Safe-D: Safety through Disruption

Webinar Archive

If you would like to be notified of future Safe-D webinars, please fill out this form.

Recording Title/Project/Date Speaker Webinar Overview
Link (YouTube) Title: Driving Risk Assessment Based on High-frequency, High-resolution Telematics Data

Project: Safe-D VTTI-00-028

Date: August 26, 2021

Dr. Feng Guo
Virginia Tech Transportation Institute

The high-frequency, high-resolution telematics driving data provide valuable information on both long-term driver behavior as well as instantaneous driving conditions. This project uses telematics driving data to modeling driving risk with three primary objectives: 1) characterize the high-frequency kinematic signatures for safety-critical events; 2) modeling driver level crash risk prediction based on kinematics features, and 3) instantaneous crash risk assessment.

The research team proposes a state-of-the-art approach for characterizing the high-frequency kinematic signatures. We developed several features representing driver behavior and underlying driving risk. These features were applied to several large-scale ride-hailing data and naturalistic driving study data to predict driving risk.

The high-frequency kinematic data coupled with the rarity of crashes demand novel modeling approaches. We developed deep learning-based models and variational inference-based rare event modeling to predict crashes from normal driving as well as predict high-risk drivers. A convolutional neural network and long short-term memory network is developed to predict crashes, near-crashes, and normal stopping behaviors. We developed a novel Variational Information for Extremal (VIE) framework for modeling rare events through deep learning models.

This project addressed key methodological challenges in predicting driving risk using high-frequency telematics data. The findings of the project will benefit driving data processing at scale, driver safety management program, and real-time risk prediction.

Link (YouTube) Title: Development of a Connected Smart Vest for Improved Roadside Work Zone Safety

Project: Safe-D 04-104

Date: July 28, 2021

Dr. Nazila Roofigari-Esfahan
Dr. Nazila Roofigari-Esfahan
Virginia Tech

Mike Mollenhauer
Mike Mollenhauer
Virginia Tech Transportation Institute

Jean Paul Talledo Vilela
Virginia Tech Transportation Institute

Roadside work zones (WZs) present imminent safety hazards for roadway workers as well as passing motorists. In 2016, 764 fatalities occurred in work zones in the United States due to motor vehicle traffic crashes. A number of factors (aging highway infrastructure, increased road work, increased levels of traffic, and more nighttime WZs) have led to an increase in WZ crashes in the past few years. Consequently, WZs are becoming increasingly dangerous for workers as well as passing motorists. The standard work zone safety signage and personal protective equipment (PPE) worn by workers at roadside WZs have not been completely effective in controlling work zone crashes.

This project was an effort in addressing this problem through designing a wearable device to accurately localize, monitor, and predict potential collisions between WZ actors based on their movements and activities, and communicate potential collisions to workers, passing drivers, and connected and automated vehicles (CAVs). The project developed a prototype of a wearable worker localization and communication device (i.e., Smart Vest) that utilizes the previously developed Threat Detection Algorithm (Safe-D project 03-050) to communicate workers’ locations to passing CAVs and proactively warn workers and passing motorists of potential collisions. The research successfully completed project objectives and developed a working prototype of a smart vest. The product was evaluated through multiple field experiments, to assess the capability of the vest in accurately localize workers and detect proximity hazards. As a result, this research is expected to significantly improve the safety conditions of roadside WZs through prompt detection and communication of hazardous situations to workers and drivers.

Link (YouTube) Title: Data Mining Twitter to Improve Automated Vehicle Safety

Project: Safe-D 04-098

Date: June 22, 2021

Dr. Tony McDonald
Texas A&M University

Dr. Bert Huang
Tufts University

Automated vehicle (AV) technologies may significantly improve driving safety, but only if they are widely adopted and used appropriately. Adoption and appropriate use are influenced by user expectations, which are increasingly being driven by social media. Prior studies have observed that major news events such as crashes and technology announcements influence user responses to AVs; however, the exact impact and dynamics of this influence are not well understood. This webinar will discuss a novel three-step process that not only measures this impact but also translates it into a set of guidelines for reporting on AV events designed to calibrate driver trust and expectations. The steps include the development of a novel search method to identify AV-relevant user comments on Twitter via a semi-supervised learning approach. A topic modeling and sentiment analysis of the identified tweets to analyze the influence of crashes and news events on user sentiment about AVs, and a mixed-methods analysis to translate these findings into guidelines for AV reporting.
Link (YouTube) Title: Examining Senior Drivers’ Adaptation to Mixed-Level Automated Vehicles

Project: Safe-D 04-103

Date: May 28, 2021

Dr. Jon Antin
Virginia Tech Transportation Institute

Advanced driver assistance systems (ADAS) may benefit senior drivers, reducing driving risk and compensating for diminished cognitive or other abilities. The degree to which such benefits can be realized depends on the facility with which drivers adapt to these features and the extent to which their use enhances mobility and driver performance. Phase I of this study included naturalistic and other data collection, investigating the attitude adaptation of eighteen senior drivers to driving ADAS- equipped vehicles for six weeks each. Participants included 18 men and women aged 70-79. The effect of exposure on driver confidence in and satisfaction with ADAS was measured.

Phase II involved the analysis of the naturalistic driving data collected during Phase I. Analyses included a comparison of mobility and safety-related kinematic data with similar data collected from older drivers in the Second Strategic Highway Research Program Naturalistic Driving Study (SHRP 2 NDS). The research team developed and employed an object detection algorithm to identify trips where Adaptive Cruise Control (ACC) was active. Kinematic data from these trips were compared with data from matched SHRP 2 trips and with data from Phase I trips where ACC was not activated.

Results indicated that ADAS-equipped vehicles may influence seniors’ driving performance both in positive as well as negative ways. Seniors generally displayed better speed management performance while driving the ADAS-equipped vehicles. However, less stable lateral control performance during ACC-use trips was also observed.

Link (YouTube) Title: Use of Disruptive Technologies to Support Safety Analysis and Meet New Federal Requirements

Project: Safe-D 04-113

Date: March 25, 2021

Ioannis (Yianni) Tsapakis
Texas A&M University

Subasish Das
Texas A&M University

Ali Khodadadi
Texas A&M University

States are required to have access to annual average daily traffic (AADT) for all public paved roads, including non-Federal aid system (NFAS) roadways. The expectation is to use AADT estimates in data-driven safety analysis. Because collecting data on NFAS roads is financially difficult, agencies are interested in exploring affordable ways to estimate AADT. The goal of this project was to determine the accuracy of AADT estimates developed from alternative data sources and quantify the impact of AADT on safety analysis. The researchers compared 2017 AADT data provided by the Texas Department of Transportation (DOT) and the Virginia DOT against probe-based AADT estimates supplied by StreetLight Data Inc. Further, the research team developed safety performance functions (SPFs) for Texas and Virginia and performed a sensitivity analysis to determine the effects of AADT on the results obtained from the empirical Bayes (EB) method that uses SPFs. The results showed that the errors stemming from the probe AADT estimates were lower than those reported in a similar study that used 2015 AADT estimates. The sensitivity analysis revealed that the impact of AADT on safety analysis mainly depends on the size of the network, the AADT coefficients, and the overdispersion parameter of the SPFs.
Link (YouTube) Title: Modeling Driver Behavior during Automated Vehicle Platooning Failures

Project: Safe-D 03-036

Date: February 23, 2021

Dr. Tony McDonald
Texas A&M University

Dr. Abhijit Sarkar
Virginia Tech Transportation Institute

The great promises of automated vehicles will only be realized if designers factor in human capabilities, which is challenging given the complexity of human behavior and its interactions with the surrounding driving environment. Driver process models offer a potential solution to this challenge because they can simulate realistic human responses to environmental conditions. These simulations can be used to calibrate design variables (e.g., following distance) to ensure a safety barrier. Despite the fact that driver models are frequently used to assess manual driving safety and advanced safety systems, there have been few attempts to extend these models to automated vehicle control transitions, particularly in platooning environments. The goal of this talk is to discuss a model development process to predict driver behavior following an unexpected prompt from the automated vehicle for a control transition. The discussion will introduce a model development approach that is grounded in contemporary cognitive theory, empirical findings, and actual on-road observations, and show that this process leads to accurate predictions of driver decisions, braking reactions, and steering behavior. The talk will close with a discussion of future directions and applications. After registering, you will receive a confirmation email containing information about joining the meeting.
Link (YouTube) Title: Implication of Truck Platoons for Roadside and Vehicle Safety Hardware

Project: Safe-D Project 01-006

Date: December 11, 2020

Dr. Chiara Dobrovolny
Texas A&M University

Dr. Costin Untaroiu
Virginia Tech

Platooning is an expansion of Cooperative Adaptive Cruise Control (CACC). It involves automated longitudinal and lateral control of vehicles moving in a tight arrangement and short following distances. Platooning, when deployed, is expected to improve fuel efficiency while maintaining occupant and roadside safety. It is unknown whether existing roadside safety devices are adequate to withstand a potential impact from an errant platoon. Additionally, the interactions of impacting trucks with roadside safety devices and the risks they pose to occupants are unknown as well.

This webinar described a strategy created to simulate multiple tractor-trailer impacts into roadside safety concrete barriers. This method was used to understand how existing concrete roadside safety devices will perform under multiple impacts due to errant truck platoons and to examine the associated injury risks for occupants within each tractor.

Finite element (FE) crash models were developed and validated against respective full-scale vehicle crash tests previously conducted at the Texas A&M Transportation Institute. Once validated, the sequential tractor-trailer collisions were simulated by using the outputs of each barrier impact as the inputs for the subsequent impact.

Next, the time history data for each node within the barrier simulations formed inputs for a detailed interior cabin simulation with occupant injury risks measured with validated Hybrid III/THOR dummy models. The Abbreviated Injury Scale (AIS) was used to classify the severity of occupant injury risk based on the simulation results. In addition, a seat position sensitivity study was performed to understand the risk severity for four different occupant seat rotation angles (−15°, 30°, 45° and 60°) to begin investigating the safety impacts of unconventional seating arrangements possibly enabled by automated systems.

Link (YouTube) Title: Analysis of an Incentive-based Smart Phone App for Young Drivers

Project: Safe-D Project TTI-01-01

Date: February 25, 2020

Russell Henk
Texas A&M Transportation Institute
Traffic crashes continue to be the leading cause of unintentional death and injury of youth across the United States. New and innovative interventions continue to be developed to address this public health issue for this high-risk driving population. This webinar will provide a variety of data associated with an incentive-based smartphone app developed by the Texas A&M Transportation Institute as part of the peer-to-peer safe driving program, Teens in the Driver Seat®. One of the core features of the app involves a reward system in which drivers earn points for miles driven without any phone interaction. The points can be redeemed for rewards and as a basis for competitions and achievement of safe driving levels. This project examined data collected from two distinct smartphone app deployments over the timespan of several months each – one in 2017 and one in 2018. The datasets include over 12,200 trips and more than 100,000 miles logged using the app. Statistical analyses were performed to assess the influence of incentives on the frequency of distracted driving. Statistically significant reductions in distracted driving (at the 95 percent confidence level) were shown to have occurred when incentives were awarded for distraction-free driving. Several other data of interest are presented herein as well. Topics discussed will also include lessons learned regarding the pros and cons of smartphone app deployment of this nature.
Link (YouTube) Title: Pavement Perspective on Automated Vehicle Safety Through Optimized Lateral Positioning Pattern

Project: Safe-D Project 02-008

Date: February 18, 2020

Fujie Zhou
Texas A&M Transportation Institute
Deployment of automated vehicles (AVs) has many benefits, such as reduction of congestion and traffic accidents, increased lane capacity, lower fuel consumption, increased transport accessibility, and reduced travel time and transportation costs. However, one aspect of AVs that is worthy of notable attention is their impact on risk of roadway hydroplaning and pavement life. Since most AVs are programmed to follow a set path and maintain a lateral position in the center of the lane, over time, significant pavement rutting will occur. This study directly measured AV lateral wandering patterns and then compared with human driven vehicles. It was found that wandering patterns of both AVs and human driven vehicles could be modeled with a normal distribution but have significantly different standard deviations. AVs wander laterally at least 3 times smaller than regular human driven vehicles. The influence of AVs (all AVs or mixed with regular human driven vehicles) on pavement rutting and fatigue life was analyzed with the Texas Mechanistic-Empirical Flexible Pavement Design System (TxME). The researchers discovered that the AVs with smaller lateral wandering would shorten pavement fatigue life by 22 percent. Meanwhile, pavement rut depth increases by 30 percent which leads to a much higher risk for AVs hydroplaning. The researchers also calculated the maximum tolerable rut depths at different hydroplaning speeds. It is noted that the AVs have a much smaller tolerable rut depth than the regular human driven vehicles due to greater water film thickness in the rutted wheel paths. To reduce the negative impact of AVs on roadway safety and pavement life, this research recommends an optimal AV wandering pattern: a uniform distribution. Not only does the uniform distribution eliminate the negative effect of AVs, but more importantly, it results in prolonged pavement life and decreased hydroplaning potential.
Link (YouTube) Title: Street Noise Relationship to Vulnerable Road User Safety

Project: Safe-D Project 02-027

Date: February 12, 2020

Greg Griffin
Texas A&M Transportation Institute
This webinar shares results of recent research that related bicycle crash rates to noise levels – measured from the bicycle handlebar in two cities. The study developed a method for evaluating street noise and documented crash rates for roadways in Austin, Texas, and Washington, D.C., in a manner that is replicable by researchers and practitioners. Researchers collected street-level noise in both cities over a range of locations, facility types, and times, and compared these against crash records, normalized by bicycle volumes, and other explanatory variables. Modeling explained 87% of the variation in crash risk in our Washington, DC Capital Area route, after controlling for infrastructure differences and nearby bicycle commute mode shares. Further explorations of street noise are needed to improve further guidance for transportation planning and design.
Link (YouTube) Title: Model Selection Heuristics based on Characteristics of Data & Rare Events Modeling

Project: Safe-D Project 01-001

Date: February 5, 2020

Ali Shirazi
Texas A&M University

Feng Guo
Virginia Tech Transportation Institute

Part 1: Model Selection Heuristics Based on Characteristics of Data. Transportation analysts usually employ post-modeling methods, such as Goodness-of-Fit statistics or Likelihood-based Ratio Tests for selecting the best distribution or model. These metrics require all competitive distributions or models to be fitted to the data before any comparisons can be accomplished. Given the continuous growth in introducing new statistical distributions, choosing the best one using such post-modeling methods is not a trivial task, especially given all theoretical or numerical issues the analyst may face during the analysis. Furthermore, and most importantly, these measures or tests do not provide any intuitions about why a specific distribution or model is preferred over another (Goodness-of-Logic). This presentation describes a methodology to design heuristics for Model Selection based on the characteristics of data, in terms of descriptive summary statistics, before the competitive models are fitted. The proposed methodology employs two analytic tools: (1) Monte-Carlo Simulations and (2) Machine Learning Classifiers, to design simple heuristics to predict the label of the ‘most-likely-true’ distribution for analyzing data.

Part 2: Rare Event Modeling. The rare event nature of crashes brings challenges in crash modeling and prediction. This study focuses on the following two aspects: 1) propose bias adjustment for more accurate estimation of the safety impact of a risk factor; 2) develop a decision-adjusted modeling framework to predict high risk drivers based on telematics data. The decision-adjusted framework optimizes predictive performance based on the objective of the study, e.g., top 1% of high risk drivers. In a case study, we developed an optimal driver level risk prediction model based on the telematics data (high G-force events) and driver demographic information using the SHRP2 NDS.

Link (YouTube) Title: Older Drivers and Transportation Network Companies: Investigating Opportunities for Increased Safety and Improved Mobility

Project: Safe-D Project 02-016

Date: January 23, 2020

Melissa Tooley
Texas A&M Transportation Institute
Transportation network companies (TNCs) such as Uber and Lyft offer an increasingly popular alternative to driving a personal vehicle. This project investigated the potential of TNCs to increase the safety and enhance the mobility of older adults who are experiencing a decline in driving ability. Interviews with commercial and non-profit transportation providers and focus groups of adults ranging from age 65 to over 85 identified attitudes and perceptions toward TNCs and related services targeting senior adults, as well as ongoing barriers to TNC use by this demographic. Barriers include insufficient familiarity and comfort with using smartphone applications, a lack of knowledge among older adults about how TNCs operate, and lack of availability of TNC services in many rural areas. Increased availability of TNC services targeted toward older adults may help to overcome some of these barriers. The project team developed outreach and education materials for older adults on how to access and use TNC services.