SAFETY ASSESSMENT OF COOPERATIVE VEHICLE INFRASTRUCTURE SYSTEM-BASED URBAN TRAFFIC CONTROL Joyoung Lee, Ph.D. Research Associate, Departent of Civil and Environental Engineering, University of Virginia, Charlottesville, VA, USA, eail: jl9hc@virginia.edu Byungkyu (Brian) Par Ph.D. Associate Professor, Departent of Civil and Environental Engineering, University of Virginia, Charlottesville, VA, USA, eail: bpark@virginia.edu Kristin Malakorn, EIT Transportation Engineer, Vanasse Hangen Brustlin (VHB) Inc., Watertown, MA, USA, eail: kalakorn@virginia.edu Jaehyun (Jason) So Graduate Research Assistant, Departent of Civil and Environental Engineering University of Virginia, Charlottesville, VA, USA, eail: js9pb@virginia.edu Subitted to the 3 rd International Conference on Road Safety and Siulation, Septeber 14-16, 011, Indianapolis, USA ABSTRACT Cooperative vehicle infrastructure syste (CVIS) or connected vehicles, forerly known as IntelliDrive, has eerged and is expected to provide unprecedented iproveents in obility and safety. A recent study developed a CVIS-based urban traffic control syste that does not require a stop-and-go style traffic signal. A siulation-based study on the CVIS-based control found significant iproveents in obility, energy (i.e., reductions in fuel consuption), and greenhouse gas eissions. However, safety was not considered in that study. This paper investigated safety aspects of the CVIS-based urban traffic control syste by applying surrogate safety assessent odel (SSAM). The purpose of this study was to assess whether safety has been ipacted and, if so, how uch safety has been coproised due to reduced tie headways between vehicles and higher acceleration/deceleration rates under the CVIS-based urban traffic control syste. A siulation-based case study was perfored on a hypothetical arterial consisted of four intersections with four traffic congestion cases covering high to low volue conditions. As a 1
result, the CVIS control, when copared to the coordinated actuated control, reduced the averages of tie to collision (TTC) and post encroachent tie (PET) by 0.69 and 1.94 seconds, respectively. Note that shorter TTC and PET indicate ore dangerous situation. However, it was discovered that the nuber of rear-end conflict events decreased by 58% under the CVIS-based control, indicating ore safer driving conditions would be achieved with the CVIS-based control syste. Keywords: Cooperative Vehicle Infrastructure Syste (CVIS), Connected Vehicle, Safety Surrogate Assessent Model (SSAM), Safety, Siulation INTRODUCTION Transportation safety is a ajor concern in the United States that ipacts the econoy, personal well being, and overall quality of life. Approxiately 40,000 people die each year in the United States due to autoobile accidents (NHTSA, 010). Transportation engineers spend a great deal of tie and energy trying to find solutions to safety issues. Connected Vehicle (CV) research is a USDOT progra that ais at iproving transportation safety, obility, and environental ipacts (Connected Vehicle Research, 011). The intension of CV applications is that they will iprove transportation safety. In the past, taking soe tasks away fro the driver was not necessarily perceived as a easure that would foster safety. A cooperative vehicle-infrastructure syste (CVIS) based on the CV environent for anaging intersection controls, naely CVIS control for short, in which assued autoation of vehicular accelerations/decelerations (i.e., taking soe driving tasks away), has shown significant obility and environental ipacts iproveents over the coordinated actuated signal control (Lee, 010; Malakorn, 010). Thus, there is a need to investigate safety ipacts due to a CVIS control syste. The purpose of this study was to assess whether safety has been ipacted and, if so, how uch safety has been coproised due to reduced tie headways between vehicles and higher acceleration/deceleration rates under the CVIS control syste. The reainder of this paper is organized into four sections. The literature review section suarizes the previous research on the CVIS control algorith (i.e., the CV-based urban traffic control syste) and the state-of-the-art safety surrogate assessent odel. The ethodology section provides little ore details about the CVIS control algorith and addresses the workflow of how to assess the safety ipacts of the CVIS control. The case study section presents the design of siulation experients and the siulation results of the obility and safety ipacts of the CVIS and the actuated controls. Finally, conclusions and recoendations regarding the safety assessents of the CVIS control are provided at the concluding rearks section.
LITERATURE REVIEW Lee (010) proposed a CVIS control algorith and assessed its potential benefits. The core of the algorith exained the predictive trajectories of vehicles that would be at risk for coing into conflict with one another at an intersection area. When ultiple vehicles on conflicting approaches are projected to cross the intersection area at the sae tie, with a safe gap constraint between two consecutive vehicles the algorith optiizes their trajectories in search of optial speeds and accelerations that will prevent the occurrence of trajectory overlaps. Coprehensive icroscopic traffic siulation-based experients covering various traffic congestion conditions were perfored on a hypothetical isolated intersection. Statistically significant benefits were observed: for obility 99% and 33% of iproveents on stop delays and travel tie, respectively, were estiated and about 34% of both CO eission reductions and fuel savings were also reported. Malakorn (010) extended the CVIS control algorith to an arterial, and exained its potential benefits of obility and sustainability. As a result, the iproveents to environental ipact were about 63% and 60% for carbon dioxide eissions and fuel consuption, respectively (Malakorn, 010). The obility benefits were greater due to increased capacity at the corridor. For exaple, total delay tie was iproved by 86% to 100% for the volue cases evaluated (Malakorn, 010). It is generally understood that the transportation safety is challenging to evaluate, especially where no post crash data are available. The ost straightforward way to evaluate safety would be through archived crash data. However obtaining such archived data would require treendous efforts or practically ipossible under the CVIS control. To overcoe such a challenge, Gettan and Head (003) proposed a siulation-based safety surrogate assessent odel (SSAM). The perforance of the SSAM progra was well validated through siulation-based case studies covering various intersection geoetries, traffic conditions, operational strategies, and deonstrated rearkable perforances (Gettan et al., 008). METHODOLOGY In order to assess the safety ipacts of the CVIS control syste, this paper incorporated two software progras: i) a CVIS siulation test-bed utilizing VISSIM, a coercial icroscopic traffic siulator (PTV, 011) and ii) the SSAM software for evaluating the safety ipacts of the CVIS controls. In this section, both CVIS control algorith and SSAM are briefly presented. CVIS Control Algorith Assuing two vehicles approaching fro conflicting streets to an intersection, Figure 1 illustrates the vehicles anticipated trajectories that would likely result in a crash in the intersection area. The length of the trajectory overlap, denoted as l, is given by Equations 1 and. With vehicles driving inforation such as locations, speeds, and acceleration/deceleration rates obtaining through connected vehicles environent, the CVIS control syste projects 3
individual vehicles traveling trajectories and identifies whether potential crashes would occur at the intersection or not by exaining the overlaps of trajectories. In case trajectory overlaps are detected as shown in Figure 1, the CVIS control syste seeks optial trajectories to avoid the crash. Figure Illustration of Vehicle Trajectory Overlap at an intersection (Lee, 010) if a n 0, otherwise, q ' l 1 x ( w) dw (1) p l ( q p) ( lw x( p)) () where: l : Length of trajectory overlap : Predicted reaining distance to the intersection stop bar of vehicle n at tie t x n (t) ( x 0) 0.5a t v t n x n (0) ( ) n n : Current (t=0) reaining distance to the intersection stop bar of vehicle n at tie t p: Arrival tie at the beginning of intersection q: Arrival tie at the end of intersection lw: Intersection length in eters a n : Acceleration or Deceleration rate of vehicle n v n : Current speed of vehicle n t : tie 4
To seek the optial trajectories, the CVIS control utilizes non-linear constraint optiization techniques, which are designed to solve an optiization proble given the Equations 3 through 6. With optial acceleration/deceleration rate for each vehicle approaching to the intersection, the overlapping trajectory for each vehicle is adjusted to safely cross the intersection without stops or the need for a traffic signal. In case no feasible solutions are found, however, the CVIS control syste runs in a recovery ode, a traffic signal-based special period designed to be quickly returned to noral optiization-based control ode (Lee, 010). It is noted that the recovery ode is not discussed in this paper. MinTL P L N i ik j jl P i1 k1 1 j1 l1 n1 L N q p ' 1 x ( w) dw (3) ik such that, vik uin v ik aik ax a,, in i, (4) xik (0) xik (0) xik ( t) uax v ik aik in aax, i, xik (0) xik ( t) (5) S.5( a a R ( a h v v ) R S 0 i, k and 1,,... N (6) 1 0 i, i1, i, i, i1, i, k Where: P : Total phase nubers i, j : Phase nuber indices (1 if phases are conflicted, 0 otherwise) l: Lane identifier, n: Vehicle identifier L i, L j : Total nuber of lanes of phase i,j, respectively N ik, N jl : Total nuber vehicles on lane k and l of phase i and j respectively. p: Arrival tie at the beginning of intersection (= ax t i, ( o), t j, l, n( o) ) q: Arrival tie at the end of intersection (= in t i, ( d), t j, l, n( d) ) t i, (o), t j,l,n (o) : Arrival ties at the beginning of the intersection of vehicle (n) on lane k(l) in phase i(j) t i, (d), tl j,l,n (d) : Arrival ties at the end of the intersection of vehicle (n) on lane k(l) in phase i(j) S.5a h v h x (0) x (0) R a 0 i, i, i, i1, 1 i1, vi 1, vi 1, ai, xi1, (0) Safety Surrogate Assessent Model (SSAM) The SSAM progra identifies conflicts by analyzing each vehicle s interaction found in the trajectory records fro the icroscopic traffic siulation software. Given the trajectory record of each individual vehicle obtained fro icroscopic traffic siulation odels, the SSAM progra evaluates i) surrogate easures such as tie to collision (TTC), post encroachent tie 5
(PET), axiu speeds, and axiu decelerations to deterine crash events, and ii) conflict angles to deterine crash types such as rear-end, lane changing, and crossing crashes, as depicted in Figure. Figure Crash type identification in the SSAM progra (Gettan et al., 008) As illustrated in the conceptual workflow in Figure 3, SSAM is a post processing-based safety surrogate easure estiator. For exaple, once the siulations of the CVIS control copleted, the resulting trajectory data of each individual vehicle run through the SSAM software to deterine what safety issues ay exist. If both the TTC and the PET of a pair of vehicles in the trajectory data are found to be within their threshold values and their conflict angle is less than 30-degree, the SSAM progra identifies it as a rear-end crash event. Note that the crash event does not indicate an actual crash but the likelihood of potential crashes. Thus, the use of proper TTC and PET threshold values are crucial as different thresholds would result in different crash estiations. This paper eployed 1.5 seconds and 5 seconds of TTC and PET threshold values, respectively, based on the previous research (Sayed et al., 1994; Hyden, 1987). Figure 3 Conceptual Workflow (Gettan et al., 008) 6
CASE STUDY Experients set-up A hypothetical arterial network consisted of four single-lane intersections was created by using the VISSIM progra. The test network has a.7-kiloeter long ajor street expanding eastbound and westbound. Each intersection along the corridor is spaced at about 400 eters with each other. Figure 4 shows the test network odeled in the VISSIM progra. Figure 4 A hypothetical test network in VISSIM To exaine the safety ipacts under varying traffic congestion conditions, four different volue cases were developed and tested. Table 1 shows specific details about each volue case. Five repetitions of each volue case were siulated. Each repetition was 35 siulation-inutes long, including a 5-inute waring-up period. To copare the perforance of the CVIS control, the coordinated actuated control syste was used for each volue scenario with the sae nuber of replications and siulation period. The tiing plans for the coordinated actuated intersection controls were developed by the Synchro progra, in which was used as a base case for coparison purpose (Husch and Albec 004). Note that the volue to capacity ratio (v/c) for each volue scenario presented in Table 1 was estiated with the optial tiing plans obtained fro the Synchro progra, assuing the corridor is operated by a coordinated actuated control. Table 1 Volue Conditions tested Scenario Major Approach Minor Approach v/c Ratio Volue, vph Volue, vph 1 900 500 0.97 800 500 0.9 3 600 500 0.88 4 400 400 0.71 7
In assessing safety perforances, this paper eployed i) a tie to collision (TTC) and ii) a post encroachent tie (PET) as safety surrogate easures. TTC is a easure of seconds that vehicles would have to continue behaving as they are to collide with one another. PET is the tie required for the lead vehicle to leave a position and the following vehicle to occupy that position. Obviously, shorter PETs are ore dangerous. Recall that the axiu threshold value of TTC for identifying a crash was set at 1.5 seconds and a PET of 5 seconds was used as a axiu threshold value in this paper. It was assued that all vehicles in the siulations of CVIS control can counicate with one another and with the infrastructure through the connected vehicle environent. In addition, all vehicles were assued to be equipped with the necessary cooperative cruise control device to allow the vehicle to anipulate its own speed, acceleration, and deceleration. Results Table suarizes the obility benefits of the CVIS control algorith applied to the hypothetical arterial network. Copared to the actuated control (AC) syste, the CVIS control draatically reduced the total delay ties between 9% and 100% depending on volue cases. Note that the total delay ties are defined as a su of the standstill ties due to congestion at the intersection. Taking into consideration that the CVIS control algorith is designed to keep vehicles crossing the intersection without any risks of crashes, such huge savings obtained fro the total delays confirs the proising benefits of the CVIS control algorith. 1 3 4 Table Mobility iproveents Case Total Delay Tie (Hour) AC 19.3 CVIS 1.5 Iproveent (%) 9 p-value 0.000 AC 18. CVIS 0.6 Iproveent (%) 97 p-value 0.000 AC 14.4 CVIS 0.0 Iproveent (%) 100 p-value 0.000 AC 9.5 CVIS 0.0 Iproveent (%) 100 p-value 0.000 While such proising benefits were realized in the obility, the CVIS control appeared to decrease the intersection safety. As suarized in Table 3, the average TTC of the CVIS control was less than that of actuated control for each volue case. Siilarly, the PET values of CVIS 8
control were all less than the actuated controls. It is noted that saller TTC and PET indicate larger dangerous situation. However, the nuber of rear-end crash events for each volue case was significantly reduced as shown in Table 3. It is also noted that the nuber of rear-end crash events eans the likelihood of potential crashes and it increases when both TTC and PET are less than the axiu thresholds, which are 1.5 and 5 seconds, respectively. It is worth noting that approxiately 0,000 traffic conflict events account for an actual crash (Gettan et al., 008). Thus, while the CVIS control incurred ore dangerous situations, its frequencies were rearkably reduced, resulting in safer conditions. This is likely because the CVIS control is designed to anipulate the aneuver of each individual vehicle to guarantee its safety condition even when crossing the intersection at high speeds. Note that the crossing events are not considered for the evaluation as they appeared insignificant in ters of the nuber of observations for both the actuated and the CVIS controls, accounting for approxiately % of the rear-end crash events. Case 1 3 4 Overall Table 3 CVIS control safety ipacts Mean TTC (Sec) Mean PET (Sec) Nuber of Rear-End Conflict Events AC 1.3 3.08 796 CVIS 0.76 1.79 536 Difference -0.47 (-38%) -1.9 (-4%) -60 (-33%) t-value 44.51 41.94 1.88 AC 1.4 3.07 679 CVIS 0.70 1.51 68 Difference (%) -0.54 (-44%) -1.56 (-51%) -411 (-61%) t-value 30.81 36. 7.36 AC 1.6 3.07 49 CVIS 0.53 0.99 109 Difference (%) -0.73 (-58%) -.08 (-68%) -383 (-78%) t-value 6.6 34.85 6.6 AC 1.8 3.13 87 CVIS 0.8 0.30 37 Difference (%) -1.0 (-78%) -.83 (-90%) -50 (-87%) t-value 4.95 56.6 18.50 AC 1.5 3.09 564 CVIS 0.57 1.15 38 Difference (%) -0.69 (-55%) -1.94 (-63%) -36 (-58%) *Tested at 95% of confidence level CONCLUDING REMARKS This paper exained the obility and safety ipacts of the CVIS control syste under the Connected Vehicles environent. The CVIS control draatically iproved both the obility and the environental perforances of the urban corridor: between 9% and 100% of delay tie reductions were estiated for the volue cases tested. Taking into consideration that these iproveents were obtained fro the adjustents of the driving aneuver of each individual vehicle to ensure high speed crossing at intersection, the 9
CVIS control would likely to result in ore dangerous situations in ters of the safety aspect as indicated by lower TTC and PET values in Table 3. However, the CVIS control reduced the frequency of such dangerous situations, resulting in 33% to 87% of rear-end crash reductions. Such huge safety iproveents obviously cae fro the anaged oveents of individual vehicles ensuring the safety gap between vehicles provided by the CVIS control. In this paper, perfect wireless counication conditions for the Connected Vehicles environent were assued such that there were no counication packet drops and no counication delays, which would not be true in real world. Given that the safety of Connected Vehicles applications would be affected by the quality of wireless counications, the aspect of counication ust be incorporated in future research for ore realistic safety assessents. Finally, while the case study was deonstrated on a corridor with a single through lane for each approach, the CVIS control can handle a generic intersection with ulti-lanes and left-turn bays as shown in the objective function in Equation (3). Although this paper did not perfor additional case studies for ulti-lanes and coordinated intersections, the ipleentations for such case studies would be feasible as future research. REFERENCES Connected Vehicle Research. Available fro: http://www.its.dot.gov/connected_vehicle /connected_vehicle.ht, Last Access: 011.4.9 Gettan, D. and Head, L. (003), Surrogate Measures of Safety fro Traffic Siulation Models. Transportation Research Record: Journal of the Transportation Research Board. 1840: p. 104-115. Gettan, D. Pu, L., Sayed, T., and Shelb, S. (008), Surrogate safety assessent odel and validation: final report, Report No. FHWA-HRT-08-051. Federal Highway Adinistration (FHWA): Washington, D.C. Hyden, C. (1987), The developent of a ethod for traffic safety evaluation: The Swedish Conflicts Technique. Departent of Traffic Planning and Engineering, Lund University: Sweden, Husch, D., and Albeck., J (004), SYNCHRO 6 User Guide. Trafficware. Lee, J. (010), Assessing the potential benefits of IntelliDrive-based intersection control algoriths. Ph.D. dissertation, University of Virginia: Charlottesville, VA. Malakorn, K. (010), Evaluation of IntelliDrive based vehicular infrastructure control ipacts using traffic siulation and life cycle assessent. M.S thesis, University of Virginia: Charlottesville, VA. 10
NHTSA (010), TRAFFIC SAFETY FACTS 009, National Highway Traffic Safety Adinistration (NHTSA), Washington D.C. PTV Aerica (009). VISSIM Traffic Siulation. Version 5.10. Sayed, T., Brown, G., and Navis, F. (1994), Siulation of Traffic Conflicts at Unsignalized Intersections with TSC-Si. Accident Analysis and Prevention, Vol. 6, No. 5, pp. 593 607, 11