gettman and head 2003

7/31/2019 Gettman and Head 2003

1/31

Surrogate Safety Measures from Traffic Simulation Models

Douglas Gettman, Ph.DSiemens Gardner Transportation Systems6375 E. Tanque Verde #170Tucson, AZ 85715(520) 290-8006 x121(520) 290-8178 [email protected]

Larry Head, Ph.D.Siemens Gardner Transportation Systems6375 E. Tanque Verde #170Tucson, AZ [email protected]

TRB 2003 Annual Meetin g CD-ROM Paper revised from original submittal.


2/31

Gettman and Head Surrogate Safety Measures from Traffic Simulation Models

Page 2

TABLE OF CONTENTS

ABSTRACT ........................................................................................ 4

INTRODUCTION................................................................................ 5

LITERATURE REVIEW........................................................................ 5

C ONFLICT S EVERITY ...............................................................................6 L ITERATURE S UMMARY .............................................................................8

TRAFFIC SIMULATION MODEL OVERVIEW ........................................ 8

MICROSCOPIC SIMULATION MODEL COMPARISONS ...................... 11

SURROGATE MEASURES DEFINITIONS AND COLLECTIONMETHODOLOGY............................................................................... 12

CONFLICT EVENT DESCRIPTIONS ................................................... 12

C ROSSING FLOWS CONFLICT POINT EVENTS .............................................. 13 M ERGING CROSSING FLOWS CONFLICT LINE EVENTS .................................... 13 F OLLOWING FLOWS REAR-END CONFLICT LINE EVENTS ...............................13 A DJACENT FLOWS LANE - CHANGING REAR-END CONFLICT LINE EVENTS ............. 13 P OTENTIAL COLLISIONS NOT REPRESENTED IN THE SURROGATE MEASURES ...............13 C ONFLICT P OINT ................................................................................. 14 C ONFLICT L INE ................................................................................... 15

R EAR - END C ONFLICT L INE ......................................................................16 S UMMARY ......................................................................................... 17

SURROGATE MEASURES DEFINITIONS ........................................... 17

S EVERITY OF CONFLICT AND SEVERITY OF RESULTING COLLISION ..........................17 S URROGATE MEASURES FOR CONFLICT POINTS ............................................... 18

Time To Collision...........................................................................18 Post-encroachment time ................................................................ 18 MaxS........................................................................................... 18 DeltaS ......................................................................................... 18

Initial Deceleration Rate.................................................................18 S URROGATE M EASURES FOR C ONFLICT L INES AND R EAR - END C ONFLICT L INES .........19

SUMMARY ....................................................................................... 19

ACKNOWLEDGEMENT...................................................................... 20

REFERENCES................................................................................... 21

TRB 2003 Annual Meeting CD-ROM Paper revised from original submittal.


3/31


Page 3

LIST OF FIGURES AND TABLES ....................................................... 26



4/31


Page 4

AbstractSafety is emerging as an area of increased attention and awareness within transportationengineering. Historically, safety has been difficult to assess for new and innovative traffictreatments, primarily due to the lack of good predictive models of accident potential and lack of consensus on what constitutes a safe or unsafe facility. This FHWA-sponsored research projectis intended to investigate the potential for deriving surrogate measures of safety from existingtraffic simulation models. These surrogate measures could then be used to support trafficengineering alternatives evaluation with respect to safety without costly accident studies or construction. In addition, the measures available from simulation models are much moredetailed that the subjective measures based on human observers. The first task was to review theliterature in surrogate safety analysis and focus the research on surrogate measures for intersection safety. The second task was to review the capabilities of existing simulation modelsthat are commercially available for producing surrogate measures. The third task was to specifysome functional requirements for a software tool that would analyze surrogate measures

produced by the simulation model(s). The final task of the research was to describe thealgorithmic approach required for obtaining surrogate measures from simulation models.

Each surrogate measure is collected based on the occurrence of a conflict event an interaction between two vehicles in which one vehicle must take evasive action to avoid a collision.Because of the limited modeling of the environment and some driver behavior assumptions insimulation models, surrogates for single-vehicle safety are not specified (e.g. risk of run-off-roadat a high-speed curve). The surrogates that are proposed as the best measures are the time tocollision, post encroachment time, deceleration rate, maximum speed, and speed differential.Time to collision, post encroachment time, and deceleration rate can be used to measure theseverity of the conflict. Maximum speed and the speed differential can be used to measure theseverity of the potential collision (along with information about the mass of the vehicles involvedto assess momentum). Definitions of the possible conflict events and algorithms for calculatingthe surrogate measures for conflict points and lines are presented. Some modifications to theinterfaces to commercial simulation models will be required to obtain the necessary data on eachconflict event. After executing the simulation model for a number of iterations, a post-

processing tool would be used to compute the statistics of the various measures and performcomparisons between design alternatives.

Word Count: 6303, 5 Figures (6303 + 5*250 = 7553 total)



5/31


Page 5

IntroductionSafety is emerging as an area of increased attention and awareness within transportationengineering. Historically, safety has been difficult to assess for new and innovative traffictreatments, primarily due to the lack of good predictive models of accident potential and lack of consensus on what constitutes a safe or unsafe facility. This FHWA research project is intendedto investigate the potential for deriving surrogate measures of safety from existing trafficsimulation models. These surrogate measures could then be used to support evaluation of trafficengineering alternatives with respect to safety without costly accident studies or construction. Inaddition, the measures available from simulation models are much more detailed that thesubjective measures based on human observers (e.g. traffic conflicts technique) and can cover more operational scenarios (i.e. times of day, days of week, months of year, special events andincidents), than the limited time that observers can be employed in the field to collect data.

The first task of the research was to review the literature in surrogate safety analysis - focusingthe research on surrogate measures for intersection safety. The second task was to review thecapabilities of existing simulation models that are commercially available for producingsurrogate measures. The third task was to specify some functional requirements for a softwaretool that would analyze surrogate measures produced by the simulation model(s) producingdistributions, graphs, charts, and other analyses. The final task of the research was to describethe algorithmic approach required for obtaining surrogate measures from simulation models.This paper summarizes the results of each task. More detail is available in the forthcomingFHWA report.

Literature Review

For the purpose of this study, safety of a traffic facility is defined as follows:

The expected number of accidents, by type, expected to occur on the entity in a certain period, per unit of time.

An accident is defined as an unintended collision between two or more motor vehicles. Notethat single-vehicle accidents are excluded from this definition. In addition, the bulk of accidentresearch and the available literature on surrogate measures neglects accidents involving morethan two vehicles given that those events are much less prevalent than accidents involving a pair of vehicles.

To estimate the safety of various traffic facilities, including facilities that have not yet been built,research in safety has focused on the establishment of safety performance functions that relatethe number of accidents or accident rate to a number of operational (e.g. AADT, averagespeed) and non-operational independent variables via a (typically complex) regressionequation(s) including AADT, occupancy, V/C ratios, products of crossing volumes, etc.(1,2,3,4,5,6,7,8,9,10,11,12) . Calibration is then required to choose the equation parameters for the best statistical fit to the available data (9,13,14) . Research has also focused on Bayesian



6/31


Page 6

methods and advanced statistical techniques (e.g. CART) for revising accident estimates basedon observations (15,16,17,18,19) and various methods for combining accident rates and other measures into safety level of service measures or common indices based on one type of accident (e.g. property-damage-only) (20,21) .

Despite the large body of safety modeling research, absolute numbers of accident and accidentrates are still difficult to predict accurately. This has lead to increased interest in obtaining

surrogate measures that reflect the safety of a facility or at least the increased probability of higher-than-average accident rates for a facility. The most prevalent literature in surrogatemeasures is related to the traffic conflicts technique (14,22,23,24,25) .

A conflict is defined as:

An observable situation in which two or more road users approach each other in time and space for such an extent that there is risk of collision if their movements remain unchanged (26).

This method has a long history of development including research on recommended datacollection methods (22,27) , definitions of various types of conflicts (24,28) , severity measures(29,30) , how conflict measures are related to accident counts and specific accident types (20) andstandards for data collection (31,32) . There is, however, some debate regarding the connection

between conflict measures and accident predictions (28) . This includes the fact that thesubjectivity of field observers induces additional uncertainty into the collection of accurate dataon conflicts. Conflict studies are, however, still continuing to be used to rank locations withrespect to safety for construction upgrades (33,34,35) .

Conflict Severity

The primary conflict severity measure that has been proposed is the time to collision (21,29,36) .Some researchers have indicated that the time to collision is the surrogate measure of safetywhile others refute that lower TTC indicates higher severity of accidents, primarily becausespeed is not included in the measure (37,38) . That is to say that lower TTC certainly indicates ahigher probability of collision, but cannot be directly linked to the severity of the collision.Some research indicates deceleration rate as the primary indicator or severity instead of TTC(39,40,41) .

Other proposed measures defining and characterizing a conflict are presented (27,42) . (27) specifies these measures primarily for left-turn conflict events and rank the above measures inoverall desirability in the following way:

1. Gap Time2. Post-encroachment time3. Deceleration rate4. Encroachment time5. Initially attempted post-encroachment time6. Proportion of stopping distance



7/31


Page 7

This ranking takes into consideration the relation to accident history, relations among the other measurements, consistency over time, relation to braking application, ease of measurement, andapplication to other conflict types. Time to Collision (TTC) was not originally included in the

list by Allen, et al. (27). In the ranking above, we consider TTC to be ranked in the compositescale used above at least as high as gap time.

Other surrogate safety measures proposed in the literature for intersections include fairlystandard measures of effectiveness: delay, travel time, approach speed, percent stops, queuelength, stop-bar encroachments, red violations, %left-turns, spot speed, speed distribution, anddeceleration distribution (43,44,45) . No attempt was made to relate these measuresquantitatively to accident rates, but rather to assert such rules-of-thumb as more stop-bar encroachments indicates higher probability of accidents, longer queues indicates higher

probability of accidents, and so on. A similar list of surrogates for two lane roads has also been published, although more non-operational variables appear in the list for two-lane roads (e.g.

super-elevation, curvature, distance since last curve) (8,43,44,45) .

The above statistics, as well as conflict measures, require field observer crews to collect the data.This is expensive and includes the difficulties of subjective observers. Collection of TTCmeasures and the other measures in Table 1 requires instrumented vehicles. Some additionalsurrogate measures proposed in the literature include:

Deceleration rate distributions Required braking power distributions Distribution of merge points (freeway travel) Merge area encroachments (freeway on-ramp merging) Gap acceptance distributions Percentage of vehicles caught in dilemma zones Speed differential between crossing movements Speed variance Red- and yellow-light violations by phase

As indicated by the above measures, microscopic simulations are generally required for generating and collecting conflict statistics and/or other surrogate measures as a substitute for field studies. These simulations include models built specifically for simulation of particular conflict types (11,21,42,46,47,48,49) and modification of multi-purpose traffic simulationmodels to include conflict statistics or other surrogates (30,41,50,51,52,53,54) . The latter category includes HUTSIM, TRANSIMS, INTRAS (now FRESIM, part of CORSIM), NETSIM,INTEGRATION, and TEXAS.

One project (SINDI) specifies including a more detailed driver-behavior model (i.e.nanoscopic simulation) into the HUTSIM microscopic simulation for representation of lapsesin driver reaction time and errors in response (51) but is still in development. The paper on



8/31


Page 8

TRANSIMS is exploratory. TRANSIMS uses macroscopic representation of vehicle movementsto simulate large-scale network (e.g. entire cities) transportation behavior and is therefore

probably not detailed enough for the level of analysis required for this project. CORSIMcurrently outputs conflict statistics by movement (left, right, through/diagonal), conflicting

movement (left, right, through/diagonal) and approach for intersections when micro-nodeanalysis is enabled, although this module is not considered viable (56). TEXAS also calculatesand outputs conflict statistics. INTEGRATION includes a TTC index for each intersectionapproach and an accident rate estimate that is linked to average speed and volume on eachapproach (54). AIMSUN has been modified to produce safety measurements of conflicts atramp merging sections (73).

Literature SummaryThere is limited quantitative research to date on surrogate measures for safety assessment. Theavailable literature has focused mainly on various aspects of traffic conflict field studies for obtaining surrogate measures. Given the technical difficulty and cost of field studies, use of

simulation models has been proposed and some previous work has developed specific models for simulating conflicts. The most notable surrogate measure of conflict severity is the time-to-collision, although other surrogates (e.g. gap time, post encroachment time) have been proposedto measure other characteristics of conflict situations. Only limited effort has been expended tomodify existing microscopic simulations for obtaining conflict or other surrogate measures or define surrogates that can be extracted from accident-free simulations that have reasonableconnectivity to safety assessment of particular facilities. The next section will discuss variousaspects of simulation models available for obtaining surrogate safety measures.

Traffic Simulation Model OverviewMicroscopic simulation models hold some promise for collecting surrogate measures of safetyfor intersections. Microscopic models typically simulate traffic systems on a vehicle-by-vehicle

basis, updating position, speed, acceleration, lane position, and other state variables on a second- by-second basis as the vehicles interact with traffic signals, signs, other vehicles and roadwaygeometrics. Some simulations allow use of even smaller time-steps for more accurate behavioralanalysis and/or use an event-driven structure for more computational efficiency. Microscopicsimulations generally also include detailed modeling of traffic signal operations, which is arequirement for derivation of surrogate safety measures. However, all simulation models weredesigned assuming that drivers behave in a safe manner, according to their particular driver

behavior characteristics (i.e. aggressiveness for gap acceptance and lane changing). As such, anyderivation of surrogate measures must account for this basic fact.

The pertinent characteristics of microscopic simulations to support this research project are:

Behavioral modeling of driver/vehicle interactions Ability to extract detailed data from the simulation (APIs, output files, open source) Ability to calibrate and select parameters of models Cost of modifications to source or outputs



9/31


Page 9

General features such as user base, longevity, stability, usability, etc.

The key elements from the scientific side are modeling of behavior and calibration and parameterization of the component models. Specific issues related to surrogate safety modeling

in these areas are detailed in the next section.

Behavioral modelingFor evaluation of surrogate safety measures, microscopic simulations must model the key driver

behaviors that produce opportunities for accidents. Those behaviors are mainly:

(a) Car-following,(b) Gap acceptance, and(c) Lane changing.

Although most, if not all, microscopic models include these behaviors, models with especially

detailed, realistic behavioral components will be more amenable for use in later phases of this project (57) .

Parameterized turning speedThe speed at which turns are made should be tunable by the user, or variable based on turningradius, number of lanes, etc. It is conceivable that the turning speed model could influencecalculation of surrogate measures.

Reaction to yellowModeling of a drivers reaction to yellow is important to measure dilemma zone performance. Itcould be important for calculation of surrogate measures if the reaction model is variable by

driver type, vehicle type, etc. Most models reviewed have reaction by driver type.

Variable Driver Reaction timeReflects the models ability to represent the delay experienced between the drivers identificationof a potential collision and the application of control measures (braking, acceleration, or lanechange) to avoid collision. In the real world, drivers reaction times vary by experience, age, etc.

Intersection box movementsFor assessment of surrogate safety measures, it is important for the simulation to modelmovement of the vehicles in the intersection with significant fidelity (Note that the initial modelchoice for this project CORSIM does not adequately model intersection box movements to

be used further without modifications) (55,56)

Variable acceleration (and deceleration) rateSimulations should include modeling of different vehicle capabilities by vehicle type.Unrealistic deceleration rates (and maximum deceleration rate distributions) may underestimatetrue statistics of surrogate measures. All models reviewed include this.



10/31


Page 10

Sight Distance LimitsModels that limit the look-ahead distance of drivers when making decisions (or model thelook-ahead distance by driver or driver-type) can more accurately model the awareness of driverson surrogate measure statistics. In addition, sight distance limits can reflect the modeling of

roadway obstructions, such as curves, crests, trees, buildings, etc. This may also apply tomodeling of in-vehicle sight restrictions, such as those that occur when following a large truck.Most of the models reviewed are lacking sophisticated sight distance limitation modeling whichis a very complex phenomena to model.

Rolling yieldAccurate modeling of yield points will be crucial for accurate collection of surrogate measures.It is hypothesized that the SSAM will be used for safety analyses of yield operations versus stopor signalized operations. Rolling yield indicates that the yield operation can occur with aslowed vehicle that does not come to a complete stop before re-entering the traffic stream.

Vehicles interact with pedestrians:Pedestrian safety is of extreme importance to traffic engineers. Simulations that model vehicleinteractions with pedestrians may have the ability to assess pedestrian safety effects of variousalternatives. Notably CORSIM does not explicitly model pedestrian movements.

Friendly mergingThe phenomenon where certain driver types slow or stop to allow vehicles to merge (more)safely, which occurs in the real world, as opposed to only modeling slowing or stopping in areactive sense. Friendly merging indicates that the following vehicle can create a gap for amerging vehicle.

Parking maneuvers:On-street parking (parallel and double) parking create conflict situations, lane-changes, etc. inthe real world and have a significant safety impact. Simulations that model on-street parkingmaneuvers are preferred.

U-turns:U-turns frequently cause conflict situations and some locations experience high enough volumesof u-turn traffic that their impact on safety should be addressed. Simulations that includemodeling of u-turns are preferred.

Variable time steps:

Simulations with tunable time step length are preferred to those with fixed time steps for evaluation of sensitivity of surrogate measures to time step size. In addition, simulations withvariable time steps have more robust behavior models. Significantly, CORSIM does not allowtunable time steps.

Time steps


11/31


Page 11

The precision of evaluating surrogate measures relies upon frequent state-variable updates. Thetime-scales of decision making for surrogate measure evaluation are on the order of fractionalseconds the simulation must allow modeling of this fidelity. The state-of-the-art simulationsall include tunable time step resolution.

Gap acceptance criteria change by delay:Many drivers in the real world change their behavior based on how long they have been waiting(i.e. they accept smaller gaps and apply larger accelerations the longer they have waited to makea particular opposed movement. Simulations that model this behavior are preferred.

Vehicle length:Safety of particular conflicting maneuvers are dependent upon the size of the vehicles involved.All of the simulations reviewed include vehicle length.

Vehicle length considered by gap logic:

Surrogate measures based on proximity of two vehicles in space and time are affectedsignificantly if the vehicles are modeled as points, rather than rectangles.

Variable headways:Different driver types maintain different headways between the car they are following basedupon their level of risk acceptance. This must be reflected in the simulation for accuraterepresentation of surrogate measures. All models include this feature to different degrees.

Variable queue discharge headway:Related to variable headways, as the queue dissipates at a traffic signal, different driver typesreact at different rates that may have an affect on surrogate measures (primarily rear-end conflict

measures).

Microscopic Simulation Model ComparisonsThese characteristics were evaluated against commonly available microscopic traffic simulationmodels. The models reviewed were CORSIM, VISSIM, SIMTRAFFIC, PARAMICS,HUTSIM, TEXAS, WATSIM, INTEGRATION, and AIMSUN (58, 59, 60, 61, 62, 63, 64, 65,66, 67, 68, 69, 70) . There are other microscopic traffic models available but used primarily for research (71). Only those that are commercially supported to some degree were evaluated.Much more detail is available in the forthcoming final report. Each simulation model reviewedhas its own strengths and weaknesses with respect to both traffic modeling in general, andsimulation of surrogate safety measures. All of the models reviewed would require some level of modification, upgrade, or enhancement to support the derivation of surrogate measures of safety

both internal enhancements to the source code and external enhancements for additional outputfile(s), statistics, and possibly new input value(s).

The significant results of the review were that CORSIM did not have adequate functionality tosupport the research further. The primary disadvantage of using CORSIM for surrogate safety



12/31


13/31


Page 13

this report, travel is assumed to be of the North American standard vehicles travel on the right-hand side of the roadway.

Crossing flows conflict POINT events

As shown in the figure, conflict points (notations 1 and 2) occur at the crossing of:

Left turn by the westbound traffic onto the minor street conflicting with the eastboundtraffic

Left turn by the minor approach traffic onto the main street to travel westbound.

These conflict points model the potential for angle collisions due to the acceptance of a gap thatis too small by the opposing traffic.

Merging crossing flows conflict LINE eventsConflict lines 3 and 4 occur at the crossing of:

Left turn by the minor street traffic conflicting with the westbound traffic Right turn by the minor street traffic conflicting with the eastbound traffic

These conflict lines model the potential for rear-end collisions (or angle collisions from the rear)due to the acceptance of a gap that is too small by the opposing traffic.

Following flows REAR-END conflict LINE eventsConflict lines 5 and 6 denote rear-end events where the leader vehicle makes a right (or left, notshown) turn causing the following vehicle to decelerate to avoid the collision. An additionalconflict line would be possible at the minor approach to the intersection, but is not shown.

Adjacent flows lane-changing REAR-END conflict LINE eventsConflict lines 7 and 8 denote rear-end events where the leader vehicle changes lanes abruptlyrequiring the vehicle in the adjacent lane to brake to avoid collision (The correspondingalternative lane changes from outer lane to inner lanes are not shown, but also possible as wellas lane changing conflict events on the minor approach).

At a four-leg (or more) intersection, additional conflict event points and lines would bereplicated for the opposite movements originating from the opposite side of the intersection. Inaddition, a crossing conflict point would be added for unsignalized intersections where vehicles

on the minor approach attempt to cross the intersection and continue straight on the minor road(i.e. northbound in the diagram).

Potential collisions not represented in the surrogate measuresAll of the conflict events represented are those that occur because of normal driving behaviors

that are observable and possible to be modeled in a computer simulation. This includesangle/turning and rear-end conflicts .



14/31


Page 14

Some collisions that are notably not included here are:

Side-swipe collisionsA vehicle in the process of changing lanes strikes an adjacent vehicle in the side because it: (a)accepts the gap too early, (b) does not see the vehicle because of obstructions, (c) the vehiclestruck has made its own maneuver simultaneously with the lane-changer, e.g. enters the link from a driveway.

Head-on collisionsWould only be represented in the intersection model as a vehicle inadvertently crossing thecenter-line. This is not modeled in current simulation codes.

Swerve-out-of-lane collisionsMuch like a head-on collision, vehicles making a right turn from a minor approach onto themajor street might veer into the opposing lanes if their speed was too high to make the turn. Thisis not modeled in current simulation codes.

U-turn-related collisionsU-turn maneuvers are particularly difficult to represent in simulation models. These conflicttypes would be evaluated at a later time when modeling of the maneuvers is better understoodand represented in simulations.

As indicated in (8), the majority of accidents at signalized and unsignalized intersections for bothurban and rural locations includes angle/turning collisions, rear-end collisions, and pedestrianaccidents (e.g. 78% of all urban 4-leg intersection accidents). Thus, neglecting the abovecollision types does not skew surrogate analysis results.

Unrepresented evasive maneuversIn addition to neglecting the above collision types, the conflict points and lines illustrated inFigure 1 also do not represent the capability of the reacting vehicles to perform the followingcountermeasures to avoid conflict events:

Change lanes or swerve,

Accelerate,

Pro-actively decelerate or change lanes [i.e. defensive driving behaviors that certain classof drivers learn from experience with a particular location], and

Abort maneuvers.

Conflict PointThe conflict point represents a fixed location in space where the crossing flow intersects with theflow proceeding straight through the intersection. In simulations where the crossing path isfixed, i.e. the turning vehicles always enter the receiving link in the same lane, this point wouldnot change for each through lane. Where there are several paths available to the turning vehiclethen there would be several conflict points defined. This would be the case if there is a shortly



15/31


Page 15

downstream driveway or intersection that the turning vehicle is desiring to make a right turn at(and the simulation models this).

It might be useful for the simulation model to pre-process the locations of these conflict points atthe beginning of the simulation and store them in a data structure for each intersection. Thiswould eliminate re-computation of the conflict points for each evaluation of a crossingmaneuver.

The timeline of a conflict point event is illustrated in Figure 2. Curve A represents the time-space trajectory of the crossing vehicle. Curve B represents the time-space trajectory of thethrough vehicle.

The times t1 t5 are defined as follows:

At time t1, the crossing vehicle enters the encroachment area (i.e. starts to turn left).

At time t2, the through vehicle realizes that a collision might occur and begins braking toavoid the collision.

At time t3, the corner of the rear bumper (either right or left rear corner, depending on thetravel direction) of the crossing vehicle leaves the encroachment point.

At time t4, the through vehicle was projected to arrive at the conflict point if the vehiclecontinued at the same speed and trajectory before it started braking, and

At time t5, the through vehicle actually arrives at the conflict point.

Conflict points also occur at the intersection of a flow from a right or left turning vehicle that proceeds in the same direction as the conflicted vehicle, but in a different lane. This situation isonly able to be evaluated in simulations where the entering path can vary by lane. For example,in the real world, many maneuvers of this type occur on purpose by drivers that want to accept a

particular gap of the size required to enter the flow. That same gap could not also includeacceleration in the same lane as the vehicle that the path of which was crossed, so a conflict pointevent is the result. This occurs even if the driver then re-enters the crossed lane after the vehiclehas passed.

Conflict LineThe conflict line represents a region of space where a preceding vehicle conflicts with afollowing vehicle in the same lane. This can be true of:

A vehicle entering the lane from a cross street in front of a vehicle proceeding straight, or

Vehicles traveling in the same direction when the leader decides to turn left or rightabruptly,

Vehicles changing lanes in front of another vehicle, causing braking by the follower tomaintain safe following distance.

The latter two cases are described in the rear-end conflict line situations in the next section.The spatial regions of conflict lines are not fixed locations, because they depend on theacceleration/deceleration characteristics of the particular vehicles involved in the conflict and the



16/31


Page 16

behavior of the driver model, i.e. how early or late the driver uses a turn signal. Thus, eachconflict line will have to be computed for each conflict event.

A timeline of a conflict line event for a vehicle turning from a minor approach onto the mainstreet in front of a vehicle progressing straight through the intersection is illustrated in Figure 5.Curve B represents the time-space trajectory of the leading vehicle (turning from the minor street). Curve A represents the time-space trajectory of the following vehicle (vehicle alreadytraveling on the main street). The times t1 t9 are defined as follows:

At time t1, vehicle B enters the encroachment area (i.e. starts to turn left into the samelane as the follower).

At time t2, vehicle A realizes that a collision might occur and begins braking to avoid thecollision.

At time t3, the next time step of the simulation is reached and state variables for eachvehicle are updated.

At time t4, vehicle B stops accelerating, meeting its intended travel speed. At time t5, vehicle A is projected to arrive at the first encroachment point if it had

continued with the same velocity before it started decelerating.

At time t6, vehicle B arrives at a reference maximum conflict evaluation distancedownstream from the starting point.

At time t7, vehicle A is projected to arrive at the second encroachment point if it hadcontinued with the same velocity at the second time step of the conflict line time period.

At time t8, vehicle A reaches the first encroachment point of the conflict line.

At time t9, vehicle A reaches the maximum conflict distance point.The reference maximum downstream distance is required for computation of surrogate measuressuch as the minimum time to collision. This is discussed further in the sections on computationof surrogate measures for conflict lines and rear-end conflict lines.

Rear-end Conflict LineRear-end conflicts are a special case of the conflict line situation. This is because either theleader or the follower could be the offending party in the generation of the near-miss collisionevent (or both). Consider that:

The leader may fail to indicate a turn signal soon enough or decelerate or turn

suddenly while the follower was initially at a safe following distance, The follower may be following too closely to react to an adequate signal indication or

safe braking speed, or

Both may be true.



17/31


Page 17

As such, the conflict line situation for the rear-end event is slightly different than the turning or lane changing events, primarily because the braking of the leader should be considered anexpected event because of the use of turn signals. Thus, the deceleration of the follower is alsoexpected to follow the application of the leaders turn signal as well. Detail is omitted here for

brevity.

SummaryConflict lines and conflict points define the situations at an intersection that are to be evaluatedfor surrogate measures of safety from simulation models. The next section identifies how safetysurrogates are defined from the conflict point and line definitions.

Surrogate Measures DefinitionsThe surrogate measures suggested for collection for each conflict event are, in order of expectedrelevance to safety:

1. Time to collision (TTC)2. Post encroachment time (PET)3. Initial deceleration rate (DR)4. Maximum of the speeds of the two vehicles involved in the conflict event (MaxS)5. Maximum relative speed of the two vehicles involved in the conflict event (DeltaS)

The approach suggested in this project is to collect all of the relevant data on all of the individualconflict events that occur for a particular scenario. The role of the analyst and the SSAMsoftware is to process this list of conflict event data into meaningful information about thesurrogate safety of the intersection scenario.

Severity of conflict and severity of resulting collisionThe size of the surrogates TTC, PET, and DR indicate the severity of the conflict event i.e. how

probable a collision could result from a conflict, such that:

Lower TTC indicates higher probability of collision Lower PET indicates higher probability of collision Higher DR indicates higher probability of collision

MaxS and DeltaS are used to indicate the likely severity of the (potential) resulting collision , if the conflict event had resulted in a collision, instead of a near-miss. Using the mass of thevehicles involved in the conflict, the MaxS and DeltaS values could also be used to calculatedmomentum values and get a better estimate of severity of the resulting collision (e.g. heavier vehicles can cause more damage on light ones the mix of the traffic stream is important data toinclude in safety analysis).

It is important to distinguish both the severity of the conflict and the severity of the resultingcollision. A location with many conflict events of severity exceeding the thresholds for TTC,



18/31


Page 18

PET, and DR but that are of low severity on the DeltaS and MaxS scales may not be of as highinterest in terms of safety implication. That is, the resulting accidents would be more likely to be

property-damage-only when MaxS and DeltaS are low. Locations that may experience fewer total conflict events but those conflict events that do occur, occur at very high resulting potential

severity (i.e. result in injury and fatality accidents) are probably of more interest to analysts andengineers deciding how to prioritize safety upgrades amongst a number of candidate locations.

The next sections identify these surrogate measures on the conflict point, conflict line, and rear-end conflict line diagrams.

Surrogate measures for conflict pointsFigure 4 illustrates the definitions of the surrogate measures for a conflict point. The nextsubsections describe each surrogate measure in the figure.

TIME TO COLLISION

As shown in Figure 4, TTC is defined uniquely for a conflict point as t4 t3. This is thedifference between the encroachment end time of the turning vehicle and the projected arrivaltime of the through vehicle with the right-of-way at the conflict point, if the vehicle hadcontinued with the same speed at the time of initial deceleration to avoid collision.

POST-ENCROACHMENT TIME

As shown in Figure 4, PET is defined uniquely for a conflict point as t5-t3. This is the time between the departure of the encroaching vehicle from the conflict point and the arrival of thevehicle with the right-of-way at the conflict point.

MAXS

MaxS is first defined for each vehicle independently as the maximum speed of the vehicle between the times t1 to t5. Then the maximum of those two maximum values for each vehiclewould be recorded as the MaxS value.

DELTAS

DeltaS is first defined for each time slice (from the beginning to the end of the conflict event) asthe difference between the velocity of the two conflicting vehicles. Then the maximum of thoseDeltaS values for each time slice would be recorded as the DeltaS value.

INITIAL DECELERATION RATE

Deceleration is the evasive action taken by the subject vehicle to avoid the collision. The initialdeceleration rate would be a useful measure to indicate the potential severity of the conflictevent. Acceleration and deceleration rates should be available directly from the simulationmodel at each time step. On Figure 4, the initial deceleration rate is the second derivative of curve B at time t2.



19/31


Page 19

Surrogate Measures for Conflict Lines and Rear-end Conflict LinesSimilar graphs describe surrogates on conflict lines and rear-end conflict except that the TTC isnot defined uniquely as for a conflict point (t4t3 in Figure 4). Because the conflict line and therear-end conflict line are a set of conflict points computed at each time step update in thesimulation, basically the TTC is defined as the minimum value that the TTC achieves during theduration of the conflict line event. This is illustrated in Figure 5. The figure shows the twovehicle trajectories with the same information as the graph of Figure 3 but with additionalinformation for the surrogate measure definitions marked on the diagram. The four trianglesadded to the curved portions of both graphs A and B near times t2 and t4are the MaxS andDeltaS candidates. The velocity of each vehicle is the derivative to the curve at those points.The brackets added between times t1-t4 and t3-t7 labeled TTC-1 and TTC-2, are the recordedTTC values for two time steps. The recorded TTC value is TTC-1, since the bracket length of TTC-1 is shorter than the bracket length of TTC-2 (the minimum TTC is retained in a conflictline event). Similarly, the recorded PET value is the minimum of all PET values recorded duringthe event. In Figure 5, this is indicated as PET-1 since the bracket length of PET-1 is shorter than the bracket length of PET-2. Other studies have proposed summary TTC statistics such asthe total TTC time or average TTC for these line events, but these types of measures are notillustrated in the Figure (72) .

SummaryThis FHWA project has identified surrogate measures that can be collected from commercialsimulation models for evaluating the relative safety of intersection design alternatives or existingfacilities. Some modifications to the interfaces to simulation models will be required to obtainthe necessary data on each conflict event. Each surrogate measure is collected based on theoccurrence of a conflict event an interaction between two vehicles in which one vehicle musttake evasive action to avoid a collision. The surrogates that are proposed as the best measuresare the time to collision, post encroachment time, deceleration rate, maximum speed, and speeddifferential. Time to collision, post encroachment time, and deceleration rate can be used tomeasure the severity of the conflict. Maximum speed and the speed differential can be used tomeasure the severity of the potential collision (along with information about the mass thevehicles involved). Definitions of all possible conflict events and algorithms for calculating thesurrogate measures for conflict points were presented.

After the surrogates are available from the simulation model for each individual conflict event, a post-processing tool can be developed to compute statistics and distributions of the surrogatemeasures according to a wide variety of accumulation methods; by approach, by movement, byconflict type, etc. Using these statistics, two or more design alternatives could be evaluated for their relative safety by comparing these cumulative distributions and aggregations. The goals of further research are to develop a standard data format for the surrogate analysis module tointeract with multiple traffic simulation models in a common way, develop a calibration andvalidation methodology, and use the validation methodology to determine the value of surrogatesafety analysis with simulation models.



20/31


Page 20

AcknowledgementThis research was funded under FHWA contract DTFH61-01-Q-00005. The authors would liketo thank Joe Bared, Forrest Council, Raj Ghaman, John Halkias, Ray Krammes, Henry Lieu,Gene McHale, Tim Neuman, and Tarek Sayed for their comments and suggestions.



21/31


Page 21

References

[1] I. Anderson, et al, 1999. Relationship to safety of geometric design consistency measures for rural two-lanehighways, Transportation Research Record, No. 1658.

[2] K. Bauer & D. Harwood, 2000. Statistical models of at-grade intersection accidents Addendum, FHWA-RD-99-094. USDOT, Washington, DC.

[3] J. Bared, et al, 1999. Safety evaluation of acceleration and deceleration lane lengths, ITE Journal, May, pp50-54.

[4] F. Council & J. Stewart, 1999. Safety effects of the conversion of rural two-lane to four-lane roadways basedon cross-sectional models, Transportation Research Record, Vol. 1665.

[5] T. Datta, 1991. Head-on, left-turn accidents at intersections with newly installed traffic signals,Transportation Research Record, Vol. 1318.

[6] G. Davis & Y. Gao, 1995, Statistical methods to support induced exposure analyses of traffic accident data,Transportation Research Record, Vol. 1401.

[7] N. Garber & S. Subramanyan, 2001. A procedure for incorporating crash risk in selecting congestionmitigation strategies: a case study: Hampton roads area, Paper presentation at 80th Annual Meeting of theTransportation Research Board, Washington, D. C.

[8] D. Harwood, et al, 2000. Prediction of the expected safety performance of rural two-lane highways, FHWA-RD-99-207. USDOT, Washington, DC.

[9] E. Hauer, et al, 1988. Estimation of safety at signalized intersections, Transportation Research Record, Vol.1185.

[10] P. Ossenbruggen, 1998. A method of identifying hazardous highway locations using the principle of individual lifetime risk, Risk, Health, and Environment. Volume 83, Winter.

[11] B. Persaud & K. Mucsi, 1995. Microscopic accident potential models for two-lane rural roads,Transportation research record. No. 1485p. 134-139.

[12] M. Zhou & V. Sisiopiku, 1997. Relationship between volume-to-capacity ratios and accident rates,Transportation Research Record, Vol. 1581.

[13] D. Malahel, 1986. A note on accident risk, Transportation Research Institute, Technion-Israel Institute of

Technology, vol. 13.

[14] C. Zegeer, 1986. Methods for identifying hazardous highway elements, NCHRP Synthesis 128.Transportation Research Board, Washington, DC.

[15] E. Hauer, 1996. Identification of sites with promise, Transportation Research Record, Vol. 1542.



22/31


Page 22

[16] J. Higle & M. Hecht, 1989. A comparison of techniques for the identification of hazardous locations,Transportation Research Record, Vol. 1238.

[17] J. Higle & J. Witkowski, 1988. Bayesian identification of hazardous locations, Transportation ResearchRecord, Vol. 1185.

[18] M. Lau & A. May, 1988. Accident prediction model development: Signalized intersections, Final Report,UCB-ITS-RR-88-7. University of California, Berkeley, CA.

[19] A. Mensah & E. Hauer, 1998. Two problems of averaging arising in the estimation of the relationship between accidents and traffic flow, Transportation Research Record, Vol. 1635.

[20] A. Kaub, 2000. Highway corridor safety levels of service based on annual risk of injury, 79 th TransportationResearch Board Annual Meeting, Washington, DC.

[21] T. Sayed, et al, 1994. Simulation of traffic conflicts at unsignalized intersections, Accident Analysis andPrevention, Vol. 26, No 5.

[22] FHWA, 1990. Traffic conflict techniques for safety and operations Course Materials. NHI course 38059.USDOT, Washington, DC.

[23] W. Glauz & D. Migletz, 1980. Application of traffic conflict analysis at intersections, NCHRP Report 219.Transportation Research Board, Washington, DC.

[24] M. Parker & C. Zegeer, 1988. Traffic conflict technique for safety and operations Engineers guide. FHWA-IP-026. FHWA, Washington, DC.

[25] D. Perkins & J. Harris, 1967. Criteria for traffic conflict characteristics, Report GMF 632. Warren, MI.General Motors Corp.

[26] F. Amundsen & C. Hyden, 1977. Proceedings of first workshop on traffic conflicts, Oslo, Institute of Transport Economics.

[27] B. Allen, et al, 1978. Analysis of traffic conflicts and collisions, Transportation Research Record, No. 667.

[28] D. Migletz, et al., 1985. Relationships between traffic conflicts and accidents, FHWA-RD-84-042. FederalHighway Administration, Washington, DC.

[29] C. Hyden, 1987. The development of a method for traffic safety evaluation: The Swedish conflictstechnique. Dept. of Traffic Planning and Engineering, Lund University, Sweden.

[30] J. Fazio & N. Rouphail, 1990. Conflict simulation in INTRAS: Application to weaving area capacityanalysis, Transportation Research Record, Vol. 1287.

[31] T. Sayed & S. Zein, 1999. Traffic conflict standards for intersections, Transportation Planning andTechnology, Vol. 22.

[32] K. Fitzpatrick, 1991. Gaps accepted at stop-controlled intersections, Transportation Research Record, Vol.1301.

[33] F. Gharaybeh, 1991. Identification of accident-prone locations in greater Amman, Transportation ResearchRecord, Vol. 1318.



23/31


Page 23

[34] N. Katamine & I. Hamarneh, 1998. Use of the traffic conflict technique to identify hazardous intersections,Road & transport research. Vol. 7, no. 3

[35] N. Salman & K. Al-Maita, 1995. Safety evaluation at three-leg, unsignalized intersections by traffic conflicttechnique, Transportation Research Record, Vol. 1485.

[36] J. Hayward, 1972. Near miss determination through use of a scale of danger, Report TTSC-7115.University Park, The Pennsylvania State University.

[37] H. Kruysse, 1991. The subjective evaluation of traffic conflicts based on an internal concept of dangerousness, Accident analysis and prevention. Vol. 23, no. 1.

[38] G. Tiwari, et al, 1995. Conflict analysis for prediction of fatal crash locations in mixed traffic streams, 29 th annual proceedings of the association for advancement of automotive medicine. Oct 16-18, Chicago, Il.

[39] J. Darzentas, et al, 1980. Simulation of Road traffic conflicts at T-junctions, Simulation, May, pp 155-164.

[40] J. Darzentas, et al, 1980. Minimum acceptance gaps and conflict involvement in a single crossing manuevre,

Traffic Engineering & Control, Vol 21, pp 58-62.

[41] D. Cooper & N. Ferguson, 1976. A conflict simulation model, Traffic Engineering and Control, Vol 17, pp306-309.

[42] M. Ghaffari, 1990. A stochastic computer simulation of traffic conflicts, Ph.D. dissertation, University of Illinois at Chicago.

[43] D. Perkins & B. Bowman, 1982. Effectiveness evaluation by using non-accident measures of effectivenessTransportation Research Record, Vol. 905.

[44] H. Thompson & D. Perkins, 1983. Surrogate measures for accident experience at rural isolated horizontalcurves, Paper presentation at 1983 Annual meeting of Transportation Research Board, Washington, DC.

[45] K. Fitzpatrick, et al, 2000. Accident mitigation guide for congested rural two-lane highways, NCHRP report440, National Academy Press, Washington, DC.

[46] V. Rao & V. Regaraju, 1998. Modeling conflicts of heterogenous traffic at urban uncontrolled intersections,Journal of Transportation Engineering, Jan/Feb.

[47] V. Hernandez, 1982. A microscopic digital computer simulation of traffic conflicts at 4-way fixed timesignalized urban intersections. Ph.D. dissertation, University of California, Los Angeles.

[48] P. Huang & P. Pant, 1996. Simulation Neural-Network model for evaluating dilemma zone problems at high-speed signalized intersections, Transportation Research Record, Vol. 1456.

[49] A. Mehmood, et al, 2001. Simulation of Road Crashes using Systems Dynamics, Paper presentation at the80th Annual Meeting of the Transportation Research Board, Washington, D.C., January 2001.

[50] J. Fazio, et al, 1995. Use of freeway conflict rates as an alternative to crash rates in weaving section safetyanalyses, Transportation Research Record, Vol. 1401.

[51] I. Kosonen & S. Ree, 2000. The potential for microscopic simulation in traffic safety and conflict studies,Proceedings of Conference on Road safety on three continents, Pretoria.



24/31


Page 24

[52] I. Kosonen, 1999. HUTSIM Urban Traffic Simulation and Control Model: Principles and Applications.Doctoral Thesis. Helsinki University of Technology, Transportation Engineering.

[53] Rioux & Lee, 1977. Texas a microscopic traffic simulation package for isolated intersections, Center for Highway research, University of Texas at Austin.

[54] A. Avgoustis, 1999. Quantifying the safety impacts of intelligent transportation systems, MS Thesis, CivilEngineering, Virginia Tech, Blacksburg, VA.

[55] Alumnus Software, 1993. Document on the micro-node logic as implemented in the fortran code of CORSIM(TSIS Version 4.01 Beta), Internal Technical Report. FHWA Turner-Fairbank Highway Research Center,McLean, VA.

[56] S. Lin, 1998. CORSIM Micronode Logic, Technical Report. FHWA Transportation Research Laboratory,Turner-Fairbank Highway Research Center, McLean, VA.

[57] H. Rakha & B. Crowther, 2002. A comparison of the greenshields, pipes, and van aerde car-following andtraffic stream models, presentation at 2002 TRB Annual meeting. Washington, DC.

[58] J. Barcelo, 2001. Microscopic traffic simulation: a tool for the analysis and assessment of ITS systems,Transport Simulation Systems Internal Report, www.tss-bcn.com .

[59] J. Barcelo, et al, 1998. Microscopic traffic simulation for ATT systems analysis a parallel computingversion, Transport Simulation Systems Internal Report, www.tss-bcn.com .

[60] J. Barcelo, et al, 1998. AIMSUN modeling issues, Transport Simulation Systems internal Report, www.tss- bcn.com .

[61] KLD & Associates, 2002. WATSIM features, Internal Report, www.kldassociates.com/faq .

[62] M. Kokkinen, 1996. HUTSIM 4.2: Quick Reference Guide, Traficon, Ltd. Helsinki, Finland.

[63] I. Kosonen, et al, 2001. A short HUTSIM description, web document.

[64] C. Lee, et al, 1993. TEXAS model for intersection traffic version 3.0 documentation updated for version3.20, Center for transportation Research, University of Texas at Austin.

[65] PTV AG, 2001. VISSIM User Manual Version 3.6. on Demo CD. Karlsruhe, Germany.

[66] Quadstone Ltd., 2002 Paramics system overview V3, internal report, www.paramics-online.com/techsupport/docum.htm .

[67] Quadstone Ltd., 2002 Paramics Developer Suite, internal report, www.paramics-online.com/techsupport/docum.htm .

[68] Quadstone Ltd., 2002 Paramics Modeller Reference V3, internal report, www.paramics-online.com/techsupport/docum.htm .

[69] Y. Wang & P. Prevedouros, 1998. Comparison of INTEGRATION, CORSIM, and WATSIM in replicatingvolumes and speeds on three small networks, Presentation at 1998 Annual Meeting of TRB. Washington, DC.

[70] M. Van Aerde, et al, 1998. INTEGRATION: Overview of simulation features, Internal Report, Dept. of Civil Engineering, Queens University, Kingston, Canada.

http://www.tss-bcn.com/http://www.tss-bcn.com/http://www.tss-bcn.com/http://www.tss-bcn.com/http://www.tss-bcn.com/http://www.tss-bcn.com/http://www.tss-bcn.com/http://www.kldassociates.com/faqhttp://www.kldassociates.com/faqhttp://www.paramics-online.com/techsupport/docum.htmhttp://www.paramics-online.com/techsupport/docum.htmhttp://www.paramics-online.com/techsupport/docum.htmhttp://www.paramics-online.com/techsupport/docum.htmhttp://www.paramics-online.com/techsupport/docum.htmhttp://www.paramics-online.com/techsupport/docum.htmhttp://www.paramics-online.com/techsupport/docum.htmhttp://www.paramics-online.com/techsupport/docum.htmhttp://www.paramics-online.com/techsupport/docum.htmhttp://www.paramics-online.com/techsupport/docum.htmhttp://www.paramics-online.com/techsupport/docum.htmhttp://www.paramics-online.com/techsupport/docum.htmhttp://www.paramics-online.com/techsupport/docum.htmhttp://www.paramics-online.com/techsupport/docum.htmhttp://www.paramics-online.com/techsupport/docum.htmhttp://www.kldassociates.com/faqhttp://www.tss-bcn.com/http://www.tss-bcn.com/http://www.tss-bcn.com/http://www.tss-bcn.com/


25/31


Page 25

[71] S. Algers, et al, 1997. Review of microsimulation models. SMARTEST Project Deliverable D3, Institute for Transport Studies, University of Leeds, Leeds, U.K. http://www.its.leeds.ac.uk/projects/smartest/deliv3.html#a2.

[72] M. Minderhoud & P. Bovy, 2001. Extended time to collision measures for road traffic safety assessment,

Accident Analysis and Prevention, Vol 33, pp89-97.

[73] A. Torday & A. Dumont, 2002. Safety indicator for microsimulation-based assessments, Swiss FederalInstitute of Technology, Laboratory of Traffic Facilities, LAVOC-EPFL. Lausanne, Switzerland.



26/31


Page 26

List of FiguresFigure 1. Conflict point and linesFigure 2. Conflict point diagramFigure 3. Conflict line diagramFigure 4. Surrogates identified on conflict point diagramFigure 5. Surrogates identified on conflict line diagram



27/31


Page 27

4

8

1

2 3

6

5

7

Figure 1. Conflict point and lines (42)



28/31


Page 28

conflict point

A

B

S p a c e

t1Time

encroachment begin

t2 t3

Vehicle begins braking

encroachment end

projected arrival at conflict pt

actual arrival at conflict pt

t4 t5

Figure 2. Conflict point diagram (42)



29/31


Page 29

A

B

Through Vehicle

Time

Distance to Conflict Point

Turning Vehicle

0

t1 t2 t3 t5

Start of encroachment

Deceleration reaction

Turning vehicle stops acceleration

Entering vehicle reachesmax distance

reference maximum distance

t4

Projected time to reachfirst conflict point

t6 t8t7

Distance of B at next time step

t9

Through vehicle reachesmaximum conflict distance

Projected time to reachsecond conflict point

Figure 3. Conflict line diagram (42)



30/31


Page 30

conflict point

A

B

S p a c e

t1Time

encroachment begin

t2 t3

Vehicle begins braking

encroachment end

projected arrival at conflict pt

actual arrival at conflict pt

t4 t5

TTC

PET

Initial Deceler ation Rate

MaxS and DeltaS

Figure 4. Surrogates identified on conflict point diagram



31/31


A

B

Through Vehicle

Time

Distanceto Con

flictPoint

Turning Vehicle

0

t1 t2 t3 t5t4 t6 t8t7

Distance of B at next time step

t9

PET-2Recorded PET value

TTC-1

TTC-2

initial decelerationrate (DR)

Recorded TTC value

speed zerono longer a

collision course

MaxS and DeltaScandidates

MaxS and DeltaScandidates

PET-1

Figure 5. Surrogates identified on conflict line diagram

gettman and head 2003

Documents