Natural Mapping of Vibration Alerts to Wheel Responses: Effects of Directional and Urgency-Based Vehicle Warnings

Abstract

Vibration alerts have not been investigated thoroughly to create intuitive warnings for takeover requests (TORs), providing directional information to drivers during partial driving automation. Using vibration alerts for TORs, the placement of the physical alerts in the system, and the orientation information provided to the driver could influence how a fast, intuitive response is chosen. The present study investigated whether vibration alerts, directional and non-directional, influence response time when choosing a direction to move around a hazard during a low and high urgency driving scenario. Unlike prior studies, which required forced correct responses, this study allowed participants to freely choose between two equally safe responses. This paradigm enabled the testing of participants’ natural response tendencies. Participants chose to steer more often using a compatible rather than an incompatible mapping based on the vibration alert location. This study recommends vibration alerts be designed using compatible mapping for accurate responses.

Introduction

Navigation directions while driving can be provided through different modalities, such as visual maps or auditory instructions. Vibrotactile cues are less common but can be used to alert the driver to lane departure or takeover requests (TORs; GMAuthority, n.d.). These vibration alerts have not been investigated thoroughly to create intuitive warnings for providing directional information to drivers during partial driving automation (Level 2 automation; SAE International, 2021). Using vibration alerts for TORs, the placement of the physical alerts in the system, and the orientation information provided to the driver could influence how a fast, intuitive response is chosen. The present study thus investigated whether vibration alerts, directional and non-directional, influence response time when choosing a direction to move around a hazard during a low- and high-urgency driving scenario.

Vibration alerts have previously demonstrated different characterizations to present warnings to a driver about upcoming hazards or TORs. The characteristics of the vibrotactile modality are limited physically (i.e., duration, intensity, and physical location; Hossain et al., 2025). Thus, directionality and orientation of vibrotactile alerts are two proposed ways to communicate effectively with a driver (Shi et al., 2023). Generally, using a generic, non-directional TOR can lead to incorrect decisions compared to an informed, directional TOR (left or right alert; Eriksson et al., 2018) due to non-directional TORs taking longer for a driver to respond as compared to directional TORs (Shi et al., 2023). Using the auditory modality, lateral (e.g., directional) warnings on the left or right physical space can represent a toward-hazard or toward-action alert. For auditory alerts, there are conflicting conclusions about whether the warning should be presented on the side with a potential hazard (toward-hazard; Chen et al., 2022) or the side where the action should occur to avoid the hazard (toward-action; Huang & Pitts, 2022). Similarly, studies using vibration alerts have found conflicting conclusions. Shi et al. (2023) found no difference in response time between toward-hazard or toward-action alerts. However, Huang and Pitts (2022) found that when introducing vibrotactile alerts with multiple modalities (visual and auditory), toward-action directional information benefited performance compared to toward-hazard alerts. In addition, Xu and Bowers (2024) did not find a benefit for performance when utilizing directional vibrotactile alerts in driving scenarios.

Toward-hazard or -action directionality for lateral warnings is opposite to each other in external space for driving scenarios. The explanation for either direction of alerts leading to faster, more automatic responses has been supported through the attention capture (favoring toward-hazard alerts) and stimulus-response compatibility (SRC) theory (favoring toward-action alerts). Attention capture suggests that a toward-hazard alert leads to faster responses due to increasing the driver’s attention to the object/collision location (Chen et al., 2022). The SRC effect describes the spatial coding between a stimulus location and a response object/location and that responses are faster when the stimulus and response are compatible (i.e., located on the same side of space) than if the pairing is incompatible (Proctor & Vu, 2006). SRC predicts that drivers will be faster to respond when using a natural mapping, such as the vibration alert being on the same side of the space where the response is completed (Proctor & Vu, 2006). Previous literature has not reached an agreement on which S-R mapping is automatically activated for mapping a vibrotactile stimulus to a response.

In addition to the characteristics vibration alerts use to convey directional information, the speed of information processing can also be facilitated when using the vibrotactile modality. Time to collision (TTC) can affect how quickly the information from an alert is processed for the driver to make an appropriate response. For vibration alerts, participants showed faster takeover times with a short TTC (4s) compared to a longer TTC (7s; Huang & Pitts, 2022). Similarly, Shi et al. (2023) varied TTC and found that the effectiveness of directional (toward-hazard or -action) and non-directional vibrotactile alerts differed for short (3 or 4 s) and long (6 or 8 s) lead times. Although the relationship between directional alerts and TTC has been investigated, previous studies did not examine how allowing participants to freely decide their response to vibration alerts might vary across scenarios with different levels of urgency. The present study examined the influence of the relationship between the side(s) on which the vibration alert occurs and the TTC level of the driving scenario on response time to decide on the direction in which they will make a wheel response. Understanding how drivers orient the vibration alert to the directionality of their wheel response can reduce confusion for drivers when responding to navigation tasks or TORs while driving.

Method

Participants (N = 37; 18–22 years) were U.S. license holders, compensated for their time through credit toward a course research requirement which was approved by the University’s Institutional Review Board. The study (30 min) was conducted in E-prime (3.0), using prerecorded driving videos from a STISIM 3.0 driving simulator (stisimdrive.com). Participants drove in the middle lane of a one-way three-lane road and used a steering wheel (Logitech G920 Driving Force) to turn either left or right when prompted by a vibration alert while seeing a gravel pile (i.e., the hazard) in the middle lane. The vibration alert(s) were presented concurrently with the gravel pile in the driving scenario. Fog (300 ft of visibility) was present in the driving scene to make the hazard more difficult to see. The vibrotactile stimuli were coded to produce vibration using an Arduino Uno Rev3 (167 Hz, DC 3 V). This frequency was chosen because it is above the range of road vibrations drivers typically experience (Fitch et al., 2007). The vibrotactile alerts were on the seat pan, placed on the outer edge of the thigh where the leg meets the chair. This setup allows the receiving body parts to remain in a fixed location while driving, as compared to on the steering wheel (Fitch et al., 2007). The vibration alerts stimulated either the left, right, or non-directional (left and right simultaneously) thighs for a duration of 200 ms. Vibration alerts were presented when the TTC was either late (3 s) or early (7 s)—chosen to represent high or low urgent scenarios, respectively (Huang & Pitts, 2022).

The present study employed a 3 (vibration direction: left, right, or non-directional) × 2 (TTC; 3 or 7 s) within-subjects design to examine reaction time (RT) and wheel response direction made by participants when responding to vibration alerts during a semi-autonomous driving scenario. Before the experimental task, participants completed a pre-task demographics survey. For the task, participants were told to “imagine they are driving a semiautonomous vehicle that is usually in self-driving mode. However, sometimes the vehicle will not know what to do in certain situations (such as a gravel pile blocking the lane), which will require a response from the driver” (similar to Chen et al., 2022). Participants were instructed to steer as fast as possible in the direction of their choice when seeing the hazard and feeling the vibration indicating the gravel pile in the middle lane. Participants were required to keep their hands at the 9 and 3 o’clock position on the steering wheel. Vibration alerts and TTC occurred randomly but had an equal amount for each experimental condition throughout the practice (18 trials) and experimental (72 trials) blocks.

Outcome

An analysis of variance (ANOVA) was conducted with vibration direction and TTC as within-subjects factors on RT. There was a significant difference in RT between late (3 s) and early (7 s) TTCs, indicating faster responses for late TTC alerts due to the higher urgency of the signal compared to the lower urgency of the early TTC alert. There was also a significant main effect of the vibration side, revealing a difference between left, right, and non-directional vibration alerts. The post hoc comparisons revealed a nonsignificant difference in RT between left and right, left and both, and right and both. There was no significant interaction between TTC and vibration alerts, suggesting TTC did not influence RT based on the directionality of the vibration.

Participants tended to respond more often in the same direction of the vibration: more left turns than right turns for left vibrations; more right turns than left turns for right vibrations. For the non-directional vibration alert, the left and right turns were similar. A mixed-effects logistic regression model was conducted with TTC and vibration side on wheel response direction. The model revealed that a right vibration side was a significant predictor for a wheel response to turn right, meaning that when a right vibration occurred, the odds of turning the steering wheel to the right increased compared to when a left vibration occurred. The non-directional (both) vibration was a significant predictor for a wheel response to turn right, meaning that when receiving a non-direction vibration compared to a left vibration, the odds of turning the steering wheel to the right increased. For early versus late TTC, TTC was not a significant predictor for steering wheel response.

Conclusion

The design of vibration alerts in ground vehicles should elicit natural, fast, and accurate responses from drivers. Unlike most prior studies, which required forced correct responses, this study allowed participants to freely choose between two equally safe responses. This paradigm enabled the testing of participants’ natural response tendencies when prompted by vibration alerts. Participants chose to steer more often using a compatible rather than an incompatible mapping based on the vibration alert location. Therefore, this study recommends that vibration alerts be designed to follow the more natural, compatible mapping to allow for accurate responses from drivers in urgent scenarios. Future studies should determine whether a lack of attention would change the current findings due to distracted driving, causing drivers to be less focused on the alerts being presented to them.