Electrical, Vibrational, and Cooling Stimuli-Based Redirected Walking: Comparison of Various Vestibular Stimulation-Based Redirected Walking Systems

The experience of being able to move freely in the VR space is one of the most basic and essential interactions that a user can have with a virtual space. In particular, when moving using natural gait motion, the user can feel a higher sense of immersion in proprioception and kinesthetics [49]. However, the technology of walking in a virtual environment using natural human gait motion is often limited in terms of its usability because of the physical limitations of real space, such as furniture, pillars, and walls, which may pose a safety threat to users who are blindfolded with an HMD.

RDW, a technology that changes a user’s direction by mapping the movement of the real and virtual environments by modulating the user movement or structure of the virtual environment without the user noticing it, was proposed to overcome this limitation of real space [54]. In particular, a technique for changing the user’s direction by providing a gain, called the redirection gain, to the user’s movement in a virtual environment is widely known. Steinicke et al. first divided redirection gain into three types depending on the user’s movement gain: rotation, translation, and curvature gains [71] (Figure 2). Rotation gain(g_R) provides gain to the degree of rotation of the user in reality such that the rotation in the virtual environment does not match the rotation in reality. Rotation gain is defined as the ratio of the virtual rotation (R_virtual) to the real rotation (R_real) (i.e., \(g_{R}=\frac{R_{virtual}}{R_{real}}\)). Translation gain(g_T) allows users in a virtual environment to move a longer or shorter distance compared to that in reality by providing a gain to the straight distance covered by the user. The translation gain is the ratio of virtual translation (T_virtual) to the real translation (T_real) (i.e., \(g_{T}=\frac{T_{virtual}}{T_{real}}\)). Lastly, curvature gain (g_C) transforms the user’s real path into a curved arc shape by rotating the map left or right for a user that goes straight in the virtual environment. By utilizing the curvature gain, users who think they are going straight in the virtual environment can move in the real world to avoid obstacles and walls such that collisions can be avoided. The curvature gain is expressed as the reciprocal of the arc curvature (r) created by the user (i.e., \(g_{C}=\frac{1}{r}\)). Because curvature gain is one of the most efficient manipulations for spatial expansion of RDW [40], we preferentially applied our vestibular stimulation RDW system to curvature gain. After the system is verified for curvature gain, we plan to study its application to other redirection gains.

RDW technology creates inconsistencies between the movement of the real environment and that of the virtual environment to overcome the physical limitations of real space. However, these manipulations inevitably lead to visual–vestibular inconsistencies, that is, the greater the manipulation of visual information for more directional change, the greater is the visual–vestibular inconsistency. This increased inconsistency causes simulator sickness, thereby making the user aware of the RDW manipulation, which causes a decrease in presence [71]. Therefore, all RDW techniques have a DT, which limits manipulation. Furthermore, DT limits the size of the direction change of RDW technology and determines the limit of avoidance performance for obstacles such as walls and furniture. Therefore, many studies have been conducted to measure and expand DT.

Steinicke et al. first presented the concept of DT and measured it using a two-alternative forced-choice (2 AFC) questionnaire, such as "Was the virtual movement smaller or greater than the physical movement?" and "Is the physical path bent left, or right?", to the participants by changing the rotation, translation, and curvature gains [71]. Participants were asked to choose one of two answers such as smaller/greater or left/right. Then, the gain value was defined as the point of subjective equality (PSE), which perceived physical and virtual movements equally based on the average participant responses. On both sides of the PSE, the points where the participant could distinguish between physical and virtual movements within a range of 75% were set as the upper DT (UDT) and lower DT (LDT).

This study used direction-based and noise-based methods to alleviate visual–vestibular inconsistencies using vestibular stimulations. Direction-based methods (directional GVS) relieve inconsistencies by providing vestibular stimulation consistent with visual information, whereas noise-based methods (noisy GVS, BCV, and CVS) alleviate inconsistencies by utilizing multisensory integration characteristics. In this section, we describe how noise-based methods can alleviate visual–vestibular inconsistencies.

Multisensory integration evaluates how information obtained from different sensory modalities is combined to influence decision-making and judgment [69]. Among the several multisensory integration studies, we focused on those based on the maximum likelihood estimation (MLE) model. In multisensory integration based on the MLE model, each modality independently influences the user’s judgment, and the weight of each modality is based on its relative reliability. The integration model to which the MLE was applied was first presented by experiments that applied visual and tactile senses [17]. In this study, the authors simultaneously provided conflicting visual and tactile information to the participants and found that the more noise there was in the visual information, the higher was the weight of the tactile information. We confirmed that this multisensory integration feature could be used to alleviate visual–vestibular inconsistency using vestibular stimulation, which applied noise to vestibular information [25, 80]. Each vestibular stimulus creates noise in the vestibular information, and vestibular information has relatively low reliability in multisensory integration compared to visual information. Therefore, vestibular information has a small weight compared to visual information, which can alleviate the inconsistencies by creating a higher dependence on visual information in visual–vestibular inconsistencies. A previous related study attempted to expand the DT by creating noise in the vestibular information using the noisy GVS [40]. A DT expansion performance of approximately 12%–16% was confirmed.

Stable gait refers to a gait where the user does not fall when a perturbation occurs [7]. If the gain stability is low, the user falls easily, increasing the risk of injury. In particular, when wearing an HMD, the user cannot observe the external environment; therefore, low gait stability is fatal to the user. According to previous studies, both VR and RDW applications induce a decrease in gait stability [27, 29, 46]. As described above, gait stability, an essential factor in RDW, was checked to verify the safety of the RDW system.

Gait stability varies depending on the intensity of the noise applied to the vestibular system. In the case of a weak noise stimulus, stochastic resonance in the vestibular organs increases the gait stability; however, with an intense stimulus, noise disturbs vestibular sensation, thereby decreasing the gait stability [77]. Noisy GVS, BCV, and CVS used in the present experiment expand the DT by perturbing the user’s vestibular organs through intense stimulation. Therefore, because stimulation may worsen a participant’s gait stability, its stability must be experimentally confirmed.

The sensor measures the user’s stability by measuring the pressure applied to the sole of the foot. There are two methods for measuring plantar pressure: the platform method, where a pressure sensor is installed on the floor, and an in-shoe method, where a sensor is installed in a shoe [53]. In this experiment, we measured plantar pressure using the in-shoe method to adapt the participant’s walks on different paths depending on the curvature gain. In the case of the in-shoe method, the stability can be measured without affecting walking [60], considering that a fabric pressure sensor is placed in the shoe without connecting cables or wearing additional equipment. In this study, anterior/posterior (A/P) and media/lateral (M/L) gait stability was measured using plantar pressure measurements [36]. According to previous studies, a stable gait center of pressure (CoP) moves from posterior to anterior. As gait becomes unstable, the time taken to move from the anterior to the posterior increases. Therefore, we measured the A/P gait instability by dividing the number of frames with A/P CoP motion toward the heel during stride by the total number of frames in the stride length. Additionally, during stable walking, the CoP starts from the middle area, moves in the lateral direction, and returns to the middle-area direction. At this time, the more unstable the gait is, the more the CoP moves to the left and right. Therefore, we measured the M/L instability when the velocity–time curves exceed ±0.5 mm/s per step after zero-crossing the M/L CoP velocity–time curve.

The GVS device used in this study was designed to provide both directional and noisy GVS stimulation of four-pole GVS (Figure 4). Four electrodes were attached to the left and right temple and mastoid. Electrical stimulation was applied at 15 V and 2 mA. The GVS device was divided into control and stimulus sections, and each part was galvanically isolated for safety. The microcontroller unit (MCU) communicates information through BLE communication with an experimental environment created using Unity. The built-in IMU sensor sends information regarding the user’s head orientation to Unity. Then, based on information received and the experimental environment, Unity calculates the type of stimulus to be applied to the user and transmits it to the GVS device. The direction and intensity of current stimulation was controlled using the H-bridge of the L293D motor driver and pulse-width modulation, respectively. All electrical stimulations were applied using a ramp function to prevent unintentional tingling, and a first-order low-pass filter with a cutoff frequency of 33 Hz was installed. This stimulus was delivered to the user via a conductive bioelectrode through a 3.5 mm connector. All devices were powered by an 8000 mAh LiPo battery and controlled by a single power switch, thereby allowing the participant to shut down quickly at any time for safety. Signals corresponding to the left and right temples can be turned on and off using the SSR relay, which is controlled to ensure that electrical signals are not sent to the temples in LDS and noisy GVS situations. All devices were designed using open-source hardware, and various sensors, such as heart-rate, electrocardiogram, and skin-conduction sensors, can be used for further research.

The noisy GVS signal uses white Gaussian noise, based on research by Matsumoto et al [40]. To prevent stochastic resonance of the noisy GVS, as described in Section 2.4.1, we provided stimulation with a current density of 0.8 A/m² or higher and below the safety standard. The average current stimulation and standard deviation were set to 0 mA and 1 mA, respectively. All noises were designed to be within the range of 2σ. The electrode used was a circular electrode with a radius of 2.5 cm; therefore, the applied current density was 1.02 A/m². The four-pole GVS can modulate the user’s vestibular information in the three directions of roll, pitch, and yaw in the user’s head coordinate system using three current-stimulation directions: LDS, SDAS, and ODAS. However, the head coordinate system of the user does not always coincide with the global coordinate system. For example, as shown in Figure 5, if LDS stimulation is applied when the user is facing the front, vestibular stimulation in the roll direction can be applied in the global coordinate system. However, when a user looks at the floor or ceiling, the same stimulus creates a yaw-direction vestibular stimulus in the global coordinate system. Therefore, we developed a system that can stimulate the user in the desired direction in the global coordinate system by receiving the user’s head-direction information using IMU data built into the device.

All GVS stimulations used in this experiment were conducted to alleviate visual–vestibular inconsistency. Noisy GVS is triggered when visual information is manipulated to alleviate inconsistency based on the characteristics of multisensory integration. The triggered device provides continuous electrical noise to the vestibular system. Then, directional stimulation of the four-pole GVS was provided to match the user’s visual and vestibular information. For example, consider a situation where the user experiences a curvature gain while moving straight in the virtual environment while looking straight ahead. At this time, the visual system transmits information to the brain that the user is going straight without curvature in the left and right directions. Simultaneously, the vestibular sense transmits information to the brain, where the body rotates toward the yaw-directed curvature gain, thereby resulting in visual–vestibular inconsistencies. In this situation, we provided the user with a vestibular stimulus in the same direction in the global coordinate system to alleviate the visual–vestibular inconsistency.

We manufactured the BCV device (Figure 6) using the bone-conduction module of a human-bone conduction speaker (Duramobi, Hong Kong), which communicates with the experimental environment in real time using Bluetooth 5.0. It can produce sound of up to 115 dB and has a sound frequency range of 200 Hz–-20 kHz. We used two units, one for each mastoid; each unit was 40 mm in diameter and weighed 35 g. Each BCV unit has an impedance of 4 Ω and an output of 3 W,and is driven by a Bluetooth audio signal amplified by the HAA2018 5W audio amplifier chip (Figure 6.(b)). An elastic band was attached to the vibration unit of the participant’s mastoid process, and the position of the module was adjusted based on the participant. In attaching the vibrator, we were careful not to place the vibrator on the electrode of the GVS attached to the same mastoid or on the hair of the participant such that the vibration was not transmitted properly. If the participant’s mastoid area was narrow and the electrode positions of the BCV and GVS interfered, the position of each stimulation device was modified according to the experimental conditions. This process corrected the position such that the stimulation under each experimental condition could be clearly felt. Moreover, before every experiment, we asked participants whether the stimuli were well delivered.

In this experiment, we triggered BCV upon visual manipulation using RDW to alleviate visual–vestibular inconsistencies. The frequency of the BCV used in this experiment was 500 Hz, which is the range that generated the largest VEMPS in previous studies. Considering that the intensity of the BCV for vestibular stimulation is significantly different based on the participant’s characteristics, the intensity of the stimulus depending on the participant [81] must be controlled. We used the following calibration method for the participant’s comfort while applying vestibular stimulation: The vibration intensity was divided into 15 stages by setting the minimum and maximum outputs to 1 and 15, respectively. After applying the minimum-vibration-intensity stimulation to each participant, they were asked if they could tolerate the strength of the stimulation. If not, the intensity was decreased by one step, and, if possible, the intensity was increased by one step. Then, the same process was repeated. Thus, we obtained the maximum vibration intensity within the comfortable range for each participant.

The devices used in this experiment were a TES1-7102 thermoelectric module (Figure 7.(c)) to stably produce cold air at 24 °C. TES1-7102 is a 20 mm-sized thermoelectric module with a power consumption of 10.3 W, and the temperature difference between the high and low-temperature parts is up to 65 °C. To ensure stable performance of the thermoelectric module, a fan driven by a DC motor was attached to the heating part of the device to release heat to the back of the participant’s neck (Figure 7.(c)). As shown in Figure 7.(a, d), the BLDC motor draws air from the vents and sends the air through the silicone tube to the thermoelectric module. The YWY-002F BLDC motor (Figure 7.(b)) was used to generate wind with low noise. The acoustic-noise level caused by cold air was approximately 65–70 dB. The cold air in contact with the cooling part of the thermoelectric module entered the participant’s external auditory meatus via a silicone tube. An elastic band was used to secure the silicone tube to the ear, and a 3D printed plastic tip was used to apply cooled air stimulation to the participant’s ear external auditory meatus. To insulate the device’s cold air and ensure the participants’ comfort, the device was coated on the outside with polymer clay (Figure 7.(e)). We conducted a preliminary test to confirm that our device can reliably produce cold air at 24 degrees. After mounting the CVS device on the head model, we attached a thermometer to the nozzle going into the participant’s ear, and checked the temperature change of the air entering the participant’s ear for 30 minutes. The preliminary test confirmed that at an external temperature of 26 °C, our device stably generates cooled air at 24 °C for 30 min, excluding the initial approximately 1 min required for cooling the thermoelectric module (All experiments in this study were conducted within 30 min). Therefore, first, the thermoelectric module was operated for 1–2 min prior to all CVS experiments to consistently create cool air.

The CVS used in this experiment was based on a previous study where cold air at 24 °C was blown into each ear through a nozzle into the external auditory meatus. The total airflow rate was set at 5 L/min. If the participant felt uncomfortable because the wind was too strong, the flow rate was reduced to 4 L/min. The outside temperature was always maintained at 26 °C during the experiment using an air conditioner. CVS is also triggered when visual information is modulated in the same manner as other noise-based vestibular stimulations.

In this study, we implemented a system that records plantar pressure in real time to measure A/P and M/L gait stability. Pressure was measured using an Opengo pressure sensor insole (Moticon, Germany) (Figure 8. We used three sensors of different sizes (approximately 240, 260, and 275 mm) depending on the participant’s foot size. The sensor had 16 capacitive pressure sensors attached to each foot, with a measurement range of 0–-50 N/cm² and a resolution of 0.25 N/cm. The area covered by the pressure sensors was 65% of the total area of the sensor insole. The sensor operates wirelessly and transmits the pressure data of each sensor via BLE communication. The transmitted data were stored at a sampling rate of 50 Hz, and the stored data were analyzed using MATLAB.

In E1, we measured DT, gain stability, simulation sickness, and discomfort when vestibular stimulation was used for the RDW. The experimental environment was configured based on the DT-measurement environment proposed by Steinicke et al [71]. A detailed description of the experimental environment is presented in this section. In E1, each participant experienced four types of vestibular stimulation (noisy GVS, directional GVS, BCV, and CVS), and a control condition in which no stimulation was applied for a total of five cases. To exclude unexpected results from wearing the equipment, the equipment was worn under all conditions. We randomized the experimental sequence of all stimulation cases using Latin squire. The participant experienced nine different curvature gains (±π/180, ±π/90, ±π/60, ±π/45, and 0) in the experimental environment. Each curvature gain was randomly applied 45 times (five times per case). The total number of participants recruited in E1 was 20 (N=20, age range: 20-–28, M = 23.05, SD = 2.33). No participant had a history of epilepsy, vestibular system disorders, migraine, brain injury, cardiovascular disorders, central-nervous-system abnormalities, or sensitive skin. Pregnant women and people with pacemakers were excluded from this study. Moreover, no participant had motion-sickness symptoms for 3D games, and there were no problems with normal walking while wearing the HMD.

Questionnaire. We used the 2-AFC questionnaire used by Steinicke et al. to measure DT in this experiment [71]. Participants answered left or right to “Is the physical path bent left or right?” on the in-game VR per trial. We used the simulator sickness questionnaire (SSQ) to measure the simulation sickness once for each stimulus [30]. Furthermore, simulator sickness was measured once in the initial state before all the experiments. To measure the user’s discomfort in the game, we used Fernandes and Feiner’s discomfort score measurement questionnaire [18]. The participant answered, “On a scale of 0–-10, zero being how you felt coming in, ten is that you want to stop, where are you now?” for discomfort with the VR controller.

Procedure. The participants changed shoes with pressure sensors to measure gait stability and answer the SSQ. After completing the questionnaire, the participants wore the stimulation device and calibrated the strength of the BCV. When the experiment started after taking sufficient rest after calibration, the participants started from the green disk shown in Figure 9 (a) and moved forward along the green path. All curvature gains were simultaneously applied after a straight path of 1.5 m to prevent the participants from noticing the gain. The participants walked along the path where a curvature gain of 5 m was applied (a total of 6.5 m walking) and stopped when the questionnaire screen appeared. The questionnaire comprised the 2-AFC questionnaire to measure DT, and Fernandes and Feiner’s discomfort score to measure discomfort. After the survey, the participants returned to the initial point indicated by the green disk shown in Figure 9 (a) and repeated the same task. When one case was completed, the participants removed the HMD, answered the SSQ for the case, and freely wrote about whether or not the stimulus was uncomfortable and the reason for discomfort. After the questionnaire, the participants were provided sufficient rest before conducting the experiment on the next stimulus case.

Through the E1 experiment, we confirmed the DT expansion performance, gait stability, simulator sickness, and discomfort felt by the users for each stimulus when each vestibular stimulus was used in the RDW. First, noisy GVS had the lowest user discomfort. However, the DT-expansion performance of noisy GVS was the worst; therefore, expecting a significant effect in space expansion or obstacle avoidance using RDW is difficult. In addition, noisy GVS has high gait instability, which may cause safety problems for the users. In contrast, directional GVS has higher user discomfort than noisy GVS. However, the DT-expansion performance was the best; therefore, using RDW to perform well in space expansion and obstacle avoidance can be expected. Furthermore, directional GVS has the lowest gait instability; hence, it can support a user in stably and safely walking in a virtual environment. Next, BCV has the advantage that it can be used as a non-electrical alternative for people who cannot use GVS, which is an electrical stimulus. A promising RDW performance can be expected because the DT-expansion performance of BCV is also high. However, similar to other methods of providing noise to the vestibular system, there is a high gait instability, which may cause safety problems for the user and has a disadvantage of high discomfort felt by the user. Finally, similar to BCV, CVS can be used as a non-electrical alternative to GVS, an electrical stimulation, but it induces significantly higher simulator sickness compared to other conditions. In addition, good usability and safety cannot be assured using CVS because of high discomfort and gait instability. We analyzed these shortcomings of CVS based on a survey of participants, which will be dealt with in the Discussion section. Based on the E1 results, the answers to H1 and H2 were obtained. For H1, we confirmed our hypothesis as BCV and CVS extended the curvature DT by 20.18% and 18.35%, respectively. Moreover, for H2, our hypothesis was confirmed because direction-based stimulation produced a more stable gait than noisy GVS, BCV in A/P gait stability, and all noise-based stimulations in M/L gait stability.

In this experiment, the experimental environment was constructed based on the environment used in the live-user experiment of RDW by Sun et al. and Hodgson et al [26, 72]. Each study used a game of finding a stick or a ball. Thus, the user could freely move around as much as possible in an environment where RDW was applied. Previous studies used fog or adjusted the transparency of the ball to encourage the user’s free movement. In the present study, the same method as that of Sun et al. was used to control the transparency of the ball [72]. The transparency of the ball decreases in proportion to the distance between the user and the ball. Therefore, in a place far away from the ball, the ball is transparent, and the user has to randomly search the map to find the ball, moving around the map in various ways. In addition, we made the ball make a sound so that the user could enjoy the game using various modalities. The ball provides a stereo sound according to the location, and the user can search the ball by inferring the location where the sound is heard. The user walked freely in a virtual space of 5 m X 5 m in total and searched for a ball for three minutes per trial. The remaining time and the number of balls found are displayed on the UI at the bottom of the participant, encouraging participants to actively participate in the game. The total number of balls found in each condition was recorded as task performance and was used to indicate how well the participant performed the task required in the game. Participants experimented with each of the four vestibular stimuli and one non-stimulus control condition, and all trials were randomized to a Latin square. A total of 27 participants were recruited in E2 (N = 20, age range 20–36, M = 23.89, SD = 3.17). As in E1, no participant had history of epilepsy, vestibular system-related migraine, brain injury, cardiovascular disorder, central nervous system abnormality, or sensitive skin. Pregnant women and pacemaker wearers were excluded from recruitment. No participant had motion-sickness symptoms related to 3D games and VR, and there was no problem with walking for a long time while wearing an HMD.

Questionnaire. In this experiment, the following questionnaire was used to measure the immersion and presence, simulator sickness, and discomfort over time. For immersion and presence, the Igroup Presence Questionnaire (IPQ) was used, and three sub-scales (spatial presence, involvement, realism) were used [57]. Spatial presence measures the sense of being physically in the virtual environment. Involvement indicates the attention devoted to the VE and the involvement experienced. Experienced realism indicates how much the participant’s experience in the virtual environment appear like the real one. Simulator sickness was measured using SSQ. IPQ and SSQ were measured once after the experiment for each condition. Moreover, SSQ was measured once before all experiments to measure the initial state. We used Fernandes and Feiner’s discomfort score measurement questionnaire every 30 seconds in-game to measure discomfort over time. The participant submitted an answer to "On a scale of 0–10, zero being how you felt coming in, ten is that you want to stop, where are you now?" about their discomfort while finding the ball. The timer for the remaining time during the questionnaire was stopped so the participant could think enough about the questionnaire.

Procedure. Participants first answered the simulator sickness in the initial state through the SSQ questionnaire. After the participant wore the vestibular stimulation device, the BCV strength was calibrated. After providing sufficient rest time after calibration, the participant stood in the middle of the room and waited for the experimental environment to start. After the experiment started, the participant explored the virtual environment using sight and sound to find as many balls as possible within the 3 min time limit. When a discomfort questionnaire appeared every 30 s during navigation, the participant stopped, answered the questionnaire, and then immediately resumed the game. After the time limit, the participant took off the HMD, answered questions, such as IPQ and SSQ, and had sufficient rest time. After sufficient rest, the participants proceeded with the experiment under the following conditions.

We used RDW with vestibular stimulation in a game scenario similar to the environment experienced by real VR users in the E2 experiment, distinct from the E1 experiment. We implemented a game in which users find objects using their visual and auditory senses to assume a situation in which users experience actual VR content. We investigated users’ immersion and presence, task performance, simulator sickness, and discomfort with vestibular stimulation RDW in an actual game environment. We confirmed that vestibular stimulation did not impair users’ sense of immersion.

First, the noisy GVS showed excellent task performance and reduced simulator sickness and discomfort. Next, the directional GVS showed a statistically similar value to the noisy GVS in terms of task performance and simulator sickness; however, user discomfort was significantly higher than in the noisy GVS. Next, BCV induced the most simulator sickness and discomfort in the game environment. We confirmed that the vibration and sound of the BCV caused significant discomfort to the users. However, BCV did not interfere with the user’s achievement of the game’s goal, nor did it interfere with the sense of immersion. Therefore, improving the vibration and noise of the BCV is expected to reduce BCV discomfort. Finally, although CVS performed poorly on the task, it was comparable to directional GVS in terms of simulator sickness and discomfort. However, characteristic wind noise of the CVS prevents the user from achieving the game’s goal using hearing. It will be possible to use the CVS as an immersive RDW system if the wind noise of the CVS can be effectively reduced.

H3 was concluded from the results of E2. Non-electrical stimulations indicated relatively high discomfort and motion sickness. However, the vibration stimulation of BCV did not cause low task performance, and BCV and CVS did not diminish users’ sense of immersion. Directional GVS induced a relatively high sense of motion sickness and discomfort among the electrical stimulations; therefore, our hypothesis H3 was false. This is analyzed in detail in the Discussion section.

We first compared electrical stimulation (noisy GVS, directional GVS) with non-electrical stimulation (BCV, CVS), and the most significant difference between the two groups was the side effects of electrical stimulation of the skin (e.g., skin, retina stimulation, eye muscle stimulation) [37, 76]. Non-electrical stimulation is a technique that can be considered for multiple users who are prone to side effects when electrical stimulation is applied to their skin [12]. Non-electrical stimulation uses vibration and cold air instead of electrical stimulation, which the participants primarily perceive as auditory noise. The BCV applies a vibration of 500 Hz within the human audible frequency. The CVS blows wind into the external auditory meatus, causing the vibration of the eardrum. Therefore, the participant experienced noise from all non-electrical stimulations, which significantly affected discomfort.

Figure 20 shows the discomfort scores for each vestibular stimulus in E1 and E2. The graphs demonstrate the difference in discomfort between E1 and E2. We observed that discomfort in the E2 condition was higher than that in the E1 condition for all non-electrical stimulations. While additional analysis is required to determine whether this is significantly attributable to the sample difference in the two conditions, the data collected in each condition represent the entire population because statistically they satisfy normality [45], and the results still have informative implications in terms of showing the difference in user experience according to the environment. We analyzed this cause as auditory noise produced by the BCV and CVS. E2 was designed to use both sight and hearing to set up a more general VR gaming environment compared with E1. Therefore, users must continuously use what they hear to enjoy the game in E2. In this situation, non-electrical stimulations continuously provide noise to users, which induces audio interferences. Humans exhibit a high cognitive load when auditory elements with multiple pieces of information are simultaneously reproduced [78], and they experience a decrease in concentration and comfort [39, 79]. These audio interferences in the E2 environment of this experiment increased the cognitive load and caused discomfort. This can be verified by the user’s answer to the reason for discomfort. The users felt “tingling” as discomfort caused by electrical stimulation, which was mostly the same for E1 and E2. However, the answers regarding the discomfort due to non-electrical stimulations were different in E1 and E2. Constant sound in E1 and other non-electrical stimulations, such as the vibration of BCV, were the primary contributors to discomfort. However, in E2, most participants answered that the sound was aurally uncomfortable while enjoying the game. This indicates interference between the auditory elements of the game and the noise induced by non-electrical stimulations, and the cognitive load induced by this interference caused the discomfort.

Audio interferences could be used to investigate why non-electrical stimulations cause relatively high discomfort at E2. However, we observed higher discomfort in the directional GVS in E2 although it was an electrical stimulation. This is because of the difference between direction- and noise-based stimulations, as described in the next section.

In this study, we alleviated the visual–vestibular inconsistencies in two ways: 1) Direction-based stimulations (directional GVS) provide vestibular information consistent with visual information, and 2) noise-based stimulations (noisy GVS, BCV, CVS) add noise to the vestibular information. The two methods exhibited a difference in gait stability, as discussed in Sections 4.2.2–4.2.3. Noise-based stimulations reduce the reliability of vestibular information, thereby alleviating inconsistency by reducing the weight of decision-making. This property restricts the use of vestibular information in gait, and, similar to previous studies, it was possible to confirm the result of inducing an unstable gait. In this section, we focus on how the presence or absence of directionality affects the discomfort and simulator sickness experienced by users in E1 and E2.

First, we analyzed discomfort. As can be observed in Figure 20, the discomfort to directional GVS, which is a direction-based stimulation, was higher in E2 than in E1. However, the reason for the discomfort answered by the user was the same, “tingling,” in both E1 and E2. Therefore, we interpreted this increase in discomfort as an increase in the tingling frequency. The participant experienced a constant curvature gain during one trial in E1. Therefore, the direction and amount of visual manipulation felt by the user during each trial were constant. By contrast, E2 constantly changed the amount and direction of the curvature gain to redirect the user to the center of the room. Therefore, the user experienced considerable changes in the direction and intensity of visual manipulation during one trial, and the change in the direction of current also occurred frequently following visual manipulation. Furthermore, the current direction changed more frequently because, in E2, the user had to turn their head in various directions to find the ball compared to the game in E1. Therefore, the current direction in E2 changed more frequently than in E1, and the user would experience more frequent tingling because of the frequently changing current direction.

Next, we analyzed the changes in simulator sickness at E1 and E2 with direction- and noise-based stimulations. As can be seen in the Figure 21, all noise-based stimulations decreased simulator sickness in E2 compared with E1, but the direction-based stimulations were almost the same. We believe that E2 had lower simulator sickness than E1 in most situations because of the difference in the time over which users experienced visual–vestibular inconsistency in RDW. The average time with RDW was 10–20 min for E1 and 3–4 min for E2. Therefore, the participants were less exposed to visual–vestibular inconsistency in E2 and showed relatively low simulator sickness compared to E1. However, direction-based stimulation induced relatively high simulator sickness at E2. We interpreted this to be because of frequent switching of the curvature gain at E2. There is a disadvantage in that it is challenging to determine how much current must be utilized to offer accurate vestibular information consistent with visual information [80]. A slight difference between the vestibular information modulated by directional GVS and visual information can significantly affect a participant’s performance and comfort [37]. Although the direction of the curvature gain is the same as that of the vestibular information applied by the directional GVS, it may cause sensory inconsistency because the intensity changes continuously in E2. Therefore, direction-based stimulation is more susceptible to simulator sickness than noise-based stimulation. However, the alleviation of simulator sickness can be expected if the intensity of direction-based stimulations can be changed to match the curvature gain.