DG RTD SEVENTH FRAMEWORK PROGRAMME THEME 7 TRANSPORT - SST SST.2011.RTD-1 GA No

Transcription

1 EUROPEAN COMMISSION DG RTD SEVENTH FRAMEWORK PROGRAMME THEME 7 TRANSPORT - SST SST.2011.RTD-1 GA No ASPECSS Assessment methodologies for forward looking Integrated Pedestrian and further extension to Cyclists Safety Deliverable No. ASPECSS D1.5 Deliverable Title Dissemination level Written By Benefit estimate based on previous studies for pre-crash bicyclist systems and recommendations for necessary changes to pedestrian test and assessment protocol Carmen Rodarius, Maurice Kwakkernaat (TNO) Mervyn Edwards (TRL) June 2014 Checked by Ma Stefanie de Hair (TNO) June 2014 Approved by Monica Pla (IDIADA) June 9 th 2014

2 Executive summary Though the main focus of the AsPeCSS project lies on pedestrian safety systems, this deliverable is dedicated to bicyclists and respective measures to also protect this group of vulnerable road users. Within this deliverable, the car-to-bicyclist accidents situation has been investigated for the UK and the Netherlands. The main differences between car-to-bicyclist and car-to-pedestrian accidents have been pointed out and general test scenarios for bicyclist safety systems have been proposed. Additionally, RCS measurements of bicyclist and pedestrian (dummies) have been conducted that can be used as baseline information for further development of a bicyclist dummy target. In a last step, the effectiveness of bicyclist helmets has been estimated based on UK data and the effectiveness of AEB systems addressing bicyclists and VRU airbags have been summarized based on SaveCAP data. SaveCAP is a project on Saving Cyclist and Pedestrian by car bounded measures ( 2/49

3 Contents 1 Introduction The EU project AsPeCSS Structure of this deliverable Objectives Data sources Cyclist casualties in Great Britain Characteristics of accidents involving an injured bicyclists Pedal bicyclist casualties Contributory factors in accidents involving at least one bicyclist casualty Defining the target population Target population characteristics Pedal bicyclist casualties Contributory factors in accidents involving at least one bicyclist casualty Accident scenarios Comparison to the Netherlands Radar cross section measurements of dummies and humans Introduction Measurement method and setup RCS signatures Pedestrian 24 GHz Pedestrian 77 GHz Cyclist 24 GHz Cyclist 77 GHz Comparison RCS signatures Preliminary Test Scenarios Differences between bicyclists and pedestrians General boundary conditions for car-to-bicyclist accidents Preliminary bicyclist test scenarios General proposed test conditions Benefit estimate based on previous studies UK study on cycle helmets based on literature review AEB systems and VRU airbag based on SaveCAP Risk Register Conclusions and Recommendations Literature Acknowledgment /49

4 4/49

5 1 Introduction 1.1 The EU project AsPeCSS The overall purpose of the AsPeCSS project is to contribute towards improving the protection of vulnerable road users, in particular pedestrians and bicyclists, by developing harmonized test and assessment procedures for forward-looking integrated pedestrian safety systems. The output of the project will be a suite of tests and assessment methods for use in future regulatory procedures and consumer rating protocols. Implementation of such procedures / protocols will encourage widespread introduction of such systems in the vehicle fleet, resulting in a significant reduction of casualties and their injury severity for vulnerable road users. The ASPECSS project is divided into five work packages (WPs) as follows: WP1 aims to identify accident scenarios of pedestrians and bicyclists and develop weighting factors. Furthermore, WP1 aims to develop a methodology for the overall assessment of pedestrian pre-crash systems and to provide a benefit estimate and a proposal for test and assessment protocols for forward looking integrated pedestrian safety systems. In addition, a benefit estimate will be provided for pre-crash bicyclist systems and recommendations will be given for the work needed to update the pedestrian test and assessment protocols to include bicyclists. WP2 will develop a test protocol for the assessment of the effectiveness of preventive pedestrian protection systems and a test protocol for quantifying unjustified system responses (sometimes referred to as false positives ). Additionally, prototype test targets will be selected. WP3 will provide specifications of new impact conditions for head- and legform impactors given by the introduction of active systems. Simulations and experimental tests will be conducted for vehicles with different levels of passive safety performance. Finally, WP3 will generate risk curves for single vehicle points as functions of vehicle impact speed for different vehicle types. WP4 is the central work package for dissemination activities. WP5 coordinates the project in terms of research quality and on time delivery of results. 1.2 Structure of this deliverable This document is structured as follows: Chapter 1 provides an overall introduction to the AsPeCSS project as well as this subtask; Chapter 2 briefly describes the data sources used for this sub task; Chapter 3 gives an overview on the bicyclist casualties in Great Britain; Chapter 4 compares the UK situation to the Dutch situation; Chapter 5 gives an overview on the conducted RCS measurements; Chapter 6 provides a general description of the proposed test scenarios for bicyclists including an overview of differences between bicyclists and pedestrians with respect to accidents involving passenger cars and recommendations for further work; Chapter 7 gives an estimate on bicyclist helmet as well as AEB and VRU airbag effectiveness based on information from previous projects; Chapter 8 includes the risk register and Chapter 9 provides overall conclusions. 5/49

6 1.3 Objectives The objectives of the work reported in this deliverable were: To provide recommendations for the additional work necessary to adopt test scenarios and extend the methodology from WP1/WP2 to include bicyclists To analyze differences between bicyclist and pedestrian accidents in terms of test scenarios (UK data) and test conditions To conduct RCS measurements on bicyclist (dummies) for comparison with pedestrian (dummy) RCS To estimate the benefit of helmets (UK data) as well as active and passive means on cars (SaveCAP data) from previous studies conducted 6/49

7 2 Data sources Data was gathered mainly from the following sources dealing with bicyclist safety: 1.) SaveCAP project ( The SaveCAP project consortium consisted of the new Ministry of Infrastructure and Environment, research institute TNO, the Dutch Cyclists Union, insurance company Centraal Beheer Achmea, and 1st Tier supplier Autoliv. The project ran until October It focused on the development and testing of vehicle systems for passenger cars, protecting bicyclists and pedestrians in case of a crash. This could be an airbag system, or a mitigation system like automatic braking. Important feature of the systems worked on, was the detection system that takes care of a timely detection and recognition of a pedestrian or bicyclist in the driving path of the vehicle. The system s output was used to trigger the safety system. The SaveCAP project consisted of the following activities: Effectiveness study Pre-development of the prototype sensor Sensor field testing in crowded urban area Pre-development prototype airbag Prototype field testing specifications System evaluation, using Beyond NCAP Protocol 2.) TNO work on bicyclist safety for the Dutch government Besides the SaveCAP project, TNO has conducted work for the Dutch Ministry of Infrastructure and Environment looking into bicyclist accidents (mainly on Dutch roads) and common accident configurations. For as far these results were released by the Ministry, they were included into this report. 3.) BRON The Bestand geregistreede Ongevallen in Nederland (BRON) is a national accident database in the Netherlands, where all traffic accidents which were reported to and processed by the police are registered. Though not being an in-depth database like GIDAS, BRON does contain a variety of parameters with respect to the accident circumstances as well as the parties involved. Data is available from 1976 onwards. The database can be considered to cover approximately 90% of all fatal Dutch traffic accidents. The degree of registration for non-fatal accidents is less, though it can be estimated in combination with other sources. It should be noted, that the database is filled based on information gathered by the police gather on the accident scene. The injury levels and therewith the severity of the accident is not always as accurate.. Therefore, SWOV (Dutch Institute for Road Safety Research) updates (if possible) the available information based on information from the Landelijke Medische Registratie (LMR) (country based medical registration). 4.) STATS19 From Road accidents on the public highway in Great Britain, reported to the police and which involve human injury or death, are recorded by police officers onto a STATS19 report form. The form collects a wide variety of information about the accident (such as time, date, location, road conditions) together with the vehicles and casualties involved and contributory factors to the accident (as interpreted by the police). The form is completed at either the scene of the accident, or when the accident is reported to the police. 7/49

8 The Department for Transport has overall responsibility for the design and collection system of the STATS19 data. The Standing Committee on Road Accident Statistics (SCRAS) is the body set up to oversee the STATS19 process for road accident data collection. The STATS19 data is the only national source to provide detailed information on accident circumstances, vehicles involved and resulting casualties and is the most detailed and reliable single source on accidents that can be used for longitudinal research in Great Britain. However, it should be borne in mind that is not a complete record of all injury accidents and resulting casualties, due to some accidents not being reported. The STATS19 data is available from the Department for Transport. Non disclosive data is also deposited with the Economic and Social Data Service. 8/49

9 3 Cyclist casualties in Great Britain Between 2008 and 2010, 661,669 casualties were injured in road accidents in Great Britain. Of these, 50,546 were bicyclist casualties. Figure 3-1 shows the breakdown of the casualties by road user type and injury severity. Figure 3-1: Casualties by road user type and injury severity ( ) 1% (330) of bicyclist casualties between 2008 and 2010 were killed, 15% (7,716) were seriously injured and 84% (42,500) were slightly injured. 3.1 Characteristics of accidents involving an injured bicyclists This section details the characteristics of accidents which involve at least one bicyclist casualty. Section looks at the characteristics of the casualties themselves, section details the most common contributory factors in these collisions and section defines the target population of interest Pedal bicyclist casualties Figure 3-2 shows the number of bicyclist casualties injured between 2008 and 2010 by age and gender. Figure 3-2: Pedal bicyclist casualties by age and gender ( ) 9/49

10 Over 80% of bicyclist casualties were male, mainly because of exposure issues. A higher proportion of these male casualties were killed or seriously injured than their female counterparts (16% KSI compared to 15%). Pedal bicyclist casualties were most commonly years old; however this age group contains the widest range of ages so this may explain some of this trend. Relatively few casualties were aged or over 80; appropriate exposure data, such as the number of vehicle-miles travelled split by age group, would be required to determine the casualty rate (or number of casualties per mile travelled) for each age group. Figure 3-3 displays bicyclist casualties by junction type. Figure 3-3: Pedal bicyclist casualties by junction type ( ) Three quarters of all bicyclist casualties were injured at a junction; most commonly at T or staggered junctions, followed by roundabouts or crossroads. The number of bicyclist casualties by speed limit and area is shown in Table 3-1. Table 3-1: Pedal bicyclist casualties by speed limit of road and area ( ) Speed limit Area Urban Rural Unknown Total Built-up ( 40mph 64 km/h) 41,716 5, ,892 Non-built-up (>40mph) 428 3, ,654 Total 42,144 8, ,546 The majority of bicyclists were injured in urban areas on roads with a speed limit of 40mph or less. Table 3-2: Pedal bicyclist casualties by road class ( ) Number of Road class casualties Motorway 1 A-road 20,992 B-road 6,201 C-road 5,122 Unclassified 18,230 Total 50,546 10/49

11 Note: UK roads are classified as follows: [15] Motorways: Motorways A-Road: A main recommended route which can be either single carriageway or dual carriageway B-road: Regional in nature and used to connect areas of lesser importance. C-road: Roads within local area often not shown on road maps. Unclassified: Local roads with no defined destination. 42% of bicyclist casualties were injured on A-roads and 36% were injured on unclassified roads; the majority of these roads had a speed limit of 30mph. Figure 3-4: Pedal bicyclist casualties by lighting condition ( ) Figure 3-4 shows that 81% of casualties were injured in daylight conditions, 17% were injured at night in locations where the street lights were lit and only 1% were injured where no street lights were lit. Table 3-3 displays the number of bicyclist casualties, injured between 2008 and 2010, by weather condition at the time of the accident. Table 3-3: Pedal bicyclist casualties by weather condition ( ) Weather condition Number of casualties Proportion of casualties Fine without high winds 43,177 85% Raining without high winds 4,130 8% Snow without high winds 150 0% Fine with high winds 489 1% Raining with high winds 434 1% Snowing with high winds 18 0% Fog or mist 150 0% Other 836 2% Unknown 1,162 2% Total 50, % The vast majority of bicycle casualties were injured in accidents when the weather was fine without high winds. 11/49

12 3.1.2 Contributory factors in accidents involving at least one bicyclist casualty In total between 2008 and 2010, 50,010 accidents involved at least one bicyclist casualty. Of these accidents 32,968 (66%) were attended by the police and had at least one contributory factor recorded. For the remainder of this section only accidents attended by the police with at least one contributory factor will be examined; this number is the divisor in any percentages reported. In the 32,968 accidents, 78,028 contributory factors were recorded: an average of 2.4 contributory factors per accident. The top ten most common contributory factors in these accidents are shown in Table 3-4. These factors can be assigned to either the bicyclist, another vehicle/driver in the accident or to a pedestrian or passenger (who may or may not be injured). Table 3-4: Top 10 most common contributory factors in accidents involving at least one bicyclist casualty CF number Contributory factor description Proportion of accidents with CF 405 Failed to look properly 70% 406 Failed to judge other person's path or speed 25% 602 Careless, reckless or in a hurry 18% 403 Poor turn or manoeuvre 17% 310 Cyclist entering road from pavement 10% 407 Passing too close to bicyclist, horse rider or pedestrian 10% 701 [vision affected by] stationary or parked vehicle(s) 7% 410 Loss of control 6% 302 Disobeyed 'Give Way' or 'Stop' sign or markings 5% 507 Cyclist wearing dark clothing at night 4% The most common contributory factor (recorded in 75% of all accidents involving a bicyclist casualty, attended by the police with at least one contributory factor) is failed to look properly. Not surprisingly, contributory factors which specifically relate to accidents involving bicyclists (e.g. bicyclist entering road from pavement, passing too close to bicyclist, horse rider or pedestrian and bicyclist wearing dark clothing at night ) also feature in the top 10 most common contributory factors Defining the target population Table 3-5 shows accidents involving at least one bicyclist casualty split by the number of vehicles involved in the accident. Table 3-5: Accidents involving at least one bicyclist casualty by number of vehicles involved ( ) Number of vehicles involved Number of accidents Proportion of accidents 1 (single vehicle bicycle accident) 1 1,587 3% 2 (bicycle + 1 other vehicle) 47,026 94% 3 (bicycle + 2 other vehicles) 1,286 3% 4 (bicycle + 3 other vehicles) 90 0% 5 (bicycle + 4 other vehicles) 15 0% 6 (bicycle + 5 other vehicles) 4 0% 7 (bicycle + 6 other vehicles) 2 0% Total 50, % 12/49

13 1 Single-vehicle bicycle accidents are likely to be under-reported. Please note that this figure will also include accidents involving a pedestrian and a bicycle. Accidents involving two vehicles (i.e. the bicycle and one other vehicle) account for 94% of all accidents involving at least one bicyclist casualty. These accidents will be examined further to determine target population : accidents involving at least one injured bicyclist where the bicycle is impacted by the front of a car or taxi. Specifically, we are interested in the number of bicyclist casualties in these accidents. In defining the target population a number of caveats are required: Only two vehicle accidents will be considered as it is often difficult to tell which vehicle(s) collided with the bicycle in an accident involving three or more vehicles. STATS19 records the first point of impact of each vehicle involved in a collision; this may refer to the first point of impact with another vehicle or with an object or pedestrian (either on or off the carriageway). For the purposes of this analysis, it will be assumed that the first point of impact recorded for the car/taxi in the collision refers to the first point of impact with the bicyclist and not with a pedestrian or object. This may result in the target population overestimating the number of bicyclist casualties in accidents where the bicycle is impacted by the front of a car or taxi. Table 3-6 displays the number of two vehicle accidents involving at least one bicyclist casualty by impact partner and first point of contact on the impact partner. The target population of accidents is highlighted in yellow. The figure in the top left hand corner represents accidents involving two bicycles. Table 3-6: Two vehicle accidents involving at least one bicyclist casualty by impact partner and first point of contact ( ) First point of impact on impact partner Impact partner Bicycle Did not impact Front Back Offside Nearside Unknown Total Pedal cycle PTW Car or taxi 1,629 19,092 2,183 7,107 10, ,529 Minibus Bus or coach ,122 Other vehicle Goods vehicle 218 1, , ,741 Unknown Total 244 2,031 21, , ,026 The target population consists of 19,092 accidents involving 19,565 casualties, of which 19,172 were bicycle casualties. The characteristics of the 19,172 bicyclist casualties will be examined in more detail in section Target population characteristics This section details the characteristics of the target population accidents (as defined in section 3.1.3). Section looks at the characteristics of the casualties themselves and section details the most common contributory factors in these collisions. The final section (3.2.3) defines the accident scenarios for these collisions. 13/49

14 3.2.1 Pedal bicyclist casualties The breakdown of bicyclist casualties in the target population by injury severity is shown in Figure 3-5. Figure 3-5: Pedal bicyclist casualties in target population by injury severity ( ) Overall, 15% of casualties in the target population were killed or seriously injured. The casualty figures show that 81% of bicyclist casualties in the target population are male; this is the same proportion of casualties which were male when all bicyclist casualties were examined (see Figure 3-2). Casualties in the target population were split by age group; Figure 3-6 displays the proportion of casualties which were killed, seriously injured and slightly injured in each of these age groups. Figure 3-6: Pedal bicyclist casualties in target population by age and injury severity ( ) A higher proportion of older bicyclist casualties (aged 50+) were killed or seriously injured compared to casualties in the younger age brackets; however these proportions are based on much smaller samples: for example, only 81 bicyclist casualties were aged 80 or over. The increase in KSI proportion between the 14/49

15 youngest and oldest groups may be due to differences in the fragility of casualties and/or differences in road user factors such as the locations where bicycle s are used. Figure 3-7 shows the percentage of casualties (in the target population) who were killed or seriously injured by junction type. Figure 3-7: Percentage of bicyclist casualties in target population killed or seriously injured by junction type ( ) Over 80% of casualties in the target population were injured in an accident at a junction. Accidents on slip roads and at multiple junctions most commonly resulted in KSI injuries. Table 3-7: Pedal bicyclist casualties in target population by speed limit of road and area ( ) Speed limit Area Urban Rural Unknown Total Built-up ( 40mph 64 km/h) 15,888 2, ,935 Non-built-up (>40mph) 176 1, ,237 Total 16,064 3, ,172 The majority of casualties were involved in accidents in urban areas on roads with a speed limit of 40mph or less. Table 3-8: Pedal bicyclist casualties in target population by road class ( ) Number of Road class casualties % KSI A-road 7,212 15% B-road 2,422 16% C-road 1,902 15% Unclassified 7,636 14% Total 19,172 15% Just under 40% of bicyclist casualties were injured on unclassified roads. In addition 7,212 (38%) were injured on A-roads. 15/49

16 Figure 3-8 shows the proportion of casualties which were killed, seriously injured and slightly injured in each of the lighting conditions. Figure 3-8: Pedal bicyclist casualties in target population by lighting condition and injury severity ( ) 78% of casualties in the target population were injured in collisions which occurred during daylight hours; 14% of these were killed or seriously injured. This compares to a KSI proportion of 15% for casualties injured in darkness when street lights were lit and 29% in darkness with no street lights. Figure 3-9: Percentage of bicyclist casualties in target population killed or seriously injured by weather condition ( ) 84% of casualties were injured in accidents when the weather was fine without high winds ; of these casualties 15% were killed or seriously injured. The highest proportion (17%) of casualties were killed or seriously injured in snow without high winds however this proportion is based on a small sample of just 69 casualties. The manoeuvre each vehicle was completing at the time of the collision is recorded. Table 3-9 shows the number of bicyclist casualties injured as the car or taxi was performing each of the listed manoeuvres. 16/49

17 Table 3-10 shows the manoeuvre of the bicyclist at the time of the collision. Table 3-9: Pedal bicyclist casualties in target population and % KSI by other vehicle manoeuvre ( ) Other vehicle manoeuvre Number of casualties % KSI Reversing 35 3% Parked % Waiting to go ahead but held up 217 8% Slowing or stopping % Moving off 2,003 11% U turn 84 4% Turning left 2,043 9% Waiting to turn left 117 5% Turning right 4,129 14% Waiting to turn right 138 8% Changing lane to left 55 16% Changing lane to right 46 17% Overtaking moving vehicle on its offside % Overtaking stationary vehicle on its offside % Overtaking on nearside 67 22% Going ahead left hand bend % Going ahead right hand bend % Going ahead other 8,426 17% Unknown 1 0% Total 19,172 15% The most common maneuvers for the car/taxi to be performing when the bicyclist was injured were going ahead other, turning right, turning left and moving off. Overtaking maneuvers and going ahead right hand bend resulted in the highest proportions of KSI bicyclist casualties. 17/49

18 Table 3-10: Pedal bicyclist casualties in target population and % KSI by bicycle maneuver ( ) Number of Bicycle manoeuvre casualties % KSI Reversing 11 9% Parked 27 4% Waiting to go ahead but held up 302 7% Slowing or stopping % Moving off % U turn 19 16% Turning left % Waiting to turn left 29 14% Turning right 1,554 18% Waiting to turn right % Changing lane to left 99 22% Changing lane to right % Overtaking moving vehicle on its offside 44 14% Overtaking stationary vehicle on its offside % Overtaking on nearside 95 9% Going ahead left hand bend % Going ahead right hand bend % Going ahead other 14,262 15% Unknown 2 0% Total 19,172 15% Almost three quarters of bicyclists were going ahead other when the collision occurred. Changing lanes to the left or right resulted in the highest proportions of KSI casualties. The number of bicyclist casualties in the target population split by impact point on the bicycle is displayed in Figure Figure 3-10: Bicyclist casualties in target population by impact point on the bicycle ( ) Bicycle s were most commonly hit on the front; however back, offside and nearside collisions were also common. 168 bicyclists were injured in the collision but were not impacted. 18/49

19 3.2.2 Contributory factors in accidents involving at least one bicyclist casualty In total there were 19,092 accidents in the target population. Of these accidents 12,585 (66%) were attended by the police and had at least one contributory factor recorded. In the 12,585 accidents, 29,747 contributory factors were recorded: an average of 2.4 contributory factors per accident. Of these 29,747 contributory factors, 29,114 were assigned to a vehicle i.e. the factor applies to a vehicle, driver/rider or to the road environment. Table 3-11 shows how the contributory factors in these accidents are assigned to each of the vehicles. Table 3-11: Contributory factor(s) assignment by vehicle type ( ) Number of Contributory factor assignment accidents Contributory factor(s) assigned to bicycle only 3,860 Contributory factor(s) assigned to car/taxi only 5,990 Contributory factors assigned to both bicycle and car/taxi 2,512 Target population accidents attended by the police with at least on contributory factor assigned to a vehicle in the accident 12,362 In just under half of these accidents the contributory factor(s) were assigned only to the car/taxi. For the remainder of this section only these 12,362 accidents are examined; this number is the divisor in any percentages reported. The top ten most common factors assigned to the car or taxi are shown in Table The top 10 assigned to the bicycle are shown in 19/49

20 Table Table 3-12: Top 10 most common contributory factors assigned to the car or taxi in target population accidents ( ) CF number Contributory factor description Proportion of accidents with CF 405 Failed to look properly 47% 406 Failed to judge other person's path or speed 15% 602 Careless, reckless or in a hurry 11% 403 Poor turn or manoeuvre 10% 407 Passing too close to bicyclist, horse rider or pedestrian 6% 302 Disobeyed 'Give Way' or 'Stop' sign or markings 5% 701 [vision affected by] stationary or parked vehicle(s) 4% 706 [vision affected by] dazzling sun 4% 707 [vision affected by] rain, sleet, snow or fog 2% 402 Junction restart (moving off at junction) 2% 20/49

21 Table 3-13: Top 10 most common contributory factors assigned to the bicycle in target population accidents ( ) CF Proportion of number Contributory factor description accidents with CF 405 Failed to look properly 28% 310 Cyclist entering road from pavement 13% 406 Failed to judge other person's path or speed 9% 602 Careless, reckless or in a hurry 8% 403 Poor turn or manoeuvre 5% 507 Cyclist wearing dark clothing at night 4% 701 [vision affected by] stationary or parked vehicle(s) 3% 506 Not displaying lights at night or in poor visibility 3% 410 Loss of control 2% 401 Junction overshoot 2% Failed to look properly was the most common contributory factor assigned to the car/taxi (47%) and to the bicycle (28%) in these accidents Accident scenarios This section splits the casualties in the target population into a number of different accident scenarios. In defining each of the accident scenarios a number of assumptions were required. The fields used to classify casualties into each of the accident scenarios were: Junction detail (at a junction, not at a junction) Car manoeuvre ( 1 ) Bicycle manoeuvre ( 1 ) Bicycle first point of impact (did not impact, front, back, offside, nearside) given that bicycle s are nearly 2D objects the first point of impact can be difficult to define; as such most of the accident scenarios defined in section specify more than one possible point of contact (the exception to this is scenario 4). Direction of vehicle travel this field records the from and to directions of travel (based on the compass points) for each of the vehicles in the collision e.g. if a vehicle travels from the North to the East this would be recorded as shown: These fields can be used to define differences in the direction of travel between the two vehicles. This is done by calculating the absolute difference between the to and from values and applying some criteria; where applicable these are described. 1 reversing, parked, waiting to go ahead but held up, slowing or stopping, moving off, U-turn, turning left, waiting to turn left, turning right, waiting to turn right, changing lane to left, changing lane to right, overtaking moving vehicle on offside, overtaking stationary vehicle on offside, overtaking on nearside, going ahead left hand bend, going ahead right hand bend, going ahead other, unknown 21/49

22 Defining the accident scenarios 1. Car turning Near-side o car turning off the road to the near side with bicyclist moving straight forward Assumptions: Junction detail: at a junction Car manoeuvre: turning near side, waiting to turn to the near side Bicycle manoeuvre: waiting to go ahead but held up, slowing or stopping, moving off, changing lane to the near side, changing lane to the far side, overtaking moving vehicle on offside, overtaking stationary vehicle on offside, overtaking on nearside, going ahead near side bend, going ahead far sight bend, going ahead other Bicycle first point of impact: front, back, nearside, offside 2. Car turning far-sight o car turning off the road to the right with bicyclist moving straight forward Assumptions: Junction detail: at a junction Car manoeuvre: turning right, waiting to turn right Bicycle manoeuvre: waiting to go ahead but held up, slowing or stopping, moving off, changing lane to the near side, changing lane to far sight, overtaking moving vehicle on offside, overtaking stationary vehicle on offside, overtaking on nearside, going ahead near sight bend, going ahead far sight bend, going ahead other Bicycle first point of impact: front, back, nearside, offside 3. Longitudinal o car travelling straight forward hitting bicyclist driving straight from behind Assumptions: Junction detail: at a junction, not at a junction Car manoeuvre: waiting to go ahead but held up, slowing or stopping, moving off, changing lane to near sight, changing lane to far sight, overtaking moving vehicle on offside, overtaking stationary vehicle on offside, overtaking on nearside, going ahead near sight bend, going ahead far sight bend, going ahead other Bicycle manoeuvre: waiting to go ahead but held up, slowing or stopping, moving off, changing lane to near sight, changing lane to far sight, overtaking moving vehicle on offside, overtaking stationary vehicle on offside, overtaking on nearside, going ahead near sight bend, going ahead far sight bend, going ahead other Bicycle first point of impact: back from direction for bicyclist - from direction for other vehicle : 0, 1, 7 4. Crossing o car and bicycle moving straight forward but at perpendicular angles to one another Assumptions: Junction detail: at a junction, not at a junction 22/49

23 Car manoeuvre: waiting to go ahead but held up, slowing or stopping, moving off, changing lane to near sight, changing lane to far sight, overtaking moving vehicle on offside, overtaking stationary vehicle on offside, overtaking on nearside, going ahead near sight bend, going ahead far sight bend, going ahead other Bicycle manoeuvre: waiting to go ahead but held up, slowing or stopping, moving off, changing lane to left, changing lane to right, overtaking moving vehicle on offside, overtaking stationary vehicle on offside, overtaking on nearside, going ahead left hand bend, going ahead right hand bend, going ahead other Bicycle first point of impact: front, back, nearside, offside from direction for bicyclist - to direction for bicyclist : 4, 3, 5 from direction for other vehicle - to direction for other vehicle : 4, 3, 5 from direction for bicyclist - from direction for other vehicle : 2, 6, 3, 5 5. Cyclist turning to far sight o bicyclist turning right while car going straight ahead Assumptions: Junction detail: at a junction Car manoeuvre: waiting to go ahead but held up, slowing or stopping, moving off, changing lane to near sight, changing lane to far sight, overtaking moving vehicle on offside, overtaking stationary vehicle on offside, overtaking on nearside, going ahead near sight bend, going ahead far sight bend, going ahead other waiting to turn far sight, turning far sight Bicycle manoeuvre: turning to far sight, waiting to turn to far sight Bicycle first point of impact: front, back, nearside, offside Casualties by accident scenario Table 3-14 shows the number of casualties in each of the accident scenarios. Table 3-14: Target population casualties split by accident scenario and injury severity ( ) Injury Severity Seriously Slightly Total Accident scenario Killed injured injured casualties % KSI 1. Car turning near sight (left) ,729 1,910 9% 2. Car turning far sight (right) ,180 3,702 14% 3. Longitudinal ,698 2,046 17% 4. Crossing ,578 5,391 15% 5. Cyclist turning far sight (right) ,133 19% Other ,253 4,990 15% Total 116 2,699 16,357 19,172 15% Other includes all those casualties not included in any of the scenarios described in section , this includes (but is not limited too) the following examples: bicyclist turning left accidents involving no impact to the bicycle longitudinal accidents where the bicycle was impacted on the front, nearside or offside head-on collisions - bicycle and other vehicle travelling from opposite directions 23/49

24 4 Comparison to the Netherlands When looking at the collision partners for bicycles, it can be found that they are most often injured either in single vehicle accidents (meaning without the involvement of any other traffic participant), in collisions with passenger cars, or in collisions with other VRU s. The distribution of collision partners for fatally and severely injured bicyclists in the Netherlands [9] is presented in Figure 4-1. Figure 4-1: Crash partners in bicyclist accidents for the Netherlands, BRON [9] Most data from the Netherlands ([2], [4], [5], [6], [8], [9], [10]) is based on BRON which is not as detailed as the German GIDAS data. In consequence, less detailed insight into the Dutch car to bicyclist accident configurations is available compared to the German situation. It should also be noted, that from the information within BRON it is not possible to see which movement was made by which crash partner. BRON data is openly accessible. Therefore, some parameters related to car to bicyclist accidents could be investigated on the newest complete dataset available which is currently dated from Some findings are presented in Figure 4-2 to Figure 4-4. From [5] (data from 2007), the following accident configurations could be identified: vehicle going straight forward, bicyclist is crossing the path also traveling straight forward (40%) one partner turned left while other went straight-on (12%), one partner crossed road laterally while other went straight-on (12%) one partner turned left while other was going straight-on for opposite direction (8%) A new analysis of 2009 BRON data provided the following configurations for car to bicyclist accidents: Side impact on crossing (26%) Other side impact (25%) Frontal without lane change (8%) Parked vehicle (7%) With respect to the boundary conditions of the accidents the following can be found: Accidents usually happen on intersections, with the bicyclist crossing a road (~65%). In more than 50% this is a side and not the main road. [4] 24/49

25 Most accidents happen during daytime (84% [4]). This is true for slight, severe and fatal accidents (Figure 4-2) Only 10% of all distances are cycled during night-time therefore the risk of being in an accident at night is higher than during day [4] Most of the fatal accident happened when it was dry (80%), severe and slight accidents only happened when it was dry in 52% and 45% of the cases, respectively. Hard gusts of wind or snow/hale are the most common impairments noted (Figure 4-2) A big number of accidents can be categorized as looked but failed to see [4] Lighting conditions fatal severe slight with injury all Weather conditions fatal severe slight with injury all daylight darkness twilight unknown dry rain fog snow / hale hard gusts of wind unknown Figure 4-2: Lighting and weather conditions from BRON (2009 data set) Road situation 639 fatal severe slight with injury straight curve roundabout crossing - 3 arms crossing - 4 arms straight road - separate lanes Figure 4-3: Road layout form BRON (2009 data set) Age of cyclist fatal severe slight with injury Age Age of car of cyclist driver fatal fatal severe severe slight slight with with injury injury Age fatal Figure 4-4: Age of bicyclist which is opponent 1 (left) or opponent 2 (right) Note: opponent 1 is the one registered by the police as main opponent; mainly the responsible partner. 25/49

26 5 Radar cross section measurements of dummies and humans 5.1 Introduction The aim of the measurements is comparing Radar Cross Section (RCS) signatures of pedestrian and cyclist dummies and pedestrian and cyclist humans in the 24 and 77 GHz frequency bands. For automotive radar two main frequency ranges can be considered, i.e. around 24 and 77 GHz. Table 1 below shows the radar frequencies that are being used in the automotive industry. Although in general both frequencies experience similar reflective behavior, i.e. geometrical reflection by objects, the large frequency difference and the resulting antenna beam change can cause significant differences in the specific reflective behavior, i.e. how good and at which point do objects reflect their energy. Radio propagation at 24 and 77 GHz is generally line-of-sight and multipath reflections. Obstruction in the optical path results in considerable attenuation. At 24 GHz, and for a distance of 100 meters, path loss is 100 db. At 76.5 GHz, and for a distance of 100 meters, path loss is 110 db. Table 1, Radar frequency allocation worldwide. 24 GHz NB 24 GHz UWB 26 GHz UWB 77 GHz 79 GHz UWB Worldwide - US/Canada - Japan - EU until US/Canada - Japan Worldwide - Singapore - EU EU: 200MHz [450MHz] US: 200MHz JP: 200MHz US: 7GHz JP+EU: 5GHz US: 1 GHz JP: 5 GHz 1 GHz JP: 500MHz 4 GHz This report describes the results of RCS signatures that are collected in the 24 and 77 GHz bands for a pedestrian and cyclist dummy and compared to a human pedestrian and cyclist. 26/49

27 5.2 Measurement method and setup Measurements have been performed in the RF Anechoic chamber in combination with the control room of TNO in The Hague, The Netherlands. Measurement targets consist of: - 4a pedestrian dummy SBV2 (central plate), presented in Figure 5-1 (left) - Human pedestrian, presented in Figure 5-1 (right) - 4a cyclist dummy SBV2 (central plate), with bicycle dummy BYV1, presented in Figure 5-2 (left) - Human cyclist on bicycle, presented in Figure 5-2 (right) Figure 5-1 4a pedestrian dummy SBV2 (left) and human pedestrian (right). Figure 5-2 (right). 4a cyclist SBV2 (central plate) and bicycle dummybyv1(left) and human cyclist on a bicycle The human pedestrian and cyclist is a male of average length (1.78m) and weight (73 Kg) dressed with average clothing (as displayed in the figure above). The cyclist dummy is based on the latest version of the pedestrian dummy. In this study, the cyclist and bicycle together are treated as one object. 27/49

28 In the measurement facility the targets are placed on a pedestal. The pedestal is fitted with a wooden plate with a diameter of about 50 cm. The load (targets) are centered on the pedestal. At a distance of 7 meter two antenna combinations are placed, one for each of the two measurement bands. The setup is shown in Figure 2.3. The distance of the antennas to the target are large enough to assume a far-field distance. The size of the targets cause the object to be in the near field. Figure 5-3 Antenna measurement setup for 24 GHz (left) and 77 GHz (right). Measurements are performed in a quasi-monostatic configuration (TX antenna and RX antenna on same side) with a single polarization using a Agilent PNA-67 GHz Network Analyzer with a maximum transmit power limited to +10 dbm. Targets are illuminated uniformly. Complex transmission coefficients (S21) are measured in the frequency domain and offline converted to RCS values. Measurements are performed at a full 360 degrees azimuth range at a constant rate of 0.9 deg/s. Time-gating and empty room compensation have been applied to minimize the effects of the anechoic chamber. In the 24 GHz band measurements are performed from GHz using two wide-beam leaky lens antennas and a LNA (1 GHz 26.5 GHz). In the 77 GHz band measurements are performed from GHz using two WR-10 horn antennas, a frequency extender ( GHz) and a LNA. To compute absolute RCS values from the measurements a reference corner reflector and reference plate reflector are also measured at the target location for the 24 and 77 GHz band, respectively. All target measurements have been performed at least twice to verify repeatability. RCS values are calculated as the frequency average RCS (in dbm2) in both 7 GHz bandwidth (24 GHz) and 5 GHz bandwidth (77 GHz) and additionally in a 500 MHz bandwidth for both bands. 28/49

29 5.3 RCS signatures Pedestrian 24 GHz Pedestrian dummy test 1 Pedestrian dummy test 2 Pedestrian human test 1 Pedestrian human test RCS (dbm 2 ) Azimuth angle (deg) Figure 5-4 Frequency averaged RCS signature for a pedestrian at 24 GHz (7 GHz bandwidth) Pedestrian dummy test 1 Pedestrian dummy test 2 Pedestrian human test 1 Pedestrian human test RCS (dbm 2 ) Azimuth (deg) Figure 5-5 Frequency averaged RCS signature for a pedestrian at 24 GHz (500 MHz bandwidth). 29/49

30 5.3.2 Pedestrian 77 GHz Pedestrian dummy test 1 Pedestrian dummy test 2 Pedestrian human test 1 Pedestrian human test RCS (dbm 2 ) Azimuth (deg.) Figure 5-6 Frequency averaged RCS signature for a pedestrian at 77 GHz (5 GHz bandwidth) Pedestrian dummy test 1 Pedestrian dummy test 2 Pedestrian human test 1 Pedestrian human test RCS (dbm 2 ) Azimuth (deg.) Figure 5-7 Frequency averaged RCS signature for a pedestrian at 77 GHz (500 MHz bandwidth). 30/49

31 5.3.3 Cyclist 24 GHz Cyclist dummy test 1 Cyclist dummy test 2 Cyclist human test 1 Cyclist human test RCS (dbm 2 ) Azimuth (deg.) Figure 5-8 Frequency averaged RCS signature for a cyclist at 24 GHz (7 GHz bandwidth) Cyclist dummy test 1 Cyclist dummy test 2 Cyclist human test 1 Cyclist human test RCS (dbm 2 ) Azimuth (deg.) Figure 5-9 Frequency averaged RCS signature for a cyclist at 24 GHz (500 MHz bandwidth). 31/49

32 5.3.4 Cyclist 77 GHz Cyclist dummy test 1 Cyclist dummy test 2 Cyclist human test 1 Cyclist human test RCS (dbm 2 ) Azimuth (deg.) Figure 5-10 Frequency averaged RCS signature for a cyclist at 77 GHz (5 GHz bandwidth) Cyclist dummy test 1 Cyclist dummy test 2 Cyclist human test 1 Cyclist human test RCS (dbm 2 ) Azimuth (deg.) Figure 5-11 Frequency averaged RCS signature for a cyclist at 77 GHz (500 MHz bandwidth). 32/49

33 5.4 Comparison RCS signatures Pedestrian - In both frequency bands the pedestrian dummy shows strong directional behaviour with dominant peaks at 90 and 270 degrees, representing the sides of the pedestrian dummy. The human pedestrian shows much more constant behaviour, without these clear peaks. The cause of the directional behaviour in the dummy is most likely due to a plate inside the dummy. - In the 24 GHz band using a measurement bandwidth of 7 GHz the differences between the pedestrian dummy and pedestrian human are clearly visible and vary roughly between 5 10 db. - In the 24 GHz band using a measurement bandwidth of 500 MHz, the differences between the pedestrian dummy and pedestrian human are also seen, but due to the smaller bandwidth larger fluctuations are visible. - In the 77 GHz band using a measurement bandwidth of 5 GHz the differences between the pedestrian dummy and pedestrian human are clearly visible at 90 and 270 degrees. The values of the dummy are roughly 5 10 db higher close to these angles. At other angles the differences are much smaller, but in the case of the human larger fluctuations are visible. - In the 77 GHz band using a measurement bandwidth of 500 MHz a similar behaviour can be observed compared to the larger bandwidth, but due to the smaller bandwidth larger fluctuations are visible. Cyclist on bicycle - In both frequency bands the cyclist with bicycle dummy and human cyclist on bicycle show directional behaviour with dominant peaks close to 90 and 270 degrees, representing the side of the cyclist on the bicycle. - In general the dummy shows about 5 db lower values compared to the human cyclist on the bicycle, except for the dominant peaks visible at 90 and 270 degrees that are about 10 db larger than the human cyclist on the bicycle. The response from the front and back of the cyclist on the bicycle matches quite well. - In the 24 GHz bands the differences between the cyclist on the bicycle dummy and cyclist human on the bicycle are smallest for the front view of the cyclist on the bicycle. Using a measurement bandwidth of 500 MHz a similar behaviour can be observed compared to the larger bandwidth, but due to the smaller bandwidth larger fluctuations are visible. - In the 77 GHz bands the differences between the cyclist on the bicycle dummy and cyclist human on the bicycle are smallest for both the front and back view of the cyclist on the bicycle. Using a measurement bandwidth of 500 MHz a similar behaviour can be observed compared to the larger bandwidth, but due to the smaller bandwidth larger fluctuations are visible. In general large RCS differences up to 10 db are observed in both frequency domains for both the pedestrian and cyclist, which means improvements are needed to have a realistic dummy RCS profile for both the pedestrian and the cyclist on the bicycle. An update of the pedestrian dummy (V3.2) has some changes in the inner side of the dummy to obtain a more realistic RCS reflectivity. A plastic plate is replaced by tubes in body, legs and arms. Based on measurements done so far (by 4a) this results in a bit higher RCS response without sharp peaks at 90 and 270 degrees. More detailed measurements are needed to confirm these improvements. For the cyclist, additional tests are needed to distinguish the observed differences, e.g. isolated measurements of the bicycle and bicycle dummy to find out what causes the main RCS difference in order to be able to judge possible improvement of bicycle dummy. The measurement method used for the comparison as described in this deliverable differs from the approach used in the sensor workshop [Dummy workshop April Fohnsdorf]. The main differences are: (this method versus sensor workshop) - Fixed distance 7 meter versus variable distance (50 to 0 meters) - Measured in isolated laboratory conditions versus measured on street (non-isolated environment; in which reflections from environment and ground play a role) 33/49

34 - Full 360 degree approach versus measurements at fixed angles (0,45,90,..315 degree) - Measured in two frequency bands (24 and 77 GHz) at different bandwidths with sensors designed for research purposes versus commercial on the market sensor systems (77 GHz). 34/49

35 6 Preliminary Test Scenarios In this chapter will be described which preliminary test scenarios are deduced from the accident scenarios. In ASPECSS the main focus was on pedestrians, which resulted In a thorough research on accidentology and test scenario definition. The test scenarios were based on a frontal looking pedestrian AEB system. For bicyclist the study was less detailed and the intervention and interaction system (whether it is AEB with frontal looking only or whether it should also take into account a blind spot scenario) was not clearly defined yet. This resulted in a less detailed accident analysis and a preliminary set of test scenarios. The analysis executed in the ASPECSS project form a good starting point and the preliminary scenario set will be worked out into more detail in the new project CATS (Project lead by TNO,joined by the automotive industry, on the topic Cyclist AEB test setup. The project will come up with a test protocol proposal 2016 for Euro NCAP ). 6.1 Differences between bicyclists and pedestrians Both, bicyclists as well as pedestrians are categorized as vulnerable road users in accidents with motorized traffic. Additionally, in almost all European countries the number of fatal pedestrian accidents is much higher than the number of fatal bicyclist accidents. This often leads to the assumption, that looking at pedestrian accidents automatically also sufficiently covers bicyclists. Cyclists do however differ in quite some aspects from pedestrians and hence should be regarded separately. Where pedestrians move relatively slow with velocities between roughly 3 and 8 km/h, bicyclists are much faster and can reach speeds of up to 50 km/h (race bike). Pedestrians always are animated as they need to move their legs and will most likely also move their arms while walking. A bicycle is able to move by itself once a certain speed is established, hence the bicyclist might be sitting statically on the bicycle without any need to move during the pre-collision phase. The size of a pedestrian can vary very significantly depending on the age and stature of the pedestrian itself. The size of bicyclists of course also varies, however less significant as for pedestrians as bicyclists are sitting on their bicycle and the difference in sitting height is less prominent than differences in standing height. On the other hand, the posture of a bicyclist might differ more depending on the bicycle used compared to the usually quite simple upright and standing posture of a pedestrian. Where a standard Dutch city bike will put its occupant in an upright seating position, race bicycle users can be inclined into an almost lying position. Where pedestrians hit their head on a car upon impact in most cases on the bonnet or the lower part of the windscreen, bicyclists tend to hit higher [5]. Looking at the accident scenarios it can be seen, that pedestrians are mainly injured crossing the road by a straight forward moving passenger car. A significant share of these pedestrians was obstructed and a significant share of fatal accidents occured during dark light conditions. For bicyclist-to-car accidents, crossing accidents with both opponents traveling straight forward are also very common. Situations where the car hits the bicyclist while turning either to the right or to the left are however also very important to consider. Longitudinal accidents with both, car as well as bicyclists traveling in the same direction are quite common in the UK (and other EU countries), however less prominent in the Netherlands [14]. The importance of night testing and (partially) obstruction of the bicyclist needs further investigation. 6.2 General boundary conditions for car-to-bicyclist accidents Information on bicyclist-to-car accidents has been made public during the last few years also for countries other than the UK and the Netherlands. For Germany for example lots of work has been done using the German in depth database GIDAS. Though the level of detail varies for the different countries based on the 35/49

36 different data sources and means of data collection, some findings are found throughout all data that seem to be true for most European countries. It was found, that in general most car to bicyclist accidents happen: On dry roads During day time Under good visibility and with good weather Not in the winter o Possible explanation for all 4 points: much more people use their bicycle when the weather is nice, temperature is not too low and visibility is good. On urban roads (in general >70%) On crossings and T-junctions followed by straights With relatively low collision speed Because either the car or the bicyclist denies right of way or due to incorrect expectations about the behaviour of the collision partner Furthermore it was observed, that head-on or rear-end collisions (for the bicyclist) are usually more severe compared to impacts to the side of the bicycle. This might be related to higher impact speeds and different impact kinematics that are also usually found for such accidents. Also, the share of rear impacts is higher on rural roads than on urban roads. The share of these accidents is relatively small in the Netherlands compared to other countries like Sweden, UK or Germany. This might be explainable by the generally good bicyclist infrastructure in the Netherlands where often a separate bicyclist lane is provided hence separating cars and bicyclists traveling in the same direction. The gender distribution of injured bicyclists is approximately 50/50 for the Netherlands whereas in the UK the proportion is almost 80/ Preliminary bicyclist test scenarios The following test scenarios are proposed based on information not only available in the UK and the Netherlands, but also from countries as Germany or Sweden as investigated by TNO for the Dutch Ministry of Transport and Environment [14]. The scenarios are defined based on accident configurations, not taking into account yet whether or not they might be addressable by current or near future safety systems to be implemented into passenger cars or currently available t est tools. It should additionally be noted, that accident research shows, that bicyclists are not solely injured in accidents involving a passenger car as shown in Figure 4-1 for the Netherlands. A significant amount of bicyclists get injured in accidents involving no other crash partner or involving a crash partner other than a passenger car. Such accidents are neglected for the scenario definition. Additionally, usually only information is available on actual accidents, but not on critical traffic situations that result in near-misses. The most critical near-miss scenarios would not necessarily need to be compliant with the most important accident scenarios. Adding such information in the future might therefore result in the need to revise the test scenarios described in this report. 36/49

37 The following test scenarios are proposed for vehicle based active bicyclist safety systems (see also Figure 6-1): 1.) City Crossing (car and bicycle moving straight forward) 2.) City turning left (car turning of the road to the left with bicyclist moving straight forward) 3.) City turning right (car turning of the road to the right with bicyclist moving straight forward) 4.) Inter-Urban Longitudinal (Car traveling straight forward hitting the bicyclist driving straight forward from behind Figure 6-1: Preliminary bicyclist scenarios from left to right: City Crossing, City turning left, City turning right, Inter-Urban Longitudinal Scenario 1 and 4 are typical Scenarios with a frontal looking AEB System, whereas the system intervention to address scenario 2 and 3 is not defined yet, hence there are no state of the art Systems on the marked for cars to address these scenarios. 6.4 General proposed test conditions Though a detailed definition of all test parameters cannot be made yet, the following general test conditions are proposed for the four scenarios presented in section 6.3: Test target: Dummy requirements need to be set according to the specifications of the tested system. Depending on the philosophy followed, the requirements can either be set to challenge the system / algorithm or to make the dummy easily recognizable. In line with current proposals for AEB car to car as well as car to pedestrian tests it is advised to choose a dummy that is easily recognizable for all systems instead of one that is designed to challenge it. The following general attributes should be taken into account: Crash-forgivingness Repeatability and Reproducibility in positioning and (if required) movement Realistic representation of characteristics needed for sensing system used by car Ideally, the pedestrian dummy that will be used for Euro NCAP AEB VRU testing in phase 1 should be positioned as bicyclist during the tests. This dummy has widely been investigated by different parties with respect to different characteristics such as realistic radar cross section (RCS), contrast issues or crushability and is likely to be widely accepted once the latest updates are finished. Based on the current status of the dummy, it will be non-articulated and simply be positioned in a cycling position on a dummy-bicycle. Articulation of the bicyclist dummy will likely need to include more sensitive joints along with sensitive components as a small motor or mechanics to make the joints move in a proper 37/49

38 manner. Such components are very likely to decrease the crushability of the dummy significantly and therefore are not taken into account for the time being. Limited information is available form accident data on the type of the bicycle. From [5]: An average city bike is most often involved in accidents in the NL. An example for such a bicycle is provided in Figure 6-2. For the definition of a dummy bicycle, several parameters such as frame set-up, saddle and steering wheel height RCS, or presence of reflective material need to be defined. Figure 6-2: Example of a typical Dutch city bike To come to a more detailed specification for the complete bicyclist and bicycle dummy more research is needed. Test speeds: It was found, that the majority of bicyclist to car accidents happen mainly on urban roads where the speed is usually limited to 50 km/h. As can be seen from Figure 6-3, a significant amount of accidents do happen also at speeds above 50 km/h, though the 75th percentile is usually covered by speeds up to 30 km/h. As injury risk for a vulnerable road user usually increases significantly with increasing vehicle speed, it is proposed to include vehicle speeds above 30 km/h for the test set up proposal. Also, in line with the test proposal currently under discussion by Euro NCAP for pedestrians, an incremental speed approach is proposed to cover a wide range of vehicle speeds. 38/49

39 Figure 6-3: German vehicle speed distribution as found from [7] for different accident configurations The following vehicle speeds are proposed: - 10 to 50 km/h for the crossing scenario - 10 to 40 km/h for both turning scenarios - 30 to 80 km/h for the longitudinal scenario 10 and 30 km/h approximately reflect the 25th percentile vehicle speed as presented in [7]. The general city speed limit of 50 km/h is chosen as upper speed limit for the crossing scenario. This upper limit is reduced to 40 km/h for the turning scenarios as turning vehicles usually travel at a reduced speed compared to vehicles traveling straight forward. The longitudinal scenario has an increased share of accidents on rural roads where car speeds are usually higher compared to within city limits. This is also reflected in the figures presented in [7] (see Figure 6-3). Therefore, car speeds are adapted to 30 to 80 km/h for this scenario only. Were bicyclist speeds are provided in literature, they are usually located around 15 km/h [12]. Hence, this speed is chosen for all of the proposed test scenarios as well. It is however believed, that bicyclists (at least in the Netherlands) usually cycle faster [13]. A speed of 20 km/h is therefore considered more realistic and might be considered as either second step for the scenarios or even as additional test speed. To confirm this value, further research is needed. Environmental conditions: As shown in section and section 4, the majority of bicyclist to car accidents happen during day time, good visibility and good weather on dry roads. Therefore, no dark or rain configuration is proposed for the crossing and turning scenarios. Night time accidents do however happen and should not be neglected completely. In line with currently considered future pedestrian AEB tests, a limited set of night time tests should be considered. It is proposed, to only consider night time tests are for the longitudinal scenario. Looking at rear end impacts (for bicyclists), it can be found that 66% of these accidents in Germany happen during daytime [7]. This leaves a share of 34% that happen under impaired lighting conditions. According to Stone [11] in the UK, approximately half of the bicyclists having an accident on a non-junction 60 mph road that happen in the dark have a rear impact. From recent BRON data it can be found, that approximately 50% of fatal rear-end car to bicyclist crashed where the first opponent was a car happen during daytime. Looking at severe and slight injury accidents, the numbers 39/49

40 change to 73% and 59%, respectively. It should however be noted, though, that the data sample for fatal accidents was very small. For 2009, the number of fatal accidents that fall under the investigated accident category was for example only 3. For this reason, it is proposed, to conduct the longitudinal test scenario not only under daytime, but also under (still to be defined) night-time conditions. 40/49

41 7 Benefit estimate based on previous studies 7.1 UK study on cycle helmets based on literature review Executive summary from [3]: Introduction: In 2008, 115 bicyclists were killed and 2,450 reported as seriously injured on Britain s roads, accounting for 9% of all killed or seriously injured (KSI) road casualties (Department for Transport, 2009). Approximately 40% of bicyclists admitted to hospital in England suffer head injuries. Cycle helmets are designed to reduce head injuries by absorbing the energy during a head impact and distributing the load. This is intended to reduce the risk of scalp laceration, cranium fracture, and severe brain injury. Cycle helmet wearing rates have increased steadily since 1994 for most bicyclist groups and in 2008 they were 34% on major roads and 17% on minor roads, up from 22% on major roads and from 8% on minor roads in This research report was commissioned to provide a comprehensive review of the effectiveness of cycle helmets in the event of an on-road accident, building on previous work undertaken for the Department for Transport (Towner et al., 2002). The objectives were to evaluate the effectiveness of cycle helmets form several perspectives: Review of cycle helmet testing and Standards (including nature and severity of head injuries that cycle helmets are designed to protect against; predicted benefit in those types of test conditions); A biomechanical assessment of the potential limitations to helmet effectiveness; A literature review of helmet effectiveness from real-world studies; and An in-depth accident data investigation to identify the potential for cycle helmets to prevent injury. This report focuses on understanding whether cycle helmets reduce the frequency and severity of injury in the event of a collision. It does not include detailed consideration of whether wearing (or not wearing) a helmet influences the likelihood of being involved in an accident, either through behavior in the rider or in other road users. Cycle Helmet Testing In most jurisdictions, cycle helmets are tested to ensure a minimum level of performance for a range of criteria that affect safety. Typically these include: Construction requirements; Impact test requirements; Retention system (Strap) strength and helmet stability; Definition of the minimum area of the head covered by the helmet; and Definition of a minimum field of view (to ensure that the helmet does not impede the vision of the wearer). Most cycle helmet standards around the world define similar types of impact test but the impact severity, pass/fail criteria and number of tests per helmet vary in different standards. This means that helmets certified to one standard may not pass the requirements of another. In addition, cycle helmet standards have changed over time and so current helmets in the UK may be quite different to those sold in other regions or in previous 41/49

42 decades. The results of real-world cycle helmet effectiveness studies must be considered in the context of these regional and temporal differences in cycle helmet standards. It was found that cycle helmets designed to the Standards currently used in the UK (EN 1078 for child and adult helmets and EN 1080 for younger child helmets) would, based on biomechanical principals, be expected to be effective in manycycle accident conditions. This effectiveness would depend on a range of factors, such as the type of accident (e.g. fall from a cycle or a collision with another vehicle), the stature and injury tolerance of the rider, and the shape and stiffness of the object struck by the head (e.g. flat road surface, a kerb, or a deformable car bonnet). Potential limitations to effectiveness The report explored a number of claims that cycle helmets may make a head injury worse than if no helmet had been worn. The report finds that a helmeted head can fall at least four times as far for the same risk of injury as an un-helmeted head, within the range to which cycle helmets are tested. There have also been concerns expressed in the literature that cycle helmets may not be effective at reducing injuries due to rotation of the head, or even that they may make such injuries worse. There are no cycle helmet standard tests for performance in rotational loading conditions. Nevertheless, no evidence was found for an increased risk of rotational head injury with a helmet compared to without a helmet. Literature Review of Cycle Helmet Effectiveness This report considered in detail the published literature on the effectiveness of cycle helmets, updating the previous review reported by Towner et al. (2002). Most of the published research into helmet effectiveness attempts to determine whether the protective effect of helmets is sufficient to affect casualty outcomes in real accidents. There are two primary forms of study into cycle helmet effectiveness: Hospital admission studies; and Population studies. The majority of hospital admission studies use a case-control design. This design matches helmeted bicyclists with unhelmeted bicyclists and attempts to discern different injury outcomes from the data that are attributable to the helmet. Population-based studies typically consider aggregate national statistics on cycle accidents and tend to be longitudinal. They compare the trend in bicyclist head injuries with the expected trend were helmets to offer a protective effect. The accurate assessment of the effectiveness of cycle helmets requires detailed and comparable data. Many of the studies reported in the literature suffered from some shortcomings in the data, examples included: No data being available on the accident characteristics leading to the head injury (e.g. the speed of an impacting vehicle and the first point of contact with the bicyclist); A lack of detailed data on the type of head injury (was it an injury a helmet could have prevented?); and The level of data reported in most of the studies reviewed [Towner et al. 2002] being aggregated to a point where it was not possible to reinterpret it to answer criticisms of study design or analysis from the published papers. Overall, there appears to be a clear difference between hospital-based studies, which tend to show a significant protective effect from cycle helmets, and population studies, which tend to show a lower, or no, effect. This is likely to be due to the difficulties in adequately controlling for confounding variables, as well as limitations regarding how representative the bicyclists are in the samples used compared with the whole cycling population. 42/49

43 Furthermore, cycle helmet designs have changed over time and it is difficult to interpret effectiveness measures from other regions in terms of cycle helmets currently on sale in the UK. As a result, it was not possible to quantify the amount of benefit offered by modern cycle helmets in the UK from the literature alone. Evidence from In-depth Accident Studies In-depth accident data were used to investigate the extent and nature of the head injuries sustained by bicyclists, which were then correlated as far as practical with the accident circumstances. In conjunction with consideration of the biomechanics of head injury and the mechanics of helmeted head impacts, this information was used to predict the potential effectiveness of cycle helmets at mitigating or preventing a proportion of the more severe types of head injury, i.e. cranium fractures and/or intracranial injury. The accident databases used were: The Hospital Episode Statistics (HES) database for England (1999 to 2005); and Police fatal file derived bicyclist database (2001 to 2006) The HES dataset contains detailed information regarding the injuries sustained but only cursory information with respect to the nature of the accident. Whereas the police fatal files provided full reconstruction evidence and allowed in most cases the cause of the head injury to be evaluated by expert assessment. Thus, an expert judgment could be made for each fatal case as to the likely potential effect a cycle helmet would have had, if worn. Therefore, the methods used and the subsequent confidence attributed to the predictions of the potential effectiveness of cycle helmets for fatalities (fatal file) and seriously injured casualties (HES database) vary. For the HES data it was not possible to state categorically the proportion of casualties which would have been prevented if all had worn cycle helmets, rather a target population was identified, or the proportion of casualties for whom a cycle helmet could have been beneficial. An in-depth review of the head injuries suffered by bicyclists who were admitted to hospital in England identified that 10% sustained serious cranium fracture and/or intracranial injuries. The majority of this group (7% of the total) only sustained these injuries and had no other head or other body region trauma. Therefore, if cycle helmets had been worn, a proportion of this 7% may not have required hospital treatment at all. A further 20% of the HES casualties suffered open wounds to the head. However, no further details regarding the location of the injury were known and therefore it was not possible to quantify how many of these may have been mitigated or prevented if a helmet had been worn. A forensic case by case review of over 100 British police bicyclist fatality reports highlighted that between 10 and 16% of the fatalities reviewed could have been prevented if they had worn a cycle helmet. This predictive analysis was undertaken by biomechanical and vehicle safety experts who excluded cases where: The cause of death was not associated with head injury; and Where the causes of the head injuries were in excess of the potential benefit a helmet could have afforded. There are limitations associated with the predictive approaches undertaken by this type of study, so conservative estimates of helmet effectiveness were assumed for different accident scenarios (10% to 50%). Further, the police fatal files reviewed were biased towards London and therefore the percentage benefit is only indicative of a national estimate. It is likely to be a conservative estimate because of the higher proportion of accidents involving large goods vehicles and crush related injuries in London compared to the national situation and fewer single vehicle accidents. 43/49

44 However, the sample contained a wide range of fatalities with respect to age, gender and accident typology. Summary of conclusions: This report focuses on understanding whether cycle helmets reduce the frequency and severity of head injury in the event of a collision. Assuming that they are a good fit and worn correctly, cycle helmets should be effective at reducing the risk of head injury, in particular cranium fracture, scalp injury and intracranial (brain) injury. Cycle helmets would be expected to be effective in a range of accident conditions, particularly: o The most common accidents that do not involve a collision with another vehicle, often simple falls or tumbles over the handlebars; and also o When the mechanism of injury involves another vehicle glancing the bicyclist or tipping them over causing their head to strike the ground A specialist biomechanical assessment of over 100 police forensic bicyclist fatality reports, predicted that between 10% and 16% of the fatalities could have been prevented if they had worn an appropriate cycle helmet Of the on-road serious bicyclist casualties admitted to hospital in England (HES database): o 10% suffered injuries of a type and to a part of the head that a cycle helmet may have mitigated or prevented; and further o 20% suffered open wounds to the head, some of which are likely to have been to a part of the head that a cycle helmet may have mitigated or prevented. Cycle helmets would be expected to be particularly effective for children, because: o The European Standard (EN 1078) impact test and requirements are the same for adult and o child helmets, both use a 1.5 m drop height test; and so Given that younger children are shorter than older children and adults, their head height would be within the drop height used in impact tests so a greater proportion of single-vehicle accidents are likely to be covered by the Standard for children. No evidence was found for an increased risk of rotational head injury with a helmet compared to without a helmet. In the literature reviewed, there is a difference between hospital-based studies, which tend to show a significant protective effect form cycle helmets, and population based studies, which tend to show a lower, or no, effect. Some of the reasons behind this were due to: o o The lack of appropriateness of the control groups used; and Limitations in the available data, such as knowledge of helmet use and type of head injury. 7.2 AEB systems and VRU airbag based on SaveCAP Within the SaveCAP project, an estimation of the benefit for VRU AEB, VRU airbag as well as a combination of both systems was provided. The results have been presented by 0 (see Figure 7-1). For this study, the aim was to estimate the potential to save bicyclists as well as pedestrian from severe (AIS3+) head injury looking first separately into auto-brake and passive deployable systems and second into a combination of both systems. From GIDAS the following sample sizes could be evaluated: VRU airbag: 54/52 pedestrians / bicyclists with AIS3+ head / face injuries Auto-brake: 431/391 pedestrians / bicyclists from GIDAS PCM with all injury levels Combined system: 11/35 bicyclists / pedestrians with data available from both sources It was concluded, that a combined system would potentially be able to reduce AIS3+ injuries by up to 70% which is a much higher potential benefit than can be expected for either system implemented without the other. 44/49

45 Figure 7-1: Effectiveness of Airbag, Braking and combination of both based on GIDAS data 0 45/49

46 8 Risk Register Risk No. What is the risk Level of risk 2 WP8.1 The cyclist accidentology analysis is based the on previous data, mainly obtained from level the UK, NL, D. WP8.2 WP8.3 In this study preliminary test scenarios are defined, based on the available data. Actually the intervention system is not yet defined, which makes it hard to define the actual test scenarios In the RCS measurements not the latest version of the pedestrian dummy was used. This was done in order to have the pedestrian dummy in line with the cyclist dummy for which no comparable upgrade was available yet. Further the cyclist and bicyclist dummy were measured as one component, which made it hard to discriminate their individual performance. Solutions to overcome the risk A more detailed study including data of other EU countries is needed to cover the accidentology of all EU countries. The bring the level of research of the carcyclist accident on a same level as done for the pedestrians) More accidentology data, real life measurements and definition of the actual intervention system is needed to define a test scenarios set (of a comparable level to the ones defined for the pedestrian in this project) Measurements of the newest dummies and of a single bicycle dummy and single real bicycle would complement the outcome of this research. 2 Risk level: 1 = high risk, 2 = medium risk, 3 = Low risk 46/49

47 9 Conclusions and Recommendations Within this report, the accident situation for car-to-bicyclist accidents was described in detail for the UK and compared to data from the Netherlands. It could be seen, that though both, pedestrians as well as bicyclists are counted amongst the group of vulnerable road users and are therefore often regarded together, there are differences between these two groups that should not be neglected. These differences were summarized and based on this data enriched with findings from other studies a proposal was made for bicyclist test scenarios to be able to include bicyclists into the tests and methodology developed in other tasks of WP1 and WP2. As this subtask did not allow for expensive reevaluation of accident data and analysis similar to what has been done for pedestrians throughout WP1, the scenarios and their parameters are not as mature yet as for the proposed pedestrian tests. The following additional work is recommended for future projects to further develop and finalize the bicyclist test scenarios: Further (detailed) investigation of bicyclist accident scenarios for countries other than UK and NL Further investigation of accident and scenario parameters (overlap, obstructions, road markings, test speeds, etc.) Further investigation of lighting conditions for night testing Development of bicyclist target based on existing pedestrian test target Adoption of test tools for bicyclist testing to accommodate turning and lateral scenarios Development of bicyclist test protocol preferably based on existing and widely accepted pedestrian test protocol Actual testing of proposed test scenarios and further refinement of test procedure 47/49

48 10 Literature [1] Fredriksson, R., Rosén, E., Potential of protection systems for vulnerable road users, ICSC 2012 [2] Hair, S. de, Versmissen, T., Schijndel, M. van, Samenwerken aan Fietsveiligheid NVVC 2012 [3] Hynd, D., Cuerden, R., Reid, S., Adams, S., The potential for cycle helmets to prevent injury a review of the evidence, Published TRL Project Report PPR446, 2009 [4] Reurings MCB, Dijkstra A, Twisk D, Vlakveld WP, Van fietsongeval tot maatregel; kennis en hiaten Leidschendam Stichting Wetenschappelijk Onderzoek Verkeersveiligheid (SWOV); [5] Rodarius, C., Mordaka, J., Versmissen, T., Bicycle safety in bicycle to car accidents, TNO report TNO-033- HM , 2008 [6] Schijndel, M. van, Hair, S. de, Rodarius, C., Fredriksson, R. Cyclist kinematics in car impacts reconstructed in simulations and full scale testing with Polar dummy, Proceedings of IRCOBI Conference, pp , 2012 [7] Helmer, Vehicle Cyclist accidents in Germany Current trends and comparison to Vehicle Pedestrian accidents, Crash.Tech Conference, 2012 [8] Schijndel, M van, et.al pre-development of vulnerable road users (VRU) airbag. Phase 1: preparation, TNO report no.: TNO-033-HM /1P, 2010 [9] Schijndel, M. van, Cyclist safety; a Dutch perspective, ActiveTest Workshop Vulnerable Road Useres, 2012 [10] Stipdong, H., Reurings, M., The effect on road safety of a modal shift from car to bicycle, Traffic Injury Prevention, 13: , 2012 [11] Stone, M., Broughton, J., Getting off your bike: cycling accidents in Great Britain in , Accident Analysis and Prevention 35, p , 2003 [12] Frederiksson R, Rosén, E., Priorities for Bicyclist Protection in Car Impacts a Real life Study of Severe Injuries and Car Sources, IRCOBI (International Research Council On the Biomechanics of Impact) Conference Dublin 2012 [13] Hendriksen, I., Engbers, L., Elektrische fiets heeft toekomst, Fietsverkeer [14] Rodarius, C., de Hair, S., Op den Camp, O., Proof-of-Concept test system for in-vehicle active cyclist safety systems (development and demonstration) TNO report, Dec [15] 48/49

49 11 Acknowledgment This project is co-funded by the 7th FP (Seventh Framework Programme) of the EC - European Commission DG Research Disclaimer The FP7 project has been made possible by a financial contribution by the European Commission under Framework Programme 7. The ation as provided reflects only the authors view. Every effort has been made to ensure complete and accurate information concerning this document. However, the author(s) and members of the consortium cannot be held legally responsible for any mistake in printing or faulty instructions. The authors and consortium members retrieve the right not to be responsible for the topicality, correctness, completeness or quality of the information provided. Liability claims regarding damage caused by the use of any information provided, including any kind of information that is incomplete or incorrect, will therefore be rejected. The information contained on this website is based on author s experience and on information received from the project partners. 49/49