This study developed and evaluated a tactile driver–vehicle interface to support non-driving-related tasks and emergency braking in conditionally automated vehicles. A within-participant experiment was conducted to assess usability, user acceptance, and safety of the proposed interface by comparing it with a conventional tactile driver–vehicle interface. Thirty drivers completed three experimental tasks. The results showed that participants completed the non-driving-related task significantly faster using the proposed interface (p = 0.039) and performed more efficient interaction behaviors (p = 0.008). Participants also reported higher satisfaction (p < 0.001) and acceptance of the driver–vehicle interface after experiencing the proposed interface. In the take-over request task, braking reaction time was significantly reduced when using the proposed interface compared to the foot brake (p = 0.001). In conclusion, the proposed interface was more efficient and acceptable than the conventional tactile interface and demonstrated potential safety benefits in conditionally automated vehicles.

An experiment was conducted to test two driver-vehicle interface (DVI) prototypes (i.e., “Handle” and conventional DVI). The experiment was a within-participant design in which all participants tried two DVI prototypes. Participants were divided into two groups (A and B) that experienced the two prototypes in different orders to rule out learning effects under specific experimental conditions.

Participant

Thirty participants took part in this study. All participants had a valid driver’s license, at least one year of driving experience, and were in good health. Participants were recruited using recruitment information from social applications. Each participant signed an informed consent form for participation in the experiment. This experiment was reviewed and approved by the Ethics Committee of the University of Tsukuba (approval number: 2023R825). Each participant received a reward (1440 yen) for participating in the experiment at the end of the experiment.

Tasks

The experiment comprised the following three tasks. Participants were asked to perform two non-driving-related tasks (NDRTs) in the context of playing a video. They were then asked to stop the simulated vehicle during a take-over request (TOR) situation.

Initially, participants followed the instructions to complete two NDRT tasks (i.e., Tasks 1 and 2).  In Task 1, participants were instructed to switch the item to the video function. In this task, the participant’s behavior of pushing the joystick to the left or clicking the left button was recorded as an executed efficient behavior.

In Task 2, the participants were asked to adjust the volume of the video. First, they clicked on the confirmation button to enter the third level. After confirming that the video was playing normally, the participants were asked to mute the volume of the video to 50% and then increase it to 80 %. The primary performance metrics of the two NDRTs were task errors, task completion time, and efficient behavior.

Participants were informed they would receive no instructions for the TOR task before the experiment. However, when the simulated vehicle arrived at the pre-programmed point of the TOR, the interface informed the participant that there was a system failure in the vehicle and instructed the driver to engage the brake. After receiving the alert, participants must hold the buttons in “Handle” or engage the foot brake pedal until the vehicle comes to a complete stop. For this task, the reaction time (RT) and subjective willingness were considered as the primary performance metrics.

Process

Before the experiment began, each participant was informed of the purpose of the study, the procedure, and the form of data collection. All participants signed an informed consent form for participation in the experiment. After completing the signature procedure, the participants were trained in the simulated vehicle. The experimenter introduced the DVIs to the participants and asked them to familiarize themselves with the environment. The training lasted approximately 10 min for each participant. They first had to try out all the functions provided by the DVIs, and then let the experimenter know that they were comfortable operating the interface. After the training, the participants completed an NDRT that was similar to the experimental tasks. The experiment only commenced if the participants completed the task without any errors. Otherwise, they were trained for an additional 5 min.

After confirming that the participant had completed all tasks using the two DVIs, the experimenter informed the participant that the experiment had concluded. The participants subsequently filled out a questionnaire and responded to one question regarding self-preference assessment to gauge their acceptance of the DVIs. The questionnaire contained ten variables representing four factors in the user acceptance model (i.e., perceived usefulness, perceived enjoyment, satisfaction, and attitude). A Likert scale (1–5) was used for all variables, ranging from “strongly disagree” to “strongly agree.” Each participant took 10–20 min to complete the questionnaire. Finally, the experiment lasted approximately 80 min for each participant. 

Measurement

Data about the defined variables (task errors, task completion time, efficient behavior, self-preference, and subjective willingness) were collected during the experiment. Owing to data loss, 29 data points were obtained with the conventional DVI, and 28 data points were obtained with “Handle” (data from 27 subjects were fully recorded after undergoing two DVIs). RT was then collected as the primary performance metric for the TOR task. However, three data points were deleted owing to brake failure (N = 54): one participant was asleep when the TOR occurred, while two individuals observed the impending accident and braked prior to the TOR (See experiment data). 

Another important part of the data obtained was the participants’ self-preferences for the two DVIs and their attitudes toward them. The data of the respondents in this study were compared with those of our previous study. Thus, the previous data were extracted from the 20–30 age group (N = 62), which constituted the predominant age group in this study (N = 28), to avoid the influence of age on the results (See questionnaire data).