Shu Ma

Audio-visual pseudo-haptic feedback can convey roughness in VR, yet the psychophysical principles governing its effectiveness remain unclear. Across two studies (total N = 82), we applied Stevens' power law (R = kSⁿ) to quantify roughness discrimination under varying sensory configurations and auditory modulations. Study 1 (N = 32) compared four perceptual conditions—real multisensory baseline, real audio-visual, virtual audio-visual, and virtual visual-only—and found that adding friction sounds in VR improved the Stevens exponent by 30%, and that baseline and real audio-visual perception were equivalent, suggesting the remaining virtuality gap was attributable to limitations in visual rendering. A crossover distortion pattern—underestimation at coarse grits, overestimation at smooth grits—revealed a visual-auditory weighting asymmetry, and higher baseline sensitivity predicted larger virtual distortion. Study 2 showed that a theory-driven selective loudness scheme restored the exponent from 0.58 to 0.84, matching baseline with uniform benefit across users, whereas a bidirectional frequency scheme produced no meaningful improvement. These findings establish Stevens' power law as a diagnostic tool for pseudo-haptic design and delineate the asymmetry and individual differences that govern roughness perception in VR.

Informative warnings can improve takeover performance in automated vehicles, but improper or excessive information may hinder it. This study investigates the optimal amount of information in urgent and non-urgent autonomous takeover transitions. Forty-five drivers were randomly assigned to groups receiving different amounts of the selected takeover warning information elements. Each group completed simulated takeover tasks in four scenarios with different urgency. Takeover performance, situation awareness, and stress levels were measured. The results indicate that, in urgent situations, single-element assistance delays takeover response, while three elements increase failure risk. However, dual-element assistance yields faster response, equivalent awareness, and minimal stress. In non-urgent situations, having three elements leads to the best awareness and lowest stress compared with having fewer elements. In conclusion, an appropriate and non-overwhelming level of information facilitates urgent takeover performance, whereas greater information is advantageous in non-urgent cases. These findings provide insights into designing appropriate warning amount assistants for automated driving takeover requests across varying time-pressure situations.

The advantages of two-stage warnings have been validated. This study investigated how drivers' expectations of automated driving system capabilities and cognitive load affect their attention allocation and takeover performance when using a two-stage warning system in a Level 3 automated driving system. Thirty-two drivers participated in a driving simulation study. The results showed that drivers under high cognitive load had longer and more frequent fixation on the road, which suggested a cautious attention strategy. The high-expectation group gazed less on the road and got greater lateral deviation and maximum acceleration. Attention allocation of the high-expectation group was similar between warning stages but was more susceptible to cognitive load within the same stage. The two-stage warnings need to be designed to direct drivers' attention effectively.

The present study aimed to explore the effects of various tactile takeover requests (TORs) (i.e., tactile sliding TOR and traditional vibration TOR) on the takeover performance in an automated driving system. A tactile sliding motor device was developed to signal the sliding TOR on the seatback of a driving simulator. Twenty-five young drivers were recruited as participants. Four types of TOR patterns were adopted in the study: ipsilateral motor rotation (IR), contralateral motor rotation (CR), ipsilateral and contralateral motor rotation (ICR), and ipsilateral motor vibration (IV). The participants were required to sit on the seat and underwent the automated driving in a low- or high-complexity scenario, then one of the four types of TORs was triggered randomly. The participants were asked to make a lane change using the steering wheel as soon as possible. Objective measures and subjective evaluations were used to assess the takeover performance. Results showed that the participants exhibited a shorter steering response time and lane change time under the three tactile sliding TORs (compared to the traditional vibration TOR). In the high-complexity scenarios and low-complexity scenarios conditions, different result patterns appeared regarding the maximum lateral acceleration and situational awareness. Our findings suggested that the tactile sliding motor is a promising way to signal a TOR in an automated driving system.