Projects

Project 1. Application of distributed driving simulation in examining complex latent hazard anticipation scenarios Mentors: Yusuke Yamani (Psychology), Kun Xie (Civil & Environmental Engineering), Sherif Ishak (Civil & Environmental Engineering)

Multi-vehicle crashes consist of 61% of traffic fatalities in the U.S, and thus it is crucial to understand how drivers interact, negotiate, and respond to other road users. However, pre-programmed ambient traffic in a standard driving simulator may not respond in a realistic way to the driver’s actions, and most driving simulators do not allow researchers to directly study the complexity of multiple vehicle crashes. Distributed simulation involves multiple drivers interacting in a single virtual reality environment similar to multi-player video gaming.

Distributed driving simulation has been used to examine pedestrian-to-driver, driver-to-driver, and pedestrian-to-automated vehicle interactions. Yet, distributed driving simulation technology has not been applied to the realm of driving training programs designed to improve the higher order cognitive skills required for road safety. One such skill is latent hazard anticipation, or the ability to anticipate hazards before they materialize. Desktop training modules for latent hazard anticipation has improved performance in young drivers, but even this improved performance is not yet perfect. One possible reason why trained drivers are still unable to anticipate all latent hazards is due to young drivers’ inability to comprehend the road environment from the perspective of other road users. For example, if an experienced driver is planning to turn onto a road, but they know that a tall hedge is blocking oncoming traffic’s view of their car, they may inch out slowly into the road to ensure that other drivers see them and slow down or stop if necessary.

The objective of this study is to examine whether experience in distributed simulation scenarios, both from the perspective of the driver encountering the hazard and from the driver who constitutes the hazard, improves latent hazard anticipation performance to a greater extent than existing desktop training programs.

Learning gain: Students will learn how to program latent hazard anticipation scenarios in a driving simulator. In addition, students will gain expertise in running participants, reducing driving behavior data, and analyzing the results.

Project 2. Operator Vigilance in Advanced Air Mobility Environment: A Human-Subject Simulator Experiment. Mentor: Yusuke Yamani (Psychology)

Advanced Air Mobility (AAM) is conceptualized as the technological development of air transportation systems that enable transportation of goods and passengers within urban and rural regions (e.g., package delivery drones, air-taxis). It is anticipated that AAM operations will incorporate high levels of automation, potentially shifting the human operator’s role from an active controller to a passive monitor. The increasingly automated systems in AAM operations will require human operators to sustain attention to perceptually demanding tasks for prolonged periods of time. However, as human operators sustain attention over time, attentional performance degrades, a phenomenon referred to as vigilance decrement. Although previous work on human performance examined vigilance decrement in perceptually demanding tasks, there is scarce work that directly examines the attentional process involved in human-automation teaming. Better understanding of the underlying psychological mechanism responsible for safe operation of air vehicles is necessary for developing an effective training program for future operators of the AAM technology.

REU students involved in this project will examine the effect of time on human-automation teaming in the AAM environment by conducting a human-subjects experiment using the newly developed Human-Autonomy Teaming Task Battery (HATTB) software. As shown in Figure 2, REU students will develop a study utilizing the multi-agent planning (MAP) task in the HATTB to collect human performance data. Upon completion, REU students will analyze human performance data and write a final report.

Learning Gain: The REU student will learn to design an experiment and collect human performance data from a newly developed software application. Also, students will be trained to perform statistical analysis using R programming and write an effective scientific paper.

Project 3. Direct observation of pedestrian distracted crossing behaviors and driver yielding responses. Mentor: Bryan Porter (Psychology)

Talking on mobile phones, texting, and listening to music or podcasts while wearing headphones has become ubiquitous on college campuses. These behaviors arguably can distract pedestrians while they make street crossings, placing them at risk for lack of attentive behavior to watch for vehicles failing to yield. In 2017, pedestrians were 16% of national traffic fatalities. Focusing on distracted behaviors that exacerbate risk for pedestrian-vehicle crashes is an important consideration. So is the focus on driver yielding/nonyielding behaviors. Both of these foci generate research questions amenable to a field, direct observation methodology. The protocol will be derived from Dr. Porter’s previous work that considered the impact of rectangular rapid flash beacons on crossings/yieldings and the prevalence of pedestrian distractions. Data collectors will observe assigned campus intersections at different times of day/days of week. This work will be long-term follow-ups to the previous studies to provide updated data to campus planners. Key pedestrian variables include: crossing locations (in/out crosswalk); with or against signals; distractions (headphones; texting; eating); looking behavior (for vehicles before crossing); and demographics. Key driver variables include yielding vs. not yielding when expected; distractions (visible handheld mobile use); and demographics.

Learning gain: Dr. Porter and his advanced student team will teach REU students about field protocols and methodological considerations and train the students to collect data for this project. The project allows students to add variables of interest or use the existing variables to create a personalized dataset producing results to present via conferences or publications.

Project 4. Modeling the effect of gusty wind force on vehicle driving performance Mentor: Sherif Ishak (Civil Engineering)

Hurricanes induced severe damage to the areas they hit hours before they make landfall. During these situations, a reliable transportation route is key to maximizing the number of evacuees as well as for post-hurricane rescue efforts. While traffic planning is an important part of a hazard mitigation plan, vehicle performance on roads especially in such harsh environments is a crucial factor that merits attention in the planning process. The goal of this project is to lay a foundation for future research into vehicle performance on roads during harsh hurricane and tropical storm conditions. The study will utilize the new RDS-1000 (driving simulator) and RDS-100 (Desktop Simulator) acquired at ODU. The driving simulator has a single seat cab with three degrees of freedom motion system, virtual dash and center stack displays, and a library of residential, urban, rural, commercial, industrial, highway, intersection and traffic signal control; autonomous, interactive ambient traffic; extensive, interactive scripted vehicle activity; variable roadway friction and weather effects; and data collection definition. The dynamics of the driving simulator can be modified within the SimCreator proprietary software tool which represents a graphical user interface that allows placement and connection of various components including extensive, scripted vehicle activity in C/C++ code components. This will provide a virtual environment that imitates real life scenarios) to reproduce the wind loadings experienced by vehicles during gusty hurricane wind events. This proposed research is exploratory to investigate the effect of wind force conditions on driving behavior. More specifically, the project will investigate how to modify the parameters of the driving simulator to replicate vehicle performance of a passenger vehicle and explore how to convert wind forces to gusty two dimensional wind loadings on vehicles using the driving simulator. An extension of this work can be applied to testing the effect of storm surge and roadway flooding on the risk associated with driving under these conditions.

Learning gain: The REU students will learn how to develop and program simulated driving scenarios using the driving simulator at ODU. The students will be trained how to build and modify existing built-in scenarios to perform the project work. This involves learning how to do programming with JavaScript and how to perform simulation experiments with human subjects.

Project 5. Analysis of seat belt use and models’ comparisons Mentor: Norou Diawara (Mathematics and Statistics)

Seat belt use is the most effective safety in a car crash. However, the percent of drivers wearing the seat belt changes. In rural, less dense areas, the seat belt use behavior is different of that in large and medium size cities. There is a myth that for large vehicles the seat belt use is less important. Other features related to the seat belt use include the types of roads, the weather and the travelling speeds. Is the driver more conscious of the seat belt use when a passenger is present or absent? Is there a gender difference associated with the seat belt use? In order to answer such questions and save lives, data on the seat belt use must be analyzed with the proper methodologies. As a binary outcome, the seat belt use must be linked to selected features (such as car type, gender and locations). The analysis can be formulated in two parts: the visualization and the modeling. In doing visualization, graphs and plots that describe the relationships will be presented. Among the methods, generalized linear models will be explored. To inform on the performance, the receiver operator characteristic curve is also evaluated. The dependence of the features under hierarchical structure is proposed. The hurdle and zero inflation models will also be implemented. We will construct spatial patterns where the features are nested within clusters and networks are formed. The project will review many of the latest modeling techniques for seat belt use, a binary data that has been collected over space and time in the last ten years.

Learning gain: the REU students will learn how to load and manipulate large data in SAS and R, select variables, visualize and analyze binary data. They will review the design and sampling collection. They will perform model comparisons with programming and coding tools. The study will be a way to use data for informed decision makings. The methods learned can be applied to many other research areas, such as in healthcare and disease control. The analysis will include segmentation of geographical behaviors and model adaptations, all written in a final report.

Project 6. Investigate human drivers’ behaviors in response to cyberattacks in a connected and automated driving environment Mentor: Kun Xie (Civil & Environmental Engineering)

Connected and automated vehicles (CAVs) utilize sensors and wireless communication to collect critical information to automate driving tasks without drivers’ involvement. Although CAVs are envisioned to drastically enhance traffic safety by eliminating crashes caused by human errors, they may underdeliver the promise, given their vulnerabilities to cyberattacks. Human drivers are those who directly interact with transportation cyber-physical systems (CPSs) and respond to changing circumstances such as CAV cyberattacks. This project aims to answer an important question – how the human drivers would respond to unexpected events in various cyberattack scenarios. This study will utilize a high-fidelity driving simulator at ODU, which has three degrees of freedom motion system, center stack (used for message displays), an autonomous driving mode, and a library of various driving scenarios. Two types of cyberattacks against CAVs will be tested: 1) intra-vehicle attacks, where a subsystem of a vehicle gets compromised and becomes remotely controllable; and 2) inter-vehicle attacks, where bogus information is created. Human drivers’ behaviors in response to cyberattacks will be investigated. The impact of cyberattacks on psychological factors including stress, awareness, and trust will also be studied. The result outcome can facilitate the development of effective human-aware mitigation strategies to secure transportation CPSs and enhance transportation safety.

Learning gain: REU students will be trained to develop virtual driving scenarios using the driving simulator at ODU. They will also learn how to collect data on vehicle motion, driving behaviors and psychological factors and how to use quantitative approaches to analyze those data.

Project 7. Investigating the relationship between physiological indices of attention and driving performance Mentor: Xiao Yang (Psychology)

Road accidents are a leading cause of death in U.S. among people under age 55, and road traffic crashes cost U.S. $871 billion. Attentional lapses among road users are a major cause of road accidents. It is a general belief that combining any other task (e.g., texting, phone conversations, and music listening) with driving will affect driving performance. However, recent studies showed that multitasking could improve driving performance. Those conflicted findings call for a better understanding of attentional processes during driving.

Physiological measures have been increasingly used to understand drivers’ attention. Event-related potential (ERP) that is derived from electroencephalogram provides attention indicators. Particularly, the ERP P300 wave has been proposed to reflect the reciprocity of attentional resources. Moreover, high frequency heart rate variability (HF HRV) that is derived from electrocardiography is thought to reflect cognitive control and mental workload and has been studied in relation to cognitive performance and human errors. Given the noninvasive nature and high temporal resolution of P300 and HF HRV, these physiological indices are promising in research on attention during driving.

The objective of the present project is to investigate the relationship between physiological indices of attention (P300 and HF HRV) and driving performance. This project will include two parts. Part 1 is to establish a multimodal assessment system using standardized laboratory cognitive tasks. Specifically, P300 and HF HRV will be examined during a dual task paradigm, which will be further analyzed in relation to task performance. Part 2 is to utilize wearable device to record P300 and HF HRV in a driving simulator. The validity of the two physiological measures in the simulator will be assessed, and the relationship between the physiological indices and driving behavioral indices (e.g., brake reaction time, time to collision, and standard deviation of lateral position) will be examined.

Learning gain: Students will obtain hands-on experiences of recording physiological measures and driving performance metrics. Further, students will learn physiological data processing and potentially use computational modeling to analyze behavioral data. In addition, students will gain experience in running experiments, conducting statistical analyses, and reporting research findings.

Project 8. Spotlighting Physiological Factors in Roundabout Traffic Accidents using Machine Learning Algorithms Mentor: Yi He (Computer Science)

Roundabouts are a unique traffic control facility tailored for increasing operational safety of intersections. However, the safety effectiveness of roundabouts only manifests in good geometric design that lends drivers sufficient time for making decisions of entering, lane-merging, and exiting. The safety performance of roundabouts thus varies significantly across different designs. As documented in 2017 Louisiana research on 19 roundabouts, their performance difference in terms of the reduction rate of accident frequency and severity can be at orders of magnitudes. Such difference escalates in the groups of novice and senior drivers, who often demand a more tolerant decision-making time span to avoid accidents. Yet, it remains unclear what physiological factors determine the drivers' performance at roundabouts, and how do the physiological factors interplay with the roundabouts' geometric designs.

In this project, we will use machine learning algorithms to identify and analyze the potential risk factors that contribute to roundabout traffic accidents. The project proceeds in two stages. At Stage 1, we will employ predictive modeling techniques to assess how certain variables such as speed, visibility, turning angle, vehicle size, and road conditions may predict the likelihood of an accident occurring in a roundabout, in order to identify the most critical factors. At Stage 2, we will leverage the graphical models to capture the correlations between the factors identified from Stage 1 and two physiological metrics, ERP (measures drivers' attention) and HF-HRV (measures drivers' mental workload), to understand how the drivers' performance varies along with the roundabout traits. Moreover, the study may include the use of data visualization tools to better illustrate and present the results of the analysis. By utilizing these methods, our students who are envisioned to be the future generation leaders can gain an understanding of what factors are at play in roundabout traffic accidents, so that they can facilitate the safety measures to be implemented to reduce the accident occurrence.

Learning gain: Students will 1) obtain hands-on experiences of running off-the-shelf machine learning models to analyze physiological datasets, 2) learn physiological data processing and visualization tools to interpret the model results, and 3) get familiar with the pipeline of data-centered research including hypothesis formulation, experiment design, results summarization and visualization, and conclusion making and report writing.

Project 9. Understanding the Potential Driving Distraction Due to Delivery Drone Operations near Roadways Mentor: Hong Yang (Electrical and Computer Engineering)

Drones are being considered as an alternative affordable and faster mode for package deliveries. The operations of these delivery drones will undoubtedly fly near roadway infrastructure, which raises distraction concerns for drivers. This study intends to investigate the potential driving distraction effects due to delivery drones flying over or near roadways. Specifically, a driving simulation study will be designed to assess the possible impact of different drone operations (e.g., with/without packages, packages of different sizes, proximity to roadway, height, etc.). Drivers with different driving experience will be invited to participate the simulation experiments. Their visual attention changes will be measured through comparative experimental results. Statistical analysis will be performed to uncover the critical factors that raise driving distraction during drone operations near roadways.

Learning gain: Students will learn how to design customized experiments in driving simulation, particularly with the use of virtual reality technology. Also, students will be able to learn how to perform statistical data analysis with the use of open-source tool R and report results scientifically.