FHWA’s Human Factors Vehicle Automation Research

With the adaptation of Vision Zero, the goal of the Federal Highway Administration is to reduce roadway deaths to zero. To help move the Nation to zero roadway deaths, FHWA has adopted a Safe System Approach for the Nation’s transportation network. The Safe System Approach takes a holistic view of the entire transportation network and considers how safety can be achieved at multiple levels rather than focusing on a single part of the system. The FHWA Human Factors team is a natural extension of the Safe System Approach because the team often considers multiple aspects of the transportation system when conducting research. One important field where FHWA’s Safe System Approach can be applied is automation research.
“Vehicle operators are exposed to ever-increasing levels of automation. FHWA’s human factors research will inform safety and operations professionals about the opportunities and challenges associated with the transition from manually driven to fully automated vehicles. It is imperative that we pursue research that addresses human adaptation toward increasing levels of automation,” says James Pol, technical director of FHWA’s Office of Safety and Operations Research and Development.
As the Nation moves toward partial and full automation in transportation, human factors research plays a significant role in assessing the overall interaction of road users, vehicles, and the infrastructure, which will provide increased safety on the Nation’s roadways. Automated vehicles provide several safety benefits from a human factors’ safety perspective, such as potentially reducing fatalities and serious injuries due to human error; however, understanding the issues and challenges that automation brings for all road users is essential. FHWA’s Human Factors team conducts research using a variety of research tools, including a highway driving simulator (HDS), field research vehicles (FRVs), and virtual reality (VR) technology. The following sections will explain each tool, the type of behaviors each tool can evaluate, and some studies that have applied these tools.

HDS
The Human Factors team has a fully-interactive high-fidelity driving simulator used to conduct research and to study drivers’ reactions in a fully controlled, simulated environment. Seven high-definition projectors generate roadway scenes with a 220 degree field of view. The view from the vehicle’s rearview and side-view mirrors are generated using a computer. The simulator has a six-degree-of-freedom motion base allowing for a realistic immersive environment. In addition to the motion-based system, the simulator’s sound system provides engine, wind, and tire noises, and other environmental sounds.
The driving simulator has a 120-hertz eye-tracking capability, which allows researchers to investigate where participants are looking when they drive through various roadway scenarios. Additionally, the HDS allows for constant monitoring of a driver’s decisions concerning speed, acceleration, braking, lane position, following distances, merging behaviors, distance from other road users (vehicles, pedestrians, bicyclists), and sign compliance. Physiological measures, such as heart rate, are also captured as a way to test variables during stressful driving situations. Finally, through the use of questionnaires, the Human Factors team can capture a driver’s perceptions, such as trust, distrust, misuse, and abuse of various automated systems. By looking at all of these behaviors, the researchers can obtain an understanding of how drivers use an automated system.
Because the HDS can simulate various levels of vehicle automation or manual driving, the HDS has been used for a variety of research related to automated vehicles. One automated vehicle study conducted using the HDS is “Response to Emergency Vehicles When Driving in a Mixed Fleet,” which evaluated the impact of connected vehicle alert messages on drivers’ responses to emergency vehicles. The participants’ vehicle connectivity was manipulated so that half the participants received an in-vehicle alert that an emergency vehicle was approaching. Connected and automated vehicle (CAV) market penetration was also manipulated to create an environment where none, some, or all of the surrounding traffic responded to a connected vehicle (CV) with a vehicle-to-vehicle emergency vehicle alert. The results of this study showed that CV alerts were effective in getting the participants to yield to an approaching emergency vehicle. Driving with CV alerts led to increased pullover rates, reduced speeds, and pulling over sooner compared to participants without CV alerts.
Another study, “Automated Vehicle Driver Behavior While Passing a Bicyclist,” was conducted in parallel to the CAV study. This study looked at how driving automation influenced the passing behaviors of drivers when approaching a bicyclist on either a shared or dedicated roadway. The study also looked at whether cooperative driving automation (CDA) messages helped support safer passing behavior. The research team anticipates this study will be published in late 2024.
One published study, “Driver Adaptation to Vehicle Automation: The Effect of Driver Assistance Systems on Driving Performance and System Monitoring,” assessed the effect of varying levels of vehicle automation on driver performance over time. Participants gained experience with driver assistance systems across four sessions in the driving simulator. The specific driver assistance systems that were manipulated between subjects included cooperative adaptive cruise control (CACC), lane-keeping assist, (LKA), a combination of CACC and LKA (CACC + LKA), and a control condition with no driving assistance features. Driver performance metrics, eye tracking, and physiological data were collected to assess how driver behavior changes as the driver adapts to automation. The results of the study showed that driver assistance technology could potentially provide useful benefits to drivers even after drivers have adapted to the technology with repeated use. Participants who used the technology were able to do so in a way that allowed them to direct more of their attention to the road ahead, and driver adaptation was not associated with impaired responses to emergency events.

A recent study evaluated the impact of infrastructure-to-vehicle messages about an upcoming lane closure on move-over behavior and safety. This study, “Lane Change Response to Infrastructure Warning About Lane Closure in a Mixed Vehicle Fleet,” specifically assessed the potential value of adding CDA technology to a changeable message sign (CMS) by comparing the responses of cooperative automated driving system (C-ADS)-equipped vehicles that receive information about an upcoming lane closure via CDA to that of conventional drivers who receive the information via CMS. The message type was manipulated, such that half of the participants received information that prompted an immediate change in C-ADS behavior and half did not receive the message. Drivers’ acceptance and trust of C-ADS behavior was examined as a function of both the source and content of the traveler information available to the driver.
FRV
The Human Factors team uses field research vehicles to conduct research on real roadways to better understand driver behavior and performance. A FRV is outfitted with a data logger with controller area network interface. The FRV is also outfitted with the CARMA Platform℠ and has the capability to conduct SAE International® Level 2™ vehicle research assessing different driver behaviors related to automation.
Similar to the HDS, each FRV is equipped with an eye-tracking system with three face cameras mounted on the dashboard of the vehicle and infrared light sources. The cameras track the head position and gaze of the driver without interfering with normal driver behavior. Three additional cameras mounted on the exterior of the vehicles’ roofs, directly above the driver’s position capture a panoramic, forward view of the driving scene. Participants’ eye gaze information is synchronized with this panoramic view to determine where the driver is looking during the driving session. The FRV can also receive signal phase and timing (SPaT) information from nearby intersections. Lastly, like the HDS, the Human Factors team can record various vehicle kinematics (speed, braking, and steering input) and can ask participants about their experience through questionnaires. The specific questions and vehicle kinematics vary depending on the aim of the research.
A study called “Cooperative Driving Automation Alerts During Rainy Weather Condition” used an FRV to evaluate the modality of a CDA alert and evaluated the most effective and preferred alert by drivers. The CDA alert informed drivers about the potentially compromised visibility and slippery road surface conditions in three modalities: auditory, visual, and a combination of both. Using speed, following distance, braking, and steering use, the researchers assessed the participants’ driving performance. On a post-experiment questionnaire, participants reported their preferred mode and utility of adaptive cruise control (ACC) and CDA technology. Results of this study are anticipated to be published in fall 2024.
In another study, “Exploring the Effects of Vehicle Automation and Cooperative Messaging on Mixed Fleet Eco-Drive,” a FRV was used to assess the behavior of drivers, either driving with ACC or manually, who followed a lead vehicle that demonstrated eco-driving strategies. Eco-driving strategies are intended to reduce fuel use by optimizing speed profiles through efficient use of acceleration, deceleration, and idling at a stop. Participants would receive CDA messages that provided SPaT information and information about the lead vehicle’s behavior. The FRV was used to create speed and acceleration profiles of the drivers to see if the various CDA messages helped create smoother driving profiles. Additionally, post-driving questionnaires were used to assess the utility and overall safety gains of the CDA messages.
A project awarded in 2023, “Enhanced Lighting Treatments for Improving Vulnerable Road User Detection Within Mixed Fleets,” will also use FRV to assess the effects of varying lighting levels on vulnerable road user detection within a mixed-vehicle fleet. This project will see how various lighting treatments can help improve various vulnerable road users’ detection not only by conventional vehicles, but also by automated vehicles. Results from this project are anticipated to be published in spring 2025.

VR Tools
VR tools are still new and emerging technology; however, they allow the Human Factors team to conduct research looking into vulnerable road users. Currently, the Human Factors team has two different ways of researching how pedestrians interact with automated technologies: using a two-lane roadway and using an omni-directional treadmill. The two-lane roadway is a full-sized two-lane road that participants can cross. To recross the roadway, the participants are turned around by the researchers. The omni-directional treadmill allows for multiple crossings. Both systems use VR headsets that put participants in an immersive, computer-generated, three-dimensional environment where they can interact. The headsets provide high-resolution organic light-emitting diode displays and provide a 110-degree field of view. Two sensors mounted on opposite corners create virtual areas participants can walk around in.
These headsets are also equipped with eye-tracking capabilities, which allow insights into where the participants are looking and how long they are looking there. In addition to eye tracking, the Human Factors team can capture information about how long it takes participants to cross an intersection, their walking pattern, gait, distance to other road users, and other relevant crossing behaviors. Lastly, like the driving simulator and FRV, the research team uses questionnaires to assess various subjective measures like trust and safety perception.
One current study, “Ensuring Cooperative Driving Automation (CDA) Vehicles and Vulnerable Road Users (VRUs) Safety Through Infrastructure: Phase 2,” will use the two-lane road to assess how pedestrians (young and old) respond to a CDA-enabled infrastructure‑based alert system. This study aims to see if providing pedestrians with a warning about an impending conflict will help keep them safe when crossing an intersection.

Conclusion
Each research tool offers its own unique range of advantages and limitations to conducting human factors research. One of the primary advantages of the HDS and VR tools is that they allow the Human Factors research team to put road users in scenarios that could be dangerous in real life. This practice enables the team to understand how a new system or technology will impact various road users’ behavior and if it has a safety benefit. The HDS and VR tools also allow the research team to test new and future automated systems, keeping the research on the forefront of innovation.
Using the HDS and VR tools has some limitations, however, such as not reflecting some drivers’ real-world behavior because it is simulated. Additionally, the participants who sign up to be in the research study may not be the most representative sample of the population. Another limitation of the HDS and VR tools is that currently all the behaviors of other vehicles and road users is coded and constant so their behavior may not perfectly reflect the choices made by real road users or automated systems. Currently, the Human Factors research team is working on reducing this limitation by integrating the CARMA Platform into the HDS so it will behave like an automated vehicle. Additionally, the research team is looking into distributed simulation, which would allow other road users (pedestrians, vehicles, and bicyclists) to actually be controlled by other, independent participants.
FRVs tend to have different advantages and limitations compared with the HDS and VR tools. One advantage of field research is that the research team is collecting data in an environment that is the same or similar to real-world driving, which increases the validity of the study. On the other hand, experimental control cannot be contained as efficiently as it can be controlled in the simulator. Other extraneous variables may have an effect on the results of the study that researchers cannot control. As far as safety is concerned, conducting data collection in the real world does not allow for a controlled environment without risk. Using simulation, virtual reality, and field equipment provides complimentary research tools for the Human Factors program that provide a unique opportunity to see how vehicle automation has an effect on all road users and serves an important role in transportation research while minimizing risk to participants and researchers.
“Human Factors research will continue to play a pivotal role in automation research. As artificial intelligence systems continue to grow, it will be imperative that these systems are designed in a manner that not only optimizes performance but is also accepted and understood by users,” says FHWA Chief Scientist Craig Thor.
Michelle Arnold leads a variety of human factors research and manages the FHWA HDS laboratory contract. She received her Ph.D. in psychology from Western Michigan University.
Jesse Eisert leads various human factors research at FHWA, including automation and how it interacts with vulnerable road users. He received his Ph.D. in psychology from George Mason University.
For more information about the Human Factors research team, see https://highways.dot.gov/research/laboratories/human-factors-laboratory/human-factors-team-members, or contact Jesse Eisert, 202-493-3284, Jesse.Eisert@dot.gov, or Michelle Arnold, 202-493-3990, Michelle.Arnold@dot.gov.
For more information about the Human Factors team research tools, see https://highways.dot.gov/research/laboratories/human-factors-laboratory/research-tools.