Browsing by Author "Guo, Feng"

Now showing 1 - 20 of 74
  • Thumbnail Image
    Analysis of the Use of Probe Vehicles for Road Infrastructure Data Analysis
    Valeri, Stephen M. (Virginia Tech, 2012-06-01)
    This thesis explores the concept of using sensors found in normal vehicles, also known as probe vehicles, to collect road infrastructure data. This concept was demonstrated by measuring vertical acceleration using in-vehicle sensors in order to describe road ride quality. Data collection was performed at the Virginia Smart Road using two instrumented vehicles. The gathered information was compared to road profile data collection, which is the current state-of-the-practice in ride quality assessment. Following the concept validation, the acceleration measurements were further analyzed for repeatability and effect of various independent variables (vehicle speed and type). A network-level simulation was completed using the robust set of measurements from the experiment. In addition, methodology for identifying rough sections and locations were established. Results show that under controlled testing conditions, roadway profile can accurately be estimated using probe vehicle acceleration data and may provide a more practical way to measure road smoothness. The analysis also showed that vertical acceleration data from a fleet of probe vehicles can successfully identify poorly-conditioned pavement areas. This suggests that instrumented probe vehicles might be a viable and effective way of implementing a network level roadway health monitoring program in the near future.
  • Thumbnail Image
    Analyzing Highway Safety Datasets: Simplifying Statistical Analyses from Sparse to Big Data
    Lord, Dominique; Geedipally, Srinivas Reddy; Guo, Feng; Jahangiri, Arash; Shirazi, Mohammadali; Mao, Huiying; Deng, Xinwei (SAFE-D: Safety Through Disruption National University Transportation Center, 2019-07)
    Data used for safety analyses have characteristics that are not found in other disciplines. In this research, we examine three characteristics that can negatively influence the outcome of these safety analyses: (1) crash data with many zero observations; (2) the rare occurrence of crash events (not necessarily related to many zero observations); and (3) big datasets. These characteristics can lead to biased results if inappropriate analysis tools are used. The objectives of this study are to simplify the analysis of highway safety data and develop guidelines and analysis tools for handling these unique characteristics. The research provides guidelines on when to aggregate data over time and space to reduce the number of zero observations; uses heuristics for selecting statistical models; proposes a bias adjustment method for improving the estimation of risk factors; develops a decision-adjusted modeling framework for predicting risk; and shows how cluster analyses can be used to extract relevant information from big data. The guidelines and tools were developed using simulation and observed datasets. Examples are provided to illustrate the guidelines and tools.
  • Thumbnail Image
    Application of Naturalistic Truck Driving Data to Analyze and Improve Car Following Models
    Higgs, Bryan James (Virginia Tech, 2011-12-02)
    This research effort aims to compare car-following models when the models are calibrated to individual drivers with the naturalistic data. The models used are the GHR, Gipps, Intelligent Driver, Velocity Difference, Wiedemann, and the Fritzsche model. This research effort also analyzes the Wiedemann car-following model using car-following periods that occur at different speeds. The Wiedemann car-following model uses thresholds to define the different regimes in car following. Some of these thresholds use a speed parameter, but others rely solely upon the difference in speed between the subject vehicle and the lead vehicle. This research effort also reconstructs the Wiedemann car-following model for truck driver behavior using the Naturalistic Truck Driving Study's (NTDS) conducted by Virginia Tech Transportation Institute. This Naturalistic data was collected by equipping 9 trucks with various sensors and a data acquisition system. This research effort also combines the Wiedemann car-following model with the GHR car-following model for trucks using The Naturalistic Truck Driving Study's (NTDS) data.
  • Thumbnail Image
    Applications of Event Data Recorder Derived Crash Severity Metrics to Injury Prevention
    Dean, Morgan Elizabeth (Virginia Tech, 2023-05-25)
    Since 2015, there have been more than 35,000 fatalities annually due to crashes on United States roads [1], [2]. Typically, road departure crashes account for less than 10% of all annual crash occupants yet comprise nearly one third of all crash fatalities in the US [3]. In the year 2020, road departure crashes accounted for 50% of crash fatalities [2]. Road departure crashes are characterized by a vehicle leaving the intended lane of travel, departing the roadway, and striking a roadside object, such as a tree or pole, or roadside condition, such as a slope or body of water. One strategy currently implemented to mitigate these types of crashes is the use of roadside barriers. Roadside barriers, such as metal guardrails, concrete barriers, and cable barriers, are designed to reduce the severity of road departure crashes by acting as a shield between the departed vehicle and more hazardous roadside obstacles. Much like new vehicles undergo regulatory crash tests, barriers must adhere to a set of crash test procedures to ensure the barriers perform as intended. Currently, the procedures for full-scale roadside barrier crash tests used to evaluate the crash performance of roadside safety hardware are outlined in The Manual for Assessing Safety Hardware (MASH) [4]. During roadside barrier tests, the assessment of occupant injury risk is crucial, as the purpose of the hardware is to prevent the vehicle from colliding with a more detrimental roadside object, all the while minimizing, and not posing additional, risk to the occupants. Unlike the new vehicle regulatory crash tests conducted by the National Highway Traffic Safety Administration (NHTSA), MASH does not require the use of instrumented anthropomorphic test devices (ATD). Instead, one of the prescribed occupant risk assessment methods in MASH is the flail space model (FSM), which was introduced in 1981 and models an occupant as an unrestrained point mass. The FSM is comprised of two crash severity metrics that can be calculated using acceleration data from the test vehicle. Each metric is prescribed a maximum threshold in MASH and if either threshold is exceeded during a crash test the test fails due to high occupant injury risk. Since the inception of the FSM metrics and their thresholds, the injury prediction capabilities of these metrics have only been re-investigated in the frontal crash mode, despite MASH prescribing an oblique 25-degree impact angle for passenger vehicle barrier tests. The focus of this dissertation was to use EDR data from real-world crashes to assess the current relevance of roadside barrier crash test occupant risk assessment methods to the modern vehicle fleet and occupant population. Injury risk prediction models were constructed for the two FSM-based metrics and five additional crash severity metrics for three crash modes: frontal, side, and oblique. For each crash mode and metric combination, four injury prediction models were constructed: one to predict probability of injury to any region of the body and three to predict probability of injury to the head/face, neck, and thorax regions. While the direct application of these models is to inform future revisions of MASH crash test procedures, the developed models have valuable applications for other areas of transportation safety besides just roadside safety. The final two chapters of this dissertation explore these additional applications: 1) assessing the injury mitigation effectiveness of an advanced automatic emergency braking system, and 2) informing speed limit selection that supports the safe system approach. The findings in this dissertation indicate that both the FSM and additional crash severity metrics do a reasonable job predicting occupant injury risk in oblique crashes. One of the additional metrics performs better than the two FSM metrics. Additionally, several occupant factors, such as belt status and age, play significant roles in occupant risk prediction. These findings have important implications for future revisions of MASH, which could benefit from considering additional metrics and occupant factors in the occupant risk assessment procedures.
  • Thumbnail Image
    Appling Machine and Statistical Learning Techniques to Intelligent Transport Systems: Bottleneck Identification and Prediction, Dynamic Travel Time Prediction, Driver Run-Stop Behavior Modeling, and Autonomous Vehicle Control at Intersections
    Elhenawy, Mohammed Mamdouh Zakaria (Virginia Tech, 2015-06-30)
    In this dissertation, new algorithms that address three traffic problems of major importance are developed. First automatic identification and prediction algorithms are developed to identify and predict the occurrence of traffic congestion. The identification algorithms concoct a model to identify speed thresholds by exploiting historical spatiotemporal speed matrices. We employ the speed model to define a cutoff speed separating free-flow from congested traffic. We further enhance our algorithm by utilizing weather and visibility data. To our knowledge, we are the first to include weather and visibility variables in formulating an automatic congestion identification model. We also approach the congestion prediction problem by adopting an algorithm which employs Adaptive Boosting machine learning classifiers again something novel that has not been done previously. The algorithm is promising where it resulted in a true positive rate slightly higher than 0.99 and false positive rate less than 0.001. We next address the issue of travel time modeling. We propose algorithms to model travel time using various machine learning and statistical learning techniques. We obtain travel time models by employing the historical spatiotemporal speed matrices in conjunction with our algorithms. The algorithms yield pertinent information regarding travel time reliability and prediction of travel times. Our proposed algorithms give better predictions compared to the state of practice algorithms. Finally we consider driver safety at signalized intersections and uncontrolled intersections in a connected vehicles environment. For signalized intersections, we exploit datasets collected from four controlled experiments to model the stop-run behavior of the driver at the onset of the yellow indicator for various roadway surface conditions and multiple vehicle types. We further propose a new variable (predictor) related to driver aggressiveness which we estimate by monitoring how drivers respond to yellow indications. The performance of the stop-run models shows improvements after adding the new aggressiveness predictor. The proposed models are practical and easy to implement in advanced driver assistance systems. For uncontrolled intersections, we present a game theory based algorithm that models the intersection as a chicken game to solve the conflicts between vehicles crossing the intersection. The simulation results show a 49% saving in travel time on average relative to a stop control when the vehicles obey the Nash equilibrium of the game.
  • Thumbnail Image
    Bayesian Approach Dealing with Mixture Model Problems
    Zhang, Huaiye (Virginia Tech, 2012-04-23)
    In this dissertation, we focus on two research topics related to mixture models. The first topic is Adaptive Rejection Metropolis Simulated Annealing for Detecting Global Maximum Regions, and the second topic is Bayesian Model Selection for Nonlinear Mixed Effects Model. In the first topic, we consider a finite mixture model, which is used to fit the data from heterogeneous populations for many applications. An Expectation Maximization (EM) algorithm and Markov Chain Monte Carlo (MCMC) are two popular methods to estimate parameters in a finite mixture model. However, both of the methods may converge to local maximum regions rather than the global maximum when multiple local maxima exist. In this dissertation, we propose a new approach, Adaptive Rejection Metropolis Simulated Annealing (ARMS annealing), to improve the EM algorithm and MCMC methods. Combining simulated annealing (SA) and adaptive rejection metropolis sampling (ARMS), ARMS annealing generate a set of proper starting points which help to reach all possible modes. ARMS uses a piecewise linear envelope function for a proposal distribution. Under the SA framework, we start with a set of proposal distributions, which are constructed by ARMS, and this method finds a set of proper starting points, which help to detect separate modes. We refer to this approach as ARMS annealing. By combining together ARMS annealing with the EM algorithm and with the Bayesian approach, respectively, we have proposed two approaches: an EM ARMS annealing algorithm and a Bayesian ARMS annealing approach. EM ARMS annealing implement the EM algorithm by using a set of starting points proposed by ARMS annealing. ARMS annealing also helps MCMC approaches determine starting points. Both approaches capture the global maximum region and estimate the parameters accurately. An illustrative example uses a survey data on the number of charitable donations. The second topic is related to the nonlinear mixed effects model (NLME). Typically a parametric NLME model requires strong assumptions which make the model less flexible and often are not satisfied in real applications. To allow the NLME model to have more flexible assumptions, we present three semiparametric Bayesian NLME models, constructed with Dirichlet process (DP) priors. Dirichlet process models often refer to an infinite mixture model. We propose a unified approach, the penalized posterior Bayes factor, for the purpose of model comparison. Using simulation studies, we compare the performance of two of the three semiparametric hierarchical Bayesian approaches with that of the parametric Bayesian approach. Simulation results suggest that our penalized posterior Bayes factor is a robust method for comparing hierarchical parametric and semiparametric models. An application to gastric emptying studies is used to demonstrate the advantage of our estimation and evaluation approaches.
  • Thumbnail Image
    Bayesian Multilevel-multiclass Graphical Model
    Lin, Jiali (Virginia Tech, 2019-06-21)
    Gaussian graphical model has been a popular tool to investigate conditional dependency between random variables by estimating sparse precision matrices. Two problems have been discussed. One is to learn multiple Gaussian graphical models at multilevel from unknown classes. Another one is to select Gaussian process in semiparametric multi-kernel machine regression. The first problem is approached by Gaussian graphical model. In this project, I consider learning multiple connected graphs among multilevel variables from unknown classes. I esti- mate the classes of the observations from the mixture distributions by evaluating the Bayes factor and learn the network structures by fitting a novel neighborhood selection algorithm. This approach is able to identify the class membership and to reveal network structures for multilevel variables simultaneously. Unlike most existing methods that solve this problem by frequentist approaches, I assess an alternative to a novel hierarchical Bayesian approach to incorporate prior knowledge. The second problem focuses on the analysis of correlated high-dimensional data which has been useful in many applications. In this work, I consider a problem of detecting signals with a semiparametric regression model which can study the effects of fixed covariates (e.g. clinical variables) and sets of elements (e.g. pathways of genes). I model the unknown high-dimension functions of multi-sets via multi-Gaussian kernel machines to consider the possibility that elements within the same set interact with each other. Hence, my variable selection can be considered as Gaussian process selection. I develop my Gaussian process selection under the Bayesian variable selection framework.
  • Thumbnail Image
    Characterizing Bicyclists Behavior in Overtaking Scenarios Over Different Road Infrastructures
    Crump IV, Eugene Raymond (Virginia Tech, 2024-06-07)
    Fatal vehicle-bicycle crashes have increased in the United States while cyclist crashes often go unreported. The underreporting of all cyclist crashes results in the overall pre-crash behavior of the cyclists being unknown. What is known is that the most fatal bicycle crash scenario occurs when a vehicle performs an overtaking maneuver. It is crucial to find effective strategies to mitigate these crashes. Vision Zero aims to eliminate all traffic fatalities and disabling injuries by the year 2050 through the implementation of the safe system approach. One of their approaches is using active safety systems like bicycle detecting automatic emergency braking. The purpose of this study was to characterize bicyclist behavior to enhance the crash avoidance potential of advanced driver assistance systems and improve safety for cyclists. An analysis on fatal crashes involving bicyclists was conducted to determine scenarios for testing bicyclist-vehicle interactions on roadways using virtual reality (VR). VR testing was conducted to capture and analyze bicyclist dynamics. Most fatal bicycle crashes occurred when motorists overtook cyclists, especially when cyclists are travelling in a travel lane in the same direction as traffic. These crashes often happen in densely populated areas with favorable weather conditions. This information was used to construct scenarios representing common fatal bicycle crash scenarios. From the analysis, four scenarios were developed. The first scenario was an overtaking scenario with the cyclist traveling in the same direction as traffic, in a travel lane without a bicycle lane or shoulder. The second, third and fourth scenarios were variations of the first to include a bike lane, shoulder, and both a bike lane and a shoulder to analyze the behavior difference due to the inclusion of each. Participants were immersed in a VR simulator that used the combination of a VR headset and a custom-built stationary bicycle. Eighteen individuals were recruited with an average age of 22.7 years. Participants experienced all four scenarios, and their speed, glance, lane position, and standard deviation of lane position were collected and analyzed. The speed for each road type and overtaking phase did not vary significantly, with an average of 4.9 m/s. In the case where there was neither a bike lane or a shoulder, the cyclists looked towards the vehicle more than the other scenarios. As for the lane position, the scenario where the cyclist had neither a shoulder or a bike lane, led to a closer vehicle-bicycle relative position than the other three scenarios. As for standard deviation of lane position, the road with neither a shoulder or bike lane had the largest interquartile range (IQR) and average and the road with both a shoulder and bike lane had the smallest IQR. This implies a lower predictability of the cyclist's movements when they are riding on a roadway with no support lane. Following the testing, participants rated the perceived realism and interactiveness of the VR world and their comfort in each road design. Most of the participants mentioned that having some allocated space felt more comfortable and lowered their sense of danger. To enhance cyclist safety, adopting Euro NCAP testing for AEB systems in the US is recommended. This form of testing could lead to improvements in AEB systems, reducing crashes with cyclists and injury severity. In terms of road infrastructure improvements increasing the number of bike lanes, adding wider shoulders, or widening lanes could also enhance cyclist safety on roadways.
  • Thumbnail Image
    Contributions to Large Covariance and Inverse Covariance Matrices Estimation
    Kang, Xiaoning (Virginia Tech, 2016-08-25)
    Estimation of covariance matrix and its inverse is of great importance in multivariate statistics with broad applications such as dimension reduction, portfolio optimization, linear discriminant analysis and gene expression analysis. However, accurate estimation of covariance or inverse covariance matrices is challenging due to the positive definiteness constraint and large number of parameters, especially in the high-dimensional cases. In this thesis, I develop several approaches for estimating large covariance and inverse covariance matrices with different applications. In Chapter 2, I consider an estimation of time-varying covariance matrices in the analysis of multivariate financial data. An order-invariant Cholesky-log-GARCH model is developed for estimating the time-varying covariance matrices based on the modified Cholesky decomposition. This decomposition provides a statistically interpretable parametrization of the covariance matrix. The key idea of the proposed model is to consider an ensemble estimation of covariance matrix based on the multiple permutations of variables. Chapter 3 investigates the sparse estimation of inverse covariance matrix for the highdimensional data. This problem has attracted wide attention, since zero entries in the inverse covariance matrix imply the conditional independence among variables. I propose an orderinvariant sparse estimator based on the modified Cholesky decomposition. The proposed estimator is obtained by assembling a set of estimates from the multiple permutations of variables. Hard thresholding is imposed on the ensemble Cholesky factor to encourage the sparsity in the estimated inverse covariance matrix. The proposed method is able to catch the correct sparse structure of the inverse covariance matrix. Chapter 4 focuses on the sparse estimation of large covariance matrix. Traditional estimation approach is known to perform poorly in the high dimensions. I propose a positive-definite estimator for the covariance matrix using the modified Cholesky decomposition. Such a decomposition provides a exibility to obtain a set of covariance matrix estimates. The proposed method considers an ensemble estimator as the center" of these available estimates with respect to Frobenius norm. The proposed estimator is not only guaranteed to be positive definite, but also able to catch the underlying sparse structure of the true matrix.
  • Thumbnail Image
    Data and methods for studying commercial motor vehicle driver fatigue, highway safety and long-term driver health
    Stern, Hal S.; Blower, Daniel; Cohen, Michael L.; Czeisler, Charles A.; Dinges, David F.; Greenhouse, Joel B.; Guo, Feng; Hanowski, Richard J.; Hartenbaum, Natalie P.; Krueger, Gerald P.; Mallis, Melissa M.; Pain, Richard F.; Rizzo, Matthew; Sinha, Esha; Small, Dylan S.; Stuart, Elizabeth A.; Wegman, David H. (Elsevier, 2019-05)
    This article summarizes the recommendations on data and methodology issues for studying commercial motor vehicle driver fatigue of a National Academies of Sciences, Engineering, and Medicine study. A framework is provided that identifies the various factors affecting driver fatigue and relating driver fatigue to crash risk and long-term driver health. The relevant factors include characteristics of the driver, vehicle, carrier and environment. Limitations of existing data are considered and potential sources of additional data described. Statistical methods that can be used to improve understanding of the relevant relationships from observational data are also described. The recommendations for enhanced data collection and the use of modern statistical methods for causal inference have the potential to enhance our understanding of the relationship of fatigue to highway safety and to long-term driver health.
  • Thumbnail Image
    Decision-adjusted driver risk predictive models using kinematics information
    Mao, Huiying; Guo, Feng; Deng, Xinwei; Doerzaph, Zachary R. (Elsevier, 2021-06)
    Accurate prediction of driving risk is challenging due to the rarity of crashes and individual driver heterogeneity. One promising direction of tackling this challenge is to take advantage of telematics data, increasingly available from connected vehicle technology, to obtain dense risk predictors. In this work, we propose a decision-adjusted framework to develop optimal driver risk prediction models using telematics-based driving behavior information. We apply the proposed framework to identify the optimal threshold values for elevated longitudinal acceleration (ACC), deceleration (DEC), lateral acceleration (LAT), and other model parameters for predicting driver risk. The Second Strategic Highway Research Program (SHRP 2) naturalistic driving data were used with the decision rule of identifying the top 1% to 20% of the riskiest drivers. The results show that the decision-adjusted model improves prediction precision by 6.3% to 26.1% compared to a baseline model using non-telematics predictors. The proposed model is superior to models based on a receiver operating characteristic curve criterion, with 5.3% and 31.8% improvement in prediction precision. The results confirm that the optimal thresholds for ACC, DEC and LAT are sensitive to the decision rules, especially when predicting a small percentage of high-risk drivers. This study demonstrates the value of kinematic driving behavior in crash risk prediction and the necessity for a systematic approach for extracting prediction features. The proposed method can benefit broad applications, including fleet safety management, use-based insurance, driver behavior intervention, as well as connected-vehicle safety technology development.
  • Thumbnail Image
    Does Eyeglance Affect Lane Change Safety: Analysis of Eyeglance Pattern Prior to Lane Change
    Guo, Feng; Han, Shu; Xu, Jingbin (National Surface Transportation Safety Center for Excellence, 2022-09-23)
    The driver’s eyeglance patterns prior to lane change can have a major impact on crash risk. This study focuses on the area-of-interest (AOI) in eyeglances related to lane changes, including rearview mirror, left/right window, left/right mirror, windshield, and over-the-shoulder (OTS) checks of corresponding lane change direction. Key AOI characteristics such as type, percentage, duration, timing, and time-varying properties were examined thoroughly. We also evaluated driver attention on the driving task and how it changed over time by event type using the AttenD algorithm to reconstruct eyeglance data into a continuous variable. The AttenD score incorporates the glance history in the profile to reflect how effectively a driver may be allocating attention and storing information about the roadway and other vehicles. A higher AttenD score indicates more attention on primary driving tasks. Baselines had drivers with significantly higher attention scores and lower variance than near-crashes and crashes. This indicates that drivers who conducted a safe lane change tended to look away from the road less often and were more consistent in allocating eyeglances forward and on the surrounding environment.
  • Thumbnail Image
    Driver Coach Study: Using Real-time and Post Hoc Feedback to Improve Teen Driving Habits
    Klauer, Charlie; Ankem, Gayatri; Guo, Feng; Baynes, Peter; Fang, Youjia; Atkins, Whitney; Baker, Stephanie Ann; Duke, Rebekah; Hankey, Jonathan M.; Dingus, Thomas A. (National Surface Transportation Safety Center for Excellence, 2017-12-08)
    Novice teenage drivers have the highest rates of fatalities and injuries on U.S. roadways compared to any other age group. This experimental research was conducted to see if presenting novice teenage drivers and their parents with feedback on teen driving performance could decrease rates of crash/near-crash (CNC) involvement. Ninety-two newly licensed teens had their vehicles instrumented with a data acquisition system (the Virginia Tech Transportation Institute’s MiniDAS) and received driving feedback in the form of a light and a tone when a potentially risky behavior was detected. Behaviors, such as swerving, speeding, lane changing without a turn signal, hard braking, hard turning, and fast starts, were used to determine when feedback was administered. Feedback continued for six months and then was turned off for one month (in the seventh month) to determine if risky behaviors returned after feedback stopped. These data were compared to a separate study (the Supervised Practice Driving Study [SPDS]) of 90 teenage drivers in the same geographic location who did not receive feedback. Parental involvement was examined by tracking which teen/parent groups checked the website and which did not. Results suggest that real-time and post hoc feedback produce a relative reduction in the rate of CNC involvement, but only when the parent is logging in to the website. If parents do not log in to the website to review the coachable events, real-time and post hoc feedback do not improve CNC rates. The analyses also indicated that once feedback was turned off in Month 7, teen CNC rates returned to baseline levels, which suggests that 6 months of feedback is not enough time to instill safe driving habits in novice drivers. This result also suggests that parental involvement in driver education must continue through the independent driving phase to improve teen CNC rates. In general, these results support previous research on monitoring and feedback, which suggest that parental involvement is critical in improving teen driving safety. These results also support current Graduated Driver’s Licensing (GDL) policies and provide research-based evidence that these policies should be strengthened.
  • Thumbnail Image
    Drivers' Visual Behavior When Using Hand-Held and Hands-Free Cell Phones
    Fitch, Gregory M.; Guo, Feng; Hanowski, Richard J.; Perez, M. P. (2014-08-25)
    This study investigated driver distraction and how the use of handheld (HH), portable hands-free (PHF), and integrated hands-free (IHF) cell phones affected the visual behavior of motor vehicle drivers. Method A naturalistic driving study recorded 204 participating drivers using video cameras and vehicle sensors for an average of 31 days. A total of 1564 cell phone calls made and 844 text messages sent while driving were sampled and underwent a video review. Baselines were established by recording epochs prior to the cell phone interactions. Total eyes-off-road time (TEORT) was examined to assess the visual demands of cell phone subtasks while driving. Percent TEORT was reported and compared against the baseline. Results Visual-manual subtasks performed on HH, PHF, and IHF cell phones were found to significantly increase drivers' mean percent TEORT. In contrast, conversing on an HH cell phone was found to significantly decrease drivers' mean percent TEORT, indicating that drivers looked at the forward roadway more often. No significant differences in percent TEORT were found for drivers conversing using PHF or IHF cell phones. The mean TEORT durations for visual-manual subtasks performed on an HH cell phone were significantly longer than the mean TEORT durations on either IHF or PHF cell phones. Practical applications This research helps to further reinforce the distinction made between handheld and hands-free cell phone use in transportation distraction policy.
  • Thumbnail Image
    Driving Risk Assessment Based on High-frequency, High-resolution Telematics Data
    Guo, Feng; Qian, Chen; Shi, Liang (SAFE-D: Safety Through Disruption National University Transportation Center, 2022-04)
    The emerging connected vehicle and Automated Driving System (ADS), the widely available advanced in-vehicle telematics data collection/transmitting systems, as well as smartphone apps produce gigantic amount of high-frequency, high-resolution driving data. These telematics data provide comprehensive information on driving style, driving environment, road condition, and vehicle conditions. The high frequency telematics data has been used for several safety areas such as insurance pricing, teenage driving risk evaluation, and fleet safety management. This report study advances traffic safety analysis in the follow aspects: 1) characterize the high-frequency kinematic signatures for safety critical events compared to normal operations; and 2) develop models to distinguish and predict crashes from normal driving scenarios based on the high frequency data. Two deep learning models were developed. The first one combines the strength of convolutional neural network (CNN), gated recurrent unit (GRU) network and extreme gradient boosting (XGBoost). The second approach is based on a novel variational inference for extremes (VIE) to address the rarity of crashes. The models proposed in this project can benefit a variety of traffic research and applications including connected vehicles and ADS real-time safety monitoring, NDS data analysis, ride-hailing safety prediction, as well as fleet and driver safety management programs.
  • Thumbnail Image
    Effect of Using Mobile Phones on Driver’s Control Behavior Based on Naturalistic Driving Data
    Zhang, Lanfang; Cui, Boyu; Yang, Minhao; Guo, Feng; Wang, Junhua (MDPI, 2019-04-25)
    Distracted driving behaviors are closely related to crash risk, with the use of mobile phones during driving being one of the leading causes of accidents. This paper attempts to investigate the impact of cell phone use while driving on drivers’ control behaviors. Given the limitation of driving simulators in an unnatural setting, a sample of 134 cases related to cell phone use during driving were extracted from Shanghai naturalistic driving study data, which provided massive unobtrusive data to observe actual driving process. The process of using mobile phones was categorized into five operations, including dialing, answering, talking and listening, hanging up, and viewing information. Based on the concept of moving time window, the variation of the intensity of control activity, the sensitivity of control operation, and the stability of control state in each operation were analyzed. The empirical results show strong correlation between distracted operations and driving control behavior. The findings contribute to a better understanding of drivers’ natural behavior changes with using mobiles, and can provide useful information for transport safety management.
  • Thumbnail Image
    The Effects of Cognitive Executive Load on Driving Crashes and Near-Crashes
    Sullivan, Keith Alexander (Virginia Tech, 2022-06-08)
    Previous naturalistic driving studies have shown that visual and manual secondary tasks increase driving crash risk. With the increasing use of infotainment systems in vehicles, secondary tasks requiring cognitive executive demand may increase crash risk, especially for young and older drivers. Naturalistic driving data were examined to determine if secondary tasks with increasing cognitive executive demand would result in increasing crash risk. Data were extracted from the Second Strategic Highway Research Program Naturalistic Driving Study, where vehicles were instrumented to record driving behavior and crash/near-crash data. Cognitive executive and visual-manual tasks paired with a second cognitive executive task were compared to the cognitive executive and visual-manual tasks performed alone. Crash/near-crash odds ratios were computed by comparing each task condition to driving without presence of any secondary task. Dual cognitive executive tasks resulted in greater odds ratios than those for single cognitive executive tasks. The dual visual-manual tasks odds ratios did not increase from single task odds ratios. These effects were only found for young drivers. These findings help validate that cognitive executive secondary task load increases crash/near-crash risk, especially in dual task situations for young drivers. Future research should be conducted to minimize cognitive task load associated with vehicle infotainment systems using such technologies as voice commands.
  • Thumbnail Image
    Emotional Impacts on Driver Behavior: An Emo-Psychophysical Car-Following Model
    Higgs, Bryan James (Virginia Tech, 2014-09-09)
    This research effort aims to create a new car-following model that accounts for the effects of emotion on driver behavior. This research effort is divided into eight research milestones: (1) the development of a segmentation and clustering algorithm to perform new investigations into driver behavior; (2) the finding that driver behavior is different between drivers, between car-following periods, and within a car-following period; (3) the finding that there are patterns in the distribution of driving behaviors; (4) the finding that driving states can result in different driving actions and that the same driving action can be the result of multiple driving states; (5) the finding that the performance of car-following models can be improved by calibration to state-action clusters; (6) the development of a psychophysiological driving simulator study; (7) the finding that the distribution of driving behavior is affected by emotional states; and (8) the development of a car-following model that incorporates the influence of emotions.
  • Thumbnail Image
    Enhanced Feature Representation in Multi-Modal Learning for Driving Safety Assessment
    Shi, Liang (Virginia Tech, 2024-12-03)
    This dissertation explores innovative approaches in driving safety through the development of multi-modal learning frameworks that leverage high-frequency, high-resolution driving data and videos to detect safety-critical events (SCEs). The research unfolds across four methodologies, each contributing to advance the field. The introductory chapter sets the stage by outlining the motivations and challenges in driving safety research, highlighting the need for advanced data-driven approaches to improve SCE prediction and detection. The second chapter presents a framework that combines Convolutional Neural Networks (CNN) and Gated Recurrent Units (GRU) with XGBoost. This approach reduces dependency on domain expertise and effectively manages imbalanced crash data, enhancing the accuracy and reliability of SCE detection. In the third chapter, a two-stream network architecture is introduced, integrating optical flow with TimeSFormer with a multi-head attention mechanism. This innovative combination achieves exceptional detection accuracy, demonstrating its potential for applications in driving safety. The fourth chapter focuses on the Dual Swin Transformer framework, which enables concurrent analysis of video and time-series data, this methodology shows effective in processing driving front videos for improved SCE detection. The fifth chapter explores the integration of corporate labels' semantic meaning into a classification model and introduces ScVLM, a hybrid approach that merges supervised learning with contrastive learning techniques to enhance understanding of driving videos and improve event description rationality for Vision-Language Models (VLMs). This chapter addresses existing model limitations by providing a more comprehensive analysis of driving scenarios. This dissertation addresses the challenges of analyzing multimodal data and paves the way for future advancements in autonomous driving and traffic safety management. It underscores the potential of integrating diverse data sources to enhance driving safety.
  • Thumbnail Image
    Evaluating the Influence of Crashes on Driving Behavior using Naturalistic Driving Study Data
    Guo, Feng; Chen, Chen (National Surface Transportation Safety Center for Excellence, 2015-07-16)
    It is hypothesized that intense events such as crashes could influence driver behavior and driving risk. This study evaluated the influences of crash events on driver behavior and driving risk using data from the 100-Car Naturalistic Driving Study, which included 51 crashes from primary drivers. Two metrics were used to measure driver behavior and risk: the proportion of baselines where the drivers were engaged in complex and moderate secondary tasks and the intensity of the near-crashes (NCs) and safety-critical incidents (SCIs). For the distraction analysis, we sampled 882 6-second baseline epochs within 15-hour windows before and after crashes. Results from a mixed binomial regression model indicated that the percentage of baselines where drivers engaged in complex secondary tasks dropped after crashes (odds ratio = 0.54; 95% CI [0.32, 0.93]). The driving risk analysis used the intensity of SCIs and NCs to measure the driving risk. Since there are typically more than one SCI and NC events before and after a crash, we developed four alternative recurrent event models to evaluate the influence of crashes based on actual driving time. The driving period was divided into several phases based on the relationship to crashes, and the intensities of these periods were compared. Results show a reduction in SCI intensity after the first crash (intensity rate ratio = 0.82; 95% CI [0.693, 0.971]) and the second crash (intensity rate ratio = 0.47; 95% CI [0.377, 0.59]) for male drivers. Females were observed to have a nonsignificant response to the first crash, but SCI intensity decreased after the second crash (intensity rate ratio = 0.43; 95% CI [0.342, 0.547]). This study indicated that crashes do have a positive effect on drivers’ behavior in terms of distraction and driving risk.