More research on safety management from a comprehensive, systems approach is desirable. Transportation Safety Management: Start to Finish is a poster session where you can interact one-on-one with the authors to discuss specific aspects of safety management.
Integrated Regional Transportation Model–Network-Based Collision Prediction Model Framework
Ali Farhan, University of CalgaryShow Abstract
Lina Kattan, University of Calgary, Schulich
Richard Tay, RMIT University
Road safety is rarely considered in the transportation planning process. Instead, each effort is typically conducted individually. For decades, transportation planners have used Regional Transportation Models (RTMs) to analyse and evaluate future transportation policies, road and transit network expansion and design options, and land use scenarios. Examples of an RTM’s outputs include future trips by transportation mode, transit ridership, and traffic patterns, volume, speed and congestion indices on road segments. Road safety is conventionally evaluated separately via statistical models that use estimated collision numbers based on historical collision data as dependent variables and that explore a variety of independent explanatory variables. Some explanatory variables are exposure variables that can be extracted from RTM models for base and future horizons, but most current Network-based Collision Prediction Models (NCPMs) are standalone models that do not interact with RTMs. The primary objective of this study is to advance transportation planning and road safety research by developing a NCPM that can be integrated with an RTM as a fifth step of the traditional four-step RTM modeling concept. The integrated RTM-NCPM framework provides estimates on both traffic demand and the number of collisions for base and future planning horizons. The City of Calgary’s RTM model is used as a case study to test various scenarios and to examine the safety implications of changes in transportation policies related to fuel price, parking fees, transit fare, and transit frequency. The results of the scenario analysis clearly show the expected reduction in collision frequency at mid-blocks and intersections upon implementation of policies designed to shift travellers’ mode choices from auto to transit. These collision-reduction policies include both incentives to encourage transit use and disincentives to discourage auto use. This study thus demonstrates how the integrated RTM-NCPM framework can help transportation planners and policy makers to incorporate a safety impact assessment as part of transportation planning process.
Hierarchical Analysis of Traffic Violations and Crashes: A Macroscopic Safety Analysis
Jaeyoung Lee, Central South UniversityShow Abstract
Mohamed Abdel-Aty, University of Central Florida
Xiaoqi Zhai, Central South University
Helai Huang, Central South University
Traffic safety has been one of the most important topics in the transportation field. Most previous safety studies have focused on analyzing traffic crash data. Crash data has been analyzed to determine the level of traffic safety along with possible candidate contributory factors. Another way to examine the safety level is to analyze traffic violations. Traffic violation and crash data of five years (2013-2017) were collected from Montgomery County, Maryland. The numbers of non-crash violations and crashes during the five years are 954,428 and 22,563, respectively. (Approximately 42:1 ratio). Using traffic, commuter, roadway, industry, socio-economic, and demographic data collected from multiple sources, three different modeling frameworks are applied to explore violations and crashes. Each framework has two components: Bayesian Poisson lognormal models for violations and crashes. The first scenario has a hierarchical structure by using the expected number of violations from the first component as the exposure variable of the crash model. The second scenario’s crash model uses the observed number of violations as the exposure. The third scenario’s crash model uses the daily miles-traveled as the exposure. The first modeling scenario shows the best performance, in terms of deviance information criterion (DIC), it is followed by the second scenario, and the third scenario performs the worst. Subsequently, hot zone identification analysis was conducted, and revealed the areas with particular problems with respect to violations and crashes. It is expected that the proposed hierarchical approach will be a useful tool to investigate traffic safety with diverse perspectives.
Development and Application of a Roadway Safety Data Integrator Tool for Highway Safety Information System Data
Seyedehsan Dadvar, CYFOR Technologies LLCShow Abstract
Young-Jae Lee, Morgan State University
Hyeon-Shic Shin, Morgan State University
The Highway Safety Information System (HSIS) is a database that maintains crash data, roadway inventory, and traffic volume data for several US states. It is an excellent source of data to highway safety research and can be used to investigate many research questions. However, to prepare a roadway safety dataset based on the HSIS or any databases that store the data in multiple different subsets and follows linear referencing, the researchers should combine multiple datasets, merge or unmerge and remove certain inconsistent records, and finally clean the dataset. The HSIS staffs are usually accommodating and eager to help, but sometimes the nature of data needs is complicated and laborious. A tool named Roadway Safety Data Integrator (RSDI) was developed for combining, segmentation, and selection of homogeneous HSIS roadway segments and also crash assignment by desired crash fields (e.g., crash severity or type). This study utilized the RSDI to enhance the study on investigation of an alternative calibration method for the Highway Safety Manual (HSM). The results of a preliminary analysis based on sample data from Maryland were validated and complemented by statewide data from Illinois and Washington. The proposed calibration methodology incorporates multiple calibration factors for different components of the HSM predictive method rather than a single calibration factor, as recommended by the HSM that only calibrates at the aggregate level. In the proposed method, the application of calibration factors expressed in both weight and power function reflects better the local conditions while still ensuring calibration at the aggregate level.
Implementing Vision Zero: A Proactive Methodology for Building Communities for Kids
Wesley Marshall, University of Colorado, DenverShow Abstract
Nick Ferenchak, University of New Mexico
With communities across North America taking the Vision Zero pledge, fresh attention and energy is being focused on improving road safety. While the goal of reducing the number of traffic fatalities and severe injuries to zero is an admirable aim, the way in which communities are attempting to reach that goal is often unfocussed and thus far, inadequate. To realize Vision Zero, we need to first accommodate our vulnerable road users; we can start by better designing communities for kids. The proactive assessment approach proposed in this paper provides a method for doing just that. By proactively identifying child pedestrian and bicyclist road safety concerns, this methodology allows for the prioritization of issues before a crash occurs. We first identify roadway characteristics that most suppress children’s walking and biking trips. We then use GIS network analyses to determine which barriers cause the most trip suppression and deserve the most attention. This approach allows us to not only reduce fatalities and injuries where children are currently walking and biking, but to also help ensure safe and comfortable mobility where children want to walk and bike. The same methods can also be applied to school siting scenario analyses. The accompanying GIS tool can be downloaded and applied to any U.S. community that is looking to provide safe, healthy, and equitable active transport for all.
Investigate Factors Affecting Driver Injury Severity in Snow-Related Rural Single-Vehicle Crashes
Runze Yuan, University of HawaiiShow Abstract
Hao Yu, University of Hawai'i, Manoa
Zhenning Li, University of Hawai'i, Manoa
Guohui Zhang, University of Hawai'i, Manoa
David Ma, University of Hawai'i, Manoa
Snow weather is consistently considered as a hazardous factor due to its potential leading to severe fatal crashes. A seven-year crash dataset including all the snow-related rural highway single vehicle crashes from 2010 to 2016 in Washington state is applied in the present study. Pseudo elasticity analysis is conducted to investigate significant impact factors and the temporal stability of model specifications is tested via a likelihood ratio test. The proposed model based on the seven-year dataset is able to capture the individual-specific heterogeneity across crash records for four significant factors, i.e., male, not impaired and no insurance for minor injury, and not impaired for serious injury and fatality. Their estimated parameters were found to be normal distribution instead of fixed value over the observations. Other significant impact factors with fixed effects are: traffic object, animal, overturn, out-of-control, snow surface, smoke surface, sleet surface, curve horizontal design, medium and high speed limits, young and old aged, impaired condition, no belt usage, pickup car type, airbag deployment. The results of temporal stability test show that the model specification is generally not temporally stable for driver injury severity model based on the years of crash data that were used, especially for longer period (more than 3-year dataset). Models that allow the explanatory variables to track temporal heterogeneity, are of great interest and can be explored in future research.
The Problem of, and a Possible Solution to, Comparison Site Selection in Scheme Evaluation
Joe Matthews, Newcastle UniversityShow Abstract
Lee Fawcett, Newcastle University
Neil Thorpe, University of Newcastle
Nicola Hewett, University of Newcastle
Karsten Kremer, PTV Group
Before-and-after studies provide by far the most common method for evaluating the treatment effect of a road safety scheme. The most common among these remain Bayesian methods, which are popular among researchers and practitioners due to their ability to account for the Regression To the Mean (RTM) effect, by using a Safety Performance Function (SPF) built from untreated comparison sites. Failing to accurately estimate the RTM effect immediately leads to a biased estimate of the treatment effect, and so ensuring a well-fitting SPF is vital. It is commonly accepted that an important part of this process is ensuring the comparison sites used to build the SPF are sufficiently similar, or exchangeable, with the treated sites being analysed. Whilst this has been accepted by many authors as intuitively true, no work has been done to numerically demonstrate the consequences of using a non-exchangeable comparison pool in a before-and after study. In this paper we use simulated data to objectively demonstrate that using non-exchangeable comparison sites directly leads to an increase in bias of RTM estimates (and hence of the treatment effect). We investigate methods of comparison site selection using categorical subsetting and propensity score matching (PSM) based methods. We finally demonstrate a new method for making most efficient use of a candidate comparison pool by weighting the SPF according to propensity score similarity, known as propensity score weighted regression (PSWR).
Meso-Level Hot Spot Identification for Suburban Arterials
Xuesong Wang, Tongji UniversityShow Abstract
Yingying Pei, Tongji University
Jinghui Yuan, University of Central Florida
Accurate identification of hotspot as well as the relationship between crashes and the influencing factors contribute to safety improvement on suburban arterials. Micro-level hotspot identification studies treat road segments and intersections as isolated units. It is not consistent with field practices because police department usually identify hotspot based on arterial-level, which consists of multiple segments and intersections. Moreover, dense access density deteriorates traffic safety on both segments and intersections, but the overall safety impact may be underestimated by analyzing segments and intersections separately. In addition, either micro-level or macro-level studies cannot capture the specific impact of road network pattern adjacent to the arterials. This study proposed a novel meso-level approach applying Full Bayesian method (FB) and potential for safety improvement (PSI) to identify hotspots for suburban arterials. In order to reduce the effect of spatial correlation, meso-level analysis units were obtained by combining intersections and their adjacent segments according to the spatial distribution of crashes. Bayesian Poisson-lognormal conditional autoregressive model (PLN-CAR) was selected as prediction model due to its strength in accounting for the spatial correlation among analysis units. The PSI value of each unit was calculated and compared with crash frequency. Results show that 1) meso-level hotspot identification can provide a reasonable reference for police department to improve traffic safety; 2) arterials with more parallel roads and less access density were associated with fewer crashes. The meso-level hotspot identification method proposed in this study are expected to be useful in field application of safety improvement on suburban arterials.
Macro-Level Traffic Safety Analysis and Model Updating in Shanghai, China
Minming Yang, Tongji UniversityShow Abstract
Xuesong Wang, Tongji University
Meigen Xue, Tongji University
Macro-level traffic crash analyses and modeling are prevalent in many countries in order to incorporate traffic safety into long-term transportation planning. Due to the burgeoning urban development and hysteretic nature of data collection, however, many existing studies might be outdated and poorly adaptable. To address the problem, this study updated a macro-level safety model for 263 traffic analysis zones (TAZs) within the urban area of Shanghai. Independent variables for 2009 and 2016 from four categories were investigated to identify specific contributing factors for traffic crashes: socio-economic factors, traffic patterns, road characteristics, and land use features. A Bayesian conditional autoregressive negative binomial (CAR-NB) model was estimated to account for the spatial correlations among TAZs. The 2016 model was developed by using the two-stage Bayesian updating method to provide informative priors for 2009 model. Results show that higher crash frequency is associated with greater population, total length of major and minor arterials, trip frequencies, and shorter intersection spacing. The fact that most variables have similar significance for the two years is indicative of the good flexibility and interpretability of Bayesian CAR-NB model. Additionally, the informative priors are capable of providing theoretically based expectations without losing flexibility. This study helps to fill the gap in formulating informative priors for independent variables in macro-level traffic safety studies. Moreover, urban policy decision makers and traffic police can benefit from this study and implement area-wide engineering, education, and enforcement countermeasures to enhance regional traffic safety.
Record Linkage of Crashes with Injuries and Medical Cost: A Case Study of Puerto Rico
Josie Bianchi, Recinto Universitario de Mayaguez Universidad de Puerto RicoShow Abstract
Didier Valdés, Recinto Universitario de Mayaguez Universidad de Puerto Rico
Héctor Colón, Universidad de Puerto Rico
Cost considerations are critical in the analysis and prevention of traffic crashes. Integration of cost data to crash datasets facilitates the crash-cost analyses with all their related attributes, but also, is a challenging task due to the availability of unique identifiers across the databases, and the privacy and confidentiality regulations. This study performed a record linkage comparison between deterministic and probabilistic approach using attributes matching techniques with numerical distance and weight pattern under the Fellegi-Sunter approach. As a result, the deterministic algorithm developed using the exact match of the 14-digit police accident record number had an overall matching performance of 52.38% of real matched records while the probabilistic algorithm had an overall matching performance of 70.41% with a quality measurement of Sensitivity of 99.99%. The deterministic approach was outperformed by the probabilistic approach by approximately 20% more of records matched. The probabilistic matching with numerical variables seems to be a good matching strategy supported by quality variables. For the proportion of non-matched records, a cost imputation was performed by regressing the personal injury insurance cost data against weekday, time and municipality of a crash, and number of claimants in the personal injury insurance records. After the record matching, a multivariate regression model was developed to estimate and identify the crash circumstances that increase the medical cost of the crash injured claimants in Puerto Rico.
Cross-Comparison and Objective-Based Crash Tree Development and Analysis for Small Counties in Florida
Roozbeh Rahmani, University of FloridaShow Abstract
Nithin Agarwal, University of Florida
Sivaramakrishnan Srinivasan, University of Florida
Ilir Bejleri, University of Florida
Xingjing Xu, University of Florida
Jia Fang, University of Florida
The Federal Highway Administration (FHWA) developed the Systemic Safety Project Selection Tool that lists six steps to integrate existing safety management practices and safety analysis tools. The first step is to identify and understand the risk factors commonly associated with the focus crash types. Crash trees have been adopted by agencies to identify the focus facility types and crash types. For most organizations and departments of transportation (DOTs), the concept behind developing a crash tree is a stepwise elimination process where higher value in the crash tree is retained and the rest of the branch is eliminated. This paper demonstrates some of the challenges with the conclusions of this traditional approach and proposes an alternative structure using a cross-comparison framework that not only compares the raw counts from the crash data but also compares focus county’s crash percentage and ranking to other similar counties or jurisdictions. This approach assists the decision-makers in understanding the intensity of overrepresentation. This study developed a tool that applied the cross-comparison crash tree approach for 27 small rural counties in Florida to determine the percentage and crash severity ranking. The results demonstrated the benefits of this approach by prioritizing the focus areas and the counties by normalizing the counts and the intensity of the overrepresentation.
Improving Driver’s Education Regarding Wrong-Way Driving Incidents
Mohammad Jalayer, Rowan UniversityShow Abstract
Kevin Takacs, Rowan University
Jason Roberts, Rowan University
Wrong-way driving (WWD) occurs when a driver, either inadvertently or deliberately, drives in the opposing direction of traffic along a high-speed, physically divided highway or its access ramp. The nature of WWD crashes, which often tend to be head-on collisions, has drawn the attention of transportation engineers over the past few decades. Several state departments of transportation have adopted three key points of interest—engineering, education, and enforcement—to mitigate this crash type. We note that numerous previous studies focused on engineering and enforcement components; however, in most cases, the education component has been underrepresented. The goal of this study is to contribute to the current research by expanding upon the education and knowledge of drivers regarding WWD incidents. Specifically, a web-based survey was designed and distributed to the Rowan University community to gauge the level of familiarity amongst drivers of different ages and experience. The results of the survey indicate that WWD incidents occur much more frequently than are reported. It is evident from the responses that drivers are much less aware of WWD incidents than other issues such as driving under the influence or distracted driving. Communities throughout the state of New Jersey should consider increasing the number of precautionary messages that highlight the dangers of WWD and consider introducing further education campaigns. These programs should then be evaluated to determine their effectiveness. Overall, these results provide valuable information for policymakers, engineers, and researchers to improve overall road safety by reducing WWD incident frequency.
Emergency Response Times for Fatal Motor Vehicle Crashes, 1975–2017
Maria Cruz, University of New MexicoShow Abstract
Nick Ferenchak, University of New Mexico
Emergency response times are an important component of road safety outcomes. While a safety analysis may identify a decrease in traffic fatalities, that decrease may be a result of improved road safety or it may simply reflect improved emergency response times. However, it is currently unclear how emergency response times have changed over the last few decades. With data from the Fatality Analysis Reporting System (FARS), we identify the national trend in emergency response times from 1975 through 2017. Results suggest that emergency response times have improved by approximately 50% over this timeframe. Findings have important implications for fatality-based traffic safety analyses.
Vehicle Occupants and Driver Behavior: An Assessment of Vulnerable User Groups
Michael Martin, Texas A&M Transportation InstituteShow Abstract
Lisa Green, Texas A&M Transportation Institute
Byron Chigoy, Texas A&M University
Eva Shipp, Texas A&M Transportation Institute
Rahul Mars, Texas A&M University
The question of whether driver behavior, and speeding in particular, differs based on vehicle occupancy requires the use of large amounts of data—some of which may be difficult to accurately obtain. Traditional methods of obtaining information on driver behavior either lack passenger information altogether (i.e., insurance companies using telematics) or rely on rough estimates of passenger age and gender obtained from blurred photos (i.e., naturalistic driving studies like State Highway Research Program 2 (SHRP 2)). This research project represents a novel, data-driven approach to this topic. Household travel survey demographic information and global positioning system (GPS) traces were linked to HERE network speed limit to study the impact of vehicle occupancy on speeding. Survey responses from 11 study areas were cleaned, merged, and ultimately used in developing binomial logistic regression models. Of particular interest were the vulnerable user groups of teenagers, adults driving with children, and seniors. The models suggest that drivers speed less when there is a passenger in the vehicle, especially adults with a child passenger.
Investigating Factors That Contributed to the Large Reduction and Subsequent Increase in Roadway Fatalities in the United States Between 2005 and 2016
Tahmida Hossain Shimu, HDRShow Abstract
Dominique Lord, Texas A&M University
Srinivas Geedipally, Texas A&M Transportation Institute
Lingtao Wu, Texas A&M University
Robert Wunderlich, Texas A&M Transportation Institute
The substantial decline in motor-vehicle fatal crashes over the period of 2008 to 2011 and a subsequent increase afterwards in the United States has been subjected to extensive research in the last few years. Following the perceptible reduction in traffic fatalities beginning in 2008, which concurred with a major recession, researchers focused on finding the relative influence of the recession on fatalities using statistical modeling. The Project 17-67 by the National Cooperative Highway Research Program (NCHRP) conducted an in-depth investigation, where the researchers developed two Poisson-gamma regression models, Model Controlling State (MCS) effect and Model Not Controlling State (MNCS) effect to analyze the factors associated with the decline in fatalities. This study sought to serve as an extension of the NCHRP Project 17-67 to provide a thorough investigation of the factors influencing fatalities during and after the 2008 recession using an updated dataset to 2016. The modeling results showed remarkable improvements, where both the MNCS and MCS models could reflect the fluctuations in fatalities over the focus period. The effect analysis revealed that the economic factors contribute as much as 84% to 86% in the reduction and subsequent increase in fatalities during and after the recession. The unemployment rate of 16 to 24 years old, median household income, and the price of gasoline were found to be the most statistically significant parameters in both the models. Changes in vehicle-miles traveled (VMT), government expenditure, and regulatory measures were not significant factors in affecting the number of fatalities over the analysis period.
Relationship Between Road Safety Pillars and the WHO Member States Mortality Rate: A Study Applying Structural Equation Models
Caio Torres, Universidade Federal do CearaShow Abstract
Xavier Vanessa, Universidade Federal do Ceara
Flávio José Cunto, Universidade Federal do Ceará
Road deaths phenomenon suggests the development of studies that consider the complex causal relationship between the factors that influence road traffic mortality at the compatible level with the definition of road safety policies. This paper analyzes the influence of 48 road safety performance indicators on 175 WHO Member States mortality rate. Structural equation models were proposed to evaluate the proposition and use of latent variables that represent five major road safety policy areas and their influence on mortality rates. The proposed model structure indicated that management has a strategic role in public policies, having an indirect influence on reducing the mortality rate through safe vehicles, user safety and safe road and mobility. The results indicated that policies aimed at encouraging safe user behavior were the ones that had the greatest influence in reducing road deaths followed by policies in safer vehicles, road safety management, safer road and mobility and post-crash response.
School Bus Routing to Allow Later School Start Times
Rana Eslamifard, University of Massachusetts, AmherstShow Abstract
Eric Gonzales, University of Massachusetts, Amherst
School districts providing busing services for students who live too far to walk to school. In many districts a fleet of school buses is used in sequence to transport high school students, then middle school students, and then elementary school students. The result is that high school classes must start much earlier in the morning than the elementary school, and buses may traverse similar routes three times each morning and afternoon. In light of recent research on the benefits of later high school start times and the need to control transportation costs, school districts are seeking efficient school bus routing plans that meet student needs at low cost. This study uses 2018 data for schools in Northampton, Massachusetts, to identify the potential to achieve two objectives: 1) start the high school classes as late as possible in the day, and 2) minimize the cost of busing. The proposed procedure makes use of existing school bus data to optimize bus routes, which can be applicable for smaller cities. A revised routing plan that mixes high school and middle school students on the same buses allows the high school to start 45 minutes later while reducing total school bus operations by 8.5 hours per day. The elementary school and high school start times could also be swapped with minimal effect on the cost of busing.
Assessing the Accuracy of “Serious Injury” Reporting with the Implementation of the New MMUCC KABCO Definition
Beau Burdett, University of Wisconsin, MadisonShow Abstract
Richard Li, University of Louisville
Andrea Bill, University of Wisconsin, Madison
David Noyce, University of Wisconsin, Madison
Across the United States large discrepancies have been found between law enforcement officer’s (LEOs) injury severity assessments and medically assessed health outcomes of crash victims. To better monitor traffic safety serious injury reporting is now federally mandated, making accurate injury severities more important. New federal KABCO injury severity definitions introduced to standardize and add clarity may help reduce inaccuracies in LEO assessments. Wisconsin implemented the new definitions January 1, 2017. Linked crash and medical data from 2009 through 2016 was compared with data from 2017 using the new definitions to determine impacts on injury severity accuracy. Large differences were evident between injuries assessed ‘A’ and ‘B’ or ‘C’ suggesting LEOs are able to differentiate between more serious injuries and less severe injuries. However, despite this difference, approximately two-thirds of crash victim’s injury severities were overestimated (assessed more severely than actual health outcomes) from 2009 through 2017. Underestimation of injury severity decreased from 3.5% to 2.5% after the KABCO definition changes. Furthermore, injuries assessed as minor by medical professionals were less often considered “serious injuries” by LEOs. LEO’s assessment of body regions with more superficial injuries, such as the face, improved. Assessments of body regions with more internal, occult injuries, such as the thorax and abdomen also improved. More accurate assessments may be due to the added clarity of the new definitions. Despite continuing issues, the definition change does suggest that injury severity assessments have improved, which in turn may lead to more accurate traffic safety data.
Road Safety Focusing Events
Ryan Archibald, University of Colorado, DenverShow Abstract
Wesley Marshall, University of Colorado, Denver
Focusing events are a concept developed and studied in the policy science field. They are described as rare events that reveal a problem to both the public and the government. For example, the events of 9/11 were a rare event that revealed security problems and resulted in policy changes. Focusing events in the road safety field are not as widely studied, nor does a method exist to determine them. In this paper, we develop a definition of a road safety focusing event derived from existing literature from the policy science field. Once defined, we propose a method to determine if a road safety incident is a focusing event or not. The method is based upon an investigation of whether or not the public and the policymakers/government were not only aware of, for instance a child pedestrian fatality, but that they were aware of an infrastructure or policy problem (e.g. no sidewalks in a neighborhood). We test the method on a subset of 2014 child pedestrian fatality data and identify six focusing events and four potentially non-focusing events. This research is important since studies on road safety focusing events are lacking. In addition, solutions that work for the United States to improve road safety are needed. Research into road safety focusing events might reveal what has and has not worked for communities. To get there, we must first develop a framework for studying road safety focusing events, which begins with defining and identifying them.
A Deep Reinforcement Learning-Based Intelligent Intervention Planning Framework for Real-Time Proactive Road Safety Management
Ananya Roy, Tokyo Institute of TechnologyShow Abstract
Yasunori Muromachi, Tokyo Institute of Technology
Moinul Hossain, Islamic University of Technology
This goal of this study is improving safety of urban expressways. A dynamic Bayesian network based real-time crash prediction model (RTCPM) and deep Q-network-based intervention is proposed to achieve the goal. Both models are highly dependent on high quality and high density traffic and crash data. These data are collected from the detectors or sensors installed on road networks. For this thesis, route 3 Shibuya (11.9 km) and route 4 Shinjuku (13.5 km) - two radial routes of Tokyo metropolitan expressway were chosen because of the availability of one minute resolution traffic data from 250 meters (approximately) spaced detectors. To make data collection flexible, uniformly spaced (150 meters) detectors (cells) were generated using a macroscopic model called cell transmission model (CTM). The CTM was then modified to incorporate the variable speed limit control in it. The CTM generated traffic data and DBN-based RTCP were fed into the DQN-based intervention model. The DQN-based intervention model then selects several VSL controls in situations when crash risk was detected to be more than or equal to 10. After several iterations the intervention model was able to learn the optimum VSL values take for related hazardous situations to reduce crash risk at the target location by about 19%.
Are Uninsured Drivers Less Likely to Request Emergency Medical Services After a Crash?
Qifan Nie, University of AlabamaShow Abstract
Xing Fu, University of Alabama
Xiaobing Li, University of Alabama
Jun Liu, University of Alabama
Emergency Medical Services (EMS) are found to be effective in dealing with injuries caused by traffic crashes, especially for severe crashes. However, some people may tend to not call the EMS after crash due to various reasons. According to the Alabama crash reports in 2017, over 10% of injured drivers declined to contact EMS. Auto insurance status may be factor than influences the decision-making of EMS requests. The objective of this study is to uncover correlates of EMS requests with a focus on the role of auto insurance. This study performed a comparative analysis for crashes that caused different damages. Considering the unobserved heterogeneity, this study employed a random-parameter modeling approach to disentangle the relationships between factors. Unexpectedly, the model results indicated that after minor vehicle damage crashes, drivers with valid auto insurance seem to be less likely to request EMS; while for crashes that caused major vehicle damages, there is no significant correlation between EMS requests and auto insurance status. Other important factors such as vehicle registration status, driver license status, driving under impairment, seat belt use were also found to be related to EMS requests. The results offer researchers and traffic incident responders an understanding of relationship between EMS requests and crash related factors, and consequently help develop effective strategies to reduce fatality rate of crashes. In addition, medical facilities may cooperate with auto insurance companies to develop an insurance-based response system in responding to uninsured crashes with diversified plans.
A Linear Poisson Autoregressive Model for Analyzing Dynamic Fatal Traffic Accident Data
Yue Zhang, Tongji UniversityShow Abstract
Yajie Zou, Tongji University
Lingtao Wu, Texas A&M Transportation Institute
Annual fatal traffic accident data often demonstrates characteristics of time series . The existing traffic safety analysis approaches (e.g., Negative Binomial (NB) model) often cannot accommodate the dynamic feature in fatal traffic accident data and may result in biased parameter estimation results. In order to consider the time series characteristics of the count traffic accident data, a linear Poisson autoregressive (PAR) model is proposed in this study. The objective of this study is to apply the PAR model to analyze the dynamic impact of traffic laws on the frequency of fatal traffic accident occurred from 1975 to 2016 in Illinois. In addition to the PAR model, the NB model and the autoregressive integer moving average (ARIMA) model are also developed and their performance and impact multipliers are compared. The important conclusions from the modeling results can be summarized as follows: (1) The PAR model outperforms the NB and ARIMA models in terms of analyzing the dynamic influences and fitting performance. The PAR model is more suitable for analyzing the dynamic impact of traffic laws on annual fatal traffic accidents, especially the instantaneous impacts. (2) The law allowing red running leads to an increase in the frequency of annual fatal traffic accidents in both the short and long term. Thus, the modeling results suggest that the PAR model is more suitable for annual fatal traffic accident data and has an advantage in estimating the dynamic impact of traffic laws.
Systemic Strategy to Mitigate Intersection Left Turn Crashes: A Regional Analysis Methodology
Margaret Herrera, Maricopa Association of GovernmentsShow Abstract
This paper presents a crash analysis methodology utilized by the Maricopa Association of Governments (MAG) to identify intersections with the best potential of benefitting from safety improvements related to creating positive offset. The Phoenix metropolitan planning area consists of 27 cities and towns and three (3) Native Nation communities with a population over five million. MAG has conducted over 70 Road Safety Assessments (RSAs) at intersections across the region. A detailed review of observations and recommendations of RSAs conducted where the left-turn crashes were overrepresented showed 26 percent of the observations related to the lack of sufficient sight distance noted as a potential causal factor. It is widely accepted that “a positive offset of left turn lanes” improves sight distance and would help mitigate this crash risk. Although a larger percentage of recommendations for mitigating left-turn crashes related to modifications to left-turn signal phasing, it was determined that the analysis should focus on the countermeasure related to sight visibility for a more meaningful screening and analysis to address a greater need in the MAG region. MAG identified a project to study a Systemic Strategy to Mitigate Intersection Left-Turn Crashes. The methodology outlined in this paper was used to screen all signalized intersections in the region. The study resulted in 1) a sample of intersections in the MAG region that would provide meaningful analysis of the left-turn crash problem as it relates to lack of sight visibility, and 2) a project assessment which included development of HSIP funding applications.
The 85 Percent Solution: A Historical Look at Crowdsourcing Speed Limits and the Question of Safety
Brian Taylor, University of California, Los AngelesShow Abstract
Yu Hong Hwang, University of California, Los Angeles
The “85th percentile rule” is commonly used to set speed limits in jurisdictions across the U.S. Modern interpretations of the rule are that it satisfies key conditions needed for safe roadways: it sets speed limits deemed reasonable to the typical, prudent driver, reduces the problematic variance in travel speeds among vehicles, and allows law enforcement to focus on speeding outliers. Authoritative publications regularly assert that the rule came about because early driving surveys often found that drivers moving at or below the 85th percentile of a speeds on a given roadway were about one standard deviation above the mean speed for that roadway and were “in the low involvement group for traffic incidents” (Research Triangle Institute, 13). This conventional wisdom about the 85th percentile rule is increasingly called into question today by both safety advocates and promoters of more “complete” urban streets. Given this emerging debate, it’s an opportune time to ask where this rule of driver-set speed limits came from and if the rule’s developers’ rationales still hold true today. While most observers trace the rule to safety research and a 1964 report, we find that it actually emerged decades earlier when “traffic service” was a preoccupation of the nascent traffic engineering profession during the first half of the 20th century, and likely a central motivation behind the development of the rule.
DISCLAIMER: All information shared in the TRB Annual Meeting Online Program is subject to change without notice. Changes, if necessary, will be updated in the Online Program and this page is the final authority on schedule information.