Project No TRACE

Transcription

1 Prject N TRACE D4.3 WP4-Summary Reprt Cntractual Date f Delivery t the CEC: 30 June 2008 Actual Date f Delivery t the CEC: 18 September 2009 (Versin 3 after 2 nd revisin) Authr(s): Menelas Pappas LMS Participant(s): LMS Validated by WP4 Leader: Ksmas Alexpuls-LMS Validated by TRACE Crdinatr: Yves Page-LAB Wrkpackage: WP4 Est. persn mnths: 1 Security: PU Nature: Reprt Versin: 3 Ttal number f pages: 36 Abstract: The aim f Wrkpackage 4 is t investigate the impact f advanced safety functins n reducing several types f accidents invlving passenger cars r mitigating accident cnsequences, cvering the secnd bjective f TRACE. The evaluatin is perfrmed frm tw different perspectives: - Assessment f the ptential prprtin f accidents that culd be avided and f the ptential prprtin f accidents whse severity culd be reduced, fr safety functins that are nt yet n the market (this is the s-called a priri effectiveness). Additinally and in strng cnnectin t WP5, the ptential limitatins/cnstraints which culd prevent the driver frm taking advantage f infrmatin added t his task by means f driving aid, and which culd lessen the effectiveness f a safety functin, have been thrughly examined by in-depth analysis in this task. - Assessment f the actual prprtin f accidents that culd be avided and f the actual prprtin f accidents whse severity culd be reduced, fr safety functins that are already in the market (this is the s-called a psteriri effectiveness) nce the cars are equipped with existing functins (Task 4.2). Deliverable 4.3 prvides a summary dcument with figures, statistics, results, r ther kind f utcmes that can be used fr the identificatin, validatin f the relevance and the evaluatin f expected r bserved effectiveness f safety functins based n technlgy that are already implemented in vehicles r culd be put in vehicles in the future (i.e. in cmmunicatin with the infrastructure r nt, c-perating between vehicles, cperating with the user, assisting r substituting the driver r nt). Keywrd list: A priri evaluatin, a psteriri evaluatin, safety functin, effectiveness, safety benefits, diver's needs Date f Delivery : September

2 Table f Cntents 1 Executive summary TRACE prject bjectives WP4 Evaluatin bjectives Deliverable bjectives 4 2 A-priri evaluatin f safety functin effectiveness Intrductin Review f crash effectiveness f Intelligent Transprt Systems (ITS) Methdlgies fr a-priri evaluatin f safety functins effectiveness Target ppulatin times efficiency Autmatic case-by-case analysis Analysis f selected in-depth cases Artificial Neural Netwrks Assumptins and bundary cnditins fr a-priri analysis Results n safety increments based n a-priri evaluatin f safety functins effectiveness Detailed effectiveness evaluatin results: Intelligent Speed Adaptin (Example) Overall a-priri effectiveness evaluatin results A human-centred apprach fr assessing drivers' needs and cntextual cnstraints f safety functins 19 3 A-psteriri evaluatin f safety functin effectiveness Intrductin Methdlgy fr a-psteriri evaluatin f safety functins effectiveness Methdlgy fr the a-psteriri evaluatin f accident avidance Methdlgy fr injury mitigatin evaluatin Assumptins and bundary cnditins fr a-psteriri analysis Results n safety increments based n a-psteriri evaluatin f safety functins effectiveness 28 4 Relatin t sci-ecnmic impact f safety functin Intrductin Link establishment with eimpact prject 33 5 Cnclusins 34 References 36 Date f Delivery : September

3 1 EXECUTIVE SUMMARY 1.1 TRACE prject bjectives In spite f cuntless amunts f research and develpment, rad safety is still ne f the main scietal cncerns tday. It is nt nly a matter f cncern fr the Eurpean Cmmissin and Natinal Gvernments but als fr the vehicle industry, insurance cmpanies, driving schls, nngvernmental rganizatins and mre generally fr every single rad user. Car manufacturers have made strng effrts and have dramatically imprved passive (and als active) safety f their vehicle fr the past 15 years. Hwever, current rad safety research has shwn that an asymptte is abut t be reached n this aspect in mst cuntries and many experts agree that preventive (preventin f accidents) and active safety (recvery f an emergency situatin) shuld nw, particularly, be brught frward. The TRACE prject has 2 majr bjectives: The first ne addresses the determinatin and the cntinuus up-dating f the etilgy (i.e. analysis f the causes) f rad accidents and injuries, and the definitin f the real needs f the rad users as they are deduced frm accident and driver behavir analyses. The secnd ne aims at identifying and assessing, amng pssible technlgy-based safety functins, the mst prmising slutins that can assist the driver r any ther rad users in a nrmal rad situatin r in an emergency situatin. S the purpse is first t bring a cmprehensive and understandable definitin f accident causatin which ges further and deeper than the usual statements. It is als t prvide the scientific cmmunity, the stakehlders, the suppliers, the vehicle industry and the ther Integrated Safety prgram participants with a glbal verview f the rad accident causatin issues in Eurpe and prmising slutins based n technlgy. 1.2 WP4 Evaluatin bjectives The aim f this wrkpackage is t investigate the impact f advanced safety functins n reducing several types f accidents invlving passenger cars r restricting (mitigating) accident cnsequences. The prpsed research investigates the effectiveness f several safety functins n different accident cnfiguratins. WP6 has prvided a list f the mst prmising safety functins that address current and future accident types n Eurpean rads. The evaluatin has be perfrmed frm tw different perspectives: Assessment f the ptential prprtin f accidents that culd be avided and f the ptential prprtin f accidents whse severity culd be reduced, fr safety functins, f passenger cars, nt already n the market (this is the s-called a priri effectiveness) (Task 4.1). This bjective has been brken dwn int three main challenges: predict the benefits f the safety systems. prvide reliable results fr future (nt yet intrduced in the market) safety functins. define the cnstraints the safety systems will have t cpe with in rder t fulfill nt nly the drivers needs but als t cmpensate the characteristics f the situatins in which these needs are met. Date f Delivery : September

4 Assessment f the actual prprtin f accidents that culd be avided and f the actual prprtin f accidents whse severity culd be reduced, fr safety functins, f passenger cars, already n the market (this is the s-called a psteriri effectiveness) nce the cars are equipped with existing functins (Task 4.2). 1.3 Deliverable bjectives The bjective f this deliverable is t prvide a summary f the main utcmes that have been generated by the wrk perfrmed in TRACE WP4 (Figure 1). Figures, tables and ther kind f utcmes that can be used fr the identificatin, validatin f the relevance and the evaluatin f expected r bserved effectiveness f safety functins based n technlgy that are already implemented in vehicles r culd be put in vehicles in the future (i.e. in cmmunicatin with the infrastructure r nt, c-perating between vehicles, cperating with the user, assisting r substituting the driver r nt). Mre specifically the fllwing tasks tk place in the frame f WP4 and they are summarized in the fllwing sectins f this reprt. Task A priri evaluatin f safety functin effectiveness Review f crash effectiveness f Intelligent Transprt Systems (ITS) [1] Literature review n accident cnfiguratins [2] Methdlgies and results n safety increments based n a-priri evaluatin f safety functins effectiveness [3], [4] A human-centred apprach fr assessing drivers' needs and cntextual cnstraints f safety functins [5] Task A psteriri evaluatin f safety functin effectiveness Methdlgies and results n safety increments based n a-psteriri evaluatin f safety functins effectiveness [6], [7] Task Relatin t sci-ecnmic impact f safety functin Link establishment with eimpact prject Evaluatin A-priri A-psteriri Capacity t meet driver s needs Safety benefits Safety benefits Figure 1: Flw diagramm f the prpsed apprach fr the evaluatin f safety functins effectiveness Date f Delivery : September

5 Initially a literature review, that tk place in strng cperatin with WP6 "Safety a functins", a lng list f 138 safety functins have been identified and described [1]. Frm this lng list, TRACE decided t evaluate the mst prmising nes based n several criteria (and eimpact interest) fcusing n passenger cars and primary safety. After detailed evaluatins and discussin 23 safety functins have finally been evaluated (Table 1): The evaluatin f the fllwing safety systems f passenger cars has been perfrmed by assessing: i) their capacity t meet the driver's needs, ii) the safety benefits based n a-priri evaluatin and iii) the safety benefits based n a-psteriri evaluatin. In Table 1 the safety functins that have been evaluated and evaluatin perspective fr each f them is presented. Safety System Functin User's needs Safety benefits Crnering Brake Cntrl Braking Systems Brake Assist Braking Systems Predictive Assist Braking Braking Systems Advanced Adaptive Cruise Cntrl Drive Safe Cllisin Warning Drive Safe Cllisin Avidance Drive Safe Vulnerable Rad Users Prtectin Drive Safe - Lane Keeping Assistant Drive Safe Lane Changing Assistant Drive Safe Blind Spt Detectin Drive Safe Intelligent Speed Adaptin Drive Safe Traffic Sign Recgnitin Drive Safe Intersectin Cntrl Drive Safe Drwsy Driver Detectin System Drive Safe Alclck Keys Drive Safe Tyre Pressure Mnitring and Warning Drive Safe Electrnic Stability Prgram Handling/Kinematics Dynamic Suspensin Handling/Kinematics Advanced Adaptive Frnt Light System Visibility Night Visin Visibility Rear Light Brake Frce Display Visibility Pedestrian Prtectin Airbag Airbags - Advanced Autmatic Crash Ntificatin Rescue - Table 1: List f passenger cars safety functins that have been evaluated in the frame f WP4 Apart frm the review fr the identificatin f the mst prmising safety systems a parallel investigatin has taken place in WP4 fr identifying (existing methds) r prpsing (nvel methds) fr the evaluatin purpses ([3], [6]). Finally, the actual evaluatin tk place bth in terms f a-priri and a-psteriri evaluatin ([4], [5], [7]). During the whle duratin f WP4 links and cmmn activities with the EU prject eimpact ( have been established fr expliting sci-ecnmic benefits f safety functins. Date f Delivery : September

6 The rest f the dcument is rganized as fllws: Chapter 2: The final results, as well as the methds used fr a-priri evaluatin f safety functins listed in Table 1 are presented. Chapter 3: The final results, as well as the methds used fr the a-psteriri evaluatin f selected safety functins are presented. Chapter 4: Outlines the relatin t sci-ecnmic impact f safety functin and especially the cllabratin activities with EU prject eimpact. Chapter 5: Overall cnclusins are drawn by cnsidering hlistically the results f WP4 wrk. The main findings that came ut frm the evaluatin f the safety functins are summarised hereafter: the safety functins that have been estimated t prvide the best safety effectiveness are: Intelligent Speed Adaptatin (17% fatalities reductin, 11% serius injuries mitigatin) Advanced Adaptive Cruise Cntrl (11% serius injuries mitigatin) Advanced Autmatic Crash Ntificatin (10,8% fatalities reductin) Cllisin Avidance (9,1% serius injuries mitigatin) Cllisin Warning (6,6% serius injuries mitigatin) Alclck Key (6% fatalities reductin, 3% serius injuries mitigatin) the safety functins that fail t prduce cnsiderable safety benefits are: Tyre Pressure and Mnitring (1,35% serius injuries mitigatin) Rear Light Brake Frce Display (0,8% serius injuries mitigatin) Advanced Adaptive Frnt Light System (0,6% serius injuries mitigatin) Predictive Assist Braking (0,2% serius injuries mitigatin) Dynamic Suspensin (0% serius injuries mitigatin) the safety functins which shw the best safety effectiveness in cmpensating fr pivtal needs are: Cllisin Avidance (14,1% capacity t meet pivtal driver's needs) Intersectin Cntrl (12,3% capacity t meet pivtal driver's needs) Cllisin Warning (8,9% capacity t meet pivtal driver's needs) Alclck Key (5,6% capacity t meet pivtal driver's needs) Electrnic Stability Prgram (4,5% capacity t meet pivtal driver's needs) Intelligent Speed Adaptin is very effective in cmpensating fr upstream needs in the driving phase (14,6% capacity t meet upstream needs) Electrnic Stability Prgram and Cllisin Avidance are very effective in cmpensating fr the emergency phase (18,9% and 7,6% capacity t meet emergency needs accrdingly) As fr n-the-market safety applicatins, the main results are the fllwing: wuld all cars be five stars, cmpared t fur stars fitted with EBA and ESC, injury accidents wuld be reduced by 19.3%, all injuries wuld be mitigated by 33.8% and severe + fatal injuries by 35.1%. wuld all cars be five stars fitted with EBA and ESC, cmpared t fur stars withut ESC and EBA, injury accidents wuld be reduced by 47.2%, all injuries wuld be mitigated by 67.8% and severe + fatal injuries by 69.5%. Date f Delivery : September

7 2 A-PRIORI EVALUATION OF SAFETY FUNCTION EFFECTIVENESS 2.1 Intrductin In depth investigatins with respect t several safety functins, were carried ut in the frame f this wrk, in rder t evaluate the ptential effects in terms f bth pssibility t avid the accident and the pssibility t reduce its cnsequences (as measured by injury severity levels). Fr executing such investigatin, it was necessary t cmbine accidentlgy results with mdeling and artificial Intelligence techniques t evaluate hw drivers cpe with safety devices that accidentlgy research has highlighted. Additinally in the frame f this wrk, and in strng cnnectin t WP5, the ptential limitatins/cnstraints which culd prevent the driver frm taking advantage f infrmatin added t his task by means f driving aid, and which culd lessen the effectiveness f a safety functin, were thrughly examined by in-depth analysis. By taking int accunt the cntexts in which the driver failures ccur, this apprach enabled the definitin f the cnstraints the safety systems will have t cpe with in rder t fulfill nt nly the needs f the drivers but als t cmpensate the characteristics f the situatins in which these needs are met. The majr challenges f this apprach are: The estimatin f the verall effectiveness f safety functins that have nt been intrduced in the market yet The predictin f the benefits f the safety systems The prductin f reliable results fr future (nn-existing) safety functins 2.2 Review f crash effectiveness f Intelligent Transprt Systems (ITS) Initially, in the cntext f Task 4.1 (in cmmn with WP6 "Safety Functins") a research wrk tk place that prvided a review f currently available Intelligent Transprt Systems (ITS) and their expected and bserved safety enhancing effects. ITS are technlgies applied t the transprt dmain that may enhance mbility, efficiency and safety, amng ther benefits. The ptential fr ITS t imprve rad user safety has been widely explred, and this wrk presented a review f what is knwn abut the effectiveness f safety-relevant ITS. Each system has been classified as in-vehicle, infrastructure-based r cperative, and further categrized as active, passive, r cmbined active and passive. The systems were described in terms f their functinal and technical characteristics, their likely crash relevance, and available literature regarding safety enhancing effects r disbenefits has been reviewed. In ttal, 138 systems have been reviewed in the frame f this wrk. Overall, it can be seen that while there may be incnsistencies amng the literature regarding the expected degree f effectiveness f these systems, there is cnsiderable evidence t suggest that ITS have had, and will cntinue t have, substantial psitive effects n rad user safety. The cllectin f all these results regarding the effectiveness f existing safety systems, as well as the upcming nes, has allwed identifying the mst prmising ITS that were cnsidered in the fllwing steps f the evaluatin prcedure. The review revealed the relevance f thse systems in terms f crash avidance/reductin, as well as in derived injuries. Sme f the systems have been fund t be especially effective (ESC, e-call, ACC), Date f Delivery : September

8 while the effectiveness f sme ther systems is nt prved t be s relevant. Nevertheless, mst f them act simultaneusly r have a similar functin (e.g. Lane Changing Assistant r Lane Departure Warning), r even they interact between the car/user and the rad (e.g. Traffic Sign Recgnitin), what makes them mre efficient. The imprtance f passive safety systems in the future has been fund t be related t active systems, which are called cmbined active and passive systems. In that field, the mst relevant systems are the nes that act just befre a crash ccurs. Sme f thse Pre-Crash systems use sme frward-facing sensrs, such as in the case f cllisin avidance systems, t determine the eminent crash and prepare the vehicle t minimize the effects f the crash. Fr example, passive safety systems like the seatbelt pre-tensiners have the bjective f diminishing the cnsequences f a crash in case it ccurs. Sme ther systems cnsist n the detectin f elements n the rad that can interfere in the trajectry f the vehicle (e.g. the Autmatic Pedestrian Detectin r the Animal Detectin System), which derives als in an imprtant reductin f crashes and injuries. Finding all the cllected infrmatin has been a difficult task (especially fr sme f the systems), and in many cases the infrmatin prvided was cntradictry. Nevertheless, it is undeniable that ITS will becme ne f the mst effective way t reduce the amunt f injury accidents regarding all types f vehicles in ur rads, under any circumstance r crash cnditin (mre details can be fund in [1]). 2.3 Methdlgies fr a-priri evaluatin f safety functins effectiveness The main bjective f this wrk was t define methdlgies in rder t estimate the ptential prprtin f injury accidents that culd be avided and/r the ptential prprtin f accidents whse severity culd be reduced, fr future and existing safety functins (r a cmbinatin f functins). The utcme f this wrk was the definitin f fur methdlgies that culd be used fr the a-priri evaluatin f the safety functins effectiveness in terms f safety increment (refer t [3] fr mre details). Based n the special characteristics f each methd, a cnnectin between the selected safety functins and the suitable methd(s) fr evaluatin was als presented. The methds range frm Expert Estimatins t Artificial Neural Netwrks. Each f the invlved partners presented a different methdlgy fr the a-priri evaluatin f safety functins effectiveness. This apprach enabled the evaluatin frm different perspective and level f details. Fr the predictin f the benefits f safety systems and functins, a variety f methds have been established (Figure 2). The fur selected methdlgies are: 1. Target ppulatin times efficiency 2. Autmatic case-by-case analysis 3. Analysis f selected in-depth cases 4. Artificial Neural Netwrks Date f Delivery : September

9 100% 90% 80% 70% 60% 50% MAIS0-1 MAIS2+ lgistic 40% 30% 20% 10% 0% 0 km/h 5 km/h 10 km/h 15 km/h 20 km/h 25 km/h 30 km/h 35 km/h 40 km/h 45 km/h 50 km/h 55 km/h 60 km/h 65 km/h 70 km/h (n=57) (n=36) (n=74) (n=71) (n=72) (n=38) (n=37) (n=20) (n=15) (n=15) (n=4) (n=3) (n=2) (n=1) (n=6) 1. Target ppulatin times efficiency 2. Autmatic case-by-case analysis 2500 Input Layer Hidden Layers Output Layer 2000 n skidding, n=2412 skidding, n=1874 Cllisin type Weather/visibility cnditin straight line curve Rad surface cnditin Rad layut Braking /steering maneuvre Max. injury severity in the passenger car Driving speed dry wet snwy/icy dry wet snwy/icy Crash speed Driver state 3. Analysis f selected in-depth cases 4. Artificial Neural Netwrks Figure 2: Methdlgies used in WP4 fr the a-priri evaluatin f the safety functins effectiveness Target ppulatin times efficiency At the beginning (r during the early phases) f the develpment f the driver assistance systems, their exact specificatin is still pen. In many cases the system s parameters have t be fund by predicting and cmparing the effectiveness fr different layuts. Fr this issue, a tw-stage apprach has been successfully used (Figure 3): In the first stage all the accidents in a database related t the studied system, are identified. The way this identificatin is dne, is depended n bth the level f details f the available data and the assistance system under discussin. Once all the relevant cases are identified, the maximum benefit f the system is als established, i.e. n system can have a larger benefit than this. Usually, the relevant cases in the database are prjected fr ne year f natinal (r Eurpean) statistics and named target ppulatin. Target ppulatin is a hard figure because the underlying filter criteria are well defined, hwever the effectiveness evaluatin needs an additinal step t accunt fr accidents that are addressed but can nt be prevented. A percentage f the addressed accidents is established that can actually be prevented by the respective system. This percentage is called efficiency and may even be btained by expert estimatins (as a last resrt). This prcess des nt prvide any hard figures because it relies n smewhat subjective expert pinin r engineering judgment, but in practice, the underlying assumptins are made cnservatively in rder t underestimate the verall benefit rather than t verestimate it. Since innvative safety functins may be hard t assess, even by a panel f experts, additinal simulatr r test track studies may be used t prvide a better feeling fr the capabilities and limitatins f a new system. Date f Delivery : September

10 30% 25% 20% 15% 10% 5% 0% Injury severity Safety systems which reduce impact severity Neutral Infrmative Driver Selec t Mandatry d Safety measures which reduce injury risk at a given impact severity 30% 25% 20% 15% 10% 5% 0% Impact severity Neutral Infrmative Dri ver Select Mandatry Accident Data Relevant Accident Cnfiguratins Safety Functin Target Ppulatin Efficiency E f f e c t i v e n e s s Figure 3: Cnceptual diagram f the Target ppulatin times efficiency a-priri evaluatin apprach A similar cncept als used fr estimating the ptential safety benefits f Intelligent Speed Adaptatin Systems in terms f avidable casualties in cars. As they are many different systems, the study fcused n the experimental French LAVIA Intelligent Speed Adaptatin system, accrding t rad netwrk and system mde, based n bserved driving speeds, distributins f crash severity and crash injury risk. The prpsed apprach invlved the fllwing steps: Estimatin f the injury risk as a functin f impact severity. Establishment f the relatinship between impact severity and the travel speed f injury crash invlved vehicles that wuld be bserved fllwing LAVIA deplyment. Cnstitutin f the distributin f the travel speeds f injury crash invlved vehicles using the real travel speed distributins frm the LAVIA field tests. Calculatin f the ptential benefits f the LAVIA system using these data. Injury Risk = f (Impact Severity) Impact Severity = f (Travel Speed) Distributin f Travel Speed ISA E f f e c t i v e n e s s Figure 4: Target ppulatin times efficiency apprach used fr the a-priri effectiveness evaluatin f the Intelligent Speed Adaptatin safety functin Date f Delivery : September

11 2.3.2 Autmatic case-by-case analysis In this apprach a scenari is defined that reflects the way in which the system wrks. In an autmated database run, every accident prcessed twice: In the first run, it is simulated just as it happened in reality, i.e. withut the system under discussin. In the secnd run, the system is simulated and the difference t the first run is calculated as benefit. Summing up individual benefit numbers (and, if desired, prjecting them t natinal r Eurpean statistics) this apprach cmes up with reliable results. Depending n the cmplexity f the system under discussin, the tw runs can either be perfrmed directly in a database query (fr straightfrward systems r activatin criteria, e.g. Brake Assist) r may require a dedicated sftware tl (Figure 5). Figure 5: Autmatic case-by-case analysis using a dedicated sftware tl, develped by VW Analysis f selected in-depth cases When the safety system itself r its interactin with the driver is t cmplex fr an autmated database run, as described abve (r if the mdel requires user interventin at runtime), a mdelbased what-if analysis may be pssible and useful. Due t the increased effrt hwever the analysis is then limited t selected individual cases where a detailed recnstructin f bth the impact itself and the pre- and pst-crash mtin f the vehicle(s) are perfrmed. Ideally this invlves the use f multi-bdy systems (e.g. ADAMS) fr the analysis f pre- and pst-crash kinematics and Finite Element Analysis Tls (e.g. PAM-CRASH) fr impact analysis. Sme Sftware packages, like MADYMO and PC-Crash allw fr bth kinds f analyses t be dne within the same simulatin envirnment. Since a reliable predictin is required, the selectin f cases needs special care and attentin t make sure that it is nt biased twards situatins that are either easy t simulate r mst prmising. In any case, this simulatin has t be preceded by statistical data analysis that prvides the characteristic parameters f the situatin as well as their verall relevance. Date f Delivery : September

12 An example f this is again Electrnic Stability Cntrl. A database analysis f skidding related accidents with severe casualties shwed that cntrary t what intuitin may suggest mst f thse accident did nt ccur in bends r n slippery rads n skidding, n=2412 skidding, n= straight line curve dry wet snwy/icy dry wet snwy/icy Figure 6: Severe casualties by rad trajectry and surface (GIDAS data) Hence, fr a realistic benefit predictin, prttypical cases had t be studied, i.e., cases that have a similar distributin f characteristics as shwn in the abve graph. These cases were then simulated using PC-Crash. A sample case n a dry straight rad with and withut ESC is shwn in Figure 7. After a number f accidents had been analyzed and the situatins re-run with and withut ESC it is then pssible t calculate a weighted average f the benefits established fr the individual situatins (i.e. the percentage f accidents that have been avided r transfrmed frm a lateral t a frntal impact). Figure 7: Recnstructin f an accident withut (left) and with ESC (right) Artificial Neural Netwrks Artificial Neural Netwrks (ANN) are sphisticated and nn analytical mdelling techniques, capable f mdelling cmplex and highly nn-linear functins, used mstly t slve prblems f predictin, classificatin r cntrl. In depth analysis with respect t several safety functins, carried ut in the frame f Task 4.1, in rder t evaluate the ptential effects which culd be induced n selected (pilt) accident cnfiguratins [2] s as t reduce its cnsequences (as measured by injury severity levels). Date f Delivery : September

13 The basic steps f the Neural Netwrks based methdlgy (Figure 8) are: 1. Acquire knwledge fr the safety functins f passenger cars t be evaluated 2. Selectin f the mdeling parameters (parameters that affect/describe an accident). 3. Accident data cllectin accrding t the selected parameters. 4. Design NN architecture fr predicting the severity level f different accident cnfiguratins based n the available data. 5. Define the relevance f safety functin t accident cnfiguratins. 6. Estimate the influence f a safety functin n different accident parameters 7. Using the NN calculate the effectiveness f the safety functin n the level f severity. 8. Estimate the effectiveness f the studied safety functin based n the calculatin f the injury severity mitigatin (in percentage). Accident Data Safety Functin Relevance/ Influence Accident Parameters/ Max Severity Build/Train Neural Nets Mdified Parameters NN Predictin S a f e t y E f f e c t i v e n e s s Figure 8: Cnceptual diagram f the Neural Netwrk based a-priri evaluatin apprach 2.4 Assumptins and bundary cnditins fr a-priri analysis Due t the nature f a-priri benefit assessment, nly a general descriptin f the safety functins / systems t be evaluated is available. In general less specific infrmatin is available as the intrductin date f the safety system is further in the future. This makes it mre difficult t develp a mdel f the safety functin. Thus, a series f implicit and explicit assumptins have t be made that culd result in a smewhat idealised safety functin. Hwever these assumptins can eventually be questinable. Fr this reasn, it is anticipated that the infrmatin abut the prcesses adpted fr this evaluatin study shuld be sufficiently transparent t be able t assess the merit f the applied assumptins, and als t re-calculate the benefits at any stage, depending upn if mre recent data r mre accurate assumptins are frthcming. The findings btained in this study shuld nt be used as a definitive assessment f the viability f the studied safety systems, but rather, as an indicatin f their ptential effectiveness in decreasing rad fatalities and serius injuries. Date f Delivery : September

14 It is als ntewrthy that this analysis assumed the absence f ther in-vehicle technlgies. Fr example during the analysis f the intelligent braking system, ther systems such as pre-cllisin systems r Electrnic Brake Assist, culd als enhance the amunt f braking time and/r braking perfrmance. A-priri analysis is valuable as it helps t fcus attentin n the ptential value in intrducing new safety technlgy in future cars and ptimize the benefits t sciety f their intrductin. Nevertheless, there are a number f limitatins in these benefit studies that need t be stated. Target ppulatin methd, partly based n the use f injury curves fr car ccupants leaves apart the calculatin f the safety benefits fr ther kind f users, particularly the vulnerable rad users such as pedal cyclists, pwered tw-wheelers and pedestrians. Insufficient accident in-depth data fr these users are the main explanatin. But the general methd als applies t them and culd be equally used in case f in-depth data availability. It is generally argued that this Target ppulatin methd nly lks int injury mitigatin and ignres accident avidance because the benefits are calculated with the help f injury curves, which is cmmnly used fr the evaluatin f passive safety measures whereas ISA is a preventive safety system. One must treat these findings with sme cautin, as a number f assumptins were necessary in undertaking this analysis. This was partially due t the paucity f in-depth pedestrian/vulnerable rad user data arund the wrld. In many instances, best estimates were made in these assumptins, which were backed by engineering judgements. Every effrt was made t make the analysis as transparent as pssible, s that with additinal infrmatin and knwledge, these calculatins culd be adjusted accrdingly. Finally, given that the analyses were based n nly furteen f the EU natins due t data availability issues, it cannt be directly assumed that the reprted results are representative f the EU as a whle, given that there is likely t be sme variability in rad safety statistics and trends between different Eurpean cuntries. 2.5 Results n safety increments based n a-priri evaluatin f safety functins effectiveness Assessing the capacity f new advanced safety functins t cmpensate fr, r mitigate, the impact f rad accidents is ne f the main purpses f the TRACE prject. This is the main purpse f the wrk dne in Task 4.1 "A-priri evaluatin f safety functins effectiveness". Different methds have been applied by the TRACE WP4 partners in rder t calculate the results. In mst cases the systems were evaluated using Eurpean data (either natinal r in EU level) and in sme limited cases nn- Eurpean data have been used. Hwever, the limitatins impsed by the use f the data and hw the results culd be 'translated' int EU level are discussed in the relevant deliverable ([4]) Detailed effectiveness evaluatin results: Intelligent Speed Adaptin (Example) Fr the purpses f this study, which is based n the Target ppulatin cncept, the results were gruped by impact type (Frntal and Side) and by rad netwrk (Urban, Inter-Urban and Mtrway). Diagrams that present the detailed results f this analysis are shwn belw (Figure 9 - Figure 14). Date f Delivery : September

15 Serius Injuries Reduced Fatalities Reduced Infrmative 4% 4% Driver activated 11% 14% Mandatry 9% 11% 0% 2% 4% 6% 8% 10% 12% 14% 16% Figure 9: Frntal Impact, Urban Serius Injuries Reduced Fatalities Reduced Infrmative 2% 5% Driver activated 3% 8% Mandatry 2% 8% 0% 1% 2% 3% 4% 5% 6% 7% 8% 9% Figure 10: Frntal Impact, Inter-Urban Serius Injuries Reduced Fatalities Reduced Infrmative 3% 7% Driver activated 6% 13% Mandatry 5% 13% 0% 2% 4% 6% 8% 10% 12% 14% Figure 11: Frntal Impact, Mtrway Date f Delivery : September

16 Serius Injuries Reduced Fatalities Reduced Infrmative 3% 4% Driver activated 1% 3% Mandatry 0 0% 0% 1% 1% 2% 2% 3% 3% 4% 4% 5% Figure 12: Side Impact, Urban Serius Injuries Reduced Fatalities Reduced Infrmative 0% 7% Driver activated 9% 17% Mandatry 6% 8% 0% 2% 4% 6% 8% 10% 12% 14% 16% 18% Figure 13: Side Impact, Inter-Urban Serius Injuries Reduced Fatalities Reduced Infrmative 0 4% Driver activated 5% 16% Mandatry 4% 16% 0% 2% 4% 6% 8% 10% 12% 14% 16% 18% Figure 14: Side Impact, Mtrway Date f Delivery : September

17 2.5.2 Overall a-priri effectiveness evaluatin results Table 2 presents the allcatin f the nineteen (19) selected safety systems t the fur (4) different methds fr the estimatin f safety ptential (a-priri effectiveness evaluatin). # Safety System Methd Target ppulatin methd Reference and future wrld scenari (VW) 1 Tyre Pressure and Mnitring X 2 Lane Keeping Supprt X 3 Lane Changing Supprt X 4 Crnering Brake Cntrl X 5 Traffic Sign Recgnitin X 6 Intersectin Cntrl X Target ppulatin - Effectivenes s evaluatin (LAB) 7 Intelligent Speed Adaptatin X 8 Blind Spt Detectin X Unit HARM (MONASH) 9 Alclck Key X 10 Advanced Autmatic Crash Ntificatin 11 Night Visin X 12 Cllisin Avidance X 13 Predictive Brake Assist X 14 Dynamic Suspensin X 15 Drwsy Driver Detectin System 16 Advanced Frnt Light System X 17 Rear Light Brake Frce Display 18 Cllisin Warning X 19 Advanced Adaptive Cruise Cntrl X Neural Netwrks Based Evaluatin (LMS) Table 2: A-priri methds used fr the evaluatin f the 19 studied safety functins The main results (estimatin f safety benefits) cming ut frm the a-priri evaluatin f safety functins effectiveness are presented in the Table 3 belw. This table shws the verall effectiveness evaluatin results fr the selected nineteen (19) primary safety systems fr passenger cars that have been studied in the frame f Task 4.1 ([3], [4]). X X X In the fllwing table the safety systems effectiveness is presented in terms f: Fatalities saved: The percentage f fatalities saved by the safety functin. Less-fatalities in the rads. Serius injuries saved: The percentage f serius injuries saved. Serius injuries that their severity is either mitigated r the serius injury accident has been avided. Date f Delivery : September

18 Accrding t the results presented in Table 3 abve (refer t [4] fr mre details), the safety functins that have been estimated t prvide the best safety effectiveness are: Intelligent Speed Adaptatin Advanced Adaptive Cruise Cntrl Advanced Autmatic Crash Ntificatin Cllisin Avidance Cllisin Warning Alclck Key On the ther hand, the safety functins that fail t prduce cnsiderable safety benefits are: Tyre Pressure and Mnitring Rear Light Brake Frce Display Advanced Adaptive Frnt Light System Predictive Assist Braking Dynamic Suspensin Safety Safety Effectiveness (%) System Functin Fatalities Saved Serius Injuries Saved Intelligent Speed Adaptin (**) Drive Safe Advanced Autmatic Crash Ntificatin (***) Rescue 10,8 N/A Advanced Adaptive Cruise Cntrl Drive Safe - 11 Cllisin Avidance Drive Safe - 9,1 Cllisin Warning Drive Safe - 6,6 Traffic Sign Recgnitin (*) Drive Safe - 5,82 Lane Keeping Assistant (*) Drive Safe - 5,67 Night Visin Visibility 3,5 4,8 Blind Spt Detectin (*) Drive Safe 2,5 4 Lane Changing Assistant (*) Drive Safe - 3,07 Alclck Key (***,#) Drive Safe 6 3 Drwsy Driver Detectin System Drive Safe - 2,9 Intersectin Cntrl (*) Drive Safe - 2,34 Crnering Brake Cntrl (*) Braking Systems - 2,29 Tyre Pressure Mnitring and Warning (*) Drive Safe - 1,35 Rear Light Brake Frce Display Visibility - 0,8 Advanced Adaptive Frnt Light System Visibility - 0,6 Predictive Assist Braking Braking Systems - 0,2 Dynamic Suspensin Handling/Kinematics - 0 *: The ptential magnitude (target ppulatin) f the effectiveness has been calculated **: The numbers are fr the 'Driver Select' ISA cnfiguratin which has been estimated as the mst effective ***: Results based n nn-eurpean data #: Fr the Alclck Key the results fr the mde "All newly registered vehicles (First full year)" with effectiveness 25% is used which gives the highest results but it is abve the average perfrmance f Alclck key N/A: Nt Applicable Table 3: A-priri evaluatin results fr the 19 studied safety functins Date f Delivery : September

19 Reductin in Serius Injuries Reductin in Fatalities Intelligent Speed Adaptin (**) Advanced Autmatic Crash Ntificatin (***) 11,00% 10,80% 17,00% Advanced Adaptive Cruise Cntrl Cllisin Avidance Cllisin Warning Traffic Sign Recgnitin (*) Lane Keeping Assistant (*) Night Visin Blind Spt Detectin (*) Lane Changing Assistant (*) Alclck Key (***,#) Drwsy Driver Detectin System Intersectin Cntrl (*) Crnering Brake Cntrl (*) Tyre Pressure Mnitring and Warning (*) Rear Light Brake Frce Display Advanced Adaptive Frnt Light System Predictive Assist Braking Dynamic Suspensin 6,60% 5,82% 5,67% 3,50% 4,80% 2,50% 4,00% 3,07% 6,00% 3,00% 2,90% 2,34% 2,29% 1,35% 0,80% 0,60% 0,20% 0,00% 9,10% 11,00% 0% 2% 4% 6% 8% 10% 12% 14% 16% 18% *: The ptential magnitude (target ppulatin) f the effectiveness has been calculated **: The numbers are fr the 'Driver Select' ISA cnfiguratin which has been estimated as the mst effective ***: Results based n nn-eurpean data #: Fr the Alclck Key the results fr the mde "All newly registered vehicles (First full year)" with effectiveness 25% is used which gives the highest results but it is abve the average perfrmance f Alclck key N/A: Nt Applicable Figure 15: Overall diagram f a-priri evaluatin results fr the 19 studied safety functins Hwever, in ttal, fr really cnsiderable safety benefits mre than ne system shuld be available n vehicles. The assessment f a cmbinatin f systems altgether and the assessment f the additinal safety benefits f a system given the presence f ther systems culd be valuable since systems are ften cnsidered as acting independently. The results delivered by the a-priri evaluatin shuld nt be used as a definitive assessment f the viability f the safety systems, but rather, an indicatin f their ptential effectiveness in decreasing related serius injuries and rad fatalities. 2.6 A human-centred apprach fr assessing drivers' needs and cntextual cnstraints f safety functins The methd that was used fr the a-priri evaluatin f the safety functins in terms f drives needs fulfillment was anther innvatin aspect f TRACE and WP4 (in clse cperatin t WP5 "Human Factrs"). Up t nw, mst evaluatin results were fcused n the estimatin f the effectiveness f safety functins in terms f safety increment. In TRACE the selected safety functins were als evaluated regarding their effectiveness in the human/driver needs fulfillment as nt all humans adapt Date f Delivery : September

20 a technlgy in the same way r capability. The ptential limitatins/cnstraints which culd prevent the driver frm taking advantage f infrmatin added t his task by means f driving aid, and which culd lessen the effectiveness f a safety functin, were thrughly examined by in-depth analysis. By taking int accunt the cntexts in which the driver failures ccur, the presented apprach enables the definitin f the cnstraints that the safety systems will have t cpe with, in rder t fulfill nt nly the needs f the drivers but als t cmpensate the characteristics f the situatins in which these needs are met. This wrk is strngly related t WP5 activities and results. This apprach enables the assessment f the effectiveness f safety functins in their capacity t cmpensate fr the malfunctins in the driving systems, which are revealed by accident analysis. It starts frm an ergnmic basis with the purpse f integrating the human functining mdes in this assessment. Frm such a perspective, it is aimed at: Defining the drivers needs in safety, and especially the mst crucial nes; Evaluating the capacity f underway safety functins t fulfill these needs; Defining the cntextual cnstraints that these functins shuld take int accunt in rder t maximise their ptential safety benefits. The methd is based n an in-depth study f a wide sample f accident cases weighted as regard as Eurpean statistical data. This study makes use f WP5 methdlgical results t analyze the ptential safety effectiveness f mst f the safety functins defined in Table 1. Accident data Cntextual parameters Malfunctin prcess Capacity t cmpensate fr cntextual cnstraints Ptential limitatins Human Functin Failures Adaptatin t drivers' needs Cnstraints t integrate Driver's needs Safety functins S a f e t y e f f e c t i v e n e s s Figure 16: Wrkflw diagramm f the prpsed methd fr assessing drivers' needs and cntextual cnstraints f safety functins The analysis f the drivers' needs and cntextual cnstraints is based n the study f a sample f 432 car drivers invlved in a rad accident. This sample was randmly extracted frm INRETS in-depth accident study database (EDA). In rder t be mre representative, this sample has been weighted as regard as Eurpean statistical facts. The needs and cnstraints diagnsed frm accidents are cnfrnted with 21 primary safety functins (see Table 1), plus 5 variants in their functining, defined in TRACE WP6 ("Safety Functins") Date f Delivery : September

21 deliverables, as the mst cnnected t safety, in rder t examine the ability f these functins t cmpensate fr the driver's difficulties. The analysis cnsists f the fllwing steps (see Figure 16): 1. Qualitative definitin and quantitative assessment f the Drivers Needs as they are expressed in accident situatins. The analysis starts frm the characterizatin f the human functinal failures (whereas perceptive, cgnitive r active) fund in accidents, as they are stated in TRACE WP5 (cf. deliverable D5.1). Mst f accidents reveal a difficulty that the driver was nt able t cmpensate fr. The needs can be cnsidered as the mirrr f these human difficulties. By s, they can be defined frm a diagnsis f the real prblems that drivers met in accidents. T be mre peratinal, they are neither inferred frm drivers' wishes nr engineers' assumptins, but frm expertise in accident prducing analysis. 2. Stressing the characteristics f the safety functins examined in TRACE prject and assessment f the Adaptatin f the safety functins t driver needs, i.e. their ptential capacity t address driver needs as they are stated in the previus step. 3. Putting frward the ptential Cntextual Cnstraints fund in accident, which culd lessen the ptimal functining f the safety systems if they are nt truly taken int accunt. S called "ptential limitatins" are defined frm the parameters characterizing the cntexts in which accident ccurred, shwing sme essential parameters t cnsider in rder t ptimize the adaptatin f the systems t accident situatins. 4. Lking at the Respnse Efficiency f the safety functins, i.e. their capacity t cmpensate fr the accident cntextual limitatins diagnsed in the previus step. 5. Prvisin f a cmprehensive result f all the previus nes. It stresses the Safety Effectiveness f the safety functins, i.e. their capacity t fulfill drivers' needs. This safety effectiveness is defined as the cmbinatin f the adaptatin f the safety functins t the needs and their respnse efficiency in cmpensating fr the cntextual limitatins fund in accident situatins. Belw are presented the main results that reflect which needs are well r less well answered by the studied safety functins. Such a view will allw shwing the prgress which have still t be dme in rder t fit mre clsely fr all the safety drivers needs at the different stages f the accident (in rupture but als upstream and dwnstream). Drivers needs which are the mst cmpensated by the safety functins are: B01 Diagnsing driver cnditin (alchl, fatigue, health, attentin, etc.) Upstream: by AK (2.8% f safety effectiveness) In rupture: by AK (5.6% f safety effectiveness) and DDS (2.7% f safety effectiveness) B03 Detecting an unexpected rad difficulty Upstream: by IC (2.3% f safety effectiveness) B04 Detecting a fixed bstacle n the rad In rupture: by CA (2.1% f safety benefit) B05 Detecting a slwly mving bstacle n the rad In rupture: by ACC+ (2.2% f safety benefit), CA (5.0% f safety benefit) and CW (2.3% f safety benefit) B07 Detecting a user n an intersecting curse Upstream: by IC (4.3% f safety effectiveness) In rupture: by IC (5.6% f safety effectiveness) B08 Detecting a user utside the frward field f visin (behind, n the side r in blind spts) Date f Delivery : September

22 Upstream: by BS (2.7% f safety effectiveness) and by LCA (2.7% f safety effectiveness) B11 Adapting speed t the rad - 1: rad gemetry Upstream: by ISA3 (2.9% f safety effectiveness) B12 Adapting speed t rad netwrk - 2: legislatin Upstream: by ISA1 (2.1% f safety effectiveness), by ISA2 (3.3% f safety effectiveness) and by ISA3 (9.8% f safety effectiveness) B17 Predicting that anther user will pull ut r fail t stp In rupture: by CA (1.9% f safety effectiveness) and by IC (1.8% f safety effectiveness) B21 Cntrlling ne s vehicle In rupture: by ESP (4.5% f safety effectiveness) Emergency: by ESP (18.9% f safety effectiveness), by CA (6.8% f safety effectiveness), by PBA (5.7% f safety effectiveness), by BA (4.2% f safety effectiveness), by ACC+ (3.6% f safety effectiveness) and by CBC (2.3% f safety effectiveness) Critical needs which are the less answered, and n which engineers will have t be attentive, are: B15 Assessing gaps when vertaking r changing lanes N safety functin has the capacity t meet this need when it appears in rupture (pivt). B02 Diagnsing vehicle cnditin This need is rarely a pivtal need. Prblems with vehicle state mst ften cntribute t anther mre imprtant failure. Anyway, nly TPMS has the capacity t meet this need. B12 Adapting speed t rad netwrk - 2: legislatin In rupture, this need is weakly cmpensated by ISA2 (0.1% f safety effectiveness) and by ISA3 (0.2% f safety effectiveness), and nt at all by ISA1 (0.0% f safety effectiveness). Figure 17 bellw shws an inverse view f the abve stated results: the capacity f the safety functins t address the pivtal, upstream and emergency driver's needs. A need is qualified as pivtal meaning its critical status as regard as the accident prductin: it figures ut the "pivt" between a cntrlled situatin and an impaired situatin (rupture phase). Needs at the stage when a human failure befrehand has favred the malfunctin encuntered at the rupture level, will be qualified as upstream needs (driving phase). The emergency needs are diagnsed when a functin failure has hindered the driver frm taking ver the situatin met at the rupture stage (emergency phase). Accrding t the results presented in Figure 17 the safety functins which shw the best safety effectiveness in cmpensating fr pivtal needs are CA, IC, CW, AK and ESP. ISA is very effective in cmpensating fr upstream needs in the driving phase, while ESP and CA fr the emergency phase. The added value n this analysis is that it brings helpful input n the real difficulties fund by drivers in accident cntexts, t which electrnic functin culd cmpensate fr. Of curse this research wrk needs additinal wrks in the purpse f prgressing tward a safe driving system, with ther data and ther methds, addressing the technlgical, ergnmic, psychlgical and scilgical aspects t cnsider when implementing a dispsal in an integrated task. Strng effrts are ntably t be put n the technical aspects t imprve fr these electrnic devices t be efficient. Fr example, a well knwn verall essential questin that is addressed t mst f the safety functins remains the capacity t get reliable infrmatin n frictin capacity (rad surface adherence) allwing t setting up the apprpriate actin t initiate n the wheels. The estimatin f the 'frictin ptential' between the car and the rad can be cnsidered as ne f the mst critical challenges fr safety functins benefit. Date f Delivery : September

23 Anther verall essential challenge fr these devices cnsists in the develpment f apprpriate HMI ergnmic wrks t develp s that the safety functins culd be a real aid withut becming an additinal burden n the driver with pssibly cunter effects n safety. In this purpse, it will be claimed that it is useful t define ergnmic standards, such as thse prpsed in the Eurpean Statement f Principles n human machine interface; but it is nt sufficient. Fr such an imprtant questin, it can be simply relying upn a catalgue. It must be cntinuusly cnsidered that each HMI has t be adapted t the users (taking accunt f their hetergeneity and weaknesses) f the device, and t the specificities f the task in which it is integrated. Capacity t meet driver's pivtal needs Capacity t meet upstream needs Capacity t meet emergency needs Cllisin Avidance 4,30% 7,60% 14,10% Intersectin Cntrl 7,80% 12,30% Cllisin Warning 0,20% 2,60% 8,90% Alclck Key 2,80% 5,60% Electrnic Stability Prgram 4,50% 18,90% Advanced Adaptive Cruise Cntrl+ 0,90% 3,60% 3,60% Vulnerable Rad Users Prtectin 0,20% 0,50% 3,10% Drwsy Driver Detectin System Lane Keeping Assistant Night Visin Intelligent Speed Adaptin -3 Advanced Adaptive Cruise Cntrl-2 Lane Changing Assistant Blind Spt Detectin 0,60% 2,80% 0,40% 0,70% 2,10% 1,00% 1,70% 0,20% 1,70% 0,50% 1,70% 2,70% 1,60% 2,70% 1,60% 14,60% Advanced Adaptive Cruise Cntrl-1 Intelligent Speed Adaptin-2 Crnering Brake Cntrl 0,50% 1,60% 0,10% 0,70% 2,30% 0,70% 5,70% Intelligent Speed Adaptin-1 0,40% 3,20% Rear Light Brake Frce Display Tyre Pressure Mnitring and Warning Predictive Brake Assist Dynamic Suspensin Traffic Sign Recgnitin 0,30% 1,00% 0,20% 0,10% 0,10% 0,50% 5,70% Brake Assist Advanced Adaptive Cruise Cntrl -3 0,00% 0,00% 4,20% Advanced Adaptive Frnt Light System 0,20% 0,00% 2,00% 4,00% 6,00% 8,00% 10,00% 12,00% 14,00% 16,00% 18,00% 20,00% The systems are ranked accrding t their capacity t meet pivtal driver's needs Figure 17: Safety functin's capacity t cmpensate fr driver's needs Date f Delivery : September