Assessment of Integrated Pedestrian Protection Systems with Autonomous Emergency Braking (AEB) and Passive Safety Components

Abstract

Objective: Autonomous emergency braking (AEB) systems fitted to cars for pedestrians have been predicted to offer substantial benefit. On this basis, consumer rating programs—for example, the European New Car Assessment Programme (Euro NCAP)—are developing rating schemes to encourage fitment of these systems. One of the questions that needs to be answered to do this fully is how the assessment of the speed reduction offered by the AEB is integrated with the current assessment of the passive safety for mitigation of pedestrian injury. Ideally, this should be done on a benefit-related basis.

The objective of this research was to develop a benefit-based methodology for assessment of integrated pedestrian protection systems with AEB and passive safety components. The method should include weighting procedures to ensure that it represents injury patterns from accident data and replicates an independently estimated benefit of AEB.

Methods: A methodology has been developed to calculate the expected societal cost of pedestrian injuries, assuming that all pedestrians in the target population (i.e., pedestrians impacted by the front of a passenger car) are impacted by the car being assessed, taking into account the impact speed reduction offered by the car's AEB (if fitted) and the passive safety protection offered by the car's frontal structure. For rating purposes, the cost for the assessed car is normalized by comparing it to the cost calculated for a reference car.

The speed reductions measured in AEB tests are used to determine the speed at which each pedestrian in the target population will be impacted. Injury probabilities for each impact are then calculated using the results from Euro NCAP pedestrian impactor tests and injury risk curves. These injury probabilities are converted into cost using “harm”-type costs for the body regions tested. These costs are weighted and summed. Weighting factors were determined using accident data from Germany and Great Britain and an independently estimated AEB benefit. German and Great Britain versions of the methodology are available. The methodology was used to assess cars with good, average, and poor Euro NCAP pedestrian ratings, in combination with a current AEB system. The fitment of a hypothetical A-pillar airbag was also investigated.

Results: It was found that the decrease in casualty injury cost achieved by fitting an AEB system was approximately equivalent to that achieved by increasing the passive safety rating from poor to average. Because the assessment was influenced strongly by the level of head protection offered in the scuttle and windscreen area, a hypothetical A-pillar airbag showed high potential to reduce overall casualty cost.

Conclusions: A benefit-based methodology for assessment of integrated pedestrian protection systems with AEB has been developed and tested. It uses input from AEB tests and Euro NCAP passive safety tests to give an integrated assessment of the system performance, which includes consideration of effects such as the change in head impact location caused by the impact speed reduction given by the AEB.

Keywords:

• pedestrian
• active safety
• crash avoidance
• brake assist
• AEB

Introduction

Annually in the European Union (EU) about 6,500 pedestrians are killed and 156,000 are injured in road traffic accidents (European Commission 2013). Around half of these are hit by the front of a car. This article reports and expands on work performed as part of a European Commission 7th framework project, called Assessment methodologies for forward looking integrated Pedestrian and further extension to Cyclists Safety Systems (AsPeCSS) (ASPECSS 2014). The main aim of this project was to contribute toward improving the protection of vulnerable road users by developing harmonized test and assessment procedures for forward-looking integrated pedestrian safety systems that incorporate both autonomous emergency braking (AEB) and passive safety systems.

AEB systems for pedestrians have been predicted to offer substantial benefit (Edwards et al. 2014; Rosén et al. 2010; Searson et al. 2014). Rosén et al. (2010) estimated an effectiveness of 44 and 33% for prevention of fatal and severe injuries, respectively, for pedestrian casualties impacted by the front of a car. An AEB system that activates the brakes one second prior to a predicted impact with a maximum deceleration of 6 m/s² was assumed. Searson et al. (2014) estimated that the fitment of a pedestrian AEB system would improve a car's European New Car Assessment Programme (Euro NCAP) performance rating by approximately one color band, which corresponds to an increase in the score of 25% of the maximum. Edwards et al. (2014) estimated an annual benefit for the EU from about €1 billion to €3.5 billion depending on the type of AEB and assuming that it was fitted to all cars.

On this basis, consumer rating programs—for example, Euro NCAP—are developing rating schemes to encourage fitment of these systems (Euro NCAP 2014). One of the questions that needs to be answered to do this fully is how the assessment of the speed reduction offered by the AEB can be integrated with the current assessment of the passive safety for further mitigation of pedestrian injury or accident avoidance. Ideally, this should be done on a benefit-related basis.

Lubbe et al. (2012) discussed the elements of existing safety assessment methods relevant for a new integrated assessment, such as Hamacher et al. (2011) and Hutchinson et al. (2012). Lubbe et al. (2012) recommended that an integrated methodology should take into account the effect that the active system has on the boundary conditions for the passive system; for example, the change in head impact location resulting from the speed reduction given by the AEB. They also recommended validation and calibration against real-world data.

The aim of the work reported was to develop a methodology to give an overall assessment of a car's pedestrian protection on a benefit basis, using the results of pedestrian AEB tests and the standard impactor tests within Euro NCAP. Weighting procedures were developed to ensure accurate representation of injury patterns observed in accident data and the independently estimated AEB benefit for a car representative of the average fleet in the accident data.

Methodology

Concept

In principle, the cost of expected injury was calculated, assuming that all casualties in the target population were impacted by the car being assessed and taking into account the impact speed reduction offered by the car's AEB (if fitted) and the passive safety protection offered by the car's frontal structure. This cost can be normalized by comparing it to the cost calculated for selected cars. It should be noted that different versions of the methodology were developed for Germany and Great Britain because of links of the methodology to the accident data and differences between the German and Great Britain accident data.

The methodology consists of five main steps as described below and illustrated in Figure A1 (see online supplement).

• Step 1: Active safety testing: Exposure–impact velocity curve shift. The exposure–impact velocity curve for the pedestrian casualty target population (i.e., pedestrians impacted by a car front) was adjusted to account for the impact speed reduction provided by the AEB system.

• Step 2: Passive safety testing: Impactor measurement and extrapolation. Following the Euro NCAP protocols, standard impactor tests and simulations were conducted at the nominal speed, which approximately represents a 40 km/h pedestrian impact. Injury criteria values recorded were extrapolated to all other speeds experienced by the pedestrian target population.

• Step 3: Calculation of injury frequency. Impact probabilities for each area of the car's front were calculated for each impactor. Using these probabilities, the injury criteria measurements from step 2, injury risk curves relating these measurements to the probability of injury, and the velocity–exposure data from step 1, injury risks for each Abbreviated Injury Scale (AIS) level were summed for tested body regions for all casualties in the target population to give injury frequency for tested body regions.

• Step 4: Calculation of socioeconomic cost. Injury frequencies for tested body regions were converted into costs using “harm” cost information for the injuries considered; that is, those related to the impactor injury criteria.

• Step 5: Vehicle assessment: Weighting and summing. The body region costs were weighted using calibration factors and summed to give the total cost of injury assuming that all pedestrians in the target population were involved in an accident with the car being assessed. This cost was also weighted using a calibration factor to account for factors such as injuries to body regions not assessed by the impactors and injuries caused by ground impact. This cost can be compared with the cost calculated for other selected cars to give a relative assessment of the car.

Step 1: Active Safety Testing: Exposure—Impact Velocity Curve Shift

As described in Edwards et al. (2014), baseline exposure–impact velocity curves were developed for Great Britain and Germany (Figures 1 and 2) using detailed and national accident data appropriate for each country. For Great Britain, On the Spot (OTS) data were weighted using STATS19 national data on the basis of the number of fatal, serious, and slight casualties. For Germany, German In_Depth Accident Study (GIDAS) data were weighted using the German national data.

Figure 1 Impact speed distribution curves developed for Great Britain.

Figure 2 Impact speed distribution curves developed for Germany.

Accident scenarios describe the typical situations in which pedestrians are struck in real-world accidents. Test scenarios were used to assess the performance of an AEB system in a laboratory (test track) environment and were designed to reproduce the relevant parameters of the accident scenarios that had a major effect on system performance. Mapping accident to test scenarios was necessary to estimate what proportion of the casualties in the target population (i.e., in the exposure–velocity curves) would experience the impact speed reductions measured in the test scenarios.

The 5 test scenarios used in the method were those developed within AsPeCSS described in Table A1 (see online supplement; Seiniger et al. 2014). The mapping shown in Table 1 was used to determine the speed reduction for the casualties in the target population. The 7 accident scenarios and the proportion of casualties in each were defined in an early part of the project and have been reported previously (Wisch et al. 2013). A summary of the accident scenarios and the proportions of casualties in them is shown in Table A2 (see online supplement). It should be noted that there was a difference in the mapping for the German and Great Britain versions of the methodology for accident scenario 6, because in the German data, these accidents usually involved the pedestrian crossing one carriageway of the road before being hit by the car and therefore from the AEB system point of view, the pedestrian was not obstructed.

Table 1 Mapping of accident to test scenarios

	Test scenario
	TS1 Walking (slow) adult,	TS2 Running adult,	TS3 Walking adult,	TS4 Walking adult,	TS5 Walking child obstructed,
Accident scenario	center impact	center impact	near-side impact	farside impact	center impact
1: Crossing straight road, near-side, no obstruction	16.5%	16.5%	48%	19%
2: Crossing straight road, farside, no obstruction	15%	15%	45%	25%
3 and 4: Crossing at junction, near- or farside, vehicle turning or not across traffic	17.75%	17.75%	38%	27%
5: Crossing straight road, near-side, with obstruction					100%
6: Crossing straight road, farside, with obstruction	15%	15%	45%	25%	100%
7: Along carriageway on straight road, no obstruction	100%
8: Not classified into scenarios 1 to 7	No speed reduction

DE = Germany, GB = Great Britain.

Step 2: Passive Safety Testing: Impactor Measurement and Extrapolation

Euro NCAP impactor test data and the manufacturer input data were used to give impactor injury criteria values for most of the car's frontal area that is likely to be hit by a pedestrian, apart from areas with a wrap-around distance (WAD) beyond 2,100 mm, such as high on the windscreen and the windscreen header rail for the headform impactor. These injury criteria values, which represent impacts at nominally 40 km/h, were extrapolated to other speeds using simple relationships found in the literature or developed empirically from simulations performed in the project. For example, for head injury criterion (HIC) the relationship shown in Eq. (1)(1) developed by Searson et al. (2009) was used. (1) where v is speed.

One of the assumptions made was not to consider the effect of an impactor “bottoming out” on stiff structures below the struck structure.

Step 3: Calculation of Injury Frequency

As mentioned above, impact probabilities for each area of the car's front were calculated and with the velocity–exposure data from step 1 used to calculate injury frequency for tested body regions.

Impact Probabilities

These were calculated, both laterally and longitudinally, for all impactors. Laterally it was assumed that the impact probability was uniform for all impactors. This assumption was supported by an analysis of GIDAS data, which showed that the weighted average distribution of impact location across the car was approximately even with a slight bias to the right side (nearside)—left 31%, center 30%, right 39%. Other work was also found to support this assumption (Barrow et al. 2014). Longitudinal impact probabilities were only considered to be relevant for the headform impactor. For this impactor, the following speed-dependent relationship between pedestrian height and the longitudinal position of head impact in terms of WAD was derived from the results of simulations with the pedestrian human body THUMS model (Mottola et al. 2013): (2) where Pedestrian_Height is in millimeters and speed v is in kilometers per hour.

Using Eq. (2)(2) and population height distributions measured for the UK (Department of Trade and Industry, UK 1996), longitudinal impact probabilities in terms of WAD were derived for the headform impactor.

Injury Risk

Injury risk curves describe the probability that a certain load will cause a specific injury (Schmitt et al. 2004). The following functions relating risk of human injury in terms of AIS to impactor injury risk criteria were selected from the literature for the appropriate impactors.

For the headform impactor, 2 sets of injury risk curves were considered, namely, curves developed by Matsui (2004) using a logistic regression type called “modified maximum likelihood method” (MMLM) and those developed by NHTSA (1995), which used data from Prasad and Mertz (1985). The Matsui MMLM curves were selected for use in the model based mainly on pragmatic reasons (Figure 3). These were that the development of Matsui injury risk curves was based on pedestrian-to-car head impact data as opposed to head-to–car interior impact data for the NHTSA curves. In addition, the Matsui curves gave head injury costs slightly closer to those calculated from accident data during calibration when both sets of curves were tried in the methodology.

Figure 3 Pedestrian headform injury risk curves (created with equations from Matsui 2004).

For the upper legform impactor, an injury risk curve at the AIS 2 level for femur and pelvis injuries as the average of the 2 risk curves developed by EEVC WG17 (2002) were adopted (Figure A2, see online supplement). These curves were developed by Rodmell and Lawrence (1998) and included 12 reconstructed accidents reported in Matsui et al. (1998).

For the EEVC WG17 legform impactor, injury risk curves from Matsui (2003) at AIS 2 level were used (Figure A3, see online supplement). These were identified by Lawrence et al. (2006) to represent the best current data.

For the flexible pedestrian legform impactor (Flex PLI), injury risk curves from Takahashi et al. (2012) were used (Figures A4 and A5, see online supplement). An alternative injury risk curve for tibia fracture based on tibia bending moments has been derived by Zander et al. (2011) alongside discussions within the Informal Group on UN GTR No. 9, Phase 2 (Zander 2012). This could be used as an alternative in future analyses.

Step 4: Calculation of Socioeconomic Cost

Injury frequencies for tested body regions were converted into costs for tested body regions using harm-type information (total monetary costs) from Zaloshnja et al. (2004) for the injuries considered; that is, those related to the impactor injury criteria.

Step 5: Vehicle Assessment: Weighting and Summing

The body region costs calculated in step 4 above were weighted and summed to give the total societal cost of injury if it was assumed that all pedestrians in the target population were involved in an accident with the car being assessed. The following calibration (weighting) factors were derived by comparing the costs calculated with those known from accident data:

• A body region calibration factor to correct the relative cost of injury calculated for the tested body regions—that is, head, upper leg, and lower leg—to represent injury cost of body regions observed in accident data.

• An overall calibration factor to correct the total cost of injury was calculated. This should help take into account injury to body regions not tested, injury caused by contacts with parts of the car not tested currently, and injury caused by ground impacts and align with an independently estimated AEB benefit for a car representative of the average fleet in the accident data. However, it does assume that the cost of these injuries is proportional to the cost of injuries to the tested body regions.

To calculate the calibration factors, impactor test results were required for a car with passive safety protection levels representative of those of the cars in the accident data sample. The median registration date for cars in the accident data samples was 1997 with a range of about 1987 to 2010. On this basis, impactor data for a car with a registration year of about 1997 were developed to be broadly representative of those in the accident data sample. This was not a straightforward exercise because it was not possible to obtain impactor test data directly for cars of this era in Euro NCAP pedestrian assessment protocol version 6.0 or higher format (i.e., color codes for all grid points for headform impact) because it did not exist, because this protocol was only introduced in 2012. It could not be generated easily either, because simulation is required to do this and simulation models of these vehicles were not available. Therefore, the following approach was used to derive the necessary impactor data:

• Available worst-case impactor test data in a different format for a car of similar age was transformed into the format required. A Volkswagen Golf V tested in 2003 was chosen for this because much research work had been performed with this car that provided information to enable this transformation. In addition, this vehicle was a popular family car well represented in the vehicle fleet.

• The transformed Golf V impactor data was scaled to be representative of good, average, and poor Euro NCAP-rated cars registered circa 1997.

Body Region Calibration Factors

The cost of injury for the head, upper leg, and lower leg body regions seen in the accident was calculated using the GIDAS (German) and OTS (Great Britain) accident databases. For casualties in the target population (pedestrians impacted by the front of a car), the AIS injuries to these body regions were converted into costs using total monetary costs from Zaloshnja et al. (2004) and summed. The distribution of these costs was compared with that calculated by the methodology described above to derive calibration factors for the German and Great Britain versions of the methodology (see Table A3, online supplement). This process was repeated using the impactor data for the good and poor performing representative cars. It was found that the calibration factors derived were not that different to those derived for the average representative car. On this basis, it was decided to use the body region calibration factors derived for the average representative car.

Overall Calibration Factor

Following application of the body region calibration factor, the overall factor was calculated from a comparison of the cost predicted by the methodology and the cost of injury estimated from the accident data. The specific data used were injury costs estimated from a benefit analysis reported in Edwards et al. (2014). In this analysis, injury costs were calculated for the following 4 potential situations:

• No AEB system fitted to cars.

• Current generation AEB pedestrian systems (2013+) fitted to all cars.

• Second-generation AEB pedestrian systems (2018+) fitted to all cars.

• Reference limit AEB pedestrian system (2023+) fitted to all cars.

Specifications of the AEB systems are described in Edwards et al. (2014). Both the Great Britain and German methodology versions were run using input data to represent the 4 situations described above. Impactor test data representative of an average car in the accident data sample and exposure–impact velocity curves from the benefit for each of the 4 situations described above were used as input data. The exposure–impact velocity curves used took into account the effect of the AEB system on the accident impact speed. The costs calculated from the benefit analysis (Edwards et al. 2014) were compared with those predicted by the methodology for each of the AEB systems to derive calibration factors for the Great Britain and German methodology versions, shown in Tables A4 and A5 (see online supplement), respectively. The calibration factors for each AEB system variant were averaged to give overall calibration factors.

Results

Input Data Used to Test Methodology

Both versions of the assessment methodology (German and Great Britain) were tested with the following input data:

• AEB input test data:

  • No AEB system; that is, zero impact speed reduction against the baseline exposure–impact velocity curves.

  • Current AEB system; that is, speed reductions measured for the 5 test scenarios described in Table A1 for a vehicle tested in Seiniger et al. (2014; vehicle C).

• Passive safety impactor test data

  • Two sets of impactor tests results representative of current good, average, and poor performing vehicles. The first set (referred to as composite vehicles) contained vehicles with different windscreen areas, whereas the second set (referred to as composite vehicles with constant windscreen area) contained vehicles with identical windscreen areas to remove this variable.

  • Because no overall good, average, and poor performing Euro NCAP-rated vehicles existed, (generally vehicles scored maximum points for legform, average for headform, and poor for upper legform), impactor tests results representative of good, average, and poor performing vehicles were developed by selecting results from a number of vehicles on a component level (i.e., headform, upper legform, and legform) and combining them into composite vehicles. Because it was found that the assessment methodology was sensitive to changes in the amount of windscreen area in the head impact assessment zone and this changed between the good, average, and poor composite vehicles, good, average, and poor composite vehicles with constant windscreen area were developed. This was achieved by scaling the head impactor results for the average composite vehicle to give poor and good constant windscreen area composite vehicles with similar Euro NCAP scores as for the poor and good composite vehicles with variable windscreen area.

• Hypothetical A-pillar airbag:

  • Impactor test results were modified to represent a hypothetical airbag fitted to the vehicle, which covered the A-pillars completely but did not cover the scuttle. This airbag was assumed to offer protection at impact velocities between 21 and 51 km/h, reducing HIC at 40 km/h from 6,000 to 400 (Fredriksson and Rosén 2014) and to follow the same HIC–velocity relation as for other vehicle structures given in Eq. (1)(1) . For this hypothetical A-pillar airbag, Euro NCAP scores were estimated by changing the rating from red (zero score) to green (full score) for test points on the A-pillar.

The test data described above were input into the German and Great Britain versions of the assessment methodology. Results are shown in terms of total casualty costs for Germany and Great Britain for the composite vehicles with no AEB system, an AEB system representative of current systems, an A-pillar airbag, and both an AEB system and an A-pillar airbag (Table 2). Costs are nominally in euros for the German version and Great Britain pounds for the Great Britain version because the models were calibrated using the results of the benefit analyses for the respective countries (see Methodology section). At the time of writing, there were approximately 1.28 euros to the pound.

Table 2 Assessment results in terms of total casualty cost for variable windscreen area composite cars with good, average, and poor Euro NCAP passive safety rating with no system, with AEB system, with airbag, and with both AEB and airbag fitted. Percentages show costs normalized to average passive safety performance with no AEB system fitted. Note that Euro NCAP scores increase for better protection, whereas methodology assessment costs decrease

	Passive safety level
Additional safety
system	Good	Average	Poor
Euro NCAP passive safety score rating
No system	32.2 (142%)	22.6 (100%)	12.2 (54%)
A-pillar airbag	33.4 (148%)	24.4 (108%)	13.3 (59%)
Methodology (German version) integrated score rating (million euros)
No system	662 (99%)	667 (100.0%)	943 (141%)
AEB system representative of current systems	559 (84%)	563 (84%)	791 (119%)
A-pillar airbag	375 (56%)	338 (58%)	661 (99%)
AEB system and A-pillar airbag	324 (49%)	333 (50%)	560 (84%)
Methodology (Great Britain version) integrated score rating (million pounds sterling)
No system	889 (99%)	895 (100%)	£1,255 (140%)
AEB system representative of current systems	753 (84%)	758 (85%)	£1,060 (118%)
A-pillar airbag	471 (53%)	482 (54%)	845 (94%)
AEB system and A-pillar airbag	405 (45%)	413 (46%)	718 (80%)

The results for the composite vehicles with constant windscreen area are shown in Table 3.

Table 3 Assessment results in terms of total casualty cost for constant windscreen area composite cars with good, average, and poor Euro NCAP passive safety rating with no system, with AEB system, with airbag, and with both AEB and airbag fitted. Percentages show costs normalized to average passive safety performance with no AEB system fitted. Note that Euro NCAP scores increase for better protection, whereas methodology assessment costs decrease

	Passive safety level
Additional safety
system	Good	Average	Poor
Euro NCAP passive safety score rating
No system	32.2 (142%)	22.6 (100%)	12.2 (54%)
A-pillar airbag	34.0 (150%)	24.4 (108%)	14.0 (62%)
Methodology (German version) integrated score rating (million euros)
No system	545 (82%)	667 (100%)	884 (132%)
AEB system representative of current systems	461 (69%)	563 (84%)	742 (111%)
A-pillar airbag	265 (40%)	388 (58%)	604 (90%)
AEB system and A-pillar airbag	231 (35%)	333 (50%)	512 (77%)
Methodology (Great Britain version) integrated score rating (million pounds sterling)
No system	736 (82%)	895 (100%)	1,181 (132%)
AEB system representative of current systems	625 (70%)	758 (85%)	997 (111%)
A-pillar airbag	322 (36%)	481 (54%)	767 (86%)
AEB system and A-pillar airbag	280 (31%)	413 (46%)	653 (73%)

Discussion

Approach

In this study, a methodology was developed to assess a car's pedestrian protection based on harm estimated from AIS. However, other injury and cost metrics exist that could possibly have been used. For example, threat to life measured by the Injury Severity Scale, quality of life year losses, long-term consequences measured by permanent medical impairment, or socioeconomic cost, or combinations of them. There was no obvious choice, so the harm method was chosen on the basis that it was one of the simplest and data to implement it were readily available. It should be noted that there is little doubt that how the injury outcome is measured will influence the result. Tingvall et al. (2013) have shown that the group of road users suffering most injuries depends to a great extent on the injury measure used. For example, passenger car occupants are the majority of Maximum Abbreviated Injury Score (MAIS) 3+ injured road users, whereas bicyclists dominate MAIS 2+ injured road users.

To illustrate what different approaches other than harm may give, the relative rating of injury outcome for the harm approach used and an approach using risk of permanent medical impairment (RPMI) were compared for single AIS injuries to the head and leg body regions. RPMI is a measure for long-term consequences of injury that has been used in Swedish studies of pedestrian protection (Fredriksson et al. 2007; Strandroth et al. 2011). It is based on Swedish insurance data available for Swedish car occupants (Malm et al. 2008).

It was found that using RPMI affects the relative ratings of AIS level injuries substantially, in particular for the head (Table A6, see online supplement). It increased the rating (cost/impairment) of an AIS 4 injury compared to an AIS 5 injury and decreased the rating of an AIS 2 injury compared to an AIS 3 injury. Which rating is best is unknown, although both have shortcomings. However, the harm cost rating was chosen because sufficient information was available for harm to differentiate between the upper and lower leg body regions and to calibrate against an independent benefit estimate, whereas for RPMI it was not. Further work is recommended to investigate different injury and cost metrics in the future.

Assessment of Passive Safety

Examination of the assessment results for the composite vehicles for “no AEB system” (Table 2) shows that the AsPeCSS assessment of the good, average, and poor performing vehicles’ passive safety performance aligns in terms of order with the Euro NCAP score rating but does not align in terms of scale. Specifically, the AsPeCSS assessment shows a large difference in rating between the poor and average vehicles and a small difference between average and good vehicles, whereas the Euro NCAP scores show large differences between both the poor and average vehicles and the average and good vehicles.

However, examination of the assessment results for the composite vehicles with constant windscreen area for no AEB system (Table 3) shows that the AsPeCSS assessment of the good, average, and poor performing vehicles’ passive safety performance aligns in terms of order with the Euro NCAP score rating and better in terms of scale; that is, there is a closer difference in the assessment (cost) between poor and average and average and good vehicles.

Head injury costs were the main cause of the different assessment results for the composite vehicles, both with variable and constant windscreen area, because they form about 80% of the total costs. Further investigation found that the main reasons for the differences between the head impact assessments for the poor, average, and good composite vehicles with variable and constant windscreen area were as follows:

• A different portion of windscreen (default green) and A-pillar (default red) in the assessment area that varied between the poor, average, and good composite vehicles with variable windscreen area but was the same for the composite vehicles with constant windscreen area, which can be seen by examining Figure A6 (see online supplement).

• The difference between the Euro NCAP and AsPeCSS assessments for severe head injury and the effective weighting of this area in the AsPeCSS assessment to account for the probability of the head strike occurring there, compared to no weighting for the Euro NCAP assessment.

The difference between the Euro NCAP and AsPeCSS assessments for severe head injury is illustrated in Table 4. This table shows head assessment results for hypothetical average composite vehicles with a constant HIC value over all the assessment area with values representing green (good), yellow, orange (average), brown, red (poor), and default red vehicles. It can be seen that the relative differences between colors are similar for the Euro NCAP and AsPeCSS assessments for green to red rated vehicles but are quite different between red and default red for values of high HIC where the Euro NCAP assessment score is constant but the AsPeCSS assessment cost increases substantially. This is caused by the injury risk curves for head injury (Figure 3), which show a substantial increase in risk for severe head injury (AIS 4+, AIS 5+) at HIC values above 1,800 and a large change from red (assumed to be HIC 1,800) to default red (assumed to be HIC 6,000). The result of this is that variations in the amount of A-pillar in the assessment area changes the AsPeCSS assessment much more than the Euro NCAP assessment.

Table 4 Comparison of Euro NCAP and AsPeCSS assessments of head injury

Head	Good		Average		Poor	Default
Assessment	(green)	Yellow	(orange)	Brown	(red)	red
HIC	600	900	1,200	1,500	1,800	6,000^a
Euro NCAP score	24	18	12	6	0	0
AsPeCSS assessment injury cost (million euros)	106	203	318	448	599	2,379
Percentage of avearge AsPeCSS cost	33	64	100	141	188	748

^aA value of 6,000 was chosen based on A-pillar test results in Fredriksson and Rosén (2014).

Due to the importance of the A-pillar area, it comes as no surprise that a hypothetical airbag reduced overall casualty cost (Tables 2 and 3). The percentage effect of an A-pillar airbag (approximately 45% casualty cost reduction) is larger in the proposed AsPeCSS assessment compared to the Euro NCAP assessment (5 to 8% score increase for the cars in Tables 2 and 3), depending on base score and number of A-pillar test points in the rating area. One should keep in mind, though, that the hypothetical airbag was assumed to deploy in all collisions in the specified speed range and that side effects such as restricted visibility for the driver after deployment may reduce the actual safety benefit in the field. The benefit of a hypothetical airbag given in Tables 2 and 3 is rather more a potential than field benefit but nevertheless illustrates how the AsPeCSS method can be used to assess such systems.

The differences caused by changes in the amount of windscreen and A-pillar in the assessment area were emphasized by the difference in the weighting of this area between the Euro NCAP and AsPeCSS assessments. In the Euro NCAP assessment there is no weighting of this area (i.e., all headform test points are equally important). However, in the AsPeCSS assessment this area is effectively weighted because of the impact probability distribution used (Figure 4). If the AsPeCSS methodology impact probability distribution curves are considered in conjunction with the impact speed distribution curves (Figures 1 and 2), it can be seen that a large proportion of fatal and serious injuries occur at speeds greater than 30 km/h, and that at these speeds areas with a WAD distance of 1,800 mm or greater are more likely to be impacted. The outcome of this is that a change in the protection offered in these areas influences the AsPeCSS assessment considerably more than the Euro NCAP assessment.

Figure 4 Impact probability distribution with WAD and impact speed used in assessment methodology.

Assessment of Active Safety and Combined Effects

If the assessment results in Tables 2 and 3 for the composite vehicles with variable and constant windscreen areas are examined it can be seen that the addition of an AEB system that has a performance representative of current systems, in terms of the assessment, is broadly equivalent to increasing passive safety from poor to average or average to good. This is an increase in the Euro NCAP rating of 2 color bands and is in alignment with the previous work of Rosén et al. (2010), who predicted a high effectiveness for pedestrian AEB systems. However, it is greater than the benefit predicted by Searson et al. (2014), who estimated that fitment of a pedestrian AEB system would improve a car's Euro NCAP performance rating by approximately one color band only.

The estimated benefits of the hypothetical A-pillar airbag (40 to 50% casualty cost reduction) exceeded the benefits of an AEB system (10 to 20% casualty cost reduction) but remained in a similar order of magnitude. The estimated benefit of A-pillar airbags seems rather high and the benefit of AEB rather low compared to Fredriksson and Rosén (2014), who estimated 20 to 30% reduction of severe AIS 3+ head injuries for A-pillar and windscreen base airbags and 10 to 50% for AEB systems. Possible explanations for these differences are as follows:

• Fredriksson and Rosén express the benefit in terms of reduction of AIS 3+ head injuries, whereas the AsPeCSS methodology expresses the benefit in terms of cost reduction. Cost reduction takes into account the higher costs of AIS 4 and AIS 5 injuries compared to AIS 3 injuries, whereas quantifying AIS 3+ injury does not. A-pillar airbags are likely to reduce these higher severity injuries to a greater extent than AEB systems, therefore helping to cause the differences in the results seen.

• Compared to Fredriksson and Rosén, for the AsPeCSS assessment, the rather optimistic assumption for the A-pillar airbag was always for deployment in the specified speed range, whereas the AEB system was rather pessimistically assumed to give no benefit in some unclassified accident scenarios (20% of all cases).

• The AsPeCSS assessment calculates the impact probability for the head for each WAD and divides this probability by the number of lateral test points for each individual WAD to calculate the impact probability for each test point. As seen in Figure A6, the highest WAD has only few test points because of the shape of the car and the marking out procedure. This overemphasizes the effect of the A-pillar at this WAD in the AsPeCSS assessment. This is because effectively the few test points are taken to be representative of the full width of the car and the windscreen area between these points, which would likely be default green, is not taken into account.

Limitations

The major limitation within the methodology is the assumption used implicitly during calibration because a simple multiplication factor is used. This is that the cost of casualty injuries to body areas, such as the thorax, not assessed by the impactors (headform, legform, and upper legform), and other casualty injuries, such as those caused by ground impact, are related linearly to casualty injuries assessed by the impactors; that is, head, upper leg, and lower leg injuries.

However, examination of the calibration results indicates that this assumption may be valid or may not overly affect the assessments made with the methodology. Specifically, the calibration factors developed using the different AEB systems are very similar with a maximum variation of less than 2% from the average for Great Britain and Germany. This is an indication that relationships are approximately linear because if they were not the calibration factor would likely change more.

Other limitations include the following:

• Accuracy of impactor criteria to speed scaling relationships and disregarding of bottoming out.

• Validity and accuracy of injury risk curves.

• Validity and accuracy of WAD relationship with speed and pedestrian height for head impact. The vehicles simulated by Mottola et al. (2013) were deemed representative of the current EU fleet but not necessarily of the future fleet. The relation between pedestrian height, impact speed, and WAD is deterministic for the proposed method, and nonnegligible variation was observed in simulation (Mottola et al. 2013) and accident data (Fredriksson and Rosén 2012; Kiuchi et al. 2014).

• Validity and accuracy of Euro NCAP assessment results for a car representative of cars within accident data used for calibration. This is because of the large number of assumptions made to derive these data.

• There was no account taken of the effect of vehicle pitching when braking.

• How well the test scenarios represent the accident scenarios that are mapped to them. For example, at present because only basic test scenarios have been developed, accident scenarios such as “crossing a straight road from near-side, no obstruction” that occur in daylight and night street conditions is mapped as a whole to a test scenario that is conducted in daylight conditions. If the performance of the AEB system is dependent on the lighting conditions, the current methodology will not show these differences.

A method to estimate the overall benefit of active and passive safety pedestrian protection was developed and successfully tested. Using 2 calibration steps, it was ensured that the assessment reflects the body region injury distribution observed in the accident data and that the indicated AEB benefit equals an independent estimate of this benefit. The method has thereby advanced integrated assessment in order to encourage and spread best possible overall pedestrian protection and is ready for use in further assessments. The indication that benefits for safety systems are of the same order of magnitude as predicted by previous research is encouraging. However, limitations exist and it remains to be seen in retrospective accident studies whether the proposed method correlates better with observed injury outcome than other assessment schemes.

Acknowledgments

The authors acknowledge the UK Department for Transport for permitting the use of the OTS (On-The-Spot) accident study. The OTS data forms part of the Road Accident In Depth Studies database; further information can be found at https://www.gov.uk/government/publications/road-accid-ent-investigation-road-accident-in-depth-studies.

Funding

The authors gratefully acknowledge the support of the European Commission and the German Federal Ministry of Transport, Building and Urban Development and thank other ASPECSS project partners for their input into this work.

Supplemental Materials

Supplemental data for this article can be accessed on the publisher's website.