Content uploaded by Bing Dai
Author content
All content in this area was uploaded by Bing Dai on Aug 12, 2016
Content may be subject to copyright.
Abstract
In China, nearly 25% of trafc fatalities are pedestrians. To avoid
those fatalities in the future, rapid development of countermeasures
within both passive and active safety is under way, one of which is
autonomous braking to avoid pedestrian crashes. The objective of this
work was to describe typical accident scenarios for pedestrian
accidents in China. In-depth accident analysis was conducted to
support development of test procedures for assessing Autonomous
Emergency Braking (AEB) systems. Beyond that, this study also
aims for estimating the mitigation of potential crash severity by AEB
systems.
The China In-depth Accident Study (CIDAS) database was searched
from 2011 to 2014 for pedestrian accidents. A total of 358 pedestrian
accidents were collected from the on-site in-depth investigation in the
rst phase of CIDAS project (2011-2014). The number of car/SUV/
VAN/microbus to pedestrian (all called hereafter referred to as
car-to-pedestrian accidents in this study) cases is n=265, which
accounts for 74% of all collected pedestrian accidents. Statistics on
all collected pedestrian crashes provided an overview understanding
of the pedestrian safety situation in China. To achieve the goal of the
study, 255 car-to-pedestrian cases were analyzed to gure out the
most frequently scenarios, which can be the reference to set the AEB
test procedure in China. Furthermore, 183car-to-pedestrian cases,
with detailed information regarding accident vehicles, pedestrians
and environment, were reconstructed using PC-Crash.
A hypothetical autonomous braking system would activate when the
pedestrian successfully detected by the sensing system and then new
impact speeds will be calculated. The study documents that the most
frequent situations in China are:(1) Unobscured, pedestrian walks out
from nearside; (2) Unobscured, pedestrian walks from far side; (3)
pedestrian walking along the road. More than 20% accidents could be
avoided with an AEB system (with an instant deceleration −8.0m/s2)
functioned at 1.0 second prior to crash. The mean of new impact
speed for un-avoided cases decreased to 24km/h from 39km/h in real
accidents.
Introduction
Requirements of legal and New Car Assessment Program (NCAP)
tests for passive pedestrian protection have been subsequently
introduced [1, 2]. Fredriksson et al. (2010) concluded that when
developing pedestrian head injury countermeasures focus should be
on the bonnet, lower windscreen, and A-pillars [3]. Pedestrian head
injury mitigating technologies, including both structural solutions
with extra space under the bonnet as well as deployable bonnets and
various airbags, were advanced [4, 5, 6, 7, 8]. Using real-life data,
several authors have pointed out that another key factor in injury
reduction is the reduction of impact speed [9-10]. Rosén and Sander
(2009) showed that reducing impact speed from 50km/h to 40km/h
would reduce pedestrian fatality risk by 50%, while a reduction from
50km/h to 30km/h would reduce the risk by as much as 80% [11].
With forward-looking sensor systems, the possibility to detect
pedestrians in advance opens up new possibilities for injury
mitigation, and has been predicted to provide considerable reductions
in pedestrian injuries. Combining autonomous braking with driver
warning or even autonomous emergency steering could further
increase effectiveness. Both driver warning and AEB systems are
already on the market. Test methods for such active safety systems
are being developed and will be implemented in NCAP tests in the
near future, e.g. 2016 in Euro-NCAP[12], and most probably in 2018
C-NCAP tests (still in consulting phase).The setting of test conditions
involves many considerations, one of which is the desirability of
subjecting the vehicles to realistic accident conditions, i.e.
circumstances that are encountered in real accidents, or at least to
understand clearly how proposed test conditions relate to the
circumstances of real accidents.
Typical Pedestrian Accident Scenarios in China and
Crash Severity Mitigation by Autonomous Emergency
Braking Systems
2015-01-1464
Published 04/14/2015
Qiang Chen, Miao Lin, Bing Dai, and Jiguang Chen
CATARC
CITATION: Chen, Q., Lin, M., Dai, B., and Chen, J., "Typical Pedestrian Accident Scenarios in China and Crash Severity Mitigation
by Autonomous Emergency Braking Systems," SAE Technical Paper 2015-01-1464, 2015, doi:10.4271/2015-01-1464.
Copyright © 2015 SAE International
Downloaded from SAE International by Bing Dai, Wednesday, July 01, 2015
Problems may appear while the trafc data-based products from
developed areas operate in complex Chinese road conditions.
Studying real-world accident data is a viable way to gain an increased
understanding of the pre-crash movements of vehicles and
pedestrians. In-depth accident analysis also provides an effective way
to evaluate products' effectiveness using real Chinese accident data.
Currently, the most detailed Chinese accident database CIDAS
includes vehicle travel and impact speeds, driver braking and steering
manoeuvres as well as detailed sketches of the accident scenes. By
combining above information it is possible to derive the pedestrian
location relative to the vehicle as a function of time during the
pre-crash phase. Such extended reconstructions can also serve to
establish the time to collision and pedestrian location at the moment
when he/she would have become detectable by a vehicle based
sensor. The aim of this paper is to describe typical accident scenarios
for pedestrian accidents based on empirical data for the reference of
AEB tests in C-NCAP. At the same time, with the purpose of
understanding of their inuence on potential system effectiveness.
Method
Case Sampling and Overview
In 2011, a China In-Depth Accident Study (CIDAS) project was
initiated by the China Automobile Technology and Research Center
(CATARC). This project aimed to collect 500∼600 on-site accidents
annually in 6 cities in China (Figure 1). In each investigated region,
approximately 100 trafc accidents were documented yearly with
1000-2000 individual pieces of data collected per accident and
entered into a database. From north to south, the 6 cities were
Changchun, Beijing, Weihai, Ningbo, Foshan and Chengdu (started in
2014).
Figure.1. CIDAS investigated cities(Five cities are shown. Cases in Chengdu
(a new started city in 2014 in southwest China) were not included in this
study.)
Specialist teams went directly to the scenes of the accidents in trafc
police vehicles to collect necessary information to complete detailed
accident reconstructions as well as medical data on how those
involved people were injured and treated. In this way, extensive
information for a wide range of research, such as ‘vehicle design for
passive and active safety’, ‘biomechanics’, ‘driver behavior’, ‘trauma
medicine’, ‘rescue services’, ‘road design’ and ‘road conditions’
could be collected. The sampling criterion for the rst phase of
CIDAS were dened as: (1) at least one 4-wheel vehicle involved in
the accidents, and (2) at least one person injured with AIS1+ injuries,
and (3) on-site information was not changed before investigators
arriving. However, in order to mitigate trafc jams, many accidents in
China (usually with slight injury, property loss and clear-cut accident
liability) were simply solved, and were thus too transient for
information collection for in-depth investigation.
From CIDAS database, 358 pedestrian accidents were searched from
July 2011 to June2014. Pedestrian accidents with (1) vehicle lost
control before collision; (2) vehicle rstly impacted with object; (3)
reverse driving were excluded. A subsample of n=337 pedestrians
was used for pre-selection. Sampling criteria for reconstruction cases
in current study were: (1) accident vehicles limited to cars, vans and
SUVs, excludingbusses and trucks; (2) principle forces during the
collision was to thecar body, thus,a pedestrian lying on the ground
prior to collision, pedestrian/e-bike sideswiped by the car and
absorbingonly a negligible impulse during the collision, were
excluded; (3) only persons taller than 150 cm or over 14. Finally
n=183car-to-pedestrian accidents were selected for reconstruction by
initially using PC-Crash to calculate impact conditions, such as
vehicle impact velocity, vehicle kinematic sequence and throwing
distance. For each accident, information on the accident spot, travel/
collision speeds and trajectories of car and pedestrian 5 seconds prior
to impact were provided by a special reconstruction database named
Pre-Crash Database (CIDAS-PCD).
In CIDAS database, injuries are coded according to the Abbreviated
Injury Scale (AIS2005). Six levels of injury severity are included,
where AIS1 denotes minor injury, 2 moderate, 3 serious, 4 severe, 5
critical, and 6 likely fatal injuries. The maximum AIS (MAIS)
indicates the severity of the worst injury if several were sustained by
the victim. In addition to AIS codes, injury severity in the CIDAS
database is also coded by ‘fatal’(dead within 7 days of the accident),
‘in-patient’ (more than 24 hours of medical treatment), ‘out-patient’
(less than 24 hours' medical treatment). Specialized medical
knowledge is necessary to identify the correct AIS code for each
injury type. To be precise, we used ‘fatal’, ‘in-patient’ and
‘outpatient’ to distinguish injury severity in this study.
In current study, lighting conditions are dened in 3 categories:
daytime, dawn and dark. According to astronomy information of the
accident city, accident investigators would identify daytime (between
sunrise and sunset), dawn (hours before sunrise) and dark (hours after
sunset) based on accident time. The duration before sunrise and after
sunset is decided by investigator according to factors such as weather
and season.
Downloaded from SAE International by Bing Dai, Wednesday, July 01, 2015
Case-by-Case Analysis
Case Information Extraction
In the current database, there is no variable to record some of the
studied items, e.g. accident scenarios for AEB test, it is necessary to
analyze each case to identify the situation if the pedestrian was
passing from far-side/near-side of the vehicle-travelling lane.
Accident causes recorded in the database up to now are roughly based
on trafc police categories, which is insufcient to understand the
real causes of the accidents. For example lots of accidents occurred
with ‘unsure safety’, which could include the situations like
‘misjudgments’, ‘unpreparedness for the accident’, ‘short distance to
other road users’ and ‘inappropriate speeds’ according to accident
description and interview.
A car-to-pedestrian accident occurring on a mix-used road in an
industrial area (dark without street light) is presented as an example
(Figure 2). A Lifan-passenger car was travelling northward when the
female pedestrian was walking in the middle of the street in the same
direction as the car moving. The driver was found impaired by
alcohol consumption. The throwing distance of the pedestrian was
about 23 meters away from the estimated collision location. The
wrap-around distance of the pedestrian's head measured 1.78meters.
According to on-site investigation, the car was stopped after 14meters
braking. Obvious damage was found on the car bumper, bonnet and
windscreen (Figure 3). The pedestrian died on-site. Car collision
speed was estimated at 55km/h with tolerance of ±5km/h.
Figure 2. Accident sketch (a) and on-site photo (b)
Figure 3. Car damage(The bumper was broken down which could relate to
pedestrian right tibiofibula fracture. The damage of bonnet could relate to the
bruise on thorax and upper limbs. Fracture of windscreen could relate to the
head injuries, which was the main fatal injury of the pedestrian by
medicolegal expertise.)
Accident Reconstruction
Reconstruction in the study included two parts: rstly collision-point
to nal-position with multi-body pedestrian model and pre-collision
phase with rigid pedestrian model. A scaled on-site sketch of the
accident scene is important for PC-Crash simulation. From the sketch
of accident we used: estimated initial impact location, rest positions
of vehicle and pedestrian, braking traces (if available) and some other
marks (such as scattering distribution, oil/water leaking trace).
Vehicle information contains the damages on accident cars, type,
model and car manufacture and so on (such as dimension, collision
mass, tyre information). Pedestrian information mainly includes the
height, weight, age, injury body regions and severity of the injuries.
Witness statements may include information about initial stance of
pedestrians at the moment of impact; however, not in all cases the
information from witnesses was available. Parametric studies
concerning the velocity of car and stance of the pedestrian, pitch
angle during the braking were performed in rene iterations to nd
the best correlations with all indications of in-depth on-site
investigations. The nal conguration that reproduced the same
impact points on the car, the same injuries and throwing distance to
the real accident was retained. Physical parameters including pre-/
post-impact velocities were calculated based on each time step for
both pre-crash and post-crash phases. The pre-crash phase was
calculated for at least 5 seconds.
Downloaded from SAE International by Bing Dai, Wednesday, July 01, 2015
Results
Overview Statistics
Environment
Of the358 cases, 291 (81%) occurred on dry road surface and 42
(12%) on wet surface, the remaining were on icy (6%) and snow
cover surface (1%)(Table 1).
Table 1. Road surface
As shown in Table 2, 54% (n=193) of the pedestrian cases occurred
in daytime, 33% in night with streetlight and 13% in night without
streetlight.
Table 2. Light condition
It is shown that more than 60% of pedestrian accidents occurred on
straight road followed by cross-section (13%) and T-junction (13%).
It is worth noting that 7% collisions happened on overhead roads
(3%) and expressways (4%), which could be special cases of China
(Table 3).
Table 3. Road section
According to on-site investigation, there was zebra crossing available
in range of 50meters in 63% of the accident location. Moreover, 18%
of pedestrian accidents occurred on zebra crossings(Table 4).
Table 4. Zebra crossing
Based on on-site investigation and driver statement, about 85%
accidents were without obstacles to drivers' view. In the remaining
cases, drivers were blocked by plants, moving/waiting vehicles,
buildings. However, according to reconstruction result, challenges
could arise to accidents with obstacles by moving/waiting vehicles to
detecting devices for crash avoidance technologies at 0.5∼1.0 second
before collision, which is regarded as the key time range for
triggering active safety products. Self-vehicle body, such as A-pillar,
was stated to block the driver view to nd the pedestrians in time in
12 cases (3%). In the evening, wrong use of high beam from the
oncoming vehicle was reported to affect the collision in 7(2%) cases
(Table 5).
Table 5. Driver view block
Table 6 shows the frequency distribution of trafc control. In more
than 73% of the cases, there was a trafc light near the collision
position and one party of the accident participants violated trafc
light sign. The second largest group is controlled by ‘right before left’
with a percentage of 13%.
Table 6. Traffic control
Accident Cause
According to accident process descriptions (police records) and
accident team investigation, six categories of accident causes listed in
Table 7 may contain 96% of those cases leading to collisions and the
remaining 4% were aggregated into ‘Others’ and ‘Unknown’. The six
categories can be describes as:(1) alcohol driving; (2) distraction
driving, such as telephone use/fatigue/thoughts etc…; (3) road
priority violation (including actions of car driver and pedestrian),
which can be classied by running yellow/red light, not giving way to
oncoming trafc with priority, stop sign violation;(4) unsure safety,
which includes misjudgments, unpreparedness for accidents (e.g.
pedestrian running behind an obstacle parking car, pedestrian
crossing street in front of bus), insufcient distance to other road
Downloaded from SAE International by Bing Dai, Wednesday, July 01, 2015
users; inappropriate speeds; (5) speed limit, which comprises over the
legal speed limit and excessive speed for safe control; (6) wrong
headlight use of approaching trafc in night.
Among the causes, the biggest category was ‘unsure safety’ (48%). In
this category, 29% drivers were unprepared for the accident, which
ranked the rst of all accident causes. The followed category was
road priority violation (26%), of which ‘not giving way to oncoming
trafc with priority’ accounted for 23% of all the accidents.
Table 7. Main accident causes
Pedestrian information
Figure 4 shows the age distribution. Pedestrians aged 40 to 50and 50
to 60 account for 22%(n=61) and 19%(n=61) respectively.
Figure 4. Pedestrian age group distribution (n=327) (n=31 unknown)
For n= 303 adult pedestrians, height ranging from 160 to 169 cm and
170-175cm comprises 80% totally (Figure.5).
Figure 5. Pedestrian height group distribution (n=303) (Only adult pedestrian
considered, n=28 unknown)
In the n= 255 car-to-pedestrian cases, n=269 pedestrians were
involved (in some cases more than one pedestrian was impacted by
one car). The number of pedestrian deaths was n= 94 (35%). Most of
pedestrians suffered inpatient injuries with number of n=135 (50%).
The remaining were with outpatient injuries (n=40, 15%).
For car-to-pedestrian accidents, n=120(47%) pedestrians can be
identied that head had obvious contact with windscreen. The contact
points were illustrated on a standard vehicle (Figure 6). For the
remaining cases, it was either without contact evidence or unsure if
the damage was caused by head or other body parts(125pedestrians
were without head contact and 10pedestrianswere indeterminable).
Figure 6. Distribution of head contact points on windscreen
Accident Scenarios
Based on the current sample, ve categories could cover all of the
cases: (1) pedestrian walks/runs from nearside; (2) pedestrian walks
out from behind obstruction;(3) pedestrian walks/runs from far side;
(4) pedestrian moves along street;(5) pedestrian walks out into the
path of the turning car. To be clear, Table 8 shows the illustration and
description of the ve categories. In this section, only car-to-
pedestrian accidents were considered.Table 9 shows the distribution
of car-to-pedestrian scenarios.
Downloaded from SAE International by Bing Dai, Wednesday, July 01, 2015
Table 8. Illustration and description of scenarios
Table 9. Distribution of car-to-pedestrian scenarios
Of the n=223 adult pedestrian accidents, n=68 were pedestrian walks/
runs from near side which account for more than 30%, of which more
than 60% (n=41) occurred in dark. Almost with the same frequency
with pedestrian from near side, n= 65 (29%) pedestrian walks/runs
from far side of which 66% (n=43) occurred in dark. Pedestrian
walking/running along the street account for 26% (n= 58), more than
50% occurred in dark. n= 20 (9%) cases can be covered by pedestrian
walks out into the path of turning car. Most of them (85%) occurred
in daytime. The situation of pedestrian walking out from behind
obstruction account for only 5% (n=12) in the current study, of which
n=2 occurred in dark. In this study, pedestrians less than14-year-old
were classied into child group. Of the n=32 child pedestrian
accidents, more than 40% (n =13) were from (walks/runs) near side.
The same percent 40% (n =13) were from far side. Of all of the child
pedestrian accidents, 47%(n=15) occurred in dark.
Reconstruction Output
Based on case-by-case analysis, n=183 car-to-pedestrian accidents
were reconstructed by the method stated in section of ‘Accident
reconstruction’. In Figure 7 (a), the cumulative distribution of car
impact speed for the sample is shown with a mean impact speed at
39km/h. For 80% cases, the collision speed was less than 60km/h.
The mean walking speed was about 5km/h(Figure 7 (b)).
Figure 7. Cumulative percent of car collision speed and pedestrian walking
speed(a),(b)
Figure 8 presents the driver behaviors in the one second before
collision based on reconstruction results. Of the n=183 cars, 66%
were braked, 27% were without braking or accelerating and 7% were
in the acceleration. Of the n=121 braked cases, Figure 9 shows the
cumulative deceleration during the process of one second before
collision. It was shown that the mean deceleration was −2.8 m/s2 for
braking cases and less than 20% were fully braked (deceleration less
than −6m/s2).
Downloaded from SAE International by Bing Dai, Wednesday, July 01, 2015
Figure 8. Driver behavior in one second before collision
Figure 9. Cumulative of mean deceleration distribution in the process of one
second before collision (n=121)
The relative positions of pedestrian to car front center at 1.0 second
and 0.5 second before crash are shown in Figure 10. Fields of view
(FOV), 40 degrees with range of 35m, 50degrees with range of 30m
and 60 degrees with 25m were outlined for each occasion. A standard
car with width of 1.66m together with a motor lane with width of
3.32m (2 times of the standard car) was illustrated. Table 10
summarizes the rate of detected cases by eld of view of 40 degrees.
It was shown that a eld of view of 40 degrees enabled detection of
93% (n=144) of pedestrians in the group of ‘on obstacle’ at the
moment of 1.0 second prior to crash. The detected cases decreased to
n=134 (86%) at 0.5 second prior to crash. Concerning the accidents
that were investigated with information of obstruction, the
pedestrians can be detected at 1.0 second prior crash if the obstruction
was coming from xed objects (such as buildings, plants) by
reconstruction results. However when the obstruction was from
temporary objects, e.g. moving/waiting vehicles, it was difcult to
identify the exact location where the obstruction were located (when
the investigator arrived at accident scene, generally the waiting/
moving vehicles had already left). To study these cases, we assumed
that the motor lane of the car travelling was not blocked by temporary
objects and the pedestrian can be ‘seen’ if he/she was in the lane.
Based on the assumption, it is shown that only 45% (5 of 11) can be
detected at both1.0 and 0.5 second before crash. With FOV
40degrees, situation of low detecting rate also appears that only 50%
(9 of 18) could be visible by sensors at both 1.0 and 0.5 second
before crash. At 1.0 second prior to collision, view detection range
30m was sufcient for pedestrians.
Figure 10. Relative position of pedestrian to the car front center (a) at 1.0
second prior to collision, (b) at 0.5 second prior to collision(Green square
presents driver view blocked by vehicle. Circle with cross inside present car
turning cases. Detection range: 25m/30m/35m; FOV: 40degrees
(red)/50degrees (blue) /60 degrees (black))
Table 10. Rate of detected pedestrian in case of FOV 40°
Crash Severity Mitigation by AEB
In Figure 11, the cumulative distribution of with/without(original
impact speed) for the sample is shown. The mitigation prediction
with autonomous braking was applied to all cases regardless of
visibility and eld of view. This represents the greatest possible level
of mitigation given the unrealistic assumption of perfect information.
The AEB was assumed to function at 1.0 second before collision with
instant deceleration of −8 m/s2 [14]. According to the prediction,
n=38 cases (21%) can be totally avoided with new collision speed
equals 0km/h. For the remaining cases (n=145), the mean collision
speed was 24km/h, which was decreased by 38% comparing to the
original mean speed 39km/h.
Downloaded from SAE International by Bing Dai, Wednesday, July 01, 2015
Figure 11. Cumulative percent of car collision speed with/without AEB (AEB
functioned at 1 second before collision with instant deceleration −8m/s2. n =
145 cases calculated and n= 38 cases avoided (collision speed = 0) for the
situation with AEB)
Discussion
For on-scene investigated accidents, a special accident investigation
team came to the accident scene upon receiving an alarm from the
emergency calling center with on-duty trafc policemen. Accident
sketches were drawn based on on-scene measurements. Detailed
vehicle damage information was inspected in a parking lot, in an area
provided for damaged vehicles in accidents. Injury information was
provided by clinical reports. It is important that accident
investigations are of high quality to assure impact speed estimations.
Systematic and random errors are expected to be kept to a minimum
level. It has been shown that in only 47% of cases pedestrian heads
contacted the car body causing obvious damage to the windscreen.
Distribution of head contact points illustrated a ‘U’ shape around the
A-pillars and the bottom of windscreen with a range of about 150mm
from the frame. In China, ‘unsure safety’ with situations of
misjudgment and unpreparedness for the accident were the most
frequent causes of pedestrian accidents. ‘Not giving way to
participants with priority’ comprises the most percentage of the cause
of road priority violation. For pedestrians, incorrectly crossing roads,
walking on freeways and trafc light violation were the most
common failures. According to the dominating accident causes,
enhancing safety consciousness of all road users and strict trafc law
enforcement could provide an effective way of decreasing accidents
and fatalities in China.
According to the studies on European and US accidents by Avery
[13], test scenarios selected to represent greatest frequency of real
world crashes were: (1) Pedestrian walks from nearside; (2)
Pedestrian walks out from behind obstruction; (3) Pedestrian walks/
runs from far side; (4) pedestrian walks along in the road; (5)
pedestrian walks out into the path of turning car. Combining Chinese
data from other international sources is shown in Table 11. It is found
that pedestrian walking along the street account for 26% in China,
which is unique comparing to developed countries. According to the
current sample, the proposal of AEB test scenarios in China could be:
(1) Unobscured pedestrian walks from nearside; (2) Unobscured
pedestrian walks from far side; (4) pedestrian walks along the road,
which comprise 85% of the pedestrian accidents scenarios. It is
necessary to underline that more than 50% pedestrian accidents
occurred in night, which should be considered in tests in China.
Table 11. Combining accident data from other international sources
Undoubtedly, reducing speed should be the primary focus for
lowering the risk of trafc fatality and injury. Currently, active
systems such as automatic braking are being rapidly developed. And
the rst systems for pedestrian safety are now on the market. The
systems consist of a pre-crash sensor that detects potentially
dangerous situations. The system will automatically brake the car if
the detected danger is unnoticed by the driver. In real-life trafc, auto
braking implies a rather abrupt intervention, which may affect the
driver's ability to maintain control of the vehicle. A discrepancy
between the need for early activation and minimum false activation is
unavoidable. Active safety products for pedestrians are now prevalent
in developed countries. However, the related products in China are
still in the debugging phase. Problems may appear while the trafc
data-based products from developed areas operate in complex
Chinese road conditions. In-depth accident analysis provides an
effective way to evaluate product efciency using Chinese data. A
previous study of pedestrian safety with Germany GIDAS data shows
that with a eld of view equal to 40°, 53 of 57 (93%) fatally/severely
injured pedestrians will be visible to the vehicle one second prior to
impact (not considering obstacles)[14]. In our study in 2014 with less
pedestrian cases (n=108) [15], the percentages of covered cases are
92% for fatal/in-patient injured pedestrians, if one second prior to
impact a similarly FOV is applied. In this study with the extended
sampling cases (n=183), similar result we got that a eld of view of
40 degrees enabled detection of 93% of pedestrians in the group of
‘on obstacle’ at the moment of 1.0 second prior to crash and the
detected percentage could decrease at 0.5 second prior to crash
(86%).With FOV 40 degrees, only 45% and 50% pedestrian could be
visible to sensors for cases that moving/waiting vehicles blocked
driver's view and when accident car was turning at 1.0second prior to
crash. The percentage would not change at 0.5 second prior to crash.
Downloaded from SAE International by Bing Dai, Wednesday, July 01, 2015
Based on the current sample, 66% drivers braked in the process of 1.0
second before collision; however, in many cases the effect of the
braking was very small with a mean deceleration −2.8 m/s2. Based on
the assumption that AEB was activated at1.0second before collision
with instant deceleration of −8.0m/s2, 21% of the studied accidents
can be avoided with the new collision speeds equal 0km/h. However,
more cases could be avoided in situations where the pedestrians could
already leave the travelling lane if car speed can be decreased. For the
remaining cases, the mean collision speed could decrease to 24km/h
and 80% of collision speeds were less than 40km/h, which could be
much safer to all kinds of pedestrian injuries.
Limitation
Sources of uncertainty for this study range from the uncertainties of
accident investigation and inaccuracy of accident reconstructions.
The data used in this study was not weighted. The reason is that in
China the only national accident statistics is not transparent and
incomplete. Momently we have no reliable representative sample to
be weighted to. However, representativeness is very important to
predictive studies, which is a disadvantage for this study. In the
current study, fatality rate for car-to-pedestrian crashes is about 35%.
Accidents with no or slight injuries were always solved by accidents
participants themselves. For these cases, it could be no report to
policeman. To avoid trafc jam, accidents with slight injuries,
property loss and clear-cut accident liability are required to take quick
on-site photos and remove the accident scene before policeman
reaching. These are part of reasons that the current database is
severely-injured biased.
Conclusion
Scenarios: (1) Unobscured, pedestrian walking out from nearside; (2)
Unobscured, pedestrian walking from far side; (3) pedestrian walking
along the road, are the most frequent situations in China. Especially
‘pedestrian walking along the road’ is quite unique comparing to
developed countries. More than 50% pedestrian accidents occurred in
dark, it is recommended to have a scenario test in dark. FOV of
40degrees with range around 30m is sufcient to detect 93%
pedestrians that the car is not in situations of turning and blocking by
moving/waiting vehicles.
Acknowledgement
We thank Dr. Yong Chen (from Volkswagen Research China) for his
support with accident reconstruction and analysis as well as paper
writing and edition. The conclusions and points stated in this paper
are only that of CATARC.
References
1. EC, 2003. Directive 2003/102/EC of the European Parliament
and of the Council of17 November 2003 relating to the
protection of pedestrians and other vulnerable road users
before and in the event of a collision with a motor vehicle and
amending council directive 70/156/eec.
2. EC, 2009. Regulation (EC) No 78/2009 of the European
Parliament and of the Council of 14 January 2009 on the Type
approval of Motor Vehicles with Regard to the Protection
of Pedestrians and other Vulnerable Road Users, Amending
Directive 2007/46/EC and repealing Directives 2003/102/EC
and 2005/66/EC.
3. Fredriksson, R., Rosén, E., Kullgren, A., 2010. Priorities of
pedestrian protection - areal-life study of severe injuries and car
sources. Accident Analysis & Prevention 42 (6), 1672-1681.
4. Fredriksson, R., Haland, Y., Yang, J., 2001. Evaluation of a new
pedestrian head injury protection system with a sensor in the
bumper and lifting of the bonnet's rear part. In: Proceedings of
17th ESV Conference, Amsterdam, Netherlands.
5. Maki, T., Asai, T., Kajzer, J., 2003a. Development of future
pedestrian protection technologies. In: Proceedings of 18th
International Technical Conference on the Enhanced Safety of
Vehicles (ESV), Nagoya, Japan.
6. Nagatomi, K., Hanayama, K., Ishizaki, T., Sasaki, S., 2005.
Development and full-scale dummy tests of a pop-up hood
system for pedestrian protection. In: Proceedings of 19th
International Technical Conference on the Enhanced Safety of
Vehicles (ESV), Washington DC, USA.
7. Oh, C., Kang, Y., Kim, W., 2008. Assessing the safety benefits
of an advanced vehicular technology for protecting pedestrians.
Accident Analysis & Prevention 40 (3), 935-942.
8. Pinecki, C., Zeitouni, R., 2007. Technical solutions for
enhancing the pedestrian protection. In: Proceedings of 20th
International Technical Conference on the Enhanced Safety of
Vehicles (ESV), Lyon, France.
9. Yao, J., Yang, J., Otte, D., 2008. Investigation of head injuries
by reconstructions of real-world vehicle-versus-adult-pedestrian
accidents. Safety Science 46 (7),1103-1114.
10. Otte, D., “Use of Throw Distances of Pedestrians and Bicyclists
as Part of a Scientific Accident Reconstruction Method,” SAE
Technical Paper 2004-01-1216, 2004, doi:10.4271/2004-01-
1216.
11. Rosén, E., Sander, U., 2009. Pedestrian fatality risk as a
function of car impact speed. Accident Analysis & Prevention
41 (3), 536-542.
12. Rosén, E., Autonomous Emergency Braking for Vulnerable
Road Users. In: Proceedings of IRCOBI 2013 Conference,
Gothenburg pp. 13-71.
13. Avery, M., AEB test development, autonomous emergency
braking, and further details at www.thatcham.org/AEB.
14. Rosén, E., Kallhammer, J., Eriksson, D., Nentwick, M.,
Frediriksson, R., Smith, K., Pedestrian injury mitigation
by autonomous braking, Accident Analysis and Prevention
42(2010) 1949-1957.
15. Chen, Q., Chen, Y., Bostrom, O., Ma, Y. et al., “A Comparison
Study of Car-to-Pedestrian and Car-to-E-Bike Accidents: Data
Source: The China In-Depth Accident Study (CIDAS),” SAE
Technical Paper 2014-01-0519, 2014, doi:10.4271/2014-01-
0519.
Downloaded from SAE International by Bing Dai, Wednesday, July 01, 2015
Contact Information
Bing Dai
daibing@catarc.ac.cn
The Engineering Meetings Board has approved this paper for publication. It has successfully completed SAE’s peer review process under the supervision of the session organizer. The process
requires a minimum of three (3) reviews by industry experts.
All rights reserved. No part of this publication may be reproduced, stored in a retrieval system, or transmitted, in any form or by any means, electronic, mechanical, photocopying, recording, or
otherwise, without the prior written permission of SAE International.
Positions and opinions advanced in this paper are those of the author(s) and not necessarily those of SAE International. The author is solely responsible for the content of the paper.
ISSN 0148-7191
http://papers.sae.org/2015-01-1464
Downloaded from SAE International by Bing Dai, Wednesday, July 01, 2015
