In Search of Severity Dimensions of Traffic Conflicts for Different Simulated Mixed Fleets Involving Connected and Autonomous Vehicles

Resea ch A icle In Sea ch o Se e i y Dimensions o T a ic Con lic s o Di e en Simula ed Mixed Flee s In ol ing Connec ed and Au onomous Vehicles Tasneem Miqdady , Roc´ ıo de Oña , and Juan de Oña TRYSE Resea ch G oup, Uni e si y o G anada, ETSI Caminos, Canales y Pue os, Campus de Fuen enue a, s/n, G anada 18071, Spain Co espondence should be add essed o Roc´ ıo de Oña; [email p o ec ed] Recei ed 13 Feb ua y 2023; Re ised 25 Ap il 2023; Accep ed 8 May 2023; Published 20 May 2023 Academic Edi o : Yanyong Guo Copy igh ©2023 Tasneem Miqdady e al. Tis is an open access a icle dis ibu ed unde he C ea i e Commons A ibu ion License, which pe mi s un es ic ed use, dis ibu ion, and ep oduc ion in any medium, p o ided he o iginal wo k is p ope ly ci ed. Tis s udy aims o es ima e he se e i y o con ic s ha may a ise om he in oduc ion o connec ed and au oma ed ehicles (CAVs) by examining he ehicle pa hs gene a ed by mic osimula ions o mixed ee s o human-d i en ehicles and CAVs wi h di e en le els o au oma ion (L1-L4 ehicles). Te s udy assesses he se e i y o con ic s using a holis ic app oach ha conside s h ee dimensions: (1) p oximi y o collision, ia he ime- o-collision (TTC) indica o ; (2) po en ial consequences o a con ic , ia single su oga e sa e y measu es such as maximum speed (MaxS) and ehicle speed di e ence (Del aS); and (3) a combina ion o bo h dimensions o assign se e i y sco es, ia TTC and eloci y ec o s. Te s udy’s ndings sugges ha mode a e pene a ion a es o L3 and L4 ehicles (35–55%) show signi can di e ences in he numbe o a c con ic s wi h a ying TTC alues. Addi ionally, high pene a ion a es o L3 and L4 ehicles (abo e 55%) esul in lowe alues o con ic consequences measu es such as MaxS and Del aS. Fu he mo e, he s udy shows ha con ic consequences dec ease i he ollowe is a L3 o L4 ehicle. Te s udy’s ndings also e eal ha he e is a conside able educ ion in high se e i y con ic s when he pene a ion a e o CAV le els eaches 50%, and he ull ope a ion o L4 ehicles esul s in a 75.5% educ ion in high se e i y con ic s. Te e o e, his s udy p o ides aluable insigh in o he po en ial se e e con ic s du ing he ansi ion pe iod om manual ehicle ope a ion o ull CAV ope a ion. O e all, he s udy’s ndings highligh he impo ance o assessing he se e i y o po en ial con ic s a ising om he in oduc ion o CAVs. By conside ing he p oximi y o collision and he po en ial consequences o con ic s, he s udy p o ides a comp ehensi e assessmen o he se e i y o con ic s. Tis in o ma ion can in o m he de elopmen o policies and s a egies o ensu e he sa e and esponsible in oduc ion o CAVs in o ou anspo a ion sys ems. 1. In oduc ion Te o hcoming in oduc ion o connec ed and au o- ma ed ehicles (CAVs) on oads has mo i a ed e- sea che s o in es iga e hei a ious implica ions, such as a c delay, conges ion, uel emissions, and a c sa e y. Al hough CAV manu ac u e s ha e p og essed om CAV esea ch o ehicle p o o ype p oduc ion wi hin se e al au oma ion le els [1], he a ailable (beha io al and c ash) da a can no su cien ly cla i y he ambigui y su ounding he c ash isks in ol ing CAVs. Acco dingly, many s udies ha e used he su oga e sa e y assessmen model (SSAM), de eloped by he Fede al Highway Adminis a ion, o analyze he ehicle ajec o ies ga he ed om a mic osimula ion pla o m o de e mine a c sa e y. Se e al su oga e sa e y measu es (SSMs) (e.g., ime- o-collision (TTC), pos enc oachmen ime (PET), and decele a ion a e) ha e been applied o es ima e he p ob- abili y o con ic . A a c con ic is an e iden ins ance in which wo o mo e oad use s o ehicles a e nea each o he in e ms o space and ime o he ex en ha he isk o collision exis s i hei mo emen s do no change [2]. In a c simula ion-based s udies, a c con ic s can be Hindawi Jou nal o Ad anced T anspo a ion Volume 2023, A icle ID 4116108, 21 pages h ps://doi.o g/10.1155/2023/4116108 de e mined by modeling a c ow acking o ex ac ehicle pa hways o e ime. Acco dingly, SSMs ha e been ex ensi ely employed o iden i y po en ial a c con ic s when CAVs sha e oads. In mos p e ious s udies, CAVs ypically ha e a high au oma ion le el (i.e., L4) [3–7]. Howe e , o he s udies ha e also included se e al le els o au oma ion [8–10]. In gene al, hey ound ha inc easing he pene a ion a es o CAV can signi can ly educe he numbe o po en ial con ic s. Al hough he impac o CAVs on a c sa e y has been widely s udied, o he bes o he au ho s’ knowledge, no s udy has ho oughly assessed he se e i y o con ic s in a c s eams esul ing om he p og essi e in oduc ion o CAVs. Te no el y o his s udy is i s comp ehensi e analysis o con ic se e i y in ol ing CAVs unde di e en simula ed mixed ee s (i.e., human-d i en ehicles (HDVs) and CAVs o di e en le els). In a e iew conduc ed by Zheng e al. [11], hey highligh ed ha i is necessa y o es ablish an adequa e a c con ic echnique o measu ing a c con ic se e i y, ap- plying a sensi i i y analysis o selec SSMs h eshold and he u iliza ion o a mul idimensional de ni ion o se e i y. Tus, his s udy conside s hese wo esea ch di ec ions o de ise a eliable echnique o assessing he a c con ic among CAVs. Te p esen app oach uses h ee dimensions o ana- lyzing con ic ’s se e i y: (1) he p oximi y o a collision; (2) po en ial con ic consequences; and (3) a combina ion o p oximi y and consequences. Te TTC h eshold is conside ed he ma gin alue o se ious con ic s [12–14]. So, ini ially di e en TTC h esholds a e es ed in his s udy wi h he in oduc ion o CAVs o a ious le els on oads. A e iden- i ying he key alues o di e en TTC h esholds, he s udy examines he consequences o con ic se e i y using some SSMs, namely maximum speed (MaxS) and ehicle speed di e ence (Del aS). Te s udy hen compa es hese alues in a ious scena ios and ypes o ehicle in e ac ion o gain insigh in o he impac o CAVs on a c sa e y. Finally, a ime- o- collision (TTC) o eloci y change a collision (Del aV) dia- g am (i.e., TTC/Del aV cha ) is de eloped o each au oma- ion le el o de i e a con ic se e i y sco e. Te emainde o his pape is o ganized as ollows. In Sec ion 2, he analysis o a c sa e y and con ic se e i y o CAVs and manually d i en ehicles epo ed in he exis ing li e a u e is discussed. Te s udy con ex modeled by Aimsun [15] and he CAV con ol algo i hms and alida ion p ocesses used a e p esen ed in Sec ion 3. In Sec ion 4, he se e i y analysis and i s esul s a e discussed. Finally, Sec ion 5 summa izes he conclusions and limi a ions o he s udy as well as he ecom- mended u u e di ec ions o CAV a c sa e y esea ch. 2. Li e a u e Re iew Tis sec ion p esen s he SSMs used o iden i y a c con ic se e i y. A e wa ds, he ex en o which he con ic se e i y in he CAV eld can be p edic ed is discussed. 2.1. SSMs and Con ic Se e i y. C ash a e and se e i y a e di ec indica o s o a c sa e y pe o mance. Howe e , c ashes a e a e, and da a on alea o y e en s leading o c ashes a e no always s a is ically su cien o s udies. Because o his, SSMs a e used o iden i y a c con ic s and es ima e hei se e i y by analyzing eco ded ideos o a eal- ime analysis [16, 17], and/o a c simula ion ou - pu s. In ac , by ex ac ing ehicle pa hways o e ime and e alua ing hei p oximi y and mo emen s, ehicles close o collisions and wi h je ky mo emen s a e conside ed o be in ol ed in mo e se e e con ic s [18]. Mos s udies ha used SSMs as a c sa e y assessmen ools implied ha a subs an ial co ela ion exis s be ween se ious con ic s and c ash se e i y [19–24]. Howe e , de- i ing con ic se e i y om SSMs as an indica o has been widely deba ed, and se e al a c con ic echniques ha e been de eloped o e he decades. P e ious esea ch has p oposed se e al SSM h esholds o delinea e isky/non isky con ic s. Uni o m and non- uni o m con ic se e i y zones we e also c ea ed ollowing a ious a c con ic echniques and SSM indica o s [25]. To p edic con ic se e i y, ime-based SSMs (e.g., TTC [26], PET [27], ime-in eg a ed TTC (TIT), and ime exposed TTC (TET) [28]), decele a ion-based SSMs (e.g., de- cele a ion a e o a oid c ash [29], maximum decele a ion a e [30], and ea -end collision isk index [31]), and ene gy- based SSMs (e.g., Del aV, ex ended Del aV [32, 33], and con ic index [34]) ha e been used [35]. Acco dingly, a c con ic se e i y has been de ned in e ms o h ee di e en ypes o SSM. Time-based and decele a ion-based SSMs de ne se e i y as he p oximi y wi h espec o a c ash. Tis is he mos p e alen indica ion o s udying a c acciden s and con ic se e i y [33]. Howe e , he ea ly decision-making c i e ia o se e e/nonse e e con ic s mainly depended on he assessmen o human obse e s by iden i ying se e e e en s based on hei p oximi y o a collision [19, 23]. Mo eo e , a ime o space h eshold ha is commonly employed o align se e e con ic s has mul iple assump ions and alida ion alues. Ene gy-based SSMs de ne se e i y by ano he di- mension: he consequences o he isk esul ing om an in e ac ion (con ic ). Te idea is ha high kinema ic o ces esul ing om ehicle in e ac ions conside ably a ec oad use s and p obably esul in se e e inju ies and a ali ies [2]. O e he yea s, esea che s ha e indica ed hei high con- dence in his ype o indica o o p edic ing c ash se e i y. Ca lson [36] a emp ed o de elop models o es ima ing he p obabili y o inju ies o a ali ies in a c ash based on a iables, such as impac speed and ehicle mass; hence, Del aV was used o p edic inju ies and a ali ies. E an [37] subsequen ly ed se e al models using Del aV o p edic inju ies and a ali ies a ising om con ic s. Ne e heless, because his indica o was no used o a c con ic analysis un il i s ecen inco po a ion in o SSAM [2], he de elopmen o new equa ions was no dis inc ly pu sued. Consequen ly, he classical E an models [37] emained in use. Finally, he hi d de ni ion o a c con ic se e i y is ela ed o he concu en p opo ioning o alues o p ox- imi y and p opensi y dimensions and gene a ing di e en se e i y le els. Con ic s wi h po en ially high consequences 2Jou nal o Ad anced T anspo a ion Table 1: Summa y o p e ious s udies abou se e i y wi hin CAV’s analysis. Re e ences Da a sou ce CAV conside ed Con ex Se e i y dimension Se e i y measu es Sinha e al. [6] Simula ion L4 2-Lane mo o way P oximi y and consequences TTC, Del a S El-Hansali e al. [43] Simula ion L4 6-Lane eeway Consequences MaxS, MaxD, MaxDel aV Rahman e al. [44] Simula ion L1, L2 A e ial (61.15 km) P oximi y and consequences TET, TIT, TERCRI, LCC, and NCJ Zhang e al. [45] Simula ion L4 4-Lane eeway (7 km) P oximi y and consequences TET, TIT, TERCRI, and LCC Lau eshyn e al. [32] Video analysis — U ban in e sec ion P oximi y, consequences and le els o se e i y T, Del aV, ex ended Del aV (T/ Del aV) Souley e e and Hochs ein [38] Simula ion — Exp essway in e sec ions Le els o se e i y TTC/MaxDel aV an de Ho s and K aay e al. [39] Manual con ic ’ obse a ion — Va ious Le els o se e i y TTC and speeds a con ic Sinha e al. [46] Field da a (c ash da a) L4 U ban ne wo k Consequences Machine lea ning classi e s Chen e al. [47] Field da a (c ash da a) L4 U ban ne wo k Consequences Machine lea ning classi e s TTC: ime- o-collision, Del aS: di e ence in ehicle speeds as obse ed a MinTTC, MaxS: maximum speed o ei he ehicle h oughou he con ic , MaxD: maximum decele a ion o he ollowe ehicle, Del aV: eloci y change a collision, MaxDel aV: maximum Del aV alue o ei he ehicle in he con ic , TET: ime-exposed- ime- o-collision, TIT: ime-in eg a ed- ime- o-collision, TERCRI: ime exposed ea -end c ash isk index, LCC: lane changing con ic , NCJ: numbe o c i ical je ks, T: he expec ed ime o he second (la es ) ehicle o a i e a he con ic poin , and ex ended Del aV: models o he in eg a ion o T/ Del aV da a. Jou nal o Ad anced T anspo a ion 3 and hose ha a e obse ed close o he occu ence o c ashes a e ound o ha e a high p obabili y o se e i y du ing he in e ac ion [32]. In he pas , a simple human decision-making app oach was employed o iden i y wo zones dis inguishing se e e con ic s om he es o he con ic s conside ing only he p oximi y h eshold alue o ime. Subsequen ly, he In e na ional Commi ee on T a c Con ic Techniques con ibu ed o he de elopmen o se e al con ic echniques ha aided in unde s anding c ash occu ence and i s po en ial se e i y manually (by obse - a ion). Te objec i e was o es ablish se e i y in e ms o se e al le els ins ead o simply spli ing i in o wo ca ego ies (se e e/nonse e e) [25, 32, 38]. Ten, he le els we e ali- da ed by s udies conduc ed ab oad. In he Du ch echnique (i.e., DOCTOR), he con ic s in which speed is high and TTC is less han he h eshold alue a e as deemed se e e [39]. In addi ion, bo h DOCTOR and he Canadian a c con ic echnique [19] inco po a e a subjec i e assessmen in which a sco e ( anging 1–5) de e mines he p obable con ic consequences based on e asi e ac ion, maneu e ing, obse ed speed, and objec i e nea ness-in- ime indica o . Te Swedish a c con ic echnique [21, 40–42] conside s bo h he p oximi y in ime and speed a which he con ic occu s o indica e se e i y and e ec he po en ial conse- quences implici ly. Equidis an pa allel se e i y zones we e es ablished by di iding he esul ing sco es in o se e al le els. Te indica o s used o de i ing he se e i y le els we e a ied (e.g., ehicle speed and dis ance om a con ic si e, equi ed decele a ion, and ic ion coe cien ) [25]. Mo eo e , se e al p oximi y- o-collision h esholds ha e been p oposed in a c con ic echnique esea ch [33]. O he app oaches ha e been employed by o he e- sea che s o indica e se e i y le els. Fo example, Souley e e and Hochs ein [38] de eloped an assessmen sco e by de- ning and adding TTC and Del aV sco es. Ten, some se e i y lines we e ed by d awing con ou s o equal assessmen sco e a eas. Simila ly, Lau eshyn e al. [32] in- co po a ed he minimum ime leading o an acciden and Del aV in a gu e, hus o e ing he ex ended Del aV alues as se e i y lines o de e mining se e i y le els. 2.2. CAV C ash/Con ic Se e i y. Table 1 p o ides a sum- ma y o p e ious s udies ha discussed he se e i y e ms and a c con ic echniques, especially, hose conside ing CAVs in hei analysis. Te e a e some unde going CAVs’ es s on public oads in se e al loca ions in he Uni ed S a es. In hose cases, some s udies a e able o analyze CAVs’ c ash se e i y based on eal da a. Sinha e al. [46] conduc ed a de ailed sa e y analysis using he da a om he Cali o nia Depa men o Mo o Vehicles (2014–2019). Te epo ed da a we e used o de- elop a ious au oma ed ehicle c ash se e i y models ha ocused on he inju ies o all c ash ypes. Howe e , owing o insu cien da a on c ashes in ol ing CAVs, he ac o s ha con ibu e o he se e i y o a CAV’s c ash a e no well de ned. Ne e heless, a ious machine lea ning app oaches ha e been used o be e unde s and CAV c ash se e i y. Chen e al. [47] used a simila app oach and ound ha among all he es ed classi e s, X eme g adien boos ing, a decision ee classi ca ion model, pe o ms be e in de ec ing inju ies occu ing in CAV c ashes. Tei ndings show ha i wo au oma ed ehicles c ash a an in e sec ion o a e unde ad e se wea he condi ions (e.g., og and snow), he se e i y o he c ash signi can ly inc eases. Fu he mo e, c ashes esul ing in inju ies a e mo e likely o occu in loca ions wi h a ious land use pa e ns. Di e se land use (e.g., esiden ial, comme cial, and public) esul s in mixed a c beha io s and changes in egional a c ow, sub- s an ially a ec ing a c sa e y. By con as , as esea che s ex ensi ely employ SSMs o unde s and he sa e y implica ions o new a c designs and al e na i e sa e y emedies be e , modeling he sa e y consequences o CAVs and hei in e ac ions wi h HDVs is a ele an applica ion o SSMs. In addi ion, owing o he limi ed in oduc ion o CAVs, a c mic osimula ion ou pu s ha e been used o p oduce SSMs a he han ana- lyzing ideos. Bo h p oximi y and consequences dimensions ha e been used o analyze he se e i y o CAV con ic s. Se e al p oximi y SSM indica o s ha e been employed, wi h TTC being he mos p e alen indica o . TIT and TET ha e been also widely employed in pa allel wi h TTC [44, 45, 48]. By con as , he dis ibu ions o eme gency b aking [49], ea - end collision isk index [44, 50, 51], sideswipe (lane-change con ic s) a c condi ion [51], and ime exposed ea -end c ash isk index [51] a e all examples o decele a ion-based SSMs o e alua ing CAV a c sa e y [51]. O he su oga e sa e y indica o s, such as s anda d de ia ion o speed [51, 52], MaxS, and Del aS [43, 46, 53], ha e been used as consequence indica o s o assess CAV sa e y implica ions. Howe e , o he bes knowledge o he au ho s, no s udies ha e combined all he se e i y dimensions in CAV a c sa e y analysis. In mos p e ious s udies, CAVs and HDVs we e assessed using he same SSMs and h esholds (e.g., TTC �1.5 s), and no speci c alues we e conside ed o CAVs’ con ic analysis [5, 43, 45]. By con as , some esea che s sugges ha in dealing wi h CAVs, he de aul TTC alue should be educed because o hei as e eac ion imes and sho e headways. Fo ins ance, Mo ando e al. [4] es ed he en- suing con ic s o L4 ehicle pene a ion using h ee TTC h esholds: 1.50 s o any con ic in ol ing HDVs and wo lowe alues (i.e., 1.00 and 0.75 s) o L4–L4 in e ac ions. Tey indica ed ha he TTC h eshold is an impo an ac o in demons a ing he bene o CAV in oduc ion in e ms o sa e y. Gu´ e iau and Duspa ic [8] and Weije ma s e al. [54] p oposed 0.75 s o de e mining con ic s in ol ing CAVs. By con as , Vi di e al. [7] used 0.50 s, and hey claimed ha ega ding hei assump ion ha he headway kep by CAVs is educed o one- hi d, hen he h eshold de ning he con ic should be also p opo ionally educed. E iden ly, o dis inguish be ween se e e and nonse e e sa e y c i ical e en s (con ic s), a su cien h eshold le el mus be de ned. Te de ni ion o his alue is a cu en challenge conce ning con ic s in ol ing CAVs ha mus be scien i cally add essed. 4Jou nal o Ad anced T anspo a ion As ega ds he esul s o p e ious s udies analyzing a c con ic se e i y in CAV a eas, Rahman e al. [44] used TTC-de i ed measu es (e.g., TET and TIT) as p oximi y indica o s in addi ion o e asi e ac ion in- dica o s (e.g., numbe o c i ical je ks and ime exposed ea -end c ash isk index) as consequence indica o s in es ima ing a c con ic se e i y when L1 and L2 ehicles en e a a c s eam. Te esul s e eal ha CAV pen- e a ion exceeding 60% signi can ly educes he con ic se e i y a a e ial segmen s and in e sec ions. Sinha e al. [6] s udied se e al SSMs (e.g., TTC, PET, MaxS) by an- alyzing hei dis ibu ions a di e en pene a ion a es and es ima ing he co esponding c ash a es o assess he con ic se e i y o L4 in oduc ion. Tei ndings showed ha a c sa e y imp o es, and con ic se e i y and c ash a es dec ease when he oads a e ully ope a ed wi h L4 ehicles. Howe e , he con ic s in ol ing HDVs did no dec ease in e ms o equency and se e i y. El- Hansali e al. [43] in es iga ed a c sa e y by compa ing HDVs and L4 ehicles ope a ing independen ly a a eeway sec ion (i.e., 100% HDVs s. 100% L4 ehicles). Con a y o expec a ions, g ea e alues o se e i y in- dica o s we e obse ed in he case o L4 ehicles han in he case o HDVs. Fo example, a highe MaxS was ob- ained o ei he ehicle ype du ing con ic s, and a highe MaxD was obse ed when only L4 ehicles occupied he oad. Zhang e al. [45] conduc ed a s udy ha ocused on oadway con gu a ion. Using p oximi y and conse- quences indica o s (i.e., TTC, TIT, TET, ime exposed ea -end c ash isk index, and lane-changing con ic s), hey in es iga ed he sa e y o lanes dedica ed o L4 ehicles wi h di e en pene a ion a es. Tey emphasized ha es ablishing e en one exclusi e lane could inc ease sa e y because he con ic se e i y was signi can ly e- duced in e ms o longi udinal and la e al mo emen s. 3. Me hodology Aimsun [15] has been used o mic osimula ion o es i- ma e ajec o ies o he di e en ypes o ehicles con- side ed. Subsequen ly, SSAM [18] was applied o ex ac su oga e sa e y indica o s o he se e i y es ima ion p ocess. 3.1. S udy Con ex . As a es co ido , he s udy a ea o a h ee-lane wo-way mo o way segmen (20.27 km o GR- 30, an impo an oad leading o G anada Ci y, Spain) (il- lus a ed in Figu e 1) was modeled using Aimsun Nex . Te geome ic design cha ac e is ics o he segmen (e.g., oad p o le, cu es, and lane de ailing) we e in oduced using an impo ed Open S ee Map o he segmen . Te chosen segmen has 14 on- amps and o - amps and wo majo en y poin s. T a c ow da a we e ga he ed using nine de ec o s ins alled in he a ea by he Gene al T a c Di- ec o a e (Di ecci´ on Gene al de T ´ a co (DGT)). Te de- ec o s egis e ins an aneous speeds, a c olumes, and ehicle ype dis ibu ions (hea y ehicles s. passenge ca s) a 15-min in e als. Te impo ed le da a om he DGT senso s o ali- da ion we e selec ed o a egula day (Tuesday) and o -peak hou (10:00-11:00 am) because his s udy modeled a ee- ow condi ion. Te a e age ins an aneous speed ange was 83–118 km/h, and he a c coun was eco ded e e y 15 min: 547−3570 pc/h and 89–260 h /h we e egis e ed o he GR-30 no hbound ehicles, and 809–3281 pc/h and 93–499 h /h we e egis e ed o he GR-30 sou hbound ehicles. 3.2. Mic osimula ion Model and Scena ios. Aimsun was selec ed o calib a e he di e en au oma ion le els o e- hicles ( om L0 o L4) because i p o ides specialized ools o CAVs. V2X ex ension was employed o model he connec i i y be ween ehicles. Te p oposed analy ical pe- iod o he mic osimula ion is 1 h; howe e , o a c alida ion ( olume and speed), his pe iod was b oken down o 15-min in e als o e ec he eal a c da a eco ded by he DGT’s de ec o s be e . T a c ope a ion da a a e gen- e a ed using a small ime s ep (i.e., 0.1 s, ollowing p e ious s udies [4, 5] o inc ease simula ion accu acy and educe he isk o losing ehicle mo emen de ails). Te wa m-up ime was se o 18 min ollowing Wunde lich e al. [55] (based on he oad sec ion leng h and he a e age speed o ehicles). Fu he mo e, he model ope a ions we e calib a ed and alida ed ollowing he modeling guidelines o Roads and Ma i ime Se ices [56]. Miqdady e al. [10] p o ide mo e de ails abou his s ep. A e checking he alidi y o he modeled ne wo k, he “ca - ollowing and lane-change models” o Gipps [57, 58] ha a e a ian s o each a el condi ion a e calib a ed o all wi hin he p oposed s ochas ic dynamic en elopes o CAVs. Fo ins ance, CAVs a e supposed o ha e sho eac ion imes, accep sho gaps, coope a e in lane changes, e c. Tables 4 and 5 (in Appendix A) show all he pa ame e s ha a e a ec ed by di e en le els o ehicle au oma ion in he ca - ollowing and lane-changing models o Gipps based on p e ious esea ch [4, 5, 8, 43–45, 54, 59] and logic. Te pa ame e de ni ions a e summa ized om he Aimsun G anada GR-30 Spain Figu e 1: S udy a ea: GR-30 oadway segmen in G anada (Spain). Jou nal o Ad anced T anspo a ion 5 use manual. Te alues summa ized in Tables 4 and 5 a e he inpu s o he mic osimula ion models. Tey a e p o- ided as means and s anda d de ia ions and a e no mally dis ibu ed as sugges ed by Gipps models o bo h passenge ca s and hea y ehicles. Te analysis a emp ed o co e a g adually in oduc ion o CAVs wi h a ious ee mixes ha he eal wo ld may encoun e . Acco dingly, as jus i ed in Miqdady e al. [10], nine mixed ee scena ios wi h di e en CAV pene a ion a es we e sugges ed. Table 2 lis s he combina ions o HDVs and ehicles wi h di e en au oma ion le els (L1–L4) in each scena io. 3.3. Se e i y Analysis. To e alua e he a c con ic se e i y in he simula ed scena ios (i.e., he po en ial ma ke in- oduc ion scena ios o CAV), his s udy conside ed he h ee ques ions p esen ed by Lau eshyn e al. [32]: (i) How can he p oximi y o a c ash be measu ed? (ii) How can he se e i y o he consequences o a po en ial c ash be mea- su ed? (iii) How can bo h dimensions be me ged? A ew s udies ha e analyzed he ex en o con ic se e i y in he CAV con ex [6, 44]. Howe e , hey ha e nei he conside ed all dimensions o se e i y combined no ana- lyzed all le els o au oma ion. Fo he nine p oposed sce- na ios, se e al SSM indica o s we e applied o de e mine bo h p oximi y and consequence dimensions a each con- ic . Te ollowing sec ion illus a es how his s udy ad- d esses he p e ious esea ch ques ions. Te de ailed amewo k is illus a ed in Figu e 2. Te nex sec ion ex- plains he app oach ollowed o explo e each se e i y di- mension and p esen s he esul s ob ained a e applying hese app oaches. 4. CAV Se e i y Dimensions 4.1. P oximi y T eshold. Te mos widely used indica o o in es iga ing a c p oximi y and con ic se e i y in HDV and CAV con ic analysis is TTC [35]. Tis indica o is de ned as “ he ime ha emains un il a collision could occu i wo successi e ehicles main ain a speed di e ence” [28]. I is gi en by he ollowing: TTCi( ) � xi−1( ) − xi( ) − li−1 i( ) − i −1( ),i i( )> i −1( ), ∞,i i( )≤ i −1( ), ⎧⎪ ⎪ ⎪ ⎨ ⎪ ⎪ ⎪ ⎩ (1) whe e TTCi( ) deno es he TTC alue o he ollowing e- hicle, i, a a ime ins an ; ,x, and deno e he ime, posi ion, and eloci y o he ehicles, espec i ely; and l i−1 ep esen s he leng h o he leading ehicle. A small TTC alue indica es a high isk o collision a a gi en ime ins an . To assess he se e i y o ehicle- ollowing e en s, a TTC h eshold mus be de ned o dis inguish be ween se e e and nonse e e con ic s [60]. Se ing an uni e sal TTC h eshold o assessing con ic se e i y has become a ma e o con- en ion, pa icula ly in he case o CAV in oduc ion. A e iew o p e ious esea ch e eals ha se e al h esholds anging 0.9–5.0 and 0.5–1.5 s ha e been p oposed o a ious HDV a c and d i ing condi ions and o CAV scena ios, espec i ely [60]. Al hough his s udy analyzes a c sa e y o CAV in oduc ion scena ios, a unique alue (i.e., 1.5 s) is p oposed o con ic s in ol ing HDVs o ehicles wi h low au oma ion (L1 and L2) as ollowe ehicles (low CAVs, LCAV). Te mos commonly used alue o HDV is 1.5 s [18]; and i is also he de aul alue used in he SSAM. Sensi i i y analysis was conduc ed o de ne a easonable h eshold o con ic s whe e a high le el o au oma ion ehicle (L3 and L4) is he ollowe (high CAVs, HCAV). Fi e di e en alues (0.5, 0.75, 1.0, 1.25, and 1.5 s) we e examined o he TTC h eshold unde each scena io o emphasize he app op ia e alue unde a ious ci cums ances. Table 3 summa izes he numbe o con ic s when applying he di e en TTC alues o de e mine whe he he e a e sig- ni can changes by using one-way analysis o a iance o each scena io. Te changes esul ing om applying any alue and he base alue (1.5 s) a e lis ed in Table 3. Table 3 shows ha (i) TTC does no p esen a signi can in uence on he numbe o con ic s a scena ios wi h low pene- a ion a es o HCAV (scena ios D o below) (ii) TTC p esen s a e y signi can in uence on he numbe o con ic s a scena ios wi h high pene- a ion a es o HCAV (scena ios G o o e ) (iii) A in e media e scena ios (E o F), ep esen ing mode a e pene a ion a es o HCAV, he numbe o a c con ic s s a s o p esen signi can di - e ences i he TTC alue is below 1.0 s. Tese esul s emphasize he impo ance o using di e en TTC alues o ob ain a eliable assessmen o a c sa e y ela ed o high pene a ion o HCAV. Mo eo e , he esul s e i y he heo e ical ision o CAV in oduc ion: when CAV pene a ion a e is high, a c ow imp o es by achie ing mo e ha monized speeds and by educing eac ion imes ha p obably ha e a di ec e ec on he TTC h eshold. Tese esul s ag ee wi h he alues sugges ed in p e ious s udies. Mo ando e al. [4] used wo TTC alues (0.75 and 1.0 s) o iden i y con ic s in ol ing CAVs; bo h alues we e assumed o be ap- p op ia e. O he s udies used 0.75 s as he TTC alue [8] o xed con ic s wi h CAV pa icipa ion, and o he s udies educed his h eshold o 0.5 s [7, 61]. Papazikou Table 2: Te s udied mixed ee s’ simula ed scena ios. Scena ios HDV (%) L1 (%) L2 (%) L3 (%) L4 (%) A 100 0 0 0 0 B 75 10 10 5 0 C 50 10 25 10 5 D 40 15 20 15 10 E 20 20 25 20 15 F 5 10 30 30 25 G 0 0 10 40 50 H 0 0 0 25 75 I 0 0 0 0 100 6Jou nal o Ad anced T anspo a ion e al. [61] claimed ha CAVs ope a ing wi h asse i e d i ing s yles could lead o di e en ci cums ances esul ing in a lowe TTC h eshold. 4.2. Se e i y Consequences Indica o s. Te p oximi y o a collision ha esul s in a sligh c ash mus no be equa ed o a c ash wi h a po en ially se e e inju y. Te e o e, he se e i y measu ed by he po en ial consequences o a c ash mus be accoun ed by some o he means [32]. Se e al SSMs can be used o ex ac he dynamic consequences o a con ic [18, 35]. Following p e ious s udies [42, 62], his esea ch uses MaxS and Del aS o measu e he esul ing se e i y o con ic s ela ed o di e en ypes o ehicles (HDVs and L1, L2, L3, and L4 ehicles). Te o me is de ned as he maximum speed o any o he ehicles h oughou he con ic , whe eas he la e is he di e ence in ehicle speed (i.e., he di e ence in he eloci y o ehicles in con ic ) obse ed a he minimum alue egis e ed o TTC. Bo h indica o s a e ou pu s o he SSAM and simula e he esul ing dynamics. Di e en ehicle in e ac ions can esul in a ied a c ow dynamics, and consequen ly, he se e i y le els di e . A he end, high MaxS and Del aS alues indica e ha he con ic s esul in high se e i y. Te a ia ions in MaxS and Del aS o di e en ehicles in ol ed in a con ic wi hin di e en a c ee scena ios a e shown in Figu e 3. Fo simplici y and cla i y in p e- sen ing he esul s, L1 and L2 ehicles a e g ouped as low CAVs and L3 and L4 ehicles as high CAVs, shown as LCAV and HCAV in Figu e 3, espec i ely. Te shown alues (o MaxS and Del aS) a e he mean alues o 15 uns in each scena io. Te blue-yellow- ed scale indica es he inc ease in se e i y owa ds he ed colo . Las ly, in each gu e, he alues a e ca ego ized by he ollowe ehicle in he con ic : -HDV, -LCAV, and -HCAV, indica ing ha he ollowe ehicle is a HDV, LCAV, and HCAV, espec i ely. Figu e 7 in Appendix B shows an example o he mic osimula ion esul s as equency dis ibu ions o MaxS and Del aS. Tese dis ibu ions show also he hea maps’ alues (i.e., he mean alues exhibi ed in Figu e 3). Examining Se e i y dimensions among he lee mixes Mic osimula ion scena ios (Ex ac ing ehicles ajec o ies) SSAM analysis (Ex ac ing SSMs) Tes ing con lic consequences Tes ing p oximi y Tes ing con lic se e i y sco e (CAV a ic con lic echnique) One-way ANOVA o es con lic s esul ed by di e en TTC h eshold (0.50 ,0.75, 1.00, 1.25, & 1.50 s) Hea maps o MaxS and Del aS by scena io and ehicle in e ac ion Ob aining TTC sco e: in lec ion poin s om he TTC cumula i e dis ibu ion a pu e ehicle ype simula ion (HDV, L1, L2,L3, o L4 ehicle, exclusi ely). Es ablishing MaxDel aV sco e. Se e i y o e all sco es cha s based on he sum o TTC sco e and MaxDel aV sco e. Classi ica ion o con lic s by se e i y sco e a he simula ed scena ios. (i) (ii) (iii) (i ) Figu e 2: F amewo k o simula ion-based a c con ic se e i y es ima ion o CAV. Jou nal o Ad anced T anspo a ion 7 Rega ding he con ic consequences ex ac ed a he di e en scena ios, Figu e 3 shows ha (i) Te highe MaxS du ing con ic s is ypically ob- se ed in scena ios in which he pene a ion a e o HCAV is om low o mode a e (less han 55%, o scena io F) (see Figu e 3(a)) (ii) By con as , high pene a ion a es o HCAV (scena ios G, H, and I) esul in lowe MaxS du ing con ic s (see Figu e 3(a)) (iii) Simila conclusions could be ob ained om Del aS’s esul s in Figu e 3(b) Sinha e al. [6] epo ed a simila pa e n. Tey ob- ained low c ash a es and a dis ibu ions o Del aS alues as he pene a ion a es o L4 ehicles inc eased. Rahman e al. [44] obse ed, using o he su oga e sa e y indica o s (e.g., TET, TIT, numbe o c i ical je ks, and ime exposed ea -end c ash isk index), ha he inc ease in he pene a ion a e o ehicles wi h low au oma ion le els (i.e., L1 and L2 ehicles) dec eased he con ic se e i y. Tey ound ha he highes educ ion in se e i y was achie ed when he pene a ion a e was 100% CAV. By con as , he educ ion was insigni can when he pene a ion a e was less han 40%. Table 3: Sensi i i y analysis o di e en alues o TTC h eshold o HCAV (-L3 and -L4 ehicles). Scena io TTC h eshold o HCAV No. o con ic s % Change A(0)∗— 3251 — B(5) 0.50 2636 −0.69 0.75 2637 −0.68 1.00 2637 −0.60 1.25 2640 −0.56 1.50 2655 — C(15) 0.50 1671 −6.08 0.75 1675 −5.86 1.00 1697 −4.62 1.25 1724 −3.11 1.50 1779 — D(25) 0.50 1131 −11.10 0.75 1137 −10.61 1.00 1156 −8.40 1.25 1200 −5.64 1.50 1272 — E(35) 0.50 890a∗∗ −16.85 0.75 900a −15.91 1.00 935a −12.69 1.25 980a,b −8.50 1.50 1071b — F(55) 0.50 628a −31.22 0.75 648a −28.29 1.00 709a,b −22.36 1.25 770b −15.66 1.50 913c — G(90) 0.50 255a −66.13 0.75 298a −60.46 1.00 415b −44.88 1.25 528c −29.99 1.50 754d — H(100) 0.50 149a −79.03 0.75 198b −72.02 1.00 341c −51.91 1.25 467d −34.06 1.50 709e — I(100) 0.50 133a −82.79 0.75 192b −75.12 1.00 365c −52.67 1.25 517d −32.99 1.50 771e — ∗Te alue in ( ) deno es o he pe cen ages o HCAVs in he scena io. ∗∗Fo each alue con aining a, b, . . ., le e in a scena io (in he no. o con ic s column), i deno es alues o s a is ically signi can di e ences (p<0.05). Two o mo e alues wi h he same le e deno e a homogeneous subg oup. No e. TTC h eshold �1.5 s is es ablished when he ollowe ehicle is a HDV o a LCAV (L1 o L2 ehicle). 8Jou nal o Ad anced T anspo a ion Vehicles in ol ed -HDV MaxS (m/s) 20 22 21 19 23 0 0 0 0 0 2124232127 0 0 0 0 1122212222 0 0 0 HCAV-HCAV LCAV-HCAV HDV-HCAV 0 5 10 15 20 25 GH IEFCBDA Scena io -LCAV MaxS (m/s) 0 0 0 21 20 22 18 12 000 24 16 20 22 23 14 00 0011 23 19 21 20 8.7 Vehicles in ol ed HCAV-HCAV LCAV-HCAV HDV-HCAV 0 5 10 15 20 25 BCDEFGHIA Scena io -HCAV MaxS (m/s) 0 0 0 11 8.6 11 8.9 7.6 0 0 0 6.2 9.6 11 13 15 8.4 0 0 0 9.1 16 19 17 7.1 7.7 0 Vehicles in ol ed HCAV-HCAV LCAV-HCAV HDV-HCAV 0 5 10 15 20 25 BCDEFGHIA Scena io (a) Figu e 3: Con inued. Jou nal o Ad anced T anspo a ion 9 Table 6: Te assigned TTC sco e by ehicle ype. TTC sco e HDV L1 L2 L3 L4 T esholds Sample size (%) T esholds Sample size (%) T esholds Sample size (%) T esholds Sample size (%) T esholds Sample size (%) 0 4.0 <TTC ≤5.0 30.0 4.2 <TTC ≤5.0 28.9 4.2 <TTC ≤5.0 30.4 4.3 <TTC ≤5.0 29.9 4.3 <TTC ≤5.0 32.8 1 2.5 <TTC ≤4.0 26.9 2.5 <TTC ≤4.2 31.9 2.5 <TTC ≤4.2 31.1 2.6 <TTC ≤4.3 33.6 2.6 <TTC ≤4.3 31.1 2 1.5 <TTC ≤2.5 27.6 1.0 <TTC ≤2.5 32.4 1.0 <TTC ≤2.5 32.3 0.75 <TTC ≤2.6 31.5 0.75 <TTC ≤2.6 31.5 3 TTC ≤1.50 15.3 TTC ≤1.0 6.6 TTC ≤1.0 6.1 TTC ≤0.75 4.8 TTC ≤0.75 4.4 16 Jou nal o Ad anced T anspo a ion 1 2 TTC sco e 1 23 4 23 4 5 01234 MaxDel aV sco e (a) MaxDel aV sub-sco e TTC sub-sco e 2 2 2 2 2 2 2 2 2 2 3 3 3 3 3 3 3 3 3 3 3 4 4 4 4 4 4 4 4 4 4 4 5 5 5 5 55 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0 1234 (b) Figu e 8: Concep ual illus a ion o conduc ing he o e all se e i y sco e: (a) o e all sco e by egions and (b) s ep-g aded lines om he subsco es. HDV 1 2 3 4 5 6 123450 TTC (s) 0 20 40 60 80 100 120 MaxDel aV (Km/h ) y = 27.273x - 76.364 y = 27.273x - 46.364 y = 25.532x - 7.6596 y = 24.194x + 29.032 y = 23.377x + 63.117 (a) LCAV (L1/L2) 0 20 40 60 80 100 120 MaxDel aV (Km/h ) 123450 TTC (s) y = 27.778x - 78.889 y = 24.324x - 31.622 y = 25x - 5 y = 25.862x + 20.69 y = 25x + 55 (b) HCAV (L3/L4) 123450 TTC (s) 0 20 40 60 80 100 120 MaxDel aV (Km/h ) y = 29.126x - 85.631 y = 23.196x - 25.979 y = 24.742x - 3.7113 y = 26.786x + 16.071 y = 24.896x + 55.519 (c) Figu e 9: Se e i y sco es (SS) o di e en ypes o ehicles: (a) o HDV, (b) o L1 & L2 ehicles (LCAV), and (c) o L3 & L4 ehicles (HCAV). Jou nal o Ad anced T anspo a ion 17 5. Conclusion Tis s udy in es iga es he ex en o con ic se e i y esul ing om he in oduc ion o CAVs in o a c s eams. I p esen s an analysis o he po en ial a c con ic s ha occu when oads comple ely ope a ing wi h HDVs an- si ion in o ull L4 ehicle ope a ion. T ee dimensions o se e i y a e examined: p oximi y o collision, consequences o collision, and p oximi y/consequence o collision classi- ed by se e i y sco e. Owing o he lack o c ash da a in- ol ing CAVs, his s udy implemen ed a a c mic osimula ion app oach ollowed by SSAM analysis. Te speci c ou pu s o he SSAM (e.g., TTC, MaxS, Del aS, and MaxDel aV) a e used o es ima e con ic se e i y. Te e- sul s o se e al mixed ee ope a ion scena ios a e compa ed o de e mine a c sa e y when he eal and cu en ex en o pene a ion o CAVs on oads is exceeded. Te key ndings o his s udy a e as ollows. Te sen- si i i y analysis o he TTC h eshold in scena ios whe e HCAV is he ollowe ehicle yields in e es ing esul s. I he p esence o HCAV on he oad is low (less han 35%), he di e ence in he numbe o iden i ed con ic s be ween he applied TTC h eshold alues (i.e., 0.5, 0.75, 1.0, 1.25, and 1.5 s) is no s a is ically signi can . By con as , he scena ios whe e HCAVs ha e mode a e sha ing pe cen ages (35%– 55%) s a o show a signi can di e ence a 1.0 s. Te scena ios whe e he ope a ion pe cen age o hese ehicles is high lead o signi can di e ences in he numbe o con ic s among all he es ed TTC alues. Te e o e, he impo ance o applying di e en TTC h eshold alues o such sce- na ios mus be ecognized. Te MaxS and Del aS alues we e discussed as con ic consequence indica o s wi hin he p oposed scena ios and se e al ehicle in e ac ions. Tese indica o s show ha he scena ios whe e 55% o mo e HCAV sha e he oad esul in con ic s wi h low se e i y (low speeds and low speed di e ences among ehicles in- ol ed in con ic s). In addi ion, he con ic s whe e HDVs a e he ollowe ehicles yield he highes se e i y con ic s, ollowed by he con ic s whe e he ollowe ehicles a e LCAV. Finally, p oximi y/consequence (TTC/MaxDel aV) cha s ela ed o di e en ehicle ypes ha e been de eloped. Tese cha s ha e been used o classi y he esul ing con ic s in o se e i y sco es in each scena io. Te esul s indica e ha inc easing he sha ed pe cen ages o CAVs ope a ing on he oad signi can ly dec eases he numbe o con ic s wi h high se e i y. When app oxima ely 100% o HCAV ope a e on oads, se e e con ic s a e an icipa ed o disappea , and hose wi h low se e i y a e educed. Tis s udy p esen ed a comp ehensi e in es iga ion o a c con ic se e i y dimensions and analyzed he con ic se e i y ela ed o se e al le els o au oma ion wi hin a ious mixed ee ope a ion scena ios. Ne e heless, his s udy p esen s some limi a ions ha should be conside ed o u u e esea ch. Fi s ly, whe he he SSMs h esholds unde con en ional a c condi ions a e applicable when mod- eling sa e y in mixed o ully au oma ed a c emains unclea . Di e en TTC h eshold alues ha e been es ed and applied o sol e his p oblem. Howe e , when eal da a become a ailable, he alidi y o SSM should be ho oughly e iewed and e i ed. Te e o e, new da a sou ces ela ed o CAV da a will be c ucial o he de elopmen o an uni e sal SSM se ha can sa is y all au oma ion le els. Secondly, o pa icula a c scena ios, he in es iga ion o SSMs, such as he la e al sa e y p o ided by lane changing and me ging maneu e s, mus be implemen ed. Bo h HDVs and CAVs can exhibi di e en le els o la e al sa e y, pa icula ly in a mixed au onomy a c. And nally, he calib a ion p ocess could be imp o ed wi h eld TTC da a [63, 64]. Tis should be conside ed in simila u u e esea ch. Appendix A. Beha io Pa ame e s Used o CAV Le els Modeling Tis appendix con ains he CAV beha io pa ame e alues as indica ed in Table 4 ( o passenge ca s) and Table 5 ( o hea y ehicles). B. A Sample o Mic osimula ion Resul s Te ollowing esul s in Figu e 7 ep esen an example o he mic osimula ion ou pu s ela ed o MaxS and Del aS ha we e gene a ed a scena io E, when HDV is he ollowe ehicle in he con ic s. MaxS and Del aS ou pu s a e p esen ed as dis- ibu ion cha s o e ec he esul ed da a mo e desc ip i ely. C. Se e i y Cha s o HDVs and CAVs Tis appendix desc ibes he p ocedu e ollowed in de- eloping he se e i y cha s (by ehicle ype) based on TTC sco e/MaxDel aV sco e. To ob ain he TTC sco e, di e en TTC h esholds we e es ablished by ehicle ype. Te p ocedu e looks o he in ec ion poin s o he TTC cumula i e dis ibu ion when pu e ehicle ype scena ios we e modeled (i.e., all he e- hicles in he simula ion a e exclusi ely HDVs, L1, L2, L3, o L4 ehicles). Speci cally, 15 mic osimula ion uns we e execu ed o each pu e scena io, and he TTC cumula i e dis ibu ion cha s we e depic ed. All he con ic s iden i ed wi h a TTC alue equal o o lowe han 5.0 s we e con- side ed o he TTC dis ibu ion analysis. Acco ding o he Hyd´ en [21] sa e y py amid, ex emely se e e con ic s a e conside ably limi ed, whe eas less se e e a c con ic s a e mo e equen . Acco ding o Souley e e and Hochs ein [38], hese se e e con ic s can be ob ained based on he in ec ion poin s o he TTC cumula i e dis- ibu ion o he pu e scena ios o each ehicle ype. Tese poin s a e used as h esholds o delinea e he ew se e e con ic s om nonse e e ones. La e , he nonse e e con ic s we e di ided in o h ee app oxima ely equal g oups. A TTC sco e was assigned o each g oup (one se e e and h ee nonse e e), which was la e used o ob ain he o e all sco e. Table 6 summa izes he p oposed TTC sco es and h esholds o de e mine he o e all sco es o he pu e ope a ion scena ios. 18 Jou nal o Ad anced T anspo a ion As lis ed in Table 6, he h esholds ha iden i y se e e con ic s (wi h a TTC sco e equal o 3) di e among he pu e ehicle ype ope a ional scena ios. P ecisely, he in ec ion poin o HCAV (L3 and L4 ehicles) was lowe han hose o he o he ehicles, indica ing hei imp o ed capabili ies. Te in ec ion poin s o he pu e scena ios o HCAV, LCAV, and HDVs we e 0.75, 1.0, and 1.5 s, espec i ely. Table 6 shows he a ia ion in he in ec ion poin a ec s o he h esholds. Mo eo e , he dec ease in a se e e con ic egion (wi h a sco e equal o 3) is clea ly achie ed by inc easing he au oma ion le el, a ming he sa e y bene o inco po a ing HCAV. Fo he consequence sco e, Souley e e and Hochs ein [38] used he equa ion o E an [37], which employs Max- Del aV o calcula e he likelihood o c ash inju ies and a- ali ies. Tei esul s showed ha MaxDel aV alues o app oxima ely 30 and 60 km/h a e key (in ec ion) alues ha signi can ly inc ease he p opensi y o se e e con ic s. Because he E an equa ion only depends on MaxDel aV and he consequences o a c ash wi h ce ain MaxDel aV alues ha e he same e ec on HDVs and CAVs (wi h di e en au oma ion le els), hese wo key alues can be conside ed he same o all au oma ion le els and in e ac ions. Te e o e, ollowing Souley e e and Hochs ein [38], he se e i y alue o MaxDel aV is di ided in o h ee sco es: (1) sco e 1, MaxDel aV anging 0–30 km/h; (2) sco e 2, Max- Del aV anging 30–60 km/h; and (3) sco e 3, MaxDel aV exceeding 60 km/h. Te nex s ep is adding bo h sco es (TTC and Max- Del aV sco es) o ob ain an o e all se e i y sco e. Te esul ed o e all sco e is ep esen ed by egions in Figu e 8(a), whe e each sco e a ea ep esen s he se e i y sco e. Howe e , se e i y sco es a e be e iden i ed by lines o cu es a he han by squa e a eas [32, 38]. Fo his eason, hese squa e a eas a e con e ed in o se e i y isolines as con ou lines (Figu e 9). P ecisely, each o e all squa e a ea is eshaped o e subsco es (wi h an inc emen o 0.2 poin s) o each majo sco e (in bo h anges o TTC and MaxDel aV sco es), see Figu e 8(b). A e wa ds, he s ep-g aded lines esul ing om he equal o e all sco es in Figu e 8(b) a e eshaped in o smoo h con ou lines o HDVs, LCAV, and HCAV, as shown in Figu e 9. In he gu e, he a ia ion o he ed colo om ligh o da k ep esen s he inc emen in Se e i y Sco e (SS). Da a A ailabili y Te da a suppo ing he cu en s udy a e a ailable om he co esponding au ho upon eques . Disclosu e Tis s udy is pa o he Resea ch P ojec PID2019- 110741RA-I00. Con lic s o In e es Te au ho s decla e ha he e a e no con ic s o in e es . Au ho s’ Con ibu ions Roc´ ıo de Oña and Juan de Oña concep ualized he s udy; Tasneem Miqdady, Roc´ ıo de Oña, and Juan de Oña pe - o med me hodology; Tasneem Miqdady, Roc´ ıo de Oña, and Juan de Oña did o mal analysis and in es iga ion; Tasneem Miqdady w o e he o iginal d a p epa a ion; Roc´ ıo de Oña and Juan de Oña w o e he a icle and e iewed and edi ed he a icle; Roc´ ıo de Oña did unding acquisi ion. Acknowledgmen s Te au ho s a e g a e ul o he Spanish Gene al Di ec o a e o T a c (DGT) o p o iding he a c ows om se e al GR-30 sec ions. Tasneem Miqdady app ecia es Aimsun o p o ide hei pos g adua e s uden license o make his wo k. Tis wo k was suppo ed by Resea ch P ojec PID2019-110741RA-I00, nanced by he Spanish S a e Re- sea ch Agency (MCIN/AEI/10.13039/501100011033). Open Access unding enabled and o ganized by CRUE-CBUA Gold. Re e ences [1] Sae, S anda d J3016: Taxonomy and De ni ions o Te ms Rela ed o On-Road Mo o Vehicle Au oma ed D i ing Sys- ems, 2014. [2] S. G. Shelby, “Del a-V as a measu e o a c con ic se e i y,” T anspo a ion Resea ch Reco d, ol. 1, pp. 1–19, 2011. [3] T. Miqdady, R. DeOña, and J. DeOña, Quan i ying he Sa e y Impac o Connec ed and Au onomous Vehicles in Mo o ways: A Simula ion-Based S udy, Lib o de Ac as), 2021. [4] M. M. Mo ando, Q. Tian, L. T. T uong, and H. L. Vu, “S udying he sa e y impac o au onomous ehicles using simula ion-based su oga e sa e y measu es,” Jou nal o Ad- anced T anspo a ion, ol. 2018, A icle ID 6135183, 11 pages, 2018. [5] A. Papadoulis, M. Quddus, and M. Imp ialou, “E alua ing he sa e y impac o connec ed and au onomous ehicles on mo o ways,” Acciden Analysis and P e en ion, ol. 124, pp. 12–22, 2019. [6] A. Sinha, S. Chand, K. P. Wijaya a na, N. Vi di, and V. Dixi , “Comp ehensi e sa e y assessmen in mixed ee s wi h connec ed and au oma ed ehicles: a c ash se e i y and a e e alua ion o con en ional ehicles,” Acciden Analysis and P e en ion, ol. 142, A icle ID 105567, 2020. [7] N. Vi di, H. G zybowska, S. T. Walle , and V. Dixi , “A sa e y assessmen o mixed ee s wi h connec ed and au onomous ehicles using he su oga e sa e y assessmen module,” Ac- ciden Analysis and P e en ion, ol. 131, pp. 95–111, 2019. [8] M. Gu´ e iau and I. Duspa ic, “Quan i ying he impac o connec ed and au onomous ehicles on a c e ciency and sa e y in mixed a c,” in P oceedings o he 2020 IEEE 23 d In e na ional Con e ence on In elligen T anspo a ion Sys- ems, Rhodes, G eece, Sep embe 2020. [9] H. Xie, E. Tanin, S. Ka unaseke a e al., “Quan i ying he impac o au onomous ehicles using mic oscopic simula- ions,” in P oceedings o he 12 h In e na ional Wo kshop on Compu a ional T anspo a ion Science, Chicago, IL, USA, No embe 2019. [10] T. Miqdady, R. De Oña, J. Casas, and J. De Oña, “S udying a c sa e y du ing he ansi ion pe iod be ween manual d i ing and au onomous d i ing: a simula ion-based Jou nal o Ad anced T anspo a ion 19 app oach,” IEEE T ansac ions on In elligen T anspo a ion Sys ems, p. 1, 2023. [11] L. Zheng, K. Ismail, and X. Meng, “T a c con ic echniques o oad sa e y analysis: Open ques ions and some insigh s,” Canadian Jou nal o Ci il Enginee ing, ol. 41, no. 7, pp. 633–641, 2014. [12] J. C. Haywa d, “Nea -miss de e mina ion h ough use o a scale o dange ,” Highway Resea ch Resea ch Reco d, ol. 384, pp. 24–34, 1972. [13] L. Zheng and T. Sayed, “F om uni a ia e o bi a ia e ex eme alue models: app oaches o in eg a e a c con ic in- dica o s o c ash es ima ion,” T anspo a ion Resea ch Pa C: Eme ging Technologies, ol. 103, pp. 211–225, 2019. [14] B. L. Allen, B. T. Shin, and P. J. Coope , “Analysis o a c con ic s and collisions,” T anspo a ion Resea ch Reco d, ol. 667, pp. 67–74, 1978. [15] Aimsun, Aimsun Nex 20 Use ´s Manual, Aimsun Nex Ve sion 20.0.2, Ba celona, Spain, 2020. [16] Y. Guo, T. Sayed, and M. Essa, “Real- ime con ic -based Bayesian Tobi models o sa e y e alua ion o signalized in e sec ions,” Acciden Analysis and P e en ion, ol. 144, A icle ID 105660, 2020. [17] Y. Guo, T. Sayed, and L. Zheng, “A hie a chical bayesian peak o e h eshold app oach o con ic -based be o e-a e sa e y e alua ion o leading pedes ian in e als,” Acciden Analysis and P e en ion, ol. 147, A icle ID 105772, 2020. [18] D. Ge man, L. Pu, T. Sayed, and S. Shelby, Su oga e Sa e y Assessmen Model and Valida ion: Final Repo , Publica ion No. FHWA-HRT-08-051, 2008. [19] G. R. B ow, “T a c con ic s o oad use sa e y s udies,” Canadian Jou nal o Ci il Enginee ing, ol. 21, no. 1, pp. 1–15, 1994. [20] K. El-Basyouny and T. Sayed, “Sa e y pe o mance unc ions using a c con ic s,” Sa e y Science, ol. 51, no. 1, pp. 160–164, 2013. [21] C. Hyd´en, Te De elopmen o a Me hod o T a c Sa e y E alua ion: Te Swedish T a c Con ic Technique Doc o al Tesis, Lund Uni e si y, Depa men o T a c Planning and Enginee ing, 1987. [22] D. Lo d, “Analysis o pedes ian con ic s wi h le - u ning a c,” T anspo a ion Resea ch Reco d, ol. 1538, 1996. [23] D. J. Migle z, W. D. Glauz, and K. M. Baue , Rela ionships be ween T a c Con ic s and Acciden s, U.S. Depa men o T anspo a ion Fede al Highway Adminis a ion, 1985. [24] E. Sacchi, T. Sayed, and P. de Leu , “A compa ison o collision-based and con ic -based sa e y e alua ions: he case o igh - u n sma channels,” Acciden Analysis and P e- en ion, ol. 59, pp. 260–266, 2013. [25] J. A che , Indica o s o T a c Sa e y Assessmen and P e- dic ion and Tei Applicai on in Mic o-simula ion Modelling: A S udy o U ban and Subu ban In e sec ionsDepa men o In as uc u e. Royal Ins i u e o Technology, S ockholm, Sweden, 2005. [26] K. Ozbay, H. Yang, B. Ba in, and S. Mudigonda, “De i a ion and alida ion o new simula ion-based su oga e sa e y measu e,” T anspo a ion Resea ch Reco d, ol. 2083, no. 1, pp. 105–113, 2008. [27] Fede al Highway Adminis a ion (Fhwa), Su oga e Sa e y Assessmen Model and Valida ion: Final Repo , Fede al Highway Adminis a ion, USA, 2008. [28] M. M. Minde houd and P. H. L. Bo y, “Ex ended ime- o- collision measu es o oad a c sa e y assessmen ,” Acciden Analysis and P e en ion, ol. 33, no. 1, pp. 89–97, 2001. [29] D. F. Coope and N. Fe guson, “T a c s udies a -junc ions - a con ic simula ion model,” T a c Enginee ing and Con ol, ol. 17, pp. 306–309, 1976. [30] F. Cun o, Assessing Sa e y Pe o mance o T anspo a ion Sys ems Using Mic oscopic Simula ion, Wa e loo, On a io, Canada, 2008. [31] C. Oh, S. Pa k, and S. G. Ri chie, “A me hod o iden i ying ea -end collision isks using induc i e loop de ec o s,” Ac- ciden Analysis and P e en ion, ol. 38, no. 2, pp. 295–301, 2006. [32] A. Lau eshyn, T. De Ceunynck, C. Ka lsson, A. S ensson, and S. Daniels, “In sea ch o he se e i y dimension o a c e en s: ex ended Del a-V as a a c con ic indica o ,” Ac- ciden Analysis and P e en ion, ol. 98, pp. 46–56, 2017. [33] A. Lau eshyn, N. Saunie , and A. Fyh i, “C oss-compa ison o h ee su oga e sa e y me hods o diagnose cyclis sa e y p oblems a in e sec ions in No way,” Acciden Analysis and P e en ion, ol. 105, pp. 11–20, 2017. [34] W. K. M. Alhajyaseen, “Te in eg a ion o con ic p obabili y and se e i y o he sa e y assessmen o in e sec ions,” A abian Jou nal o Science and Enginee ing, ol. 40, no. 2, pp. 421–430, 2015. [35] C. Wang, Y. Xie, H. Huang, and P. Liu, “A e iew o su oga e sa e y measu es and hei applica ions in connec ed and au oma ed ehicles sa e y modeling,” Acciden Analysis and P e en ion, ol. 157, A icle ID 106157, 2021. [36] W. L. Ca lson, “C ash inju y p edic ion model,” Acciden Analysis and P e en ion, ol. 11, no. 2, pp. 137–153, 1979. [37] L. E ans, “D i e inju y and a ali y isk in wo-ca c ashes e sus mass a io in e ed using New onian mechanics,” Acciden Analysis and P e en ion, ol. 26, no. 5, pp. 609–616, 1994. [38] R. R. Souley e e and J. L. Hochs ein, De elopmen o a Con ic Analysis Me hodology Using SSAM, 2012. [39] R. an de Ho s and R. J. K aay, Te Du ch Con ic Technique −DOCTOR, ICTCT Wo kshop, Budapes , 1986. [40] P. G˚ a de , “Kon ik s udie i lands ¨ agsko sninga (in Swed- ish) Con ic s udies in u al in e sec ions,” Bulle in, ol. 42, Lund, Ins i u ionen ¨ o T a k eknik, 1982. [41] L. Shbeeb, De elopmen o a T a c Con ic s Technique o Di e en En i onmen s - a Compa a i e S udy o Pedes ian Con ic s in Sweden and Jo dan, Doc o al hesis,Uni e si y o Lund, Lund Ins i u e o Technology, Depa men o Tech- nology and Socie y, T a c Enginee ing, 2000. [42] A. S ensson, “A me hod o analysing he a c p ocess in a sa e y pe spec i e,” Bulle in, Lund Ins i u e o Technology, Lund Uni e si y, ol. 166, 2010 [43] Y. El-Hansali, S. Fa ag, A. Yasa , E. Shakshuki, and K. Al- Ab i, “Using su oga e measu es o e alua e he sa e y o au onomous ehicles,” P ocedia Compu e Science, ol. 191, pp. 151–159, 2021. [44] M. S. Rahman, M. Abdel-A y, J. Lee, and M. H. Rahman, “Sa e y bene s o a e ials’ c ash isk unde connec ed and au oma ed ehicles,” T anspo a ion Resea ch Pa C: Eme ging Technologies, ol. 100, pp. 354–371, 2019. [45] J. Zhang, K. Wu, M. Cheng, M. Yang, Y. Cheng, and S. Li, “Sa e y e alua ion o connec ed and au onomous ehicles’ exclusi e lanes conside ing pene a e a ios and impac o ucks using su oga e sa e y measu es,” Jou nal o Ad anced T anspo a ion, ol. 2020, A icle ID 5847814, 16 pages, 2020. [46] A. Sinha, V. Vu, S. Chand, K. Wijaya a na, and V. Dixi , “A c ash inju y model in ol ing au onomous ehicle: in- es iga ing o c ash and disengagemen epo s,” Sus ain- abili y, ol. 13, no. 14, p. 7938, 2021. 20 Jou nal o Ad anced T anspo a ion [47] H. Chen, H. Chen, Z. Liu, X. Sun, and R. Zhou, “Analysis o ac o s a ec ing he se e i y o au oma ed ehicle c ashes using XGBoos model combining POI da a,” Jou nal o Ad- anced T anspo a ion, ol. 2020, pp. 1–12, 2020. [48] Y. Li, Z. Li, H. Wang, W. Wang, and L. Xing, “E alua ing he sa e y impac o adap i e c uise con ol in a c oscilla ions on eeways,” Acciden Analysis and P e en ion, ol. 104, pp. 137–145, 2017. [49] Z. Zhong, J. Lee, and L. Zhao, “T a c ow cha ac e is ics and lane use s a egies o connec ed and au oma ed ehicles in mixed a c condi ions,” Jou nal o Ad anced T anspo a ion, 2021. [50] Y. Li, Y. Tu, Q. Fan, C. Dong, and W. Wang, “In uence o cybe -a acks on longi udinal sa e y o connec ed and au o- ma ed ehicles,” Acciden Analysis and P e en ion, ol. 121, pp. 148–156, 2018. [51] M. S. Rahman and M. Abdel-A y, “Longi udinal sa e y e alua ion o connec ed ehicles’ pla ooning on exp essways,” Acciden Analysis and P e en ion, ol. 117, pp. 381–391, 2018. [52] Y. Tu, W. Wang, Y. Li, C. C. Xu, T. Xu, and X. Li, “Longi- udinal sa e y impac s o coope a i e adap i e c uise con ol ehicle’s deg ada ion,” Jou nal o Sa e y Resea ch, ol. 69, pp. 177–192, 2019. [53] A. D. Tibljaˇ s and T. Giu , “In oduc ion o au onomous ehicles: oundabou s design and sa e y pe o mance e alu- a ion,” Sus ainabili y, ol. 10, no. 1060, pp. 1–14, 2018. [54] W. Weije ma s, A. Hula, A. Chaudh y e al., “LEVITATE: oad sa e y impac s o connec ed and au oma ed ehicles,” 2021, h ps://le i a e-p ojec .eu/downloads/. [55] K. Wunde lich, M. Vasude an, and P. Wang, T a c Analysis oolbox olume III: Guidelines o applying a c mic o- simula ion modeling so wa e 2019 upda e o he 2004 e sion, 2019. [56] Roads and Ma i ime Se ices, T a c Modelling Guidelines, Roads and Ma i ime Se ices, Sydney, 2013. [57] P. G. Gipps, “A beha iou al ca - ollowing model o compu e simula ion,” T anspo a ion Resea ch Pa B: Me hodological, ol. 15, no. 2, pp. 105–111, 1981. [58] P. G. Gipps, “A model o he s uc u e o lane-changing decisions,” T anspo a ion Resea ch Pa B: Me hodological, ol. 20, no. 5, pp. 403–414, 1986. [59] A kins, Resea ch on he Impac s o Connec ed and Au ono- mous Vehicles (CAVs) on T a c Flow Summa y Repo De- pa men o T anspo , 2016. [60] S. Das and A. K. Mau ya, “De ning ime- o-collision h esholds by he ype o lead ehicle in non-lane-based a c en i onmen s,” IEEE T ansac ions on In elligen T anspo a ion Sys ems, ol. 21, no. 12, pp. 4972–4982, 2020. [61] E. Papazikou, M. Zach, H. C. Boghani e al., “De ailed lis o sub-use cases, applicable o ecas ing me hodologies and necessa y ou pu a iables,” Deli e able D4.4 o he H2020 P ojec LEVITATE, 2020. [62] D. Ge man and L. Head, “Su oga e sa e y measu es om a c simula ion models,” Repo No. FHWA-RD-03-050, 2003. [63] Y. Guo, M. Essa, T. Sayed, M. M. Haque, and S. Washing on, “A compa ison be ween simula ed and eld-measu ed con- ic s o sa e y assessmen o signalized in e sec ions in Aus alia,” T anspo a ion Resea ch Pa C: Eme ging Tech- nologies, ol. 101, pp. 96–110, 2019. [64] Y. Guo, T. Sayed, L. Zheng, and M. Essa, “An ex eme alue heo y based app oach o calib a ion o mic osimula ion models o sa e y analysis,” Simula ion Modelling P ac ice and Teo y, ol. 106, A icle ID 102172, 2021. [65] G. Mesionis, M. B acks one, and N. G a e , “Mic oscopic modeling o he e ec s o au onomous ehicles on mo o way pe o mance,” T anspo a ion Resea ch Reco d, ol. 2674, no. 11, pp. 697–707, 2020. [66] L. Ye and T. Yamamo o, “E alua ing he impac o connec ed and au onomous ehicles on a c sa e y,” Physica A: S a- is ical Mechanics and I s Applica ions, ol. 526, A icle ID 121009, 2019. [67] D. S anek, E. Huang, R. Milam, and A. Wang, Measu ing Au onomous Vehicle Impac s on Conges ed Ne wo ks Using Simula ion, 2017. [68] J. Ka jan o, N. Yuso , J. Te ken, and F. Delb essine, Simula ing Au onomous D i ing S yles: Accele a ions o T ee Road P o les, ol. 1–16, 2017. [69] A. K. Bakhshi and M. M. Ahmed, Accoun ing o Human- Rela ed Unobse ed He e ogenei y in he Sa e y Pe o mance o Connec ed Vehicles: An Inco po a ion o Bayesian Hie a chical Nega i e Binomial in o Simula ed Wo k Zone Wa ning Ap- plica ion, IATSS Resea ch, 2021. Jou nal o Ad anced T anspo a ion 21