incident data were
analyzed in the present study.
12
Near-miss Incident (walking)
13
Near-miss Incident (running)
14
Macro Data from ITARDA
Macro Data: All recorded accident data reported to police in Japan
Injury level
Fatal, Serious, Minor
Most severely
injured part
Overall, Head, Face, Neck, Chest,
Abdomen, Waist, Arm, Leg
Cause
e.g., Steering, Glass, Ejection from a car
Weather
Sunny, Cloudy, Rain, Foggy, Snow
Category
Vehicle-pedestrian, Vehicle-vehicle,
Vehicle by itself
Road circumstance
Straight road, Intersection
Vehicle travel vel.
Halt, up to 10km/h, up to 20km/h
15
Similarity of Car Traveling Velocity
between Accident and Near-miss Incident
Real world accident: 2008
Fatal
Serious
Minor
100
90
80
70
60
50
40
30
20
10
0
(n=10,257)
)
(n=1,721)
)
(n=163)
)
Minor
Serious
Nearmiss
Fatal
010
11
-2
0
21
-3
0
31
-4
0
41
-5
0
51
-6
0
61
-7
0
71
-8
0
81
-9
0
91
-1
00
Cumulative Percentage (%)
(n=61,010)
)
Near-miss incident
Near-miss
Car traveling velocity (km/h)
Matsui et al. ESV 2011
The car travel velocity of near-miss incident was similar to
that of serious accident.
16
Moving Directions Between Vehicle and Pedestrian
(1) On a straight road
(2) At an intersection
B
C
C
B
D
D
D
D
A
A
C
C
B
A
B
B
Cross in front of the moving car
Go to the same direction as the moving car
17
Similarity of Moving Directions
between Accident and Near-miss
100
Ratio (%)
80
D
C
18%
21%
16%
5%
60
B
29%
0
38%
39%
Fatal
accident
Near-miss
incident
Daytime
30%
A B
68% Cross in
front of the
moving car
78%
74%
67%
A
31%
32%
35%
40
20
19%
C D
Go to the same
direction as the
moving car
46%
38%
Fatal
accident
Near-miss
incident
Nighttime
Hatching:
intersection
Matsui et al. ESV 2011
About 70% pedestrians were crossing roads in front of the forward
moving cars.
• Similarities are observed between accidents and near-miss incidents.
• We could estimate accident situations from near-miss incident data.
18
Classification of Near-miss
Incidents
19
Definition of Time in Vehicle TTC and Pedestrian TTV
Focused object
Vehicle
Pedestrian
Time
Vehicle time to collision
(Vehicle TTC)
Pedestrian time to vehicle
(Pedestrian TTV)
v
L
Ld
Definition
V
Vehicle TTC =
Reference
L
V
Matsui et al. ESV 2011
Pedestrian TTV =
Ld
v
Matsui et al. Traffic Injury
Prevention 2013
20
Vehicle TTC from Near-miss Incident Data
101 near-miss incident data: Pedestrians were crossing roads.
A drive recorder can capture forward motion pictures.
Vehicle TTC =
L
V
L
At a moment when a
pedestrian appeared
initially in front of a
car in the video frame
V
Assumption
The worst situation was assumed that a car was
moving toward a pedestrian without braking due to a
car driver’s carelessness.
21
Classified Four Pedestrians Stepping in Patterns to
Find Out a Severe Condition using Vehicle TTC
(1)
Unobstructed
view
(2)
From behind
a building
(4)
(3)
From behind
From behind
a parked vehicle a moving vehicle
Car
installing
drive
recorder
22
Results of Vehicle TTC in the Four Stepping
in Patterns
2.0
1.5
30.3km/h
2.0sec
16.2m
1.0
24.3km/h
1.4sec
8.6m
The vehicle TTC
(1)
(n=55)
(2)
(n=8)
35
1.2sec
11.1m
15
10
25.8km/h
1.3sec
8.8m
0.5
0
32.9km/h
(3)
(n=28)
(4)
(n=10)
30
25
20
5
0
Avg Velocity of a car (km/h)
2.5
Average car traveling velocity
Average forward distance of a car & a pedestrian
Avg forward distance of a car
& a pedestrian (m)
Avg Vehicle TTC (sec)
Average Vehicle TTC
the longest in (1) from unobstructed view
the shortest in (4) from behind a moving vehicle
23
Definition of Time in Vehicle TTC and Pedestrian TTV
Focused object
Vehicle
Pedestrian
Time
Vehicle time to collision
(Vehicle TTC)
Pedestrian time to vehicle
(Pedestrian TTV)
v
L
Ld
Definition
V
Vehicle TTC =
Reference
L
V
Matsui et al. ESV 2011
Pedestrian TTV =
Ld
v
Matsui et al. Traffic Injury
Prevention 2013
24
Calculation of Pedestrian TTV (Time to Vehicle)
v
Ld
Pedestrian TTV =
v
Ld
At a moment when a
pedestrian appeared
initially in front of a car
in the video frame
Assumption
The worst situation was assumed that a pedestrian
was moving toward a forward moving car line.
25
Results of Pedestrian TTV in the Four Stepping
in Patterns
2.6m/s
2.0
1.5
1.8sec
1.0
1.8m/s
2.0m
0.8sec
0.5
(1)
(n=55)
Unobs tructed view
The pedestrian TTV
(2)
(n=8)
From behind a
2.2m
2.1m/s
1.1sec
(3)
(n=28)
1.8m
1.8m/s
2.5
2.0
1.5
1.1sec 1.0
(4)
(n=10)
0.5
0
Avg walking speed (m/s)
3.5
3.0
3.2m
2.5
0
Average lateral distance of a car & a pedestrian
Average walking speed
Avg lateral distance (m)
Avg pedestrian TTV (sec)
Average pedestrian TTV
From behind a parkedFrom behind a moving
the longest in (1) from unobstructed view
the shortest in (2) from behind a building
26
Comparison between Vehicle TTC and Pedestrian TTV
in 4 Classified Stepping in Patterns
Average vehicle TTC
Average pedestrian TTV
Avg TTC, TTV (sec)
2.5
2.0
2.0s
1.8s
1.4s
1.5
1.0
1.3s
1.1s
1.2s1.1s
0.8s
0.5
0
*Matsui et al. ESV 2013
(1)
(n=55)
(2)
(n=8)
(3)
(n=28)
(4)
(n=10)
Each of vehicle TTC and pedestrian TTV was similar
in the 4 classified patterns.
27
Pedestrian Injuries and
Collision Speeds
28
Flow for Relations of Fatality Risks and
Impact Velocities
Impact
velocity
micro
data
Impact
velocity
Travel
velocity
MACRO
DATA
Fatality
risk
Fatality
risk
29
Combined Results between Travel Velocity and
Impact Velocity for Sedan - Micro Data
Minor injury
Fatal
Serious injury
Impact velocity
(km/h)
(km/h)
100
100
y = 0.8929x
0.9484xx
x
2
80 y =y 2=0.95
80 y = 0.89
R = 0.7255
R = 0.8189
60
60
40
40
20
20
n = 25
n = 48
0
0
0 20 40 60 80 100
0 20 40 60 80 100
Travel velocity (km/h)
Travel velocity (km/h)
Impact velocity
Impact velocity
(km/h)
100
y = 0.7496x
2 x
80 y = 0.75
R = 0.6312
60
40
20
n = 50
0
0 20 40 60 80 100
Travel velocity (km/h)
Linear Regression Coefficient
Injury level
Mean
SD
Number
P-value
Minor
Serious
Serious
Fatal
0.75
0.03
50
0.89
0.03
48
0.89
0.03
48
0.95
0.03
25
0.004*
0.180
Matsui, Oikawa and Ando STAPP Jnl 2013
30
Vehicle Types
Category
Vehicle types
Ordinary
automobile
(a) Sedan
Light
automobile
Engine
displacement
(d) Light passenger
≤ 660 cc
car
(b) Mini van
(c) Box van
(e) Light
cargo van
16%
(a)
(d) Light
Sedan
1,089
passenger
41%
(e) Light cargo van
car
26%
(b) Mini
van
(c) Box van 4%
13%
Pedestrian fatal accidents in 2009 in Japan
31
Fatality Risk as Functions of Travel Velocity/ Impact Velocities
0 10 20 30 40 50 60 70 80
Vehicle velocity (km/h)
(b) Mini van
0 10 20 30 40 50 60 70 80
Vehicle velocity (km/h)
(d) Light passenger car
Fatality risk (%)
100
80
60
40
20
0
100
80
60
40
20
0
Impact velocity
(estimated)
0.95
Fatality risk (%)
(a) Sedan
Fatality risk (%)
100
80
60
40
20
0
Fatality risk (%)
Fatality risk (%)
Travel velocity
(macro data)
0 10 20 30 40 50 60 70 80
Vehicle velocity (km/h)
100
80
60
40
20
0
100
80
60
40
20
0
(c) Box van
0 10 20 30 40 50 60 70 80
Vehicle velocity (km/h)
(e) Light cargo van
0 10 20 30 40 50 60 70 80
Vehicle velocity (km/h)
In case of fatal, driver did not brake enough so speed did not
reduce drastically.
32
Fatality Risk as Functions of Impact Velocities
Vehicle
impact
velocity
(km/h)
Fatality (%)
Sedan
Mini van
Box van
Light
Light
passenger
cargo van
car
10
0
0
1
0
1
20
0
1
4
1
2
30
3
3
5
2
5
40
9
10
11
9
12
50
22
30
31
22
25
60
38
42
38
For the five types of vehicles,
≤ 30 km/h: The fatality risks are less than or equal to 5%.
We would like to propose the specification of CDMBS: ≤ 30 km/h
Matsui, Oikawa and Ando STAPP Jnl 2013
33
Performance of AEB for
Pedestrian Protection
34
Experiments for AEB Performance
Frame material: FRP( Fiber Reinforced Plastics)
Avoidance
case
Sensor
Collision
case
35
Test Conditions
Items
Test vehicle
Detection sensor
Conditions
A, B, C
(A) (C) laser, camera and radar
(B) stereo camera
Test speed
5km/h 〜60 km/h (interval 5km/h)
Environment
Surface
Dummy position
day, night
dry, wet
vehicle center , offsets
Dummy orientation
front, side
Dummy color
black, white, gray
Standard condition
36
Scattered Results
70
Collision speed (km/h)
60
50
40
Unstable
30
20
10
0
0
10
20
30
40
50
60
70
Vehicle speed (km/h)
37
Binomial Test Results
Collision
Vehicle
speed
AEB
Avoid Collision
exp(β 0 + β1 x)
pθ (c = 1 | x) =
1 + exp(β 0 + β1 x)
θ = (β 0 , β1 )
c = 1 : collision
c = 0 : avoid
38
AEB Performance
by Binomial Logistic Regression
Probability of collision
1
0.8
0.6
0.4
Vehicle A
0.2
Vehicle B
0
0
10
20
30
40
50
60
70
Vehicle speed (km/h)
39
Pedestrian/vehicle fatalities (2006-2010)
2.500
fatality
Pedestrian fatalities
2.000
1.500
1.000
500
0
0-10
10-20 20-30 30-40 40-50 50-60 60-70 70-80 80-90
90-
Hazard recognition speed(km/h)
40
Reduction of Fatality by AEB
Pedestrian fatalities
2.500
2.000
Reduction by Vehicle A
AEB
Fatalities
1.500
1.000
500
0
0-10 10-20 20-30 30-40 40-50 50-60 60-70 70-80 80-90
90-
Vehicle speed (km/h)
41
Three Groups of Test Results
Severe Collision
Vehicle
speed
AEB
Minor Collision
Threshold
Avoid Collision
42
Total Risk of Severe Injury of Pedestrian
Total Severe Injury Risk: P (head and chest)
P (head or chest) = P (head)+P (chest)- P(head)*P (chest)
Collision velocity
20 km/h
30 km/h
40 km/h
50 km/h
(1) Medium Sedan
(2) Minicar
(3) SUV
2%
2%
2%
5%
4%
4%
16%
5%
8%
35%
9%
95%
43
Risk Distribution (Vehicle B)
Avoidance
Minor collision
Severe collision
1
0.9
0.8
Probability
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
10
20
30
40
50
Vehicle speed (km/h)
60
44
AEB Effects on the Distribution of
Pedestrian Fatalities (Vehicle B)
2,500
Avoidance
Minor collision
Severe collision
Number of pedestrian
2,000
1,500
1,000
500
0
0-10
10-20
20-30
30-40
40-50
50-60
60-70
70-80
80-90
90-
Vehicle speed (km/h)
45
46
Thank you
47
Fidelity Test : Radar Reflection
• Human
• Dummy
48
Deterministic vs. Probabilistic
Caution!
In some objects or environment, the pre-crash brake
cannot recognize objects or does not work.
Drive carefully, without relying on the pre-crash brake
too much.
49