U.S. Department of Transportation
Federal Highway Administration
1200 New Jersey Avenue, SE
Washington, DC 20590
Federal Highway Administration Research and Technology
Coordinating, Developing, and Delivering Highway Transportation Innovations
|This report is an archived publication and may contain dated technical, contact, and link information|
|Publication Number: FHWA-HRT-11-040 Date: November 2012|
Publication Number: FHWA-HRT-11-040
Date: November 2012
V2I communication for safety enables vehicles with 360-degree awareness to inform drivers of hazards and situations they cannot see. The following levels of action are envisioned:
The V2I communication focuses on applications in which safety can be enhanced through connectivity that enables the exchange of information from a vehicle to the roadway infrastructure, from the infrastructure to a vehicle, or from the infrastructure to some other wireless-enabled device. Both original equipment and aftermarket solutions are being considered.
The objective of this study was to conduct and document crash data analyses of the primary V2I for safety application areas being evaluated by the Federal Highway Administration (FHWA) Safety Program. The goal was to determine which subsets of crashes are potentially treatable with currently identified V2I for safety application areas and which additional subsets could be treated either with modifications to the current application areas or with new application areas. The primary application areas of interest were intersection safety, speed, vulnerable road users, and others (i.e., applications that cannot be classified in the aforementioned areas of interest).
The specific objectives of the crash data analyses were as follows:
This report documents the most salient findings of this effort. It includes a description of the data used, the methods employed, and results by V2I application area. Supporting materials are provided in the appendix.
Throughout this report, crashes are described as either targeted crashes or unaddressed crashes with respect to V2I for safety application areas. Targeted crashes are crashes that could potentially be eliminated through the deployment of a specific V2I application or set of applications (i.e., researchers should determine the potential benefit of an application area, assuming 100 percent effectiveness and 100 percent deployment). The actual number of crashes mitigated depends on the effectiveness of the application and the extent of deployment. Unaddressed crashes are those that are not eliminated even if a V2I application, or set of applications, is 100 percent effective and fully deployed. That is not to say that unaddressed crashes cannot be mitigated by V2I application areas. Rather, currently identified application areas and potential extensions do not target these crashes. Unaddressed crashes might be covered through the development of new V2I application areas or by V2V and AV applications.
The primary analysis involved an examination of the NASS GES database and several State databases from HSIS.(1,2) These databases and analysis methodologies are briefly described in the following sections.
NASS GES contains data on a representative random sample of thousands of reported minor, serious, and fatal crashes involving passenger cars, pickup trucks, vans, large trucks, motorcycles, and pedestrians.(1) It is based on cases selected from a sample of police crash reports within randomly selected areas of the country.
The crash reports are chosen from 60 urban and rural areas that are representative of the geography, roadway mileage, population, and traffic density of the United States. Data collectors make weekly visits to approximately 400 police jurisdictions in the 60 areas throughout the United States and randomly sample approximately 50,000 police crash reports each year. Weights are provided so the sample data can be weighted to a national estimate. NASS GES data from 1988 to 2008 (crash, vehicle, and person files) are available online from the National Highway Traffic Safety Administration (NHTSA).(1)
This study analyzed the most recent 4 years of crash data from NASS GES, which was from 2005 through 2008. The analysis used the following five datasets for each year:
Analyses were conducted using both raw and imputed variables in the database. Raw data reflect the probability sample of police-reported crashes. The NHTSA report Imputation in the NASS General Estimates System describes imputation as the process of fabricating data when data are unknown.(4) Imputed variables are used to fill in unknown values. This is done because of historical precedence, convenience, consistency of data, and potential reduction in bias. Weighted variables are national estimates of crash characteristics based on weights established by NHTSA. According to the NASS GES Analytical User's Manual 1988-2008, the weight is the product of the inverse of the probabilities of selection at each of the three stages in the sampling process and is used to produce national estimates from the data.(5) Information on national estimates can be found in the National Accident Sampling System General Estimates System Technical Note.(6)
Crashes have multiple characteristics that can be grouped in an almost infinite number of ways. The research for this report used work conducted by The John A. Volpe National Transportation Systems Center (Volpe) to categorize each crash in the NASS GES database in a pre-crash scenario. The pre-crash scenarios were assigned using a Statistical Analysis Software® (SAS®) program developed by Volpe. Detailed criteria for assigning pre-crash scenarios are summarized in Pre-Crash Scenario Typology for Crash Avoidance Research.(7) Volpe classifies crashes by 38 single-vehicle pre-crash scenarios and 46 multi-vehicle pre-crash scenarios. During the 2005-2008 analysis period, there were observed crashes for 32 single-vehicle pre-crash scenarios and 44 multi-vehicle pre-crash scenarios. Distributions of crashes by pre-crash scenarios and vehicle type are presented in the appendix.
The Volpe SAS® program assigns one of six vehicle types to each vehicle involved in a crash. The vehicle types are light vehicle, transit vehicle, specialty vehicle, single-unit truck, combination-unit truck, and other. These vehicle types are assigned based on the NASS GES data variables for vehicle body type (bdytyp_h), special use type (spec_use), and trailer type (trailer) (see table 3). Definitions for vehicle body, special use, and trailer types can be found in the National Automotive Sampling System (NASS) General Estimates System (GES) Analytical User's Manual 1988-2008.(5) Motorcycles are included in the "other" and "specialty" vehicle types.
Table 3. Vehicle type assignment criteria.
|Vehicle Type||Assignment Criteria|
|Light vehicle||If (((1 ≤ bdytyp_h ≤ 22) or (28 ≤ bdytyp_h ≤ 41) or (45 ≤ bdytyp_h ≤ 49)) and spec_use = 0)|
|Transit vehicle||If (bdytyp_h in (25, 58, 59)) and (spec_use < 1 or spec_use = 3 or spec_use = 8 or spec_use = 9 or spec_use > 12)|
|Specialty vehicle||If ((80 ≤ bdytyp_h ≤ 89) or bdytyp_h in (23, 24, 42, 50, 65, 93, 97)) and ((4 ≤ spec_use ≤ 7) or spec_use = 2 or spec_use = 12)|
|Single-unit truck||If ((bdytyp_h in (60, 64, 66, 78, 79)) and (trailer in (1, 6)))|
|Combination-unit truck||If (bdytyp_h in (60, 64, 66, 78, 79)) and ((2 ≤ trailer ≤ 5))|
Note: NASS GES data variables and codes used in these assignment criteria are defined in the National Automotive Sampling System (NASS) General Estimates System (GES) Analytical User's Manual 1988-2008.(5)
Crash costs developed by Volpe were employed in this study as part of the economic analysis to ensure consistency between these results and those from past studies. All costs associated with NASS GES crash costs in this report were based on 2007 U.S. dollars.
The crash costs associated with each pre-crash scenario were calculated based on procedures used in a previous crash typology study, Heavy Vehicle Pre-Crash Scenario Typology for Crash Avoidance Research.(8) The conversion from crashes to economic costs was based on the severity of the crash. Specifically, DaSilva et al. used the maximum abbreviated injury scale (MAIS).(8) The crash costs used in this study are based on MAIS (see table 4).
Table 4. MAIS comprehensive crash costs (based on 2007 U.S. dollars).(8)
|Consumer Price Index||Factor||MAIS 1||MAIS 2||MAIS 3||MAIS 4||MAIS 5||MAIS 6|
|1.204077||Emergency medical services||$117||$255||$443||$999||$1,026||$1,003|
|1.277512||Quality-adjusted life years (QALYs)||$9,118||$186,525||$262,189||$784,777||$2,674,628||$4,889,799|
|New comprehensive costs||$22,420||$271,780||$500,045||$1,232,893||$4,086,149||$6,128,666|
Note: MAIS severity levels are as follows: MAIS 0 = no injury, MAIS 1 = minor, MAIS 2 = moderate, MAIS 3 = serious, MAIS 4 = severe, MAIS5 = critical, and MAIS 6 = fatal.
NASS GES does not provide detailed information regarding injury severity based on the MAIS coding scheme. Instead, it records injury severity by crash victim based on the KABCO scale as follows:
NASS GES also provides information for "injury severity unknown" and "died prior." Because there were not many "died prior" crashes, they were not considered in the analysis. DaSilva et al. used a conversion matrix to estimate MAIS injuries from the KABCO scale.(8) The series of multiplicative factors that were applied to convert injury severity from KABCO to MAIS designations were obtained from the NHTSA report, Preliminary Regulatory Evaluation.(9)
The conversion factor was multiplied by MAIS subtotal dollar values in table 4 to obtain a weighted cost for each severity (travel delay and property damage are not included in this value). For MAIS 0 (no injury), $2,423 was used based on unpublished values provided by Volpe. It should be noted that the goal of the current research was not to determine the exact dollar values associated with various crash types and pre-crash scenarios. Rather, the goal was to identify the relative magnitude of the crash problem for specific scenarios. As such, it was acceptable to use the unit costs in 2007 dollar values from the report by DaSilva et al.(8)
The resulting crash cost per maximum severity as reported in NASS GES was calculated using the procedure previously described and is shown in table 5. The crash cost per maximum severity was applied to the imputed crashes for each single-vehicle and multi-vehicle pre-crash scenario by maximum severity from 2005 to 2008.
Table 5. Cost per maximum severity (based on 2007 U.S. dollars).
|NASS GES Code||Description||Cost|
|5||Injured, unknown severity||$118,770|
For single-vehicle crashes, the total cost from 2005 to 2008 was $657 billion, with an annual average cost of $164 billion. The three highest-cost single-vehicle pre-crash scenarios represented 73 percent of the total costs for single-vehicle crashes ($478 billion of the $657 billion total) and included the following:
For multi-vehicle crashes, the total cost from 2005 to 2008 was $700 billion, with an annual average of $175 billion. The three highest-cost multi-vehicle pre-crash scenarios represented 45 percent of the total costs for multi-vehicle crashes ($315 billion of the $700 billion total) and included the following:
Detailed cost information for each pre-crash scenario by injury type is presented in the appendix (see table 24 for single-vehicle crashes and table 25 for multi-vehicle crashes).
Many application areas required the identification of crashes related to various types of intersections and non-intersection segments. The location of the crash was characterized by the location where the first harmful event occurred. The first harmful event is defined as the occurrence of injury or damage involving a motor vehicle in transport, which can result from an impact or non-collision event. The variable imputed relation to junction (RELJCT_I [A09I]) specifies whether the crash occurred at a junction or non-junction area and whether it occurred at an interchange or non-interchange area. This variable was used to identify intersection-related and segment-related crashes.
Crashes were identified as either intersection, intersection-related, or segment crashes. Intersection crashes were crashes that occurred within the intersection, while intersection-related crashes occurred on the approach to or exit from an intersection and resulted from an activity, behavior, or control related to the movement of traffic through the intersection. Crashes were identified as intersection or intersection-related if the relationship to the junction was coded as any of the following:
All other crashes were considered segment crashes, which included crashes that occurred at non-interchange areas (i.e., non-junctions, driveways or alley accesses, bridges, and crossovers) or at interchange areas (i.e., non-junctions, ramp exits/entrances, and other interchange locations). While crashes associated with driveways and other access points are similar to crashes at intersections, the V2I application areas discussed in this report did not target driveway crashes.
Table 6 presents single-vehicle and multi-vehicle crashes by location. Intersection and intersection-related crashes were analyzed by type of traffic control, which was identified using the crash-level NASS GES data variable imputed traffic control device (TRFCON_I[A16I]) and, in some cases, the similarly named vehicle-level variable. Intersection crashes were classified by traffic control device as signalized intersection crashes (TRFCON_I = 1), stop-controlled intersection crashes (TRFCON_I = 21), and other intersection crashes. NASS GES determines the intersection control by the control affecting every vehicle in the crash. In a situation where two vehicles crash on an uncontrolled approach of a two-way, stop-controlled intersection, the intersection control for the crash is coded as uncontrolled.
Segment-related crashes were analyzed by presence of curvature (i.e., tangents [ALIGN_I = 1] versus curves [ALIGN_I = 2]). The majority of single-vehicle crashes occurred on tangent sections (67 percent) and curve sections (20 percent). The primary locations of multi-vehicle crashes were tangent sections (42 percent) and signalized intersections (28 percent).
Table 6. Distribution of crash location-average annual national crashes.
|Location of Crash||Single-Vehicle||Multi-Vehicle||Total|
|Intersections and intersection-related||Signalized||67,520||4||1,147,720||28||1,215,240||20|
Table 7 shows the assignment criteria and variable description for the area types defined in the NASS GES data. Area type was determined using the crash-level NASS GES data variable land use (Land_Use[A05]). Crashes that occurred within areas with a population greater than 25,000 people were considered urban, and areas reported as "other area" were considered rural.
Table 7. Area type assignment criteria.
|Area Type||Assignment Criteria||Variable Description|
|Urban||Land_use = 1||Area population of 25,000-50,000|
|Land_use = 2||Area population of 50,000-100,000|
|Land_use = 3||Area population of 100,000 or more|
|Rural||Land_use = 8||Other area|
|Unknown||Land_use = 9||Unknown|
Several application areas required the identification of crashes that were related to speeding. Speed-related crashes for single- and multi-vehicle pre-crash scenarios were identified using the NASS GES data speed-related variable (SPEEDREL [D9N] = 1). This variable was coded at the vehicle level to indicate whether speed was a contributing factor in the crash. If speed was coded as a factor for any of the involved vehicles, the crash was considered speed related.
Some application areas required the identification of crashes that occurred during adverse driving conditions, which were defined based on the crash-level NASS GES data variables for roadway surface condition (Sur_Cond[A15]) and weather (Weather[A20]). Crashes were considered to be related to adverse conditions if conditions for the crash were coded as any of the following:
HSIS is a roadway-based system maintained by FHWA that provides quality data on crash, roadway, and traffic variables linked to homogeneous sections of the highway system under State control.(2) It is the only multi-State database that allows for the safety analysis of roadway design factors through its file system and that has the capability to link roadway inventory and exposure data to crash data for a large sample of primary route mileage. It is also the only file system that includes both roadway sections with and without crashes. Currently, seven States are part of HSIS: California, Illinois, Maine, Minnesota, North Carolina, Ohio, and Washington. Historical data from Michigan and Utah are also available, but updated data are no longer captured. This study analyzed crash data for the most recent 3 years for California, Illinois, Minnesota, and Washington, which was from 2005 through 2007.
There are six types of data files available within HSIS, and all States maintain three basic files: a crash file, a roadway inventory file, and a traffic volume file. Additional roadway geometry files are also available within selected States, including a horizontal curve file (Illinois, Ohio, and Washington) and a vertical grade file (Illinois and Washington). Intersection and interchange data are also available for a limited number of States.
California and Minnesota were selected for detailed intersection analyses since these States provide intersection datasets. Illinois and Washington were used to conduct detailed analyses of curves and curve crashes since both States have a curvature file.
It is important to note that HSIS data are only available for State-maintained roadways in each State. As such, HSIS represents more rural than urban areas since roadways in urban areas are often maintained by a municipality. Quality data are largely unavailable for municipalities but would be helpful to better define the magnitude of the safety problems and potential impacts of V2I application areas in urban areas.
Previous studies have provided the foundation for this study, including the development of pre-crash scenarios and the investigation of the potential benefits of specific V2I for safety application areas. The following sections review select studies that were used as a basis for the analyses conducted as part of this current research. A brief overview of each study is provided, including the types of crash analyses performed and any gaps that were addressed as part of this project.
Najm et al. analyzed 2004 NASS GES data to develop a new typology of pre-crash scenarios for all police-reported crashes that involved at least one light vehicle (e.g., passenger car, sports utility vehicle, van, minivan, and light pickup truck).(7) A total of 37 pre-crash scenarios were defined based on two existing typologies: the 44 crashes typology developed by General Motors® and the pre-crash scenarios typology developed by USDOT.(10,11) Najm et al.'s new typology defines pre-crash scenarios that describe the vehicle movements, vehicle dynamics, and critical events that occurred immediately prior to the crash.
Each pre-crash scenario was ranked by three measures: crash frequency, functional years lost, and economic cost. Functional years lost and economic costs were estimated for each pre-crash scenario based on the severity of each crash assigned to that scenario. A summary table was provided for each pre-crash scenario that included the number of vehicles and people involved as well as the distribution of crashes by severity using the KABCO injury scale and the abbreviated injury scale (AIS). AIS uses a score of 1 (minor injury) through 6 (unsurvivable injury) to describe the injury severity of the crash victim. A detailed summary also described the typical scenario for a crash, factors that are overrepresented, dynamic variations of the scenario, and the general severity of crashes.
Single light-vehicle crashes resulted in an estimated economic cost of about $37 billion and 1.1 million functional years lost. The top three scenarios accounted for about two-thirds of all single light-vehicle crashes. In terms of economic costs and functional years lost, the top three pre-crash scenarios were as follows:
Two-vehicle crashes involving at least one light vehicle resulted in an estimated economic cost of about $69 billion and 1.4 million functional years lost. The top three pre-crash scenarios accounted for about 40 percent of all two-vehicle crashes. In terms of economic costs, the top three scenarios were as follows:
Multi-vehicle light-vehicle crashes (i.e., crashes involving more than two vehicles where at least one is a light vehicle) resulted in an estimated economic cost of about $14 billion and 292,000 functional years lost. The top three pre-crash scenarios accounted for 68 percent of all multi-vehicle crashes and were mostly related to rear-end crashes. In terms of economic costs, the top three scenarios were as follows:
These findings will help establish research priorities and provide a framework for a more consistent approach to identify interventions. The research also consolidates existing crash typologies into a single set of pre-crash scenarios from which all police-reported crashes can be categorized. While the study provided a basis for the current research, it was only based on a single year of data from 2004 and did not associate pre-crash scenarios or crash costs with specific V2I for safety application areas. The study also focused on light vehicles and did not specifically address crashes and costs related to other vehicle types (e.g., heavy vehicles and motorcycles). Crashes related to other vehicle types were included in the analysis but only when a single vehicle was involved. There is still a need to identify crashes and costs related to specific vehicle types.
Similar to Najm et al.'s study on light vehicles, DaSilva et al. developed a new typology of pre-crash scenarios involving single-unit and combination-unit heavy vehicles (gross vehicle weight more than 10,000 lb) based on NASS GES data from 1996 through 2005.(8) A total of 46 pre-crash scenarios were defined by describing the vehicle movements, vehicle dynamics, and critical events that occurred immediately prior to the crash.
Each pre-crash scenario was ranked by three measures: crash frequency, functional years lost, and economic cost. Functional years lost and economic costs were estimated for each pre-crash scenario based on the severity of each individual crash assigned to that scenario. A summary table was provided for each pre-crash scenario, including the number of vehicles and people involved and the distribution of crashes by severity using both KABCO and AIS. A detailed summary was also provided describing the typical scenario for each crash, factors that were overrepresented, dynamic variations of the scenario, and the general severity of the crashes.
Heavy vehicles accounted for approximately 6.5 percent (411,000 crashes) of all police- reported crashes annually. Of those, single-unit trucks accounted for an annual average of 214,000 crashes, and the remaining 197,000 crashes per year were associated with combination-unit trucks. The most common crash types for single-unit and combination-unit trucks were "off-the-roadway" and "changing lanes," respectively. Annually, approximately 974,000 people were involved in a crash with a heavy truck, and approximately 14 percent were injured or killed in those crashes. The annual economic cost of crashes involving heavy trucks, both single-unit and combination-unit, was estimated at $10.4 billion. More than 292,000 functional years were lost annually due to death and injury.
Single-unit truck crashes involving only one vehicle accounted for nearly 29 percent of all single-unit truck crashes, with an economic cost of approximately $877 million and 23,800 functional years lost. The top three pre-crash scenarios in terms of economic costs and functional years lost were as follows:
Two-vehicle single-unit truck crashes accounted for 67 percent of all single-unit truck crashes. These crashes resulted in an annual economic cost of $3.1 billion and 77,100 functional years lost. The top three pre-crash scenarios in terms of economic costs and functional years lost were as follows:
Multi-vehicle (i.e., three or more vehicles) single-unit truck crashes accounted for 4 percent of all single-unit truck crashes. These crashes resulted in an annual economic cost of $458 million and 11,800 functional years lost. The top three pre-crash scenarios in terms of economic costs and functional years lost were as follows:
Single-vehicle combination-unit truck crashes represented 22 percent of all combination- unit truck crashes. These crashes resulted in an annual economic cost of $1 billion and 32,000 functional years lost. The top three pre-crash scenarios in terms of economic costs and functional years lost were as follows:
Two-vehicle combination-unit truck crashes accounted for 73 percent of all combination- unit truck crashes. These crashes resulted in an annual economic cost of $4.3 billion and 125,400 functional years lost. The top three pre-crash scenarios in terms of economic costs and functional years lost were as follows:
Multi-vehicle combination-unit truck crashes represented 5 percent of all combination-unit truck crashes. These crashes resulted in an annual economic cost of $712 million and 21,600 functional years lost. The top three pre-crash scenarios in terms of economic costs and functional years lost were as follows:
These findings will help establish research priorities and provide a framework for a more consistent approach to identify countermeasures to target heavy truck crashes. The analysis identified the magnitude of the problem for 46 different pre-crash scenarios.(5) While the study provided a basis for the current research, it did not associate pre-crash scenarios or crash costs with specific V2I for safety application areas.
Najm et al. conducted a high-level analysis of potential collisions impacted by three general IntelliDrive safety systems.(12) The report employed the crash typologies developed in the Pre-Crash Scenario Typology for Crash Avoidance Research to estimate the potential safety effects of V2V, V2I, and AV communication systems.(7) The study looked at all light vehicles, heavy vehicles, and all vehicles combined. The analyses were based on NASS GES statistics from 2005 through 2008.
Target crashes were measured by the number of police-reported crashes that involved all vehicle types. Target crashes represented the maximum potential safety benefit if the fully deployed system was 100 percent effective in reducing target crashes. To avoid double counting, target crashes were first determined for a primary system category (e.g., V2V), and the remainder of the crash population was later assigned to the other two system categories (e.g., V2I and AV). Several analyses were reported, allowing each system to represent the primary countermeasure.
The study presented the results for each individual system and included tables and figures comparing the total percentage of crashes that could be targeted. Comparing each system individually as the primary countermeasure, V2V had the greatest potential to address crashes (4,409,000 target crashes), representing 74 percent of all police-reported crashes. AV systems represented the second greatest potential to address crashes (3,591,000 target crashes), accounting for 60 percent of all police-reported crashes. V2I was ranked third with respect to the potential to address total crashes (1,465,000 target crashes), representing 25 percent of all police-reported crashes.
The study also included results for the combination of the V2V and V2I systems. This combination accounted for the highest percentage of potential crashes targeted (4,503,000 crashes and 75 percent of all police-reported crashes).
The study also estimated potential benefits based on annual police-reported crashes involving at least one light vehicle and at least one heavy vehicle. Light vehicles include all passenger cars, vans, minivans, sports utility vehicles, and light pickup trucks with a gross vehicle weight rating less than 10,000 lb. Heavy vehicles included pickup, single-unit, and multi-unit trucks with a gross vehicle weight rating greater than 10,000 lb.
Results showed that the combination of V2V and V2I systems targeted the highest percentage of all light-vehicle crashes (77 percent) and crashes involving one heavy vehicle (71 percent). The V2V system alone targeted 76 percent of all light-vehicle crashes when it was considered the primary countermeasure. When considered individually as the primary countermeasure, AV systems targeted 60 percent of all light-vehicle crashes, and V2I systems targeted 25 percent of all light-vehicle crashes. The V2V system targeted 70 percent of all heavy-vehicle crashes when considered the primary countermeasure. When considered individually as the primary countermeasure, AV systems targeted 64 percent of all heavy-vehicle crashes, and V2I systems targeted 14 percent of all heavy-vehicle crashes.
These findings will help establish research priorities with respect to the various systems. While the study compared various systems in general, it did not associate pre-crash scenarios or crash costs with specific V2I for safety application areas.
Chang et al. estimated the magnitude of fatal and injury crashes as well as comprehensive crash costs associated with signalized and stop-controlled intersections.(13) Specifically, the study estimated the potential benefits of Cooperative Intersection Collision Avoidance Systems (CICAS) in terms of comprehensive crash costs. Three CICAS application areas were investigated: violation (CICAS-V), signalized left-turn assist (CICAS-SLTA), and stop sign assist (CICAS-SSA).
National estimates of crashes at signalized and stop-controlled intersections were identified based on data from FARS, NASS GES, and the Crashworthiness Data System.(3,1,14) Crashes were assigned to various crash types for both signalized and stop-controlled intersections corresponding to the CICAS application areas. For signalized intersections, crashes were assigned to violations (crossing path or non-crossing path) and signalized left-turn assist (left-turn across path/opposite direction (LTAP/OD) or left-turn and pedestrian). For stop-controlled intersections, crashes were assigned to violations (crossing path or non-crossing path) and stop sign assist (various crossing path crashes). Crash costs were assigned to each crash by allocating a unit cost to people and vehicles involved in a crash depending on four different categories: fatalities, injured people, non-injured people in injury vehicles, and property damage only (PDO). These costs were assigned based on the level of injury as indicated by MAIS.
In 2000, 43,000 fatalities were reported, with 9,500 occurring at intersections. In the same year, 2.7 million people were injured in vehicle crashes, of which, 1.3 million were related to crashes at intersections. A total of 8.9 million vehicles were involved in PDO crashes, of which, 4 million involved crashes at intersections. Intersection-related crashes and costs were further analyzed to identify the potential benefit of the CICAS application areas. The following results were reported for signalized intersections:
The results for stop-controlled intersections were as follows:
These findings will help establish research priorities with respect to the various CICAS systems. While the study compared specific CICAS systems, it did not identify the potential benefits of other V2I for safety application areas. The results also did not identify contributing factors or specific locations (e.g., urban/rural) where the CICAS applications could be most effective.
A 2000 FHWA study explored the potential to reduce crossing path intersection crashes using advanced technology.(15) The primary objective of this study was to define and evaluate infrastructure-only concepts complementary to AV and V2I cooperative concepts to reduce the number of intersection crashes. High-priority intersection locations were chosen in three States (California, Minnesota, and Virginia), and a crash analysis was conducted at each location to identify the types of crossing path crashes and the potential causes of those crashes.
The three States provided the selection of high-priority intersections and hard copies of crash reports at these intersections. In total, 61 intersection locations were studied (21 locations in California, 20 locations in Minnesota, and 20 locations in Virginia). Crash data covered a 3-year period from 1997 through 1999 in California and from 1998 through 2000 in Minnesota and Virginia.
Crash reports showed that more than 50 percent of all crashes analyzed at intersection locations were crossing path crashes. Results from the crash analysis revealed that of all the crossing path crashes identified, LTAP/OD crashes were the predominant crash type at urban intersections, while straight cross path (SCP) crashes were the dominant crash type at rural intersections. As part of the crash analysis, two main causes of those crashes were identified: traffic control violations and insufficient gap. For insufficient gap crashes, the predominant causes were as follows:
The remaining 5 percent of the insufficient gap acceptance crashes were varied in their causes.
A 2005 FHWA study investigated the magnitude and distribution of crossing path crashes at intersections.(16) A diverse set of data was used to conduct the analyses, including data from two States (Maine and California) and two cities (Detroit, MI, and San Francisco, CA). State-level crash and intersection data were obtained from HSIS. City-level data were obtained from a database developed as part of a previous effort; the data only included signalized intersections.(2)
For the State-level analyses, the study identified the number of signalized and stop-controlled intersections by area type (i.e., rural or urban). The total number of crossing path crashes and crossing path crashes by severity for each intersection type and area type were also identified. In both States, crashes were overrepresented at signalized intersections. While the vast majority of intersections were stop-controlled (98 percent rural and 76 percent urban), the percentage of crashes at stop-controlled intersections was substantially less (61 percent rural and 56 percent urban). Crashes were also more prevalent in urban areas, accounting for 88 to 95 percent of crashes at signalized intersections and 57 to 65 percent of crashes at stop-controlled intersections.
For the city-level analyses, the study identified the number of crashes, injuries, and fatalities for various crossing path crash types, including left-turn, straight, and right-turn crossing paths. Left-turn crossing path was the leading crossing path crash type in both cities, followed by straight crossing path and right-turn crossing path. However, the severity was greatest for straight crossing path crashes.
While this study provided a basis for estimating the potential benefits of specific V2I for safety application areas, namely the CICAS application areas, it did not provide estimates at the national level. This study was also limited in that it only researched crossing path collisions and did not produce crash cost estimates for other scenarios.
Previous research efforts have defined pre-crash scenarios and quantified the relative safety issues in terms of crashes and economic costs. Research studies have explored potential safety impacts of V2I for safety applications. Some studies provided general comparisons of various ITS-related systems (i.e., V2I, V2V, and AV), while others focused on a specific subset of V2I applications (e.g., CICAS). While previous efforts have helped establish general research priority areas and have provided a foundation for subsequent efforts, they were either too general to identify benefits of specific applications or too specific (i.e., only focused on a subset of applications). Many of the studies focused on light vehicles with limited analysis of trucks, motorcycles, and pedestrians.
This study provides a comprehensive evaluation of the V2I for safety applications by the FHWA Safety Program to fill in gaps identified from the review of previous studies. Specifically, this report quantifies the national safety issue with respect to crashes and economic costs and shows the potential safety impacts of currently identified applications as well as the locations (e.g., rural/urban) where those applications could be deployed. It also identifies unaddressed crashes (i.e., those not targeted by current V2I applications). The results are presented in terms of all potential crashes targeted, with details for specific vehicle types and user groups when appropriate.