Investigating the Effects of Driver-Vehicle Availability in Human-Autonomous Vehicle Teaming (HAVT)

Introduction

Human–Autonomy Teaming (HAT) has gained considerable attention within the last decade. As technology continues to evolve, people and organizations are tasking intelligent, advanced technologies to carry more agency and autonomy in their actions. This is done, in part, to aid and promote more seamless interactions with their human counterparts. This evolution means that the interactions between humans and advanced technologies may begin to resemble more of a teaming relationship, which brings about the critical need to better understand how humans “team” with intelligent agents. Within human-machine collaboration, given that the knowledge and expertise of the human operator may be different from that of the intelligent system, “the highest team performance is often reached when they both know how and when to complement one another” (Ren et al., 2023).

Traditionally, intelligent entities, for example, drones and robots, have been the focal point of investigation within HAT. However, as automated vehicles become more equipped with self-driving features that emulate a teammate, such as Tesla’s Autopilot navigation and Smart Summon (Tesla, 2025), the idea of teaming with a semi-autonomous vehicle, as opposed to viewing it as a tool, is gaining substantial traction.

Approach

An in-lab driving simulation study was conducted to answer the following research question: Can the (un)availability of a driver and/or vehicle influence teaming perception and performance in an automated driving context? Data from 14 participants (Mean age = 25.57 years; two Female) was collected (IRB # 2020-755). Inclusion criteria required each participant to have had previous experience driving a semi-AV at least once. This study used a Driving Safety Research Institute (DSRI) fixed-base medium-fidelity driving simulator, miniSim™. There was an additional touchscreen monitor integrated to serve as an in-vehicle infotainment display.

Participants were asked to commute to and from work while riding in a semi-autonomous vehicle with SAE L3 capabilities. The road environment was a three-lane highway with a roadway topography consisting of mainly flat segments with occasional hills. There was mild traffic present with a series of construction sites along the commute. Each construction site was either active or inactive, that is, workers present/not present. During their commute, the driver and vehicle agents had a shared task of arriving to and from work timely, while safely navigating all construction work sites and the overall driving environment. The vehicle occasionally asked the driver about potential roadway dangers via visual question prompts, and the driver responded to the vehicle’s queries via touch input on the infotainment display. Given the time-sensitive nature of navigating the roadway, if the driver did not respond before a set distance from the construction site, the vehicle made its own decision regarding how to proceed. Additionally, individual goals of the vehicle and driver varied. The vehicle’s goal was to successfully manage the dynamic driving task. The driver’s goal was to complete a work-related document-editing task.

The independent variable (see Table 1 for more details) was driver/vehicle availability, which consisted of three levels: (1) the driver and the vehicle were both available (DVa), (2) the driver was available, but the vehicle could not be (Vu), or (3) the vehicle was available, but the driver chose not to be (Du).

OPEN IN VIEWER

Table 1.

Independent Variable (Three-Levels): Agent Availability Status.

Availability status	Scenario interaction description
Driver and Vehicle agents are both available (DVa)	Each agent is working on their individual task but are both available to alert/ask and answer questions about the road environment.
Vehicle agent is unavailable (Vu)	Vehicle’s ability to intake information from the environment has decreased (sensor degraded) and cannot receive communicative input from the driver during this time.
Driver agent is unavailable (Du)	Driver signals to the vehicle that they do not want to be disturbed during this time.

Here, vehicle unavailability resulted from environmental sensing degradation due to dense fog, while driver unavailability was determined by the experimenter but set by the driver. The availability status of the driver and vehicle was controlled via the infotainment touchscreen display.

This paper reports preliminary findings from select analysis. Dependent measures include: voluntary takeover observations, driver perception of the semi-AV as a teammate before and after the interaction, driver trust in the semi-AV before and after the interaction, driver preference for availability status, and team communication effectiveness (drivers’ correct input vs. vehicle’s receipt of drivers’ input). All analyses included descriptive (percentages) and qualitative trends/themes. Given the preliminary nature of the results, no statistical analysis was conducted.

Outcome

Of the 14 drivers, 11 decided to take control of the vehicle at least once, with the highest number of takeovers within the Vu availability status and lowest in the Du availability status. When the vehicle was unavailable, drivers regained control of the vehicle 45% of the time. Participants were asked about their perception of teaming before and after their interaction with the vehicle. Figure 1 illustrates the changes in perception status. Trust was similar before and after the interaction, with only a 4.6% increase. Drivers felt the teaming dynamic the most (79%) when the vehicle asked for their input to aid in its decision-making (vehicle’s inquiries about the environment—DVa), and secondly (57%) when they were able to do work uninterrupted while the vehicle drove (Du). During the DVa availability status, drivers answered the vehicle’s queries with 93% accuracy in assessing the environment, but given the time sensitivity of the input, the vehicle received the driver’s input only 64% of the time. Having individual goals during the driving experience was associated with an overall sense of being more efficient during the drive.

OPEN IN VIEWER

Figure 1.

Alluvial plot depicting the perception of the semi-AV as a teammate before and after driving interaction.

Conclusion

Preliminary findings from this research are expected to contribute to the knowledge base in HAT by providing an understanding of drivers’ perspectives and behaviors regarding teaming with an autonomous vehicle both now and in the future. Driver/vehicle availability afforded the opportunity to complete individual goals, which was associated with a sense of drive efficiency and a positive shift in the perception of the semi-AV being the driver’s teammate. Also, communication between the driver and vehicle was associated with an overall sense of teaming, and is known to be an important factor in effective teaming (Lee et al., 2023). However, further investigation is needed to determine how to do so effectively. Although the number of voluntary takeover instances was the highest when the vehicle was unavailable, rather than viewing this situation as a failure, participants perceived it as “stepping in” to aid the vehicle in navigating the environment. Srinivasan and Takayama (2016) found that people are more willing and quicker to help a robotic system if they believe that the robot is working autonomously. In other words, if they believed that there was no other human available to help the robot, but understood the robot was not infallible, they would be more willing to be the “human helper.”

This study limitations include: a limited sample size, convenience sampling, and drivers with some, but limited, experience with semi-AV interactions. Also, given that this study represents ongoing efforts, inferential statistical analysis as well as additional formal thematic coding for subjective findings will be included in future analysis and dissemination. Nonetheless, these study results can be used to understand the current mental models of drivers of vehicles with SAE L3 capabilities and their willingness to adopt behaviors toward the perspective of teaming with their vehicles. This work can also be used to aid AV designers in developing improved in-vehicle communications that support drivers in understanding and executing their roles within the DDT.