The definition, production and validation of the …...THE DEFINITION, PRODUCTION AND VALIDATION OF THE DIRECT VIS ION S TANDARD (DVS ) FOR HGVS C onsultation exercise Progress report-

T HE D E F INIT IO N, P R O DUC T IO N AND

VAL ID AT IO N O F T HE D IR E C T V IS IO N

S T ANDAR D (DVS ) F O R HG VS

C ons ultation exercis e - P rogress report

P repared on behalf of T rans port for L ondon

P repared by: Dr S teve S ummers kill Dr R uss ell Mars hall

Dr Abby P atterson Anthony E land

Dr J ames L enard L oughborough Des ign S chool, L oughborough Univers ity

T he definition, production and validation of the direct vis ion s tandard (D V S ) for HG Vs S eptember 2017

T A B L E O F C O N T E N T S

1 Preface ........................................................................................................................... 3

2 Executive Summary ........................................................................................................ 4

3 Acknowledgments .......................................................................................................... 5

4 Introduction ..................................................................................................................... 5

5 Background .................................................................................................................... 6

5.1 Summary of the Lds team’s previous experience in understanding direct vision from HGVs ......................................................................................................................... 6

5.2 Summary of the ‘Definition of Direct Vision Standards for Heavy Goods Vehilces’ project ................................................................................................................................ 9

6 Project aims and objectives .......................................................................................... 11

7 Stakeholder consultation .............................................................................................. 11

8 Accident Data ............................................................................................................... 12

8.1 Results for the national STATS 19 accident database: severity of accidents between HGVs and VRUs ................................................................................................ 13

8.2 Results for the national STATS 19 accident database: Causation data ................ 13

8.3 Results for the national STATS 19 accident database: First point of contact and vehicle manoeuver being performed prior to the accident for accidents where blind spot was reported as the causation factor ................................................................................. 1

8.4 Results for the national STATS 19 accident database: age of the VRU .................. 4

8.5 Summary for the accident data analysis .................................................................. 5

9 Development of the Direct Vision Standard (DVS) ......................................................... 5

9.1 Virtual CAD Based Assessment Methodology ......................................................... 5

9.1.1 The Projection of Direct Field of View .............................................................. 6

9.1.2 Volumetric Assessment of Field Of view ........................................................ 10

9.2 Vehicle Sample ..................................................................................................... 10

9.2.1 Vehicle Classification N3 vs N3G ................................................................... 11

9.2.2 Cab Mounting Height...................................................................................... 12

9.2.3 3D Vehicle Data ............................................................................................. 13

9.3 Vehicle Set Up ....................................................................................................... 15

T rans port for L ondon 1 L oughborough D es ign S chool ©


9.3.1 Eye Point Definition – Provisional .................................................................. 16

9.4 The Definition of the Assessment Volumes ........................................................... 20

9.4.1 The Vertical Specification of the Assessment Volume ................................... 20

9.4.2 The Horizontal Specification of the Assessment Volume ............................... 22

9.5 Results of the Volumetric Projections .................................................................... 24

9.5.1 Volumetric Results For All Vehicles Using Candidate 1 (Untrimmed) Assessment Zone Volume ............................................................................................ 24

9.5.2 Quantifying Volumetric Results Against Real World Performance ................. 25

9.5.3 Correlation Results of Candidate 1 (Untrimmed) Assessment zone Volumes Against VRU Distance .................................................................................................. 27

9.5.4 Volumetric Results For All Vehicles Using Candidate 2 (Trimmed) Assessment zone Volumes ............................................................................................................... 28

9.5.5 Correlation Results of Candidate 2 (Trimmed) Assessment Zone Volumes Against VRU Distance .................................................................................................. 29

9.5.6 Exploring Weighting of the Assessment Zones .............................................. 29

9.5.7 Exploring Ambinocular Projections ................................................................. 31

9.5.8 Summary of Results ....................................................................................... 32

9.6 The definition of the method by which the star ratings would be produced ........... 32

9.6.1 The detailed approach used to find the DVS rating of each vehicle and define the star rating boundaries considering options 1-4 ....................................................... 35

9.6.2 Results for Option 1........................................................................................ 37




9.6.6 Final Selected DVS Option Subject to Consultation ....................................... 40

9.6.7 Summary of the results for the four options of terms of the vehicles analysed which were included in each star boundary .................................................................. 43

9.6.8 Impact of the Lower Door Window on the DVS Rating ................................... 43

10 Validation exercise ....................................................................................................... 44

11 Discussion .................................................................................................................... 44



1 P R E F A C E

T he following report defines the methodology and process for the T ransport for L ondon (T fL ) D irect V is ion S tandard for trucks .

T his report was provided in draft form for review by vehicle manufacturers and other stakeholders in S eptember of 2017.

S ubsequently a key is sue was identified with one of the fundamental variables used in the testing of the DVS . T his is sue was associated with the potential variability in the eye point that is used to determine what is vis ible to a driver. T his needs to be revised and will subsequently require the reanalys is of all vehicles to produce revised scores for the direct vis ion s tandard. T his is described in section 9.3.1.

T herefore there are some aspects of the report which indicate the proposed methodology and s ome aspects which will change based upon the current redefinition process and consultation with manufacturers .

T he report has been edited to highlight the key aspects of the DVS that are currently seen as fixed, but are still open for consultation, and the key aspects which are currently being revised by the L DS team through direct consultation with the T fL s takeholder group, including all vehicle manufacturers .

T he following s ections of the report have been updated to reflect current s ituation in the project.

• E xecutive summary • S ection 9.3.1 • Discus s ion



2 E X E C U T IV E S U MMA R Y

T his report pres ents res earch performed by L oughborough D es ign S chool (L D S ) on behalf of T ransport for L ondon (T fL : tfl_scp_001570). T he res earch has been conducted agains t a background of over repres entation of heavy good vehicles (HG Vs ) being involved in road traffic accidents with vulnerable road us ers (V R Us) where ‘fa iled to look properly’ and ‘vehicle blinds pot’ are often reported as the main casual factors in the accident data. P revious work by L D S on driver’s vis ion from HG Vs has identified the need to reduce reliance on indirect vis ion via mirrors through the s pecification of a direct vis ion s tandard (D V S ) for HG V s . R ecent work commis s ioned by T fL and performed by the T rans port R es earch L aboratory (T R L ) res ulted in a draft D V S . T his draft D V S has been evaluated and reworked by the L D S team to produce a viable and robus t method to quantify direct vis ion performance of an HG V together with a means to rate that vis ion performance agains t a s tar rating s tandard. T hroughout this proces s s ignificant s takeholder cons ultation has been us ed to s upport the development of the D V S .

A total of 21 vehicles repres enting the majority of the current E uro 6 N3 HG V fleet have been modelled in C AD . Where data were available thes e have been mounted at the highest, lowes t and mos t s old heights to produce a s ample of 41 tes t vehicles . A methodology has been developed that utilis es volumetric projection of the field of view of the driver via the windows in the cab. T his projection is then inters ected with an as s es sment volume. T he res ult is a volumetric repres entation of the s pace around a HG V cab that the driver can s ee to the front, driver and pas s enger s ides . T he volume of this s pace can be calculated to provide a rating of direct vis ion performance. An iterative des ign process was followed that explored different s pecifications of the as s ess ment zone around the cab, factoring in the collis ion data with V R Us and the use of weightings to prioritis e what needs to be seen. T wo weighting s chemes were evaluated one prioritis ing the volumes vertically, recognis ing the importance of being able to s ee closer to the ground, and a second prioritis ing the volumes directionally to address the prevalence of accidents being greater to the front and pas s enger s ide when compared to the driver’s s ide. T he final s pecification of the volumetric as s es s ment cons is ts of a s ing le, unweighted zone around the cab, informed by the current coverage of mirrors s pecified in UNE C E regulation 46. T his was done to fos ter direct vis ion that aims to remove the reliance on mirrors and thus should focus on providing direct vis ion of the areas currently covered by mirrors . T he vehicle sample was then evaluated for its performance us ing this assessment, providing a volumetric score for each vehicle. T hese volumetric s cores were then quantified by correlating them with a VR U s imulation. T hirteen 5th % ile Italian female VR Us were placed around the vehicle and moved laterally to a point at which their head and s houlders could be seen. T his served to provide context for the volumetric results such that a particular volume could be equated to an average dis tance at which the small E uropean adult could be s een. F urthermore, the VR U s imulations provided a means to trans late the volumetric performance into s tar ratings .

F our star rating specifications were produced following an absolute (based on risk/safety) and a relative (based on the performance of the current fleet) approach. F or both absolute and relative, two iterations were proposed: 1. the VR U s imulation dis tances were used to establish a thres hold value, 2. the median volumetric result was used to es tablish a threshold value. T he final option taken forwards used the VR U s imulation dis tances for a 5th % ile Italian female to define the 1 star boundary. Vehicles able to provide direct vis ion of the VR Us at an average of <2m to the front, <4.5m to the passenger s ide and <0.6m to the driver’s s ide achieved a s tar rating 1 s tar or above. A ll others achieved a rating of zero star. S tar ratings from 1 to 5 star were sub divided equally.

T he final result cons is ts of three main outcomes:

1. A robust, repeatable and validated method for the volumetric analys is of direct vis ion performance us ing a C AD based process



2. A process to map a volumetric s core for a g iven vehicle onto the 5 star rating scale to produce a DVS rating for any vehicle.

3. S tar ratings for the majority of the E uro 6 N3/N3G HG V fleet showing that of the 41 configurations analys ed, two vehicles are rated 5 s tar, no vehicles are rated 4 star, five vehicles are able to achieve 3 s tar, three vehicles are able to achieve 2 star, and s ix vehicles are able to achieve 1 s tar, the remainder 25 vehicles were rated as zero s tar.

S ubsequently the L DS team identified an is sue with the definition of the eye point location which is used to as the orig in for the projections of vis ible space for the driver. T he process relied upon the definition of a driver eye point as per the s tandard UNE C E reg 46, which in turn relies upon the manufacturers applying the definition of a driver hip point cons istently. An analys is of the hip point locations that were provided by manufacturers identified that there is variability in how the s tandard hip point was being defined. T his led to the exploration of other options for the eye-point definition which has involved a s takeholder meeting held on the 16th of O ctober 2017, along with a number of propos al documents which have been sent for manufacturer review, and further direct contact with vehicle des ign team members . At the time of writing (08-11-17) the L DS team have been through two cons ultation rounds with proposals . T his report presents methodology that has been used to date, and highlights the aspects of DVS definition which are cons idered to be fixed, but open to cons ultation, and the aspects which will vary. S ee section 9.3.1 for a description of the eye point definition options .

T he report uses anonymised results as a method to demonstrate the outcomes of the various methods that have been explored to define a workable DVS which is acceptable to stakeholders . Whils t the redefinition of the eye point will affect these results , the key proces ses for the production of a vehicle rating will be same in the future vers ions of the standard, and so these provis ional, anonymised results have been included in this report for illus trative purposes .

3 A C K N O W L E D G ME N T S

T he project des cribed in this report would not have pos s ible without with cons iderable s upport provided by vehicle manufacturers . E ach manufacturer has a llowed acces s to their vehicles to a llow data to be captured, and provided additional information that relates the vehicle s pecifications . We would like to thank the following manufacturers and their representatives for their s upport, D AF , D ennis , IV E C O , MAN, Mercedes , R enault, S cania and V olvo.

4 IN T R O D U C T IO N

T he following report details the production of a sys tem that can be used to rate the performance of Heavy G oods Vehicles in terms of their ability to allow drivers to see vulnerable road users (VR Us ) s uch as cyclis ts and pedestrians in close proximity to the cab. T his method is known as the D irect V is ion S tandard. In addition to this , the D irect V is ion S tandard has been applied to the asses s ment of 38 vehicle makes and models , and the results of these have been analysed to define a ‘s tar rating sys tem’. T he rating s ys tem us es 6 s tar ratings from 0 to 5 where 0 des ignates a poor direct vis ion performance and 5 des ignates an excellent direct vis ion performance. In this way, the performance of a particular vehicle can be communicated in a method that is s imple to unders tand for operators and other stake holders such as members of the general public.

T his project has been funded by T ransport for L ondon with the aim of improving safety on the s treets of L ondon by limiting acces s to the city to vehicles with the lowest DVS score. In addition, the



aim is to fos ter the us e of vehicles which have improved direct vis ion, and to encourage HG V des ign teams to show further cons ideration of direct vis ion in the des ign process .

T he case for direct vis ion has been made by T fL through a series of research projects which it has funded to explore the reasons for the disproportionate number of accidents which occur between VR Us and HG Vs. T hese include, most notably, previous work by the L DS team which explored the s ize and location of blind spots for a range of exis ting trucks , leading to the recommendation of a direct vis ion s tandard for trucks , and a project by AR UP 1 and L eeds Univers ity, which demons trated that direct vis ion is more effective than the use of mirrors in allowing HG V driver’s to identify the presence of VR Us around a truck cab.

O ver the past 3 years , HG Vs were involved in 20% of pedestrian fatalities and over 70% of cyclis t fatalities , despite HG Vs only making up 4% of road miles in L ondon.

T he following report describes the process that has been undertaken to des ign, tes t and apply the D irect V is ion S tandard.

5 B A C K G R O U N D

T he following section describes a series of research projects which led to the requirement for a D irect vis ion s tandard (DVS ) for trucks .

5 .1 S U MMA R Y O F T H E L D S T E A M’S P R E V IO U S E X P E R IE N C E IN U N D E R S T A N D IN G D IR E C T V IS IO N F R O M H G V S

T he project to define and test the DVS is part of a series of vehicle des ign and assessment projects that have been performed by the L DS team over a period of 15 years including projects with S hanghai Automotive, J aguar L and R over, Honda, T he Department for T ransport (DfT ), Nis s an T echnical Development C entre E urope, T ransport and E nvironment (T &E , 2017) and T rans port for L ondon. As an example of the kind of work that is performed by the team, the project with Nis s an, funded by Innovate UK and completed in 2015 involved the des ign and production of a prototype electric vehicle that allows a novel, more space efficient driving posture to be adopted. T his work and other projects performed by the L DS team involve the use of the only UK based D ig ital Human Modelling s ys tem, S AMMIE (S ys tem for A iding Man Machine Interaction E valuation), which has been developed by the L DS team at L oughborough Univers ity s ince 1980. T he use of dig ital human modelling allows vehicles to be s imulated along with the interactions between drivers , vehicles and the wider environment.

1 AR UP (2016). E xploring the R oad S afety B enefits of D irect vs Indirect V is ion in HG V C abs . http://content.tfl.gov.uk/road-s afety-benefits -of-direct-vs -indirect-vis ion-in-hgv-cabs -technical.pdf



F ig ure 1. T he projec tion of the C las s V (look down) mirror (the red v olume) and the s pac e v is ib le throug h the pas s eng er window (the orang e v olume)

T he L DS team first used S AMMIE to model HG V driver’s vis ion in 2010, in a project for the DfT which resulted in the identification of a key HG V blind spot. T his led to the L DS team supporting the DfT by presenting the research to the 100th meeting of the United Nations E conomic C ommiss ion for E urope G eneral S afety C ommittee which led to the revis ion of UNE C E regulation 46, improving the coverage of the C las s V HG V mirror. T his change came onto force in J uly of 2015, with all new trucks in E urope being required to be fitted with an improved C lass V mirror. T he DfT project report produced by the L DS team in 20112 discussed the potential difficulty that HG V drivers face in the use of s ix mirrors and upwards of three windows in gaining s ituational awareness of VR U locations around the cab, and advocated the need for improved direct vis ion in HG Vs . T he C L O C S ‘Modelling of HG V B lind S pots ’ project followed on from the DfT research, us ing a further developed vers ion of the S AMMIE C AD DHM sys tem which allows the volume of space that is vis ible to the vehicle driver through windows and mirrors to be visually represented and numerically quantified. S ee F igure 1.

2 S UMME R S K IL L , S . ...et al., 2015. Understanding direct and indirect driver vis ion in heavy goods vehicles : F inal report prepared on behalf of T ransport for L ondon . L oughborough: L oughborough Univers ity, pp. 1-432.



F ig ure 2. Imag es of the v ehic le mak es and models that were produc ed during the projec t

T his technique was been used in conjunction with 19 vehicle models (see F igure 2) that have been produced by the L DS team which represent the most sold configurations of HG Vs in the UK along with cabs with ‘high vis ion’ des igns based upon low entry cabs . T he combination of the unique vis ualisation features of S AMMIE and the 19 vehicle models , which have been specifically produced to benefit from these features , has resulted in the analys is of the areas around a vehicle cab which can and cannot be seen by a driver through the use of windows (direct vis ion) and mirrors (indirect vis ion). T hese data and vis ualis ations allow vehicle operators to identify vehicles which have improved direct vis ion compared to the res t of the vehicle sample and allow vehicle manufacturers to identify how they perform in comparison to competitors . T he further analys is of the results identified the key features of vehicle des igns which contribute to the s ize of areas around the vehicle which cannot be seen by the driver (blind spots). T he predominant factor was shown to be the mounting height of the vehicle cab above the ground which is affected by the vehicle specification. F or example, an N3G cab will generally have a higher maximum height compared to the N3 variant. However, it was also s hown that the wors t performing vehicle was a category N3 des ign, illus trating that N3G configurations are not necessarily the wors t performing on the road when cabs are ass essed at their most sold height. O f the 10 wors t performing vehicles 5 were N3 and 5 were N3G . S ee F igure 3.



F ig ure 3. T he c omparis on between the driv er’s eye heig ht and the dis tanc e away at whic h a c y c lis t adjac ent to the pas s eng er door c an be hidden from the direc t v is ion of the driv er

T he variability in the des ign of vehicle cabs with regard to these factors illus trated the need for a direct vis ion s tandard with the aim of providing a mechanism for vehicle des igners to optimise their des igns . T his led to the current project to develop the DVS .

5 .2 S U MMA R Y O F T H E ‘D E F IN IT IO N O F D IR E C T V IS IO N S T A N D A R D S F O R H E A V Y G O O D S V E H IL C E S ’ P R O J E C T

T he recommendation for a D irect V is ion S tandard by the L DS team led to T fL commiss ioning the T ransport R esearch L aboratory (T R L ) to define a D irect V is ion S tandard protocol in 2015. T his draft vers ion3 of the D irect V is ion S tandard established (T R L Draft DVS ) a base principle of defining a volume of space around the vehicle, with the proportion of that volume that is not vis ible to the driver being the performance metric. F igure 4 shows the definition of the T R L ‘assessment volume’ placed around the truck (top left, right and bottom right images) and the result of the subtraction of the volume that can be seen by the driver (bottom left). In addition, T fL prescribed that the results of the DVS would be demonstrated through the mechanism of a five s tar rating sys tem. At the s tart of the project the definition of the s tar rating sys tem was as per the lis t below;

• Z ero star – Vehicles with this rating would be banned from L ondon in the year 2020 • 1 s tar - Vehicles with this rating would be banned from L ondon in the year 2024 • 2 s tar - Vehicles with this rating would be banned from L ondon in the year 2024 • 3 s tar - Vehicles would be acceptable for use in L ondon • 4 s tar - Vehicles would be acceptable for use in L ondon • 5 s tar - Vehicles would demonstrate best in class direct vis ion performance

T he L DS reviewed the outcomes of the ‘Definition of D irect V is ion S tandards for Heavy G oods Vehicles ’ project and identified a number of areas where the definition warranted further refinement.

3 R obins on et al (2015). Definition of Direct V is ion S tandards for Heavy G oods Vehicles (HG Vs ) T echnical R eport



F ig ure 4. T he T R L draft v ers ion of the D irec t v is ion s tandard

T hese were;

1. T he analys is and testing of the DVS had been performed us ing vehicle C AD data sourced from internet s ites that provide visualisation models . T herefore the validity of the C AD data could not be determined

2. T he number of vehicles used in the process included three vehicle des igns compared to, for example, the 19 vehicle des igns tested in the ‘Modelling of HG V B lind S pots ’ L DS project, and so a wider sample was required to ensure the definition of the standard was applicable to the whole fleet. T he three vehicles were modified in the T R L study by removing the das h board to provide more conditions for the tes ting, a des ign change that would not be poss ible. T he results for the three vehicles were used to define the star rating boundaries , which again required wider cons ideration for the whole fleet.

3. T he assessment zone did not cover the offs ide of the vehicle 4. T he assessment zone did not cover the full height range down to the floor. Whils t the

accident data analys is does not indicate a high proportion of VR U accidents associated with children, wheel chair users , or scooter users , the potential to include these VR Us of all heights in the assessment zone was seen as necessary by the L DS team.

5. T he assessment zone definition in terms of the dimens ions used was based upon a mix of data from different nationalities (North American, for the dimens ion 0.3m shown in F igure 4, and UK data for the dimens ions 1.87m, 1.41m and 0.93m in F igure 4). G iven the population of L ondon and UK , data which reflects the anthropometric variability of the E uropean population would be preferred.

6. T he plan view dimens ions of the assessment zone were defined by the average walking speed of pedestrians , and the average cycling speed (which defines the 10m dimens ion in F igure 4). T his was done in an effort to define the s tarting point of pedestrians and cyclis ts at the point at which they should be seen by a driver, including the need for a zone which could not currently ever been seen through direct vis ion by the driver. T he specific scenarios us ed covered a limited number of accidents , and a limited range of actual vulnerable road us er capability to move at a certain speed. T herefore, the L DS team determined that another approach should be taken to the definition of the assessment volume.

7. A method which subtracts the volume that is not vis ible to the driver from the assessment volume produces a contiguous volume (see F igure 4 bottom right image) that cannot be s ubdivided to determine the performance of the vehicle to the front, left and right separately.



T he L DS team were then commiss ioned by T fL to perform a project which would redefine the DVS from base principles and to validate and test this s tandard us ing a sample of vehicles that reflect the UK vehicle parc. In addition, the L DS team would redefine the method by which the star ratings are defined. T he following section presents the formal aims and objectives of the projective.

6 P R O J E C T A IMS A N D O B J E C T IV E S

T he project requirements were lis ted in the project IT T document as follows .

• G ain an ins ight into the profile of exis ting HG Vs (manufacturer and model) that are currently operating on G B roads to es tablish a representative sample to be modelled agains t the DVS

• R eview, validate and test the DVS assessment protocol to ensure the star rating thres holds are realis tically calibrated, redefining the DVS as required

• E ns ure the DVS assessment protocol is legally defens ible and stands up to the scrutiny of regulators , vehicle manufacturers and representative bodies

• Model a baseline representative sample of HG Vs agains t the DVS s tar rating that is relevant to their prevalence in the current G B operating market

• S ecure the support of the eight principal HG V manufacturers that supply vehicles to the G B market, to encourage application and adoption of the DVS

Upon the winning the contract the L DS team further detailed the project objectives as shown below.

• T o perform an analys is of the accident data both nationally and with respect to accidents in L ondon to es tablish the area of greatest risk around the truck

• T his analys is was specified to use the UK S T AT S database with the additional us e of ‘causation’ data which is collected at the scene

• T o explore the potential of us ing the accident data and the areas of greates t risk to ‘weight’ the volumetric scores for particular areas around the truck cab

• T o develop a C AD based methodology from base principles and develop a range of candidate assessment volumes to be tested with a full sample of vehicles

• T o define the required sample us ing vehicle regis tration data • T o engage with vehicle manufacturers and other s takeholders to explain the DVS process

and secure access to C AD data for the vehicles , or access to vehicles for 3D scanning • T o test the candidate rating schemes against the sample and refine the assessment

methodology based upon the results that are achieved • T o define a method that quantifies the abstract nature of the assessment volume scoring

s ys tem us ing a methodology that reflects the real-world problem • T he use of the combination of all data to define a s tar rating sys tem

T he following sections section describe how each of the objectives have been achieved.

7 S T A K E H O L D E R C O N S U L T A T IO N

T he project has been supported by the vehicle manufacturers and other stakeholders through a s takeholder consultation process . A range of engagement activities have been performed to ens ure that the L DS and T fL teams were aware of the des ign processes and constraints involved in vehicle des ign, aware of the efforts that certain manufacturers already place into the des ign of vehicles to improve direct vis ion and aware of range of types of trucks that are produced.



Many of the manufacturers had reviewed the draft T R L DVS prior to the project s tarting, and s o it was important to clarify with manufacturers that this was not the definitive vers ion, and inform them of the further work that L DS team would be performing to move the DVS forward. T he following in a chronological lis t of vis its and meetings that have been held during the course of the project.

1. Dr S teve S ummerskill vis ited the S C ANIA development centre in S tockholm, S weden, in S ept of 2016 and presented the L DS previous work, along with the draft DVS , to the whole development team of engineers and human factors/HMI engineers

2. Dr S teve S ummerskill (L DS ), Hannah White (T fL ) and B en P lowden (T fL ) and Dr R uss ell Marshall (L DS ) vis ited the Volvo development team (headed by Avedal C laes ) in G othenburg, S weden, in December of 2016. R epresentatives from R enault T rucks were als o present. T his vis it included a review of how the Volvo s taff had tried to implement the draft T R L DVS and had identified the is sues discussed in section 9.5.7 relating to the difficulty associated with combining ambinocular projections , and the complexity of the eye point projection used in the DVS .

3. T he DAF development team (headed by J ohan B roeders ) vis ited the L DS team at L oughborough Univers ity. T fL s taff Hannah White and G len Davies were also in attendance. T he DAF team raised a number of is sues for cons ideration including the same is sues rais ed by the Volvo/R enault teams , and the is sues of how the DVS would deal wind screen wipers .

4. Dr S ummerskill V is ited P hil R ootham (S cania, Milton K eynes ) in F ebruary of 2017. 5. A s take holder event took place at L oughborough Univers ity in March of 2017 during which

progress regarding the accident data analys is and the definition of candidate DVS schemes was presented to representatives of all manufacturers , T he UK Department for T ransport, T he S MMT , and the F T A.

6. Dr S teve S ummerskill (L DS ) and Hannah White (T fL ) vis ited the Daimler cab development team (L ed by S tefan Huegin) in S tuttgart in March of 2017 in an effort to secure C AD data for the Mercedes vehicles . T his was a successful meeting which improved the understanding of the Mercedes range by the L DS team and led to access to C AD data.

7. A s takeholder event was held in L ondon in J une of 2017 with representatives of all manufacturers , T he S MMT , T he UK Department for T ransport, AR UP and the F T A. During this meeting a draft vers ion of the DVS process and results were presented. T his presentation was widely accepted with a pos itive attitude from manufacturers with s ome s uggestions for improvement in the process including the use of a DVS rating to all s ides of the vehicle as discussed in section 8.

In addition to these face to meetings there have been regular phone calls and emails to manufacturer representatives and members of des ign teams . T his included a valuable validation exercise performed with one manufacturer in S eptember of 2017 as discussed in section 10.

8 A C C ID E N T D A T A

T he accident data analys is s tarted with the sourcing of the S T AT S 19 data base results from the UK Department for T ransport (DfT ).. T his analys is was performed for all accidents between 2010 and 2015 due the quality of the data that is available for that date range compared to earlier available data. Nationally this involves 2443 accidents . E ach accident is categorised and recorded by a police officer us ing the S T AT S 19 accident recording form which is used when someone has been injured or killed on the highway. T here are numerous accident categories including data on accident causation which had to specially requested by the accident data analys t. T he accident databas e that we have in the UK is widely regarded as being the most detailed in E urope. T he data contains a



wide range of fields which can be used to explore specific is sues . F or this project we have us ed the following fields ;

• Accident causation data (e.g. B lind spot, did not look properly etc. ) extra layer of data that must be requesting from DfT

• T he severity of the accident (F atal, serious or s light) • T he vehicles and other people involved (e.g. HG Vs above 7.5 tonnes, pedestrian, cyclis t) • T he P olice force which has captured the data which allows us to compare the data in L ondon

and nationally • F irs t point of contact between the vulnerable road user and the vehicle • C ategory of vehicle (e.g . rig id or articulated) • Vehicle manoeuvre being performed when the accident occurred (e.g. turning left, going

s traight on) • J unction type • Vehicle make and year of firs t registration • S peed limit on the road where the accident occurred • L ighting and weather conditions • Age of the casualty

T he analys is of the UK accident database (S T AT S 19) for accidents between Vulnerable road us ers and HG Vs above 7.5 tonnes has been performed T he process ing of the accident data allows the analys is of s pecific scenarios , such as the number of accidents that occur between HG Vs and cyclis ts or pedestrians , the severity of those accidents , the firs t point of contact between the vehicle and the VR U, and the causation that has been ass igned by the attending police officer. T he following s ections describe the manner in which the data has been interrogated and the results which were derived.

8 .1 R E S U L T S F O R T H E N A T IO N A L S T A T S 19 A C C ID E N T D A T A B A S E : S E V E R IT Y O F A C C ID E N T S B E T W E E N H G V S A N D V R U S

T he analys is of the res ults initially highlighted the severity of the accidents that occurred between VR Us and HG Vs through a comparison to the whole S T AT S 19 database for all accident types .

F or accidents between HG Vs and cyclis ts and HG Vs and pedestrians the following figures were derived for the number and percentage of accidents in the sample that were rated as fatal, serious or s light.

• C yclis ts : 93 fatal (8% ) – 336 s erious (27% ) – 773 – s light (64)% • P edes trians : 226 fatal (18% ) – 362 serious (29% ) – 653 – s light (53)%

T herefore 35% of accidents between HG Vs and cyclis ts and 47% of accidents between HG Vs and pedes trians involve some being killed or serious ly injured (K S I). T his was compared to the whole S T AT S 19 database for all accident types which showed the following results .

• 1732 fatal (1% ) – 22137 S erious (12% ) – 162340 s light (87% ) T herefore 13% of all accidents recorded in S T AT S 19 involve someone being killed or serious ly injured. C learly the consequences of an accident between an HG V and a VR U are more severe.

8 .2 R E S U L T S F O R T H E N A T IO N A L S T A T S 19 A C C ID E N T D A T A B A S E : C A U S A T IO N D A T A



T he caus ation data that has been requested from the DfT is shown in the table this section.

T able 1. T he c aus ation data for ac c idents between HG Vs and P edes trians

T able 2. T he c aus ation data for ac c idents between HG Vs and C y c lis ts

T able 1 shows the causation data for all accidents between HG Vs and pedestrians . T his shows that the predominant causation factor linked to these accidents were “F ailed to look properly” and " Vehicle blind spot”. T able 2 shows the causation data for all accidents between HG Vs and cyclis ts . T his shows that the predominant causation factor linked to these accidents was “F ailed to look properly” with a high occurrence of " Vehicle blind spot”.

Whils t it is pos s ible that s ome of the cases labelled as “F ailed to look properly” involved a blind s pot without the knowledge of the P olice officer, the L DS team have not made such assumptions .

T able 3 shows the causation data for accidents between HG Vs and cyclis ts for L ondon in the stated period. Here we see that the main causation factor was " Vehicle blind spot” with a high occurrence of “F ailed to look properly”.



T able 3. T he c aus ation data for ac c idents between HG Vs and C yc lis ts in L ondon

In either the case for the national data or the L ondon data, a key is s ue that has been highlighted by the causation data is that a lack of ability to see the VR U can be seen a contributory factor in accidents .

8 .3 R E S U L T S F O R T H E N A T IO N A L S T A T S 19 A C C ID E N T D A T A B A S E : F IR S T P O IN T O F C O N T A C T A N D V E H IC L E MA N O E U V E R B E IN G P E R F O R ME D P R IO R T O T H E A C C ID E N T F O R A C C ID E N T S W H E R E B L IN D S P O T W A S R E P O R T E D A S T H E C A U S A T IO N F A C T O R

F ollowing on the from the analys is of the severity of the accidents , the firs t point of impact data were interrogated. T hese data are important as they highlight the location of the VR U in respect to the HG V at the point of impact. In the case of pedestrians they highlight the locations around the vehicle in which the pedestrian should have been seen by a driver us ing either direct vis ion through the windows or indirect vis ion through the mirrors . F or cyclis ts , the potential variation of the speed of the cyclis t make the interpretation of the cyclis t location seconds before impact more uncertain. F or example, if the firs t point of impact is the left hand s ide of the cab adjacent to the cab there are two example accident scenarios cons idered. T he firs t is that the cyclis t was s tationary and next to the vehicle at the point of impact (not seen by the driver in the C lass V mirror or windows). T he s econd is that the cyclis t was approaching the HG V from the rear at speed, and down the left hand s ide of the cab. T he firs t s ituation can potentially be improved by better direct vis ion, the second will rely upon the ability of the driver to look in the left hand C lass II and C lass IV mirrors at the correct time to see the cyclis t.

F or the national data, firs t point of impact for cyclis ts = 16% to the front, 74% to passenger s ide and 11% to the driver’s s ide

F or the national data, firs t point of impact for pedestrians = 56% to the front, 35% to passenger s ide and 9% to the driver’s s ide



P edestrian firs t point of contact

C yclis t firs t point of contact C ombined P ercentage

F ront 54 22 76 32 Driver’s s ide 9 15 24 10 P as s enger s ide 34 104 138 58

T otal 97 141 238 F ig ure 5. Number of ac c idents with a firs t point of c ontac t to the front, offs ide and nears ide of the

v ehic le

T he cyclis t and pedestrian data were combined to produce the percentage of accidents occurring in each of three zones as shown in F igure 6.

F ig ure 6. T he c ombined % of ac c idents whic h oc c ur to the front, left and rig ht of a rig ht hand driv e v ehic le in the UK

0

20

40

60

80

100

120

F ront O ffs ide Nears ide

Number of accidents with a firs t point of contact to the front, driver's s ide and passenger of the vehicle

pedestrian firs t point of contact C yclis t firs t point of contact



During the initial s tages of the project it was anticipated that these percentages could be used as weighting factor for the volumetric res ults for trucks . S ection 9.5.6 shows the exploration of the benefits of these weightings .

T he vehicle manoeuvers that are being performing during accidents with vulnerable road users are presented below for the cases where a blind spot was specified for the causation factor for national data.

Vehicle manoeuver prior to accident: C yclis t No of incidents Manoeuver % for each

type 71 T urning left 46.71 20 G oing ahead other 13.16 19 Moving off 12.5

13 O vertaking moving vehicle - offs ide 8.55

10 T urning right 6.58 6 P arked 3.95 3 C hanging lane to left 1.97

3 G oing ahead left-hand bend 1.97

2 Waiting to go - held up 1.32 1 R evers ing 0.66 1 S lowing or s topping 0.66 1 U-turn 0.66

1 O vertaking static vehicle - offs ide 0.66

1 O vertaking - nears ide 0.66 T able 4. Vehic le manoeuv er being performed during the ac c ident: C yc lis t

Vehicle manoeuver prior to accident: P edestrian No of incidents

Manoeuver % for each type

52 Moving off 41.6 25 R evers ing 20 25 G oing ahead left-hand

bend 20

16 G oing 12.8 15 T urning left 12 9 T urning right 7.2 2 Waiting to go - held up 1.6 2 Waiting to turn left 1.6 2 G oing ahead left-hand

bend 1.6

2 G oing ahead other 1.6 1 U-turn 0.8 1 G oing ahead right-hand

bend 0.8

T able 5. T he c aus ation data for ac c idents between HG Vs and C yc lis ts in L ondon



T he data for the manoeuvre that the HG V is making during the accident indicates that for cyclis ts the s ituation with the most casualties is associated with the vehicle turning left at junctions . B ased upon the firs t contact data it can be inferred that left turning accidents where the cyclis t is on the left hand s ide of the vehicle are common accident s cenarios .

T he data for the manoeuvre that the HG V is making during the accident indicates that for pedes trians the s ituation with the most casualties is associated with the vehicle moving off from a s tand still. Again this matches with the firs t contact data which shows that the front of the vehicle is the mos t common firs t point of contact.

8 .4 R E S U L T S F O R T H E N A T IO N A L S T A T S 19 A C C ID E N T D A T A B A S E : A G E O F T H E V R U

T he final analys is explored the age of the VR Us involved in accidents . T his a finding of interes t. T he graph in F igure 7 shows a peak in the data for cyclis ts around the age of 25, as might be expected for the age of cyclis ts that are willing to commute in the capital. T here is also an increase in pedes trian casualties for pedestrians above the age 60. However, there is no reason that pedes trians should have a higher prevalence of over 60’s . It can be inferred that people over the age of 60 are disproportionally represented when one cons iders the most common point of contact and manoeuvre, i.e. to the front, and moving off respectively. T herefore, there is the potential for older people to be unable to move out of the way of truck that is moving off at a pedestrian cros s ing or junction.

F ig ure 7. T he number of VR U inv olv ed in ac c idents where blind s pot was attributed as the c ontributory fac tor by ag e

0

5

10

15

20

25

30

0 10 20 30 40 50 60 70 80 90 100

C yclis t and pedestrian casualities by age National 2010-2015 where accident is attributed to blind spot

Number cyclis t casualties by age2010-2015 (B lind S pot) Number pedestrian casualties by age 2010-2015 (B lind S pot)

Age

num

ber o

f VR

Us



8 .5 S U MMA R Y F O R T H E A C C ID E N T D A T A A N A L Y S IS

T he accident data analys is has highlighted the following is sues;

• T he s everity of an accident between VR Us and an HG V is higher than the severity of accidents in general with 35% of accidents between HG Vs and cyclis ts and 47% of accidents between HG Vs and pedestrians involve someone being killed or serious ly injured, compared to 13% for all accidents

• F or the national data as sociated with cyclis t and pedestrian accidents the most common causation factors are associated with not seeing the VR U

• T he frontal area is the most common firs t contact point for pedestrians • T he P as s enger s ide area is the most common firs t contact point for cyclis ts • T he combination of the cyclis t and pedestrian firs t contact points creates data which can be

used to determine the areas of greatest risk around the vehicle, with the nears ide having 58% of accidents , the front have 32% of accidents . 10% of accidents have a firs t contact point on the offs ide

o T hes e data have the potential to be applied as weightings to the volumetric data if the volumetric data is subdivided to the front, right and left of the cab

• T he most common manoeuvre being performed during an accident is moving off for pedestrians and turning left for cyclis ts

• T he age of pedestrian casualties is disproportionately skewed to the older population. T hes e summary points have been used in the development of the DVS as described in the following sections .

9 D E V E L O P ME N T O F T H E D IR E C T V IS IO N S T A N D A R D (D V S )

As introduced in S ection 5.2 the initial method for es tablishing the DVS exhibited the following characteris tics :

• A virtual as s es sment • A volumetric ass es sment of the field of view • T he volume would be limited to key areas around the vehicle

T he following sections detail the development of the DVS from these characteris tics through to the final proposal.

9 .1 V IR T U A L C A D B A S E D A S S E S S ME N T ME T H O D O L O G Y

T he Draft T R L DVS protocol developed by T R L 4 that details the initial DVS approach outlines the use of a C AD bas ed methodology that provides a virtual assessment of direct field of vis ion performance. P revious work by L DS including blind spot evaluations for the UK Department for

4 R obins on, T ., et al., (2015). Definition of D irect V is ion S tandards for Heavy G oods Vehicles (HG Vs ) T echnical R eport



T ransport5 and understanding direct and indirect vis ion from HG Vs for T fL 6 demonstrated the efficacy of this virtual approach. B uilding upon that work the DVS would also exploit a virtual assessment us ing a C AD based method that would allow the direct field of view from a 3D model of any HG V to be evaluated and rated agains t the standard. It was als o deemed to be important to develop a method that would be access ible by the industry. In previous work the DHM tool S AMMIE , a tool developed by the L DS team, was used for the majority of evaluations . However, it was cons idered important for this project to develop generic methods us ing industry s tandard tools that would be applicable by all s takeholders .

9 .1 .1 T H E P R O J E C T IO N O F D IR E C T F IE L D O F V IE W

T he evaluation of field of view in a C AD environment cons ists of a number of key elements . T hes e elements cons is t of: an eye point, a series of apertures and / or mirrors from which to project the field of view, and a means to extrapolate from the eyepoint, through the apertures (or mirrors ) to produce a projection of the vis ible area or volume of space afforded the driver as shown in F igure 8.

F ig ure 8. Volumetric projec tion of the nears ide window illus trating the projec tion of the v is ib le v olumes around the obs c uration c aus ed by the mirrors .

F or the DVS the projections would only be assess ing direct vis ion (via windows) and so mirrors would not be projected as part of the field of view, though they would be cons idered as part of the obscuration to direct vis ion where they impinge on the field of view from the windows . As already dis cussed it was important to develop an access ible methodology that could be implemented us ing equivalent tools in industry s tandard C AD sys tems . T he process developed below illus trates the

5 C ook, S . E ., et al., (2011). T he development of improvements to drivers ' direct and indirect vis ion from vehicles . P has e 2. L oughborough Univers ity. R eport for Department for T ransport. 6 S ummerskill, S ., et al., (2015). Understanding direct and indirect driver vis ion in heavy goods vehicles . L oughborough Univers ity. R eport for T rans port for L ondon.



method developed for the volumetric projection of direct vis ion demonstrated in the C AD software R hino7.

T he process begins with the definition of an eye point. T he process for defining an eye point is detailed later in S ection 9.3.1. F rom this eye point an outline must be produced that represents the limits of a g iven aperture through which the driver would be able to see. F igure 9 shows the view from the eye point through the windscreen. F or the projection, the inner path of the vis ible portion of the windscreen g lass needs to be traced taking account of any obscuration from mirrors , s teering wheel, windscreen wipers etc. F igure 10, F igure 11, F igure 12 and F igure 13 show how this is performed in the R hino C AD software. T his process uses a tool called MeshO utline to automatically trace the relevant area. T he way in which this is performed is not relevant in this ins tance but it is important that the continuity of the curves is high with sufficient points to capture the complex curvature accurately. In addition, care must be taken to outline all vis ible areas such as thos e between the wipers and the frit and through the s teering wheel. T iny areas such as those shown in F igure 11 caused by small gaps in the wiper arms can be ignored as their contribution to direct vis ion is effectively zero.

F ig ure 9. V iew from the ey e point throug h the front winds c reen

F ig ure 10. R elev ant g eometry s elec ted for outlin ing

7 R hinoceros . R obert McNeel & As s ociates (2017). Available at: https ://www.rhino3d.com/



F ig ure 11. O utline of relev ant g eometry after us e of the R hino Mes hO utline tool

F ig ure 12. O utlines reduc ed to the v is ib le areas us ing the R hino C urv e B oolean tool

F ig ure 13. Illus tration of the v is ib le outlines mapped onto the C A D g eometry

O nce the vis ible area has been defined this is converted into the volumetric projection. In the example tool R hino this involves extruding the curves back to the eye point as shown in F igure 14 and F igure 15. O nce the bas ic projection has been produced it can be scaled from the eye point to extend the projection if required.



F ig ure 14. Volumetric pro jec tion of the v is ib le outline from the ey e point us ing the R hino E x trude P lanar C urv e to P oint tool

F ig ure 15. Illus tration of the v olumetric projec tion ex tending from the winds c reen area of the c ab

T he process is then repeated for each window including the driver’s s ide, passenger s ide and any lower door window that may be present as shown in F igure 16.



F ig ure 16. C omplete s et of projec tions for the three windows of this v ehic le

9 .1 .2 V O L U ME T R IC A S S E S S ME N T O F F IE L D O F V IE W

In order to provide a volumetric evaluation of the field of view from the vehicle the projections as detailed in S ection 9.1.1 would need to be intersected with an assessment volume. In contrast to the method proposed in the T R L Draft DVS , the process here would evaluate the vis ible volume, rather than the volume that cannot be seen. T he advantages of this approach are that three dis crete volumes are produced that allow weighting to be applied for front, left and right if required. In addition, they produce much more access ible results in which visualisations are much more meaningful if s takeholders are presented with an image of what can be seen rather than a more abstract image of what cannot be seen as shown in F igure 17. T he projection and the assess ment volume are intersected us ing a common C AD operation called a B oolean operation. T he res ulting geometry can then have its volume calculated to provide a rating. T he development of the assessment volumes is detailed in S ection 9.4

F ig ure 17. E x ample of the v olumetric as s es s ment proc es s with an as s es s ment s howing what c annot be s een on the left and what c an be s een on the rig ht

9 .2 V E H IC L E S A MP L E

T he definition of the sample of HG Vs to be used in the project had two main requirements . T he firs t was to ensure that the sample was sufficiently broad to ensure that any method developed for the evaluation of direct vis ion performance was capable of assess ing all configurations of vehicles . T he second was to as s es s the current mos t recent (E uro 6) HG Vs on UK roads . During the course of



the project these two requirements were combined and it was decided that the most expedient approach was to evaluate the broadest range of vehicles repres enting all of the major manufacturers . T he L DS team sourced sales data from the S ociety of Motor Manufacturers and T raders (S MMT ) with the summary of sales number being produced as per T able 6.

Manufac turer S ales D AF T R UC K S 9,092 ME R C E D E S 5,989 S C ANIA 5,837 VO L VO 5,281 MAN 3,570 R E NAUL T T R UC K S 2,589 IVE C O 1,291 D ennis 821 O ther 111

T able 6. 2014 s ales data for C ateg ory N3 truc k s from S MMT

T he final vehicle s ample lis t included vehicles from all of the manufacturers lis ted in T able 6 and represented over 99% of the HG V fleet for C ategory N3 vehicles . T he full lis t is shown below:

• DAF X F • DAF C F • DAF L F • Dennis E agle E lite 6 • Iveco • MAN T G X • MAN T G S • Mercedes 2.5m C ab (Actros , Arocs , Antos ) • Mercedes 2.3m C ab (Actros , Arocs , Antos ) • Mercedes Atego • Mercedes E conic • Volvo F H • Volvo F M • Volvo F MX • Volvo F L • Volvo F E • Volvo F E L E C • R enault T • R enault K • R enault C 2.5m C ab • R enault C 2.3m C ab • R enault D • S C ANIA R • S C ANIA G • S C ANIA P

Note that the project only assessed R ight Hand Driver (R HD) vehicle configurations .

9 .2 .1 V E H IC L E C L A S S IF IC A T IO N N 3 V S N 3 G



O ne of the s tarting points of this research concerned the over representation of construction vehicles in K S I accidents between HG Vs and vulnerable road users . F or HG Vs those des igned for a construction role are des ignated as N3G . After a thorough review of the vehicle fleet it was determined that in all cases the cab of an N3G vehicle is identical to its N3 (dis tribution) variant in the features that would affect direct vis ion e.g. seating pos ition (eye point), window apertures and das hboard. T he only differences demonstrated by some manufacturers is the inclus ion of modified mirrors and mirror mounting arms between the N3G and N3 variants . However, the largest differentiator between N3G and N3 vehicles was found to be their cab mounting height as shown in F igure 18.

F ig ure 18. Mounting heig ht differenc e between the lowes t pos s ible D A F C F N3 (left) v s the lowes t pos s ible D A F C F N3G (rig ht)

9 .2 .2 C A B MO U N T IN G H E IG H T

F or the majority of HG Vs the cab can be mounted at different heights from the ground. T here are a number of variables including axle configuration, suspens ion type, and tyre profiles that can affect this height. In addition, whether the vehicle is laden or unladen also has an effect. As an example, a s ingle vehicle may have the ability to vary by 320 mm in height due to these variables .

P revious work by the L DS team as part of the Unders tanding D irect and Indirect V is ion from HG Vs project demonstrated that height correlates directly with direct vis ion performance as shown in F igure 19.



F ig ure 19. T he c omparis on between the driv er’s ey e heig ht and the dis tanc e away at whic h a c y c lis t adjac ent to the pas s eng er door c an be hidden from the direc t v is ion of the driv er

T hus , it was important that cab mounting height was addressed in the method and ultimately the DVS with an appropriate analys is s ample selection. F or expediency it was initially proposed to primarily evaluate the vehicle sample at their most sold height. T his approach had been successfully utilised previous ly providing a means to manage the range of potential vehicle mounting heights and thus the large number of evaluations that may need to be performed for each vehicle in the sample and a way of representing the most prolific vehicles on the road. In addition, the maximum and minimum mounting heights were to be evaluated for one cab from two manufacturers to explore the impact of the mounting height on performance and to ensure that the method was sufficiently flexible to accommodate mounting height variability.

During the project it was found that cab mounting height played such a s ignificant role that it was deemed important to evaluate all vehicles at the maximum and minimum mounting heights in addition to their most sold configuration. T his provided a much greater unders tanding of the performance of the fleet and was ultimately important for defining the DVS .

When referring to maximum and minimum mounting heights the project used the unladen values . T his was deemed appropriate as it represented the wors t case configuration. F or the purpos es of rating a vehicle agains t a DVS this would result in a vehicle being rated such that any changes to cab height due to load would only improve its performance (i.e. the vehicle would get lower due to load). It would be inappropriate to have a scenario where a vehicle could be rated and then its direct vis ion performance decrease in normal use (i.e. the vehicle would get higher when load was removed).

9 .2 .3 3D V E H IC L E D A T A

T he vehicle s ample assessed during the project was ultimately obtained us ing two methods: 3D C AD data directly from manufacturers and data collected from 3D scanning of real vehicles . Ideally C AD data would have been obtained for all of the sample however data of this nature is commercially very sens itive and it was understood that not all manufacturers would be willing or able to share thes e data. Manufacturers that were able to provide C AD data include: DAF , Dennis ,



Iveco, Mercedes , R enault and Volvo, manufacturers for which 3D scan data were obtained include: MAN and S cania. T he following table highlights the type of data used for each manufacturer.

Make & model C AD DAT A Height data DAF C F N3, N3G (Highest, L owest & Most S old) C AD data provided A ll height range data DAF X F (Highes t lowes t & most s old) C AD data provided A ll height range data DAF L F C AD data provided A ll height range data Dennis E agle E lite 6 C AD data provided A ll height range data MAN T G S N3, N3G (Highes t, lowes t & most sold) S can data used B ody builder data used MAN T G X N3, N3G (Highes t &lowes t)) S can data used B ody builder data used Mercedes 2.3 (Highest & lowest) C AD data provided Min/max only Mercedes 2.5 (Highest & lowest) C AD data provided Min/max only Mercedes Atego ((Highest & lowest) C AD data provided Min/max only Mercedes E C O NIC C AD data provided s ingle height R enault D 2.3 (Highest & L owest) C AD data provided Min/max only R enault C 2.5 (Highest & lowest) C AD data provided Min/max only R enault K (Highest & lowest) C AD data provided Min/max only R enault T R M (Highest & lowest) C AD data provided Min/max only S cania P (Highest & lowes t) S can data used Min/max only S cania R (Highest & lowes t) S can data used Min/max only Volvo F E L E C (Most S old) C AD data provided Min/max only Volvo F H (Highest & lowes t) C AD data provided Min/max only Volvo F M (Highest & lowest) C AD data provided Min/max only Volvo F MX (Highest & lowest) C AD data provided Min/max only Iveco S tralis C AD data provided No data supplied T able 7. S ummary of the data obtained for the s ample

9 .2.3.1 MA N U F A C T U R E R C A D D A T A

F or the purposes of evaluation only the necessary C AD data were requested. T hese included:

• T he internal vehicle cab s tructure with window apertures and seats . Including anything which may obscure the view of the driver e.g . the windscreen frit (fade off), windscreen wipers , dashboard, s teering wheel, etc.

o T he seat reference point (S gR P ) or the R point o T he s teering wheel to be in its neutral adjus tment pos ition

• T he external vehicle cab s tructure including the bas ic external panels , mirrors and mirror mounting arms

F igure 20 shows an example of the data requested.



F ig ure 20. E x emplar C A D model of a HG V c ab inc luding the nec es s ary elements for ev aluation (left interior, rig ht ex terior)

9 .2.3.2 3D S C A N N E D D A T A

S can data had to be obtained where manufacturer C AD data were not available. F or the two manufacturers , MAN and S cania, scan data were already available for the E uro 6 models from previous work (see F igure 21 left). However, it was deemed important to refine these exis ting models through the additional capture of high-resolution data of the interior. T hus dashboards , A and B -pillars and relevant interior fittings (grab handles , rain sensors ) were captured us ing a combination of Artec E va and Artec S pace S pider light-based scanners (see F igure 21 right). T he additional data were integrated into the exis ting models to produce final ass essment vehicles .

F ig ure 21. E x is ting s c an data (left) aug mented with additional interior s c an data (rig ht)

9 .3 V E H IC L E S E T U P

O nce data had been obtained either from manufacturers or via the scanning process each vehicle had to be prepared for evaluation. S etup included ensuring that the steering wheel was in its mid/neutral/50th % ile adjus tment pos ition, that passenger seats were fully rearwards , and wipers were in their parked pos ition. T he final and most critical element of the set-up process was the definition of an eye point.



9 .3 .1 E Y E P O IN T D E F IN IT IO N – P R O V IS O N A L

T here are potentially many ways in which an eye point could be es tablished for the projection of the field of view. P revious work by the L DS team utilised dig ital human models in the S AMMIE DHM sys tem of different s izes including: 5th % ile UK female and 50th and 99th % ile UK Male, with their hip point located at the standard seating reference point (S gR P ) and placed in a driving posture. T hes e eye points were subsequently validated against R eed’s eyellipse data (20058) for HG Vs with height adjus table s eats . T he initial approach cons idered for the DVS was to employ R eed’s method again, however with input from relevant s takeholders it was decided that wherever poss ible exis ting s tandards should be utilised as these would be familiar to those in the industry and are already factored into vehicle des ign and evaluation by manufacturers . F urthermore, any specification of an eye point should be as s imple as poss ible to minimise error and aid ease of use.

T he definition of an eye point required cons ideration of three main factors :

• Vertical (Z ) pos ition relative to the S gR P to account of the height of the driver • Horizontal (X /Y ) pos ition relative to the S gR P to account for the direction of view of the driver • Whether a s ingle (monocular) or double eye ((am)binocular) point(s ) would be used

F or the vertical pos ition it was decided to adopt the eye point definition derived from UNE C E R egulation 469 which s tates a drivers ocular point(s ) as being “635mm vertically above the point R [S gR P ] of the driver’s seat”. F or the horizontal pos ition it was decided to adopt the eye point definition derived from S AE J 105010 which allows for a “head turn about a vertical axis (of) a maximum of 60 degrees to the left or to the right from the s traight ahead pos ition.” as shown in F igure 22. In order to pos ition the 60 degree rotation a neck point was also defined, again us ing S AE J 1050. T his defines the neck point (P ) as 98mm rearwards of the mid point between the left and right eye points (E ) shown in F igure 22.

F ig ure 22. S ig ht lines and eye point rotation from S A E J 1050.

Updated c ontent: After the full analys is of the vehicle sample it was noted that the definition of the S gR P (S eating R eference P oint) that is defined by UNE C E reg 46, varies between manufacturers . i.e. traditionally in the automotive des ign process the S gR P refers to the hip location of a 95th% ile

8 R eed, M.P . (2005). Development of a New E yellips e and S eating Accommodation Model for T rucks and B uses . T echnical R eport UMT R I-2005-30. Univers ity of Michigan T rans portation R es earch Ins titute, Ann Arbor, Michigan. 9 UN E C E R egulation 46 – R evis ion 6 (2016). Uniform provis ions concerning the approval of devices for indirect vis ion and of motor vehicles with regard to the installation of thes e devices . UN E C E Vehicle R egulations . 2016. 10 S AE J 1050 (2009). Describing and Meas uring the Driver's F ield of V iew. S AE S urface Vehicle R ecommended P ractice. S AE International.



driver in the lowest rearmost s eat adjus tment pos ition (see S AE J 941). However, three manufacturers applied the S gR P in a location that is forward of the rearmost lowest seating pos ition, and one manufacturer used a location than is higher than standard. T his is shown below in F igure 23. T he S gR P defines the location of the eye point that is used in the DVS .

F ig ure 23. T he differenc e in the relativ e pos ition of the S g R P within the H-P oint env elope demons trated by the data prov ided by v ehic le manufac turers

T he degree of variability in the use of S gR p was unexpected and led to inequalities in the assessment proces s which were cons idered to be unacceptable. T his is sue is currently being addressed by the L DS team through cons ideration of potential options for the definition of the hip point and eye point, in parallel with s takeholder consultation. A resolution is expected by the end of November 2017 lates t.

9.3.1.1 T H E U S E O F MO N O C C U L A R O R A MB IN O C C U L A R E Y E P O IN T S IN T H E A S S E S S ME N T

B oth R egulation 46 and J 1050 define two ocular points (eye points) and so the impact of us ing binocular, ambinocular or monocular vis ion had to be carefully cons idered. An ambinocular assessment is arguably the truest representation of human s ight and has the potential to recognis e des igns of features such as a-pillars that may have been optimised for binocular vis ion. However, the concern raised by stakeholders was the complexity of the projection process when combined with an ambinocular process .

With two eye points and three head pos itions (forward, left and right) a total of s ix projections are required. T hese projections would then need to be combined to create a compos ite projection. In evaluating this pos s ibility two is s ues arose:

T he firs t is sue concerned what would be projected from each eye point. In its s imples t form, this would res trict each of the three eye points to only project through the window to the relevant s ide. T hus the eye point at the pos ition of 60 degrees to the right would only be used to project through the driver’s s ide window. T he eye point s traight ahead only used for the windscreen and the eye point 60 degrees to the left used to project through the passenger s ide window. However, in limiting the eye points in this manner much of the benefit of binocular vis ion is removed. Ideally the field of view from an eye would be the limiting factors for what is projected. F igure 24 shows the three views for an exemplar vehicle. T his image shows that in each case it is poss ible to see through adjacent window apertures . T hus , for the left view the projections could be produced for both the passenger s ide window, but also for part of the windscreen. T his raises two difficulties , the further proliferation of projections , and also how to project a partial view. T he projection could be cons trained by the angular field of view of an eye and / or the limits of movement of a human eye.



T he res ult is a potentially complex process for generating a projection, for which s tandard C AD tools are not currently available.

F ig ure 24. L eft, front and rig ht v iews from the three propos ed ey e points for ev aluation

T he second is sue is then the management of multiple projections . If three projections are us ed from a s ingle eye point in each case, and limited to the aperture in the direction of view, the projections produce three discrete non-intersecting projections as shown in F igure 25.

F ig ure 25. T hree projec tions from three ey e points , one eac h for the pas s eng er s ide, winds c reen and driv er’s s ide

If more projections are produced from two eye points the projections can intersect as shown in F igure 26. O nce projections intersect the complexity increases as the projections have to be combined to generate a compos ite projection in order to calculate a s ingle volume that does not include areas of overlap and thus double counting of volume. In tests performed by the L DS team and by one of the manufacturers , the process of combining the projected volumes was extremely complex and time cons uming. In s ome cases projections could not be combined us ing standard C AD tools and had to be manually rebuilt us ing surface modelling.



F ig ure 26. S ix projec tions from s ix ey e points , two eac h for the pas s eng er s ide, winds c reen and driv er’s s ide

T he result of this evaluation was the decis ion to prioritise a monocular projection. However, it was s till important to attempt to quantify the differences between a monocular and an ambinocular evaluation in the DVS . T hus a binocular assessment was compared to a monocular assessment. T he res ults of which are detailed in S ection 9.5.7.

T he final eye point definition used in the method to evaluate the direct field of view is shown in F igure 27.

F ig ure 27. E y e point defin ition for the v olumetric field of v iew as s es s ment methodolog y

F ront eye point (winds c reen)

R ig ht eye point (driver’s s ide window)

L eft eye point (pas s eng er s ide window)

S g R P

635mm

60°



9 .4 T H E D E F IN IT IO N O F T H E A S S E S S ME N T V O L U ME S

T he volumetric projections can theoretically extend for any dis tance beyond the cab of the vehicle. In the Draft T R L DVS the initial proposal was to constrain the projections to a defined assess ment volume. A key part of this project was to explore potential assessment volumes and jus tify the most appropriate candidate.

9 .4 .1 T H E V E R T IC A L S P E C IF IC A T IO N O F T H E A S S E S S ME N T V O L U ME

T he Draft T R L DVS proposed a vertical arrangement of the assessment volume that cons is ted of two zones that extended from a point 0.93m from the floor to 1.41m from the floor and from 1.41m from the floor to 1.87m from the floor as shown in F igure 28.

F ig ure 28. V ertic al s pec ific ation of the as s es s ment zone as propos ed in the D raft T R L D VS

Whils t this approach would cover the majority of vulnerable road users it was cons idered to be limiting in that the assessment zone did not extend to the floor. T he lower zone was cons idered important to provide assessment of vis ion performance nearer to the ground where VR Us such as children may be present. F urthermore an increased assessment zone in the most challenging area for HG Vs and would both foster and provide a means to evaluate new des igns that maximised direct vis ion. T o account for these is sues two alternate zones were proposed.

T he firs t option was a s ingle volume that extended from 1.602m to the ground. T he rationale for this was to provide an assessment of vis ibility from the ground a height representative of the talles t E uropean population. T he shoulder height of the Dutch Male population was used on the premis e that seeing the head and shoulders of a person would allow them to recognised by the driver. T his resulted in an upper limit value of 1.602m as shown in F igure 29.



F ig ure 29. F irs t option for the v ertic al s pec ific ation of the as s es s ment zone

T he second option divided option 1 into a series of layers . T he layers were introduced to allow weightings to be subsequently explored in the assessments . F or example the layers would allow more emphas is to be placed on being able to see lower and therefore rewarding vehicles with better direct vis ion performance. T he layers represented a differentiation in equal segments between the s houlder height of the smalles t population in E urope (Italian females ) through to the same upper limit represented by the talles t E uropean population (Dutch males ). F igure 30 shows the zones in this option with a large zone extending from the floor to 1.177m from the floor and then five equal zones between 1.177m and 1.602m. T he figure also shows how this is representative of other % iles and populations from across E urope.

F ig ure 30. S ec ond option for the v ertic al s pec ific ation of the as s es s ment zone

1.602



9 .4 .2 T H E H O R IZ O N T A L S P E C IF IC A T IO N O F T H E A S S E S S ME N T V O L U ME T he Draft T R L DVS proposed a horizontal arrangement of the assessment volume that cons is ted of an area extending across the passenger s ide and front of the cab as shown in F igure 31. T he s pecification of this zone included a number of features :

• T he zones are offset from the s ides of the cab by 0.3m representing the hip breadth of a large US female

• T he zone extends to the nears ide 3.5m to cover a s tandard road lane width • T he zone extends to the front by 4.7m based upon the evaluations carried out by T R L on the

ass essment vehicles

F ig ure 31. Horizontal s pec ific ation of the as s es s ment zone as propos ed in the D raft T R L D V S

T his approach was cons idered to be limited for a number of reasons as outlined in S ection 5.2 however the primary is sues were the lack of driver s ide zone, the large assumptions made about the movement of VR Us about the vehicle, and the offset of the zone from the s ide of the vehicle us ing hip breadth as a measure for where the vis ible part of a VR U would be pos itioned. T o account for these is sues two alternate zones were proposed.

T he firs t candidate was a large 20m by 20m zone centred on the driver’s eye point laterally s uch that 10m extended to the left and right of this point. T he zone extends 1m to the rear of the eye point and there is no gap between the outer surfaces of the cab and the zone to both the front and s ides as s hown in F igure 32. T he premise for this zone was to ensure that all vehicles no matter their direct vis ion performance would produce projections that intersected the assessment volume. T he large s ize would als o account for direct vis ion performance where blind spots are known to exis t at cons iderable dis tances from the cab due to is sues such as A-pillar obscuration. In addition, the candidate accounted for front, driver and passengers s ides , and allowed all of the poss ible area around the cab to be evaluated with no offset.



F ig ure 32. C andidate 1 for the horizontal s pec ific ation of the as s es s ment zone

T he second candidate was effectively a trimmed vers ion of candidate 1. T he zone was limited to the passenger s ide by 4.5m, to the driver’s s ide by 2m and to the front by 2m as shown in F igure 33. T he premis e for this zone was to limit the evaluation to areas around the vehicle in close proximity in which accidents with VR Us are known to occur. Whils t it is acknowledged that this would not then account for direct vis ion is sues further away from the cab it prioritised the main focus of the res earch on direct vis ion of VR Us in the area of greates t risk, the zones adjacent to the cab as identified in the accident data. T he limits to the left, right and front of the cab were defined by the coverage of the mirrors , as defined in R egulation 46. T he C lass V mirror has a requirement (for mounting heights above 2.4m) to be able to see a zone 4.5m from the passenger s ide of the cab at ground level. T he C lass V I mirror needs to be able to see a zone 2m to the front of the cab. T here is no equivalent mirror for the driver’s s ide but it was cons idered appropriate to extend the 2m requirement from the front to also cover this area. Whils t it may seem counter-intuitive to use an indirect vis ion requirement to specify a direct vis ion s tandard it provides a rationale for how far direct vis ion s hould be as s essed from the vehicle. If current vehicles are des igned such that mirrors are required to cover up to 4.5m from the passenger s ide of the cab it follows that direct vis ion should be afforded beyond this dis tance. T he direct vis ion s tandard aims to remove the reliance on mirrors and thus s hould focus on providing direct vis ion of the areas currently covered by mirrors .

20m

10m

Eye point

10m

1m



F ig ure 33. C andidate 2 for the horizontal s pec ific ation of the as s es s ment zone

9 .5 R E S U L T S O F T H E V O L U ME T R IC P R O J E C T IO N S

T he following sections present the results from the application of the volumetric assessment of field of view as described in S ection 9.1.1and S ection 9.1.2 applied to the assessment zones described in S ection 9.4.

9 .5 .1 V O L U ME T R IC R E S U L T S F O R A L L V E H IC L E S U S IN G C A N D ID A T E 1 (U N T R IMME D ) A S S E S S ME N T Z O N E V O L U ME

F igure 34 shows the results for the candidate 1 assessment volume (see F igure 32) with s ingle vertical section (see F igure 29).

2m

2m

Eye point

1m

4.5m



F ig ure 34. R es ults for the c andidate 1 (untrimmed) as s es s ment zone v olumes with s ing le v ertic al s ec tion

At this s tage in the evaluations it was clear that a method was required to place the volumetric results into context relevant to the direct vis ion of VR Us in proximity to the cab. In order to provide this a VR U s imulation s imilar to that applied in previous work by the L DS team was used11.

9 .5 .2 Q U A N T IF Y IN G V O L U ME T R IC R E S U L T S A G A IN S T R E A L W O R L D P E R F O R MA N C E

C ontinuing the methodology of the need to be able to see the head and shoulders of a VR U introduced in S ection 9.4.1 a process to quantify the volumetric results agains t real world performance was developed. T his quantification took the form of a number of VR U s imulations (human models of a g iven s tature) placed around the vehicle and moved laterally to a point at which their head and shoulders is vis ible to the driver as shown in F igure 35.

11 S ee S ummers kill et al., (2015). Unders tanding D irect and Indirect Driver V is ion in Heavy G oods Vehicles , F inal R eport. S ection 3.4 concerning VR U vis ibility analys is .



F ig ure 35. Inters ec tion of the VR U s imulations with the projec tion v olumes . VR Us pos itioned s uc h that the head and s houlders inters ec t with the projec tions and thus head and s houlders are v is ib le in thes e loc ations .

A number of variants of this method were evaluated including nine VR U s imulations (three to the front, and three to both the left and right of the cab), thirteen VR U s imulations (three to the front and five to both the left and right of the cab), VR U s imulations with a 5th % ile Italian female stature (1500mm high) and VR U s imulations with 50th % ile UK male s tature (1755mm high).

After an evaluation of the various options the configuration shown in F igure 36 was adopted us ing 5th % ile Italian female VR Us to ens ure that performance is evaluated and rated agains t a thres hold that represents over 99% of the E uropean population.

F ig ure 36. V R U s imulations pos itioned around the v ehic le to prov ide quantific ation of the v olumetric analys is



9 .5 .3 C O R R E L A T IO N R E S U L T S O F C A N D ID A T E 1 (U N T R IMME D ) A S S E S S ME N T Z O N E V O L U ME S A G A IN S T V R U D IS T A N C E

F ig ure 37. G raph s howing the v olumetric res ults from c andidate 1 as s es s ment zone (untrimmed v olumes ) with s ing le v ertic al s ec tion c orrelated with the av erag e d is tanc e of the 13 VR U s imulations

As shown in F igure 37 the use of the VR U s imulations provided a quantification of the volumetric res ults . F or example V19 exhibited a volumetric result of 5.22x1011 mm3. T he volume itself is then placed into a real-world context such that the average dis tance at which the head and shoulders of a 5th % ile Italian female is vis ible to the driver is 0.6m. C onversely V18 at its highest mounting height (H) achieved a volumetric result of 4.18x1011 mm3 which equates to a VR U performance such that the average distance at which the head and shoulders of a 5th % ile Italian female is vis ible to the driver is 5.1m. T his quantification addressed not only the is sue of access ibility of the results but als o provided a means to later determine thres holds for s tar ratings .

O ne of the observations of the results was that whils t the correlation of volume to VR U distances was s uccessful for quantifying the volumetric results , there were some characteris tics of the data that were not ideal. F igure 38 shows the same graph as shown in F igure 37 with two pairs of vehicles highlighted. In both of these cases two vehicles can be observed to have s imilar volumetric performance and yet different VR U performance. It was acknowledged that this is due to the resolution of the two assessment methods. T he volumetric assessment performance is cons idered to be a high resolution evaluation taking account of all of the field of view afforded the driver. However, the VR U assessment could be argued to be a low resolution evaluation having effectively only 13 data points . T he phenomenon observed in F igure 38 is therefore not surpris ing but does pose some potential challenges when ultimately determining and jus tifying s tar rating boundaries . In order to attempt to improve the data to differentiate between vehicles a range of evaluations were performed exploring the options outlined in previous sections with the candidate 2 assessment zone and volume weighting.



F ig ure 38. Volumetric res ults from c andidate 1 as s es s ment zone (untrimmed v olumes ) with s ing le v ertic al s ec tion c orrelated with the av erag e dis tanc e of the 13 VR U s imulations . H ig hlig hted v ehic les hav e s imilar v olumetric performanc e but different VR U performanc e.

9 .5 .4 V O L U ME T R IC R E S U L T S F O R A L L V E H IC L E S U S IN G C A N D ID A T E 2 (T R IMME D ) A S S E S S ME N T Z O N E V O L U ME S

F igure 39 shows the results for the candidate 2 assessment zone volumes (see F igure 33) with s ingle vertical s ection (see F igure 29).

F ig ure 39. R es ults for c andidate 2 as s es s ment zone (trimmed) v olumes with s ing le v ertic al s ec tion



What can be seen from the results of candidate 2 assessment zone on the volumetric evaluation is much greater differentiation between vehicles than that demonstrated by candidate 1 assessment zone shown in F igure 34.

9 .5 .5 C O R R E L A T IO N R E S U L T S O F C A N D ID A T E 2 (T R IMME D ) A S S E S S ME N T Z O N E V O L U ME S A G A IN S T V R U D IS T A N C E

F igure 40 shows the correlation of candidate 2 assessment zone volumetric results to the average of the 13 VR U dis tances . T he is sues observed in F igure 38 were largely removed by trimming candidate 1. T he correlation augments the volume graph s hown in F igure 39 by highlighting the performance of the best vehicles which are s ignificantly dis tanced from the majority of the fleet.

F ig ure 40. G raph s howing the v olumetric res ults from the c andidate 2 as s es s ment zone (trimmed) with s ing le v ertic al s ec tion c orrelated with the av erag e dis tanc e of the 13 VR U s imulations

9 .5 .6 E X P L O R IN G W E IG H T IN G O F T H E A S S E S S ME N T Z O N E S



In an attempt to further refine the spread of the volumetric res ults a range of weighting options were explored. T hese included:

• Weighting the multiple vertical layers shown in F igure 30. T he weightings applied halved the importance of each layer from the bottom to the top.

C olour name

Weig hting C olour s hown

R ed 1/64

O range 1/32

G old 1/16

Y ellow 1/8

G reen 1/4

Dark G reen 1/2

• Weighting by direction, taking into account the areas of greatest risk shown in F igure 6

T he result of the vertical weighting applied to the candidate 2 assessment zone volumes is shown in F igure 41. T he weighting further exaggerated the differences between the best and wors t performing vehicles but also bunched the majority of vehicles to the left of the graph. T his reduced the ability to differentiate between vehicles and so was ultimately rejected.

F ig ure 41. G raph s howing the v olumetric res ults from c andidate 2 as s es s ment zone (trimmed) with multip le v ertic al s ec tions weig hted of dec reas ing importanc e from bottom to top, c orrelated with the av erag e dis tanc e of the 13 VR U s imulations

T he result of the weighting by direction based upon the accident data and the area of greatest ris k applied to the candidate 2 assessment zone volumes is shown in F igure 42. As seen with the vertical weight, the weighting by direction also led to the majority of vehicles to be bunched to the



left of the graph, though to a lesser extent. Again, this reduced the ability to differentiate between vehicles and so was rejected.

F ig ure 42. G raph s howing the v olumetric res ults from the c andidate 2 as s es s ment zone (trimmed) with s ing le v ertic al s ec tion weig hted by direc tion (front, left and rig ht) bas ed upon ac c ident data, c orrelated with the av erag e dis tanc e of the 13 VR U s imulations

9 .5 .7 E X P L O R IN G A MB IN O C U L A R P R O J E C T IO N S

As dis cus s ed in S ection 9.3.1 the projections us ed in the evaluations could be generated us ing a s ingle or pair of eye points . T o evaluate the impact of us ing an ambinocular projection three vehicles were assessed us ing a pair of eye points as specified in S AE J 1050. T he projection volumes us ing the two methods on two sample vehicles : V16 and the V11 are shown in T able 8 below.

T otal Volume (mm3)

Monocular Ambinocular D ifference (mm) D ifference (% )

D A F X F (MS ) 3.69E +09 4.30E +09 6.13E +08 16.61

MA N T G S (N3 MS )

3.74E +09 4.43E +09 6.84E +08 18.26

R enault T (H) 3.84E +09 4.41E +09 5.74E +08 14.97 T able 8. Volumetric res u lts c omparing monoc ular projec tion to ambinoc ular projec tion

T he results in T able 8 show that for the vehicles assessed there is a 15.0 – 16.6% increase when moving to an ambinocular as s es s ment. Whils t this increase can be cons idered substantial what is als o clear is that both results increased, and they both increased by a s imilar amount. T he vehicles



s elected for assessment were selected due to their s imilarity in their volumetric results and proximity in the correlation graph as shown in F igure 43 (left). C ritically the ambinocular results do not change the order of the results as shown in F igure 43 (right) and when the increase is compared to the total candidate 2 assessment zone the increase is only 0.01 % .

T his lack of change in the order of the results combined and the small increase In volume when compared to the assessment volume, the cons iderable additional work in generating ambinocular projections was not cons idered to be worthwhile addition to the method for evaluating the performance of an HG V.

F ig ure 43. S ample of the full c orrelation of the v olumetric res ults from the c andidate 2 as s es s ment zone (trimmed) with s ing le v ertic al s ec tion with the av erag e dis tanc e of the 13 VR U s imulations – hig hlig hting the v ehic les ev aluated for ambinoc ular projec tions . L eft imag e s hows the monoc ular v olumes . R ig ht imag e s hows the ambinoc ular v olumes .

9 .5 .8 S U MMA R Y O F R E S U L T S

T he result of the evaluation of the options presented in the previous sections was that the candidate 2 (trimmed) assessment zone shown in F igure 33 provided the best result when combined with only a s ingle vertical zone as s hown in F igure 29. ‘B es t’ in this ins tance was determined by having the maximum differentiation between volumetric performance of the vehicles assessed.

Having es tablis hed the volumetric res ults for all of the vehicles in the sample the performance of the vehicles had to be mapped onto thresholds that would determine the star rating boundaries .

9 .6 T H E D E F IN IT IO N O F T H E ME T H O D B Y W H IC H T H E S T A R R A T IN G S W O U L D B E P R O D U C E D

With an as sessment volume and VR U approach defined the next phase of the project was to explore how the s tar rating boundaries of the DVS would be set. O ne of the key benefits of the VR U approach was that it could be used to define a minimum threshold of acceptability regarding the pos ition(s ) around a vehicle in which a VR U can be located and be vis ible to the driver. T hrough dis cuss ion with the stakeholder group and the T fL team a process was defined for the production of four poss ible ways in which the DVS could be defined. T hese four candidates would have two variants each of an ‘Absolute approach’ and a ‘R elative approach’.



• T he Absolute approach: In which the star rating boundaries would be defined based upon a minimum performance which was inclus ive of the whole E uropean population and was cons idered to represent a des ired level of direct vis ion from a risk perspective

• T he R elative approach: In which the s tar rating boundaries would be defined based upon the c urrent performanc e of ex is ting v ehic les in the fleet

T he relative and absolute approaches were defined to allow a comparison of the outputs in terms of the s afety benefit, and the cos t to the haulage industry, manufacturers and other stakeholders that would be determined in an impact assessment s tudy performed s eparately. In addition, it was agreed that for each approach there would be a vers ion of the star rating based sys tem based upon the VR U dis tance technique to define the minimum requirement, and also a vers ion which was based upon the median volumetric score. T hat is , one vers ion based upon VR U dis tance, and one vers ion based upon the volumetric data alone.

T he four different approaches were there defined as the following;

• R elative approach

O ption 1: T es ting the dis tance away that an average s ized UK male (50th% ile = 1755mm tall) could be seen to the front, offs ide and passenger s ide of the truck

O ption 2: T he median volumetric score defines the 3 star boundary

• Absolute approach

O ption 3: T es ting the dis tance away that a small Italian female (5th% ile = 1500mm tall) could be seen to the front, offs ide and P assenger s ide of the truck

O ption 4: T he median volumetric score defines the 1 star boundary

T he two different s izes of human were initially selected to allow an exploration of whether there were fundamental differences between the way that vehicles performed close to the vehicle and further away from the vehicle, based upon the fact that a taller person would be able to stand closer to a vehicle and s till be seen. T his tes t was done for the full sample and there were no fundamental differences in the pattern of dis tances for 13 VR Us for taller and shorter human models .

F ig ure 44. Imag es s how av erag e dis tanc es VR Us c an be s een to eac h s ide of the c ab for three v ehic les

A ll vehicles were tes ted to find the distance away that the VR Us could be placed (as per F igure 44. Images show average dis tances VR Us can be seen to each s ide of the cab for three vehicles ) whils t being able to see the head of shoulders of the human as discussed in S ection 9.5.2. T hese res ults were then processed to see how many vehicles allowed the two different VR U s izes to be seen to inters ect with volumetric projection within the assessment zone. T he concept here was that if a VR U cannot be s een within the assessment zone (which also covers the area that can be seen in mirrors ) then there would be locations around the vehicle that would allow the VR U to be invis ible to the driver, as they would potentially be between the volume of space that is vis ible to the driver through windows and the mirrors . E ffectively they would be in a blind spot. F igure 45 shows an example of a



truck cab with a low volumetric projection total (anonymised) which does not allow the head and s houlders of the VR U to be seen. What this means is , a small person cannot be seen directly through the windows when they are 4.5m to the left of the driver’s cab, or 2m in front of the driver’s cab, and they cannot be seen through the mirrors . If this person moved closer to the truck, they still would not be seen through the windows , but they may be seen in the close proximity mirrors (C lass V and V I), if the driver looks in those mirrors at the right time. T he s ituation where a blind spot of this s ort exis ts was seen as unacceptable, and therefore any vehicle which cannot allow, as a minimum the head and shoulders of a small Italian female (5th% ile) to be vis ible at the edge of the asses sment volume will be cons idered inappropriate to use in an Urban environment and therefore will have a rating of zero s tar. T his es tablis hes a limit of acceptability.

In F igure 46 we see a vehicle with a high volumetric projection total (anonymised) which allows the head and shoulders of small Italian female (5th% ile) to be vis ible when she is 450mm from the front of the truck and 200mm from the passenger s ide of the truck. T his demonstrates the result for a high vis ion cab which can allow s mall adults to be vis ible to the driver even at close proximity to the s ide of a vehicle. T he res ults for the whole s ample varies between these two extremes . T he following s ection des cribes how the s tar rating sys tem was defined us ing the VR U dis tances as a meas ure of direct vis ion effectivenes s us ing the limits of acceptability defined above.

F ig ure 45. A n ex ample of a c ab des ig n whic h does not allow a s mall Italian F emale (5th% ile) to be v is ib le. T he red projec tion s hows the v olume of s pac e v is ib le to a driv er throug h the window. T he purple and orang e pro jec tions s how the v olume of s pac e v is ib le to a driv er throug h the c las s V and c las s V I mirrors res pec tiv ely .

F ig ure 46. A n ex ample of a c ab des ig n whic h does allow the head and s houlders of a s mall Italian F emale (5th% ile) to be v is ib le within the as s es s ment v olume. T his s hows the res ult for the c ab whic h ac hiev ed the hig hes t rating in the D V S as s es s ment



9 .6 .1 T H E D E T A IL E D A P P R O A C H U S E D T O F IN D T H E D V S R A T IN G O F E A C H V E H IC L E A N D D E F IN E T H E S T A R R A T IN G B O U N D A R IE S C O N S ID E R IN G O P T IO N S 1 -4

O ptions 1 and 3 cons ider the location of the VR Us as way to quantify the volumetric scores , with the allocation of the boundary vehicles to a 1 star or 3 s tar boundary based upon the absolute or relative approach respectively. O ptions 2 and 4 cons ider the volumetric data alone, and apply either a 1 star or a 3 s tar the median volumetric value based upon the absolute or relative approach respectively. T he following des cription of the methodology followed describes the approach for option 3. T his option was selected as the approach that would be adopted for the DVS by the T fL B oard in August of 2017.

T he approach that has been defined utilis es and combines the results for the three directions of view (front, offs ide and nears ide) that have been used in the volumetric assessment process , as suggested by manufacturers in the J une s take holder event discus sed in section 7. Initially a threshold vehicle was identified which jus t meets the test of acceptability defined above for each of the three viewing directions that have been included in the volumetric measurement of direct vis ion performance i.e. the front, P assenger s ide and offs ide of the truck.

T his was done by taking the average of the VR U dis tances from the front, P assenger s ide and offs ide of the truck. F or example, for the nears ide, the truck which had an average VR U dis tance value clos es t to the 4.5m cut off shown in F igure 33 was identified. In this case the value was 4.437m. All other trucks produced average VR U dis tances where the head and shoulders could be seen for the 5 nears ide VR Us that were below 4.43m, or above 4.5m. T he volume of the projection for the P as s enger s ide for this one vehicle was then found from the data. T his set a threshold volume for the P assenger s ide which was equated to the one s tar boundary for O ption 3. T he volumetric difference between the 1 s tar boundary limit and the highest volume was then subdivided equally to create the 2 star, 3 s tar, 4 s tar and 5 star boundaries . In the case of option 3, this subdivided the data between the threshold and the best performing vehicle’s volume into five. T he divis ion could have been into four equal sections , to represent the performance between 1-2, 2-3, 3-4 and 4-5 s tar. However, five divis ions were cons idered to be more appropriate to reduce five s tar performance being reliant on a s ingle vehicle; to ensure there was an opportunity for there to be more than one vehicle in the fleet that could be rated as five star, and encourage manufacturers to develop improved future vehicles aspiring to a five star rating. T his is shown in F igure 47 (anonymised).



F ig ure 47. A g raph s howing the A v erag e VR U dis tanc e to the P as s eng er s ide ag ains t the v olumetric s c ore for the P as s eng er s ide projec tions . T he red line s hows that V32 defined the 1 s tar boundary , with the other boundaries being c reated by an s ubdiv is ion of the v olume between V32 and the bes t

performing v ehic le, V10.

T he s tar rating was then tabulated for each vehicle by seeing within which boundaries each vehicle falls as per F igure 47. T his process was recreated for the front and offs ide projection data providing three separate s tar ratings for each vehicle. It was a requirement to do each s ide separately in this manner, as the front, offs ide and nears ide all had different offsets which determined acceptability, (e.g. 4.5m to the nears ide, and 2m to the front).

With the three star ratings produced it was then a requirement to combine the data to form an overall rating for each vehicle. T he s tar ratings to the front, P assenger s ide and driver’s s ide were then averaged to g ive and overall rating for each vehicle. T he vehicle with the lowest volume result that achieves an average s tar rating of 1star was identified. T he vehicle with the 1 star rating for the average of all s ides , was then used to identify the volumetric result associated with 1 star for the combination of all three volumes . T he volumetric difference between the 1 s tar boundary limit and the highes t volume was then subdivided equally to create the 2 s tar, 3 s tar, 4 s tar and 5 star boundaries .

T herefore a vehicles final s tar rating is defined by the combined performance to all three of the areas around the cab, with the 1 s tar rating boundary being defined by direct vis ion performance in the VR U tes t.

T he benefits of this approach are that the VR U dis tances have been used as a performance metric to identify a volumetric score which equates to a s tar boundary for each the three directions separately. T his us es the VR U data to help quantify the results , but does not require VR U tes ting in the future application of the s tandard. T his approach is summarised in F igure 48.



F ig ure 48. A flow c hart illus trating the proc es s us ed to c reate the s tar boundaries

T he results for each of the four options are presented in the following sections .

9 .6 .2 R E S U L T S F O R O P T IO N 1 .

T he process ing of the results for option 1 follows the method described in section 9.6.1 except that where boundary vehicles were ass igned to a 1 star rating in O ption 3, they were ass igned a 3 s tar rating in O ption 1.

F R O NT

Vehicle with a VR U dis tance closes t to 2m

Average of the three ratings used to identify the lowest performing 1 star vehicle . T he volumetric score of this one

vehicle is used to identify the combined volume score for the 1

P assenger s ide

Vehicle with a VR U dis tance

closes t to 4.5m

Driver’s s ide

Vehicle with a VR U dis tance

closes t to 0.6m

Defines the 1 star volumetric boundary



O ther boundaries defined and a star

rating ass igned


rating ass igned


rating ass igned

S tar rating for every vehicle

identified for the


identified for the


identified for the

T he other boundaries are identified through the subdivis ion of the difference between the lowest scoring 1 s tar vehicle and the

best scoring vehicle



O ption 1 – R elative. Where the 3 s tar performance equates to the ability to see the head of an average s ized UK male (50th% ile)

• 2m to the front of the cab, 600mm to the right and 3.25m to the left of the cab

• 3.25m was selected as the mid point between the two poss ible class V mirror coverage zones (2m or 4.5m) based upon C lass V mirror height (see UNE C E R egulation 46)

• T hese criteria were used to produce a s tar rating for each of the s ides which was then averaged to provide an overall rating which then defined the 3 s tar boundary in the graph below

F igure 49. A graph s howing the placement of the s tar boundaries for O ption 1.


O ption 2 – R elative. Where the median volume for all trucks was used to define the 3 s tar boundary and the volume to the left and right of this boundary was equally divided



F ig ure 50. A g raph s howing the plac ement of the s tar boundaries for O ption 2


O ption 3 – Absolute. Where the 1 s tar performance equates to the ability to see the head and s houlders of a 5th% ile Italian female

• 2m to the front, 600mm to the right and 4.5m to the left of the cab, linked to the C las s V and V I Mirrors

• T hese criteria were used to produce a s tar rating for each of the s ides which was then averaged to provide an overall volume which then defined the 3 star boundary in the graph below





O ption 4 – Absolute. Where the median volume for all trucks was used to define the 1 s tar boundary and the volume to the left and right of this boundary was equally divided


9 .6 .6 F IN A L S E L E C T E D D V S O P T IO N S U B J E C T T O C O N S U L T A T IO N

As des cribed in S ection 9.6.1 the option selected by the T fL board was option 3. O ption 3 utilis es candidate 2 assessment zone with no weightings to generate the volumetric projection scores . T he 1 s tar boundary is defined by the ability to see the head and shoulders of a 5th% ile Italian female at an average dis tance of 4.5m to the passenger s ide, 2m to the front, and 0.6m to the driver’s s ide of the cab. F igure 53 shows the total volumes plotted agains t the average VR U dis tances , F igure 54 s hows the total volumes only and T able 9 shows the raw data.

NO T E : the data presented in previous figures was correct at the point in time in which the various options were evaluated during the iterative development process . As such previous data may have been supers eded by the data presented below. Due to updates in data supplied by manufacturers , clarifications and different process ing options between previous vers ions and the final vers ion, the data in F igure 53, F igure 54 and T able 9 should be cons idered definitive subject to any final cons ultation.



F ig ure 53. G raph s howing the s elec ted option (3) for the D VS s tar rating boundaries . D ata s how the v olumetric res ults from the c andidate 2 as s es s ment zone (trimmed) with s ing le v ertic al s ec tion ag ains t the av erag e dis tanc e of the 13 VR U s imulations

F ig ure 54. G raph s howing the s elec ted option (3) for the D VS s tar rating boundaries . D ata s how the v olumetric res ults from the c andidate 2 as s es s ment zone (trimmed) with s ing le v ertic al s ec tion



R ating Mak e, Model (and heig ht) T otal v olume (mm 3)

5 V44 1.8586E +10

V43 1.7200E +10 5 s tar 1.6126E +10

4 s tar 1.3665E +10

3

V42 1.2626E +10 V41 1.2499E +10 V40 1.1657E +10 V39 1.1647E +10 V38 1.1538E +10 V37 1.1288E +10 3 s tar 1.1205E +10

2

V36 1.0886E +10 V35 1.0567E +10 V34 8.9837E +09 V33 8.7471E +09 2 s tar 8.7444E +09

1

V32 8.6699E +09 V31 8.4338E +09 V30 8.0425E +09 V29 7.4329E +09 V28 7.2307E +09 V27 6.4397E +09 V26 6.2840E +09 1 s tar 6.2840E +09

zero

V25 6.1701E +09 V24 6.1269E +09 V23 5.8856E +09 V22 5.6232E +09 V21 5.5026E +09 V20 5.2701E +09 V19 5.1287E +09 V18 4.9763E +09 V17 4.4802E +09 V16 4.4225E +09 V15 3.8350E +09 V14 3.7448E +09 V13 3.6884E +09 V12 3.3002E +09 V11 3.1121E +09 V10 3.0626E +09 V9 2.9646E +09 V8 2.8296E +09 V7 2.7535E +09 V6 2.3151E +09 V5 2.2894E +09 V4 2.0558E +09 V3 1.9629E +09 V2 1.5691E +09 V1 1.3782E +09

T able 9. T able detailing v olumetric res ults from the c andidate 2 as s es s ment zone (trimmed) with s ing le v ertic al s ec tion and the s tar boundaries



9 .6 .7 S U MMA R Y O F T H E R E S U L T S F O R T H E F O U R O P T IO N S O F T E R MS O F T H E V E H IC L E S A N A L Y S E D W H IC H W E R E IN C L U D E D IN E A C H S T A R B O U N D A R Y

T he table below summarises the number of vehicles (configurations ) in the sample that are included in each star rating.

O ption 1 O ption 2 O ption 3 O ption 4

R elative VR U

R elative Median

Absolute VR U

Absolute Median

Z ero s tar

6 6 25 16

1 S tar 7 6 6 6

2 S tar 3 3 3 4

3 S tar 12 10 5 3

4 S tar 5 4 0 0

5 s tar 2 2 2 2 T able 10. T he number of v ehic les (c onfig urations ) inc luded in eac h boundary for the four options c ons idered

F ig ure 55. G raph s howing the s elec ted option (3) for the D VS s tar rating boundaries . Mounting heig ht rang es for eac h model h ig hlig hted.

9 .6 .8 IMP A C T O F T H E L O W E R D O O R W IN D O W O N T H E D V S R A T IN G

F or a number of the vehicles assessed in the sample a lower door window was present. In s ome cases this is s tandard fitment, in others it was an option and certain vehicles were assessed with and without the lower door window. In order to explore the impact of a lower door window the



volume increase for an example lower door window was calculated. T able 11 shows an example us ing the Volvo F M. T he specification of a lower door window increases the volumetric projection res ults between 6-15% based upon the mounting height of the vehicle. C ritically this represents between 27-40% of the volume between s tar rating boundaries . T hus , fitting a lower door window in the Volvo F M could change the s tar rating by a maximum of one star e.g . from 2 star to 3 s tar at its lowes t mounting height as shown in F igure 55. E qually, depending on the mounting height, the fitment of a lower door window way have no effect on the star rating.

L OWE R DO O R WINDO W (L DW) C O NT R IB UT IO N

T otal Volume (mm3)

Increase (mm3)

% Increase

% of S tar R ating

Volvo F M (High) 6.4397E +09

Volvo F M (L ow) 1.0886E +10

Volvo F M with L DW (High) 7.4329E +09 9.9323E +08 15.42 40.37

Volvo F M with L DW (low) 1.1540E +10 6.5407E +08 6.01 26.58 T able 11. T able s howing the inc reas e due to a lower door window in v olume, as a perc entag e and as a perc entag e of a s tar rating (e.g . % of v olumetric rang e between 1 and 2 s tar)

10 V A L ID A T IO N E X E R C IS E

As part of the DVS development process one manufacturer offered to recreate the DVS volumetric ratings for a specific vehicle in an effort to see how repeatable the L DS results were.

T his involved the manufacturer applying the DVS analys is protocol us ing the same C AD data that they had s upplied to the L DS team. T he manufacturer wished to use their direct vis ion asses sment tool, which is a plugin called C AVA for the C AT IA C AD tool. T his was seen as a pos itive approach to determining if the process of the DVS could be recreated in another C AD sys tem with an analys is of any volumetric differences in the results .

T his process highlighted the importance of the detailed description of key stages of the DVS protocol, including the alignment of the assessment volume. Initially there a difference in placement of the assessment zone of 0.446mm to the front of the cab between the L DS and manufacturer implementations which resulted in a 1% difference in the volumetric results .

T his was corrected which led to a percentage difference between the results of the two C AD sys tems of 0.07% . A further analys is us ing the data to higher number of decimal places produced a result percentage difference of 0.01% . T his difference was cons idered to be neglig ible by all parties involved.

T his process improved the definition of the DVS protocol and demonstrated that the technique could be applied in two separate C AD sys tems with the same result. F urther validation is expected through the consultation process that has been defined for the Autumn of 2017.

11 D IS C U S S IO N

T he methodology and results in this report present an overview of what was a highly iterative proces s . T he method of assessment and in particular the quantification of the volumetric res ults and ultimately the use of these data to produce the star rating boundaries was produced in parallel with



data collection. T he collection of data as summaris ed in T able 7 was beholden on the ability of manufacturers to supply the data in C AD form or provide access to vehicles . In some cases this needed cons iderable internal negotiation to obtain the clearance to make this sens itive data available to support the project. As such the sample was increas ing as the methodology evolved.

In addition, as can be seen from the volumetric results and more clearly, the VR U s imulation dis tances the results are extremely poor. T he vas t majority of the fleet cannot see a small adult via direct vis ion within an average of 2m of the cab. B ecause of this , and its potential impact on the industry many alternative solutions were cons idered. Whils t it is not the purpose of this report to detail every option cons idered in full, some of the cons iderations evaluated are summarised below:

• It was acknowledged that the trimmed assessment zone of 4.5m to the passenger s ide could s till be cons idered to be too large g iven the focus on VR Us . When cons idering the scenarios of accidents to the passenger s ide it is likely that the pedestrian or cyclis t would need to be s een much closer to the vehicle. However, reducing the assessment zone on the pass enger s ide to less than 4.5m would have left most vehicles in the s ample (and the fleet) with almost no projected volume and thus differentiation would have been much more difficult.

• T he 5th % ile Italian female lower threshold was used in an attempt to provide a lower limit bas ed on population data. T he 5th % ile female provides context in terms of E urope by being representative of more than 99% of the adult population. T he results show that this thres hold proved to be extremely challenging for the vehicles in the sample with many performing very poorly. As discussed earlier alternatives were cons idered. A 50th % ile UK male was als o evaluated. T his produced improved but s till largely poor results in terms of the dis tance away that a VR U could be seen. Us ing a larger VR U s imulation also eroded the mess age regarding safety. If the assessment is based upon an ‘average’ s ized threshold value all it really does is suggest that it is not important to be able to see half of the population. In reality defining the head and shoulders of a small adult s till potentially excludes the vis ion identification of wheel chair users , children and scooter users . T his has been cons idered through review of the accident data which highlights that these VR Us are rarely involved in accidents with HG Vs. A compromise has been derived which allows the vehicle des ign to be differentiated.

• T he use of head and shoulders for vis ibility of the VR U s imulations was a challenging part of the project. In previous work the L DS team had deliberately avoided the complexity of determining how much of a VR U was enough to enable recognition by a driver by pos itioning the VR U s imulations at a point where they were jus t not vis ible. T his is particularly challenging in the more complex areas of the field of view, such as around the mirrors when ‘most’ of a head and shoulder could be vis ible but not all. T here is no data to be able to determine how much of a VR U is enough to enable recognition as the research has not been done and thus a cons istent approach was taken in the placement of VR Us in these s ituations . However, it should be noted that in adverse weather conditions , poor lighting, dark clothing, s ituations of high driver workload through busy traffic, job pressures etc. that it could s till be poss ible that a driver may miss the head and shoulders of a VR U in proximity to the vehicle.

• T he VR U s imulations were introduced to provide a quantification of the volumetric res ults . In addition, the VR U s imulations provide context and access ibility of the volumetric results . In isolation the volumetric results are rather arbitrary and have little meaning in an absolute s ense. However, they a very high resolution evaluation of the direct vis ion performance of a vehicle. In contras t the VR U results have high impact, they clearly demonstrate the capability of a vehicle. However, the VR U results are relatively low resolution, with arguably



only 13 data points , compared to the many thousands of data points that the complexity of the projection paths produced demonstrate. As such they are an approximation of the high resolution volumetric data and thus are only used to augment the volumetric assessment, not replace it.

• A number of the vehicles assessed can be fitted with lower door windows that have been added to vehicle from the s tart of the vehicle des ign process . S ome vehicles were therefore evaluated with and without the door window. T he lower door window can therefore make a direct contribution to improving the volumetric result as described in S ection 9.6.8. However, the lower door windows were not cons idered in the VR U s imulations . As the VR U s imulations required head and shoulders to be seen the lower door windows in the vehicles assessed played no part in affording this view. In some instances the lower door window provided a view of the legs or torso of the VR U however this was cons idered to be too difficult to evaluate in terms of the benefit to the driver in terms of recognition of a VR U. T he result is that lower door windows are cons idered to be a pos itive offering for direct vis ion and thus they are included in the volumetric results and will contribute to improving a vehicle’s performance and potentially its s tar rating but they were not used for determining the s tar boundaries .

• O nce the threshold vehicle and thus threshold volume was determined in the production of the star rating boundaries the remaining data were equally subdivided. In the case of option 3, this subdivided the data between the threshold and the best performing vehicle’s volume into five. T he divis ion could have been into four equal sections , to represent the performance between 1-2, 2-3, 3-4 and 4-5 star. However, five divis ions were cons idered to be more appropriate to reduce five s tar performance being reliant on a s ingle vehicle; to ensure there was an opportunity for there to be more than one vehicle in the fleet that could be rated as five s tar, and encourage manufacturers to develop improved future vehicles aspiring to a five star rating.

• T he use of an equal subdivis ion of the volumetric results to define the star rating boundaries beyond 1 s tar (in the case of option 3) was a pragmatic decis ion to aid unders tanding and in acknowledgement of the lack of data to inform anything more s trategic. Ideally, all of the thres holds would have been based on real world performance via the VR U s imulations . However, as already discussed the data to suggest exactly what needs to be seen, and where in proximity to the vehicle, is not unders tood. As such, performance values in a manner that would allow differentiation to five levels cannot be reliably and meaningfully determined. In absence of these data, a s imple even subdivis ion of the data was cons idered to be appropriate.

• T he complexity of the assessment methodology and its repeatability on different C AD sys tems was a key factor. As has already been discussed some cons iderations were ultimately discounted due to complexity. T he resulting method is a s implification of the field of view afforded to a driver in the real world. However, a proliferation of eye points in terms of both the head direction and (am)binocular vis ion increases the complexity of the evaluation exponentially. In some cases the evaluation becomes nearly imposs ible in any s ens ible timeframe. F or the example of ambinocular projection described in S ection 9.5.7, additional vehicles beyond the two presented were attempted but could not be completed as the projections would have needed to have been recreated manually to produce the combined ambinocular projection. T his process ing would have taken days of effort to produce only one result. T his complexity was also highlighted by the s takeholders who advocated s implicity wherever poss ible.



• T he project has resulted in two main outcomes, a rating methodology for the evaluation of the direct vis ion performance of an HG V together with the definition of a s tar rating system that together form the DVS . T he evaluation methodology of a vehicle has been detailed in a DVS protocol that will allow anyone to evaluate a vehicle with the appropriate data and plot the results agains t the DVS s tar ratings . T he detail of the protocol is likely to evolve through the consultation phase that is planned in the Autumn of 2017 when manufacturers will apply the protocol in a range of C AD sys tems . T his may result in variable performance. T he main is sue concerns accuracy in a number of forms . T he firs t is the accuracy to which the boundaries of the projections are created. T he tools used convert the mathematical curves of the C AD geometry into line segments . T he degree to which the line segments are an approximation of the curves is dependent on the number of points . It is not trivial to place figures on how many points should be used and thus specifying this level of detail could prove to be problematic. In addition, where scan data have to be us ed it is almost imposs ible to validate the accuracy of those data to the real world condition. Arguably this is also true of the manufacturer’s C AD data that may be a s omewhat idealis ed vers ion of the real world performance. Again, specification of achievable and verifiable metrics in this regard may be problematic. T he definition of the external surfaces of the cab from which to offset the assessment zone may also be challenging. T he methodology used in the project used most external contiguous surfaces around the cab, but also ignored any small protrus ions . T his will need to be carefully defined in the protocol but accounting for all the variety of vehicle des ign may be challenging to word a s pecification that achieves a cons is tent result. As described earlier some elements of vehicle geometry such as the windscreen wipers result in mathematically determined ‘apertures ’ that could be projected to contribute to the volumetric result. However, some are so small as to not contribute to direct vis ion performance. T his could be addressed by specifying a limit to the area that can be counted but there is no data to inform such a limit. T he danger is that des igns could be envisaged in which volumetric performance is augmented through mesh like surfaces . T he collective small apertures may result in s ubstantial volumetric results but afford poor direct vis ion. T his needs to be avoided and there has to be a baseline of a spirit of the DVS protocol in addition to letter of the DVS protocol.

• At the start of the project the use of accident data to identify the areas of greatest risk and to then apply weightings to the DVS volumetric results to reflect the importance of direct vis ion to certain locations around the cab was seen as a valuable approach. T he project has highlighted that this approach did not improve the applicability of the DVS scores and s o the data has not been weighted.

T he DVS project and the final DVS sys tem definition is the result of a concerted effort by T fL and L DS s taff in the involvement of all key stakeholders in the des ign process . T he process has included vis its to the des ign teams for VO L VO , Daimler, S C ANIA, and DAF . T he open and constructive approach taken by the vehicle manufacturers has been invaluable to the project teams . T his has resulted in a s tandard which has real potential to improve the safety of the urban environment in L ondon and further afield. In addition the s tandard has the potential to influence the future des ign of truck cabs with regard to the detailed cons ideration of direct vis ion.

Updated c ontent: T he need to redefine the eye point for the assessment process as discuss ed in section 9.3.1 has allowed for further consultation with manufacturers on the key is sues associated with the definition of a DVS sys tem. T he options being cons idered for the redefined eye point which



will affect the results preliminary that have been presented in this report. T he reanalys is will be complete in J anuary of 2018.

T he T fL and L DS teams would like to thank the following s takeholders for their valuable contributions to the project.

D A F Merc edes Volvo S C A NIA J ohan B roeders R os s P aters on C laes Avedal P hillip R ootham

P hilip Moon R upert B arnard J ohn C omer J orge S oria G alvarro

P iet K uylaars Martin T ebbe Degerman Hanna Hanna S taf R oger B osmans S tefan Huegin Michael Dahl IVE C O MA N R enault B urgess Hannah S tuart B adcock L es B ishop Andrew S cott E riksson Mari Martin F lach Mike S tringer


Documents

The definition, production and validation of the …...THE DEFINITION, PRODUCTION AND VALIDATION OF THE DIRECT VIS ION S TANDARD (DVS ) FOR HGVS C onsultation exercise Progress report-