Effects of Image Scale and System Time Delay on Simulator Sickness within Head-Coupled Virtual Environments.
Simulator sickness is a variant of motion sickness that often results from exposure to simulated environments such as flight simulators, driving simulators, and virtual environments (VEs). Whereas motion sickness refers to the adverse consequences of exposure to environments that physically displace an individual, simulator sickness is mainly the result of technological limitations in simulating dynamic environments (Kennedy, Hettinger, & Lilienthal, 1990; Pausch, Crea, & Conway, 1992). Computer simulations of a dynamic environment are necessarily imperfect because (a) simulations can reproduce only a subset of all the information resident in the real environment; (b) this information is often changed or distorted during the simulation process; and (c) additional, superfluous information is often unavoidably included because of current technological and modeling limitations.
Thus current-generation head-coupled VEs are imperfect simulations of human-world interactions (Durlach & Mayor, 1995; Peli, 1995). Conflicting cues of self-motion and spatial orientation occur as unavoidable side-effects of existing technology. Examples of these imperfections include optical deficiencies, image scale factor magnifications, system time delays, limited display field of view (DFOV), head tracker inaccuracies, and so on. Many of these imperfections, when combined with the high-bandwidth multisensory coupling between human and computer, are theorized to induce simulator sickness (McCauley & Sharkey, 1992; Regan, 1995; Rushton, Mon-Williams, & Wann, 1994). The resulting symptoms include headache, nausea, dizziness, stomach discomfort, eyestrain, oscillopsia, and postural ataxia (Kennedy et al., 1990).
Poorly designed virtual interfaces can decrease performance in VEs and have health, safety, and legal ramifications (Peli, 1995; Viirre, 1994). Discomfort experienced during exposure to these environments may also result in decreased time spent in the environment, decreased desire to repeat the experience in the future, and degradations in transfer of training. A thorough analysis of how humans are influenced by particular virtual interfaces is critical before virtual interfaces can be properly designed and administered (Durlach & Mavor, 1995). Specific stimulus rearrangements that commonly appear in virtual interfaces should be evaluated for their relative potential to drive adaptation processes and induce simulator sickness. In the following sections, we summarize a few theories of motion/simulator sickness, list some of the causal factors previously identified, and describe the potential causal factors investigated in the experiments detailed here.
Motion Sickness/Simulator Sickness Theories
A universally accepted theory of motion sickness/simulator sickness does not currently exist. Current theories lack the specificity required to make solid, testable predictions. With that in mind, this section briefly describes some of the predominant simulator sickness theories in order to illustrate the current level of theory development and to provide a framework for discussing the results of the following experiments.
The most widely recognized theory of motion sickness (and, by extension, simulator sickness) is the sensory rearrangement theory. This theory, as presented by Reason (1978), states that "all situations which provoke motion sickness are characterized by a condition of sensory rearrangement in which the motion signals transmitted by the eyes, the vestibular system and the nonvestibular proprioceptors are at variance either with one another or with what is expected based upon previous experience" (p. 820). Applications of this theory to VEs often focus on mismatches involving visual and vestibular cues to self-motion.
Sensory rearrangement theory offers explanatory value but little prediction regarding which sensory rearrangements will result in sickness symptoms and which will not, and it is unclear as to what operationally constitutes a "sensory rearrangement" (Ebenholtz, Cohen, & Linder, 1994; Griffin, 1990; Stoffregen & Riccio, 1991). This theory also does not answer the question of why sensory rearrangements should be nauseogenic, though Treisman (1977) presented one possible explanation with his evolutionary hypothesis (i.e., that motion sickness is an accidental triggering of a system that rids an individual of ingested neurotoxins).
A variant of the sensory rearrangement theory, termed the subjective vertical theory, attempts to define a priori which sensory rearrangements are nauseogenic. This theory asserts that all situations which provoke motion sickness involve a rearrangement, or variance, between the sensed vertical (as determined by integrated information from the eyes, the vestibular system, and non-vestibular proprioceptors) and the subjective vertical (Bles, Bos, de Graf, Groen, & Wertheim, 1998). Thus sensory conflicts resulting from yaw motions/rotations are not assumed to be provocative stimuli. Though this theory adds a level of detail beyond that provided by the sensory rearrangement theory, ambiguity remains regarding what operationally constitutes a provocative sensory rearrangement.
An alternative to the sensory rearrangement theory is the ecological theory of motion sickness (Riccio & Stoffregen, 1991), which states that environments creating an extended period of postural instability will produce sickness symptoms. This theory focuses on behavior of the individual instead of patterns of sensory inputs, and it assumes that early-onset postural instability is a causal factor and not a preliminary symptom of simulator sickness. Additionally this theory does not readily predict which specific environments will create extended periods of postural instability and, by extension, simulator sickness. Thus in its current state this theory has limited utility as a tool for applied interface design.
Previous Simulator Sickness Research
Previous research has identified many factors that influence the occurrence and severity of simulator sickness. Three significant variables include past experience with the simulated environment, exposure time, and individual factors. Other identified factors include age, gender, display field of view, binocular viewing, degree of motion control, and acceleration characteristics. Detailed descriptions of known factors and their potential modulating effects can be found in Kennedy et al. (1990) and Pausch et al. (1992).
As stated earlier, VE applications cannot fully replicate the veridical perceptual dynamics associated with self-motion. Regardless of any specific underlying theory, the resulting imperfections in the simulated environment are believed to be potential causal factors for simulator sickness (Kennedy et al., 1990). Two such imperfections that are common in current virtual interfaces are image scale magnifications and system time delays.
Image scale magnifications. Before a virtual image can be displayed, a computer graphics system must first ascertain which part of the virtual world is to be viewed by the observer. In perspective projection, geometric field of view (GFOV) essentially defines the horizontal and vertical boundaries of the scene along with its aspect ratio. GFOV is the angle subtended by the viewport from the center of projection in virtual space or, equivalently, the angle subtended by the viewing frustum (Danas, 1995; Foley, van Dam, Feiner, & Hughes, 1996).
The image projected onto the viewport in virtual space will subsequently be rendered on the physical display of the virtual interface. This is where GFOV and display field of view (DFOV) are integrated, creating an image scale magnification factor. As illustrated in Figure 1, if the GFOV angle is less than the corresponding DFOV angle, the viewport image will appear magnified on the physical display because of the requirement for the image to fill a larger subtended angle in real space versus virtual space. In head-coupled interfaces, visual optic flow rate in response to head movement will also increase proportionately. Conversely, if the GFOV is greater than the DFOV, the resulting displayed image will appear miniaturized and the visual optic flow rate will decrease proportionately. Finally, if GFOV is equal to DFOV, the image on the viewport will map correctly from virtual space onto the physical display such that the spatial relationships in the VE are maintained.
GFOV manipulations (while holding DFOV constant) thus provide a controllable optical distortion resulting in novel visual-vestibular patterns in head-coupled virtual interfaces. For clarity, the term image scale is used to describe the ratio DFOV/GFOV. This ratio can be thought of as a magnification factor. If this ratio equals 1.0, the individual will perceive a spatially accurate image, as defined by the spatial dimensions of the virtual space model. If the image scale deviates from 1.0, the result will be a change in optic flow velocity during rotational motion and either image magnification (if GFOV [less than] DFOV) or image minification (if GFOV [greater than] DFOV) of the scene.
Although researchers have investigated the effects of GFOV on performance and level of presence experienced (Boyer & Wickens, 1994; Hendrix, 1994), few studies have explicitly considered the relationship between GFOV and DFOV. However, Danas (1995) did examine the effects of DFOV/GFOV ratios on auditory perception in VEs.
System time delays. Time delays exist between movements made by the user (which are tracked by a position-sensing device) and the response of the virtual interface system to those movements (Durlach & Mayor, 1995; Wloka, 1995). The most conspicuous time delay in head-coupled virtual reality (VR) is the delay between a participant's head motion and the corresponding response of the virtual scene. The measured time delay between head movement and scene response is the result of several smaller time delays within the system (Wloka). Much of the system time delay is fixed and determined by the particular hardware chosen. For instance, mechanical trackers have total throughput delays of approximately 3 to 10 ins, whereas electromagnetic trackers can range from 10 to 250 ms or more, depending on choice of filtering, configuration, and so on (Meyer, Applewhite, & Biocca, 1992). A portion of the system time delay is variable, however. For example, when the visual scene becomes more complex, the speed of the image-ren dering process will likely decrease, increasing system time delay.
System time delays are hypothesized to be among the areas most in need of improvement in simulators and virtual interface design (Ebenholtz, 1992; Kennedy et al., 1990). Large time delays between control input and visual scene update adversely influence task performance. So and Griffin (1995) found that head-tracking performance was significantly degraded at system time delays of 80 ms. Furthermore, time delays can affect user acceptance of the VE (Pausch et al., 1992).
The time delay between head movement and virtual image response also stimulates a rearrangement of visual and vestibular cues of motion. This rearrangement is hypothesized to be a significant contributor to simulator sickness and reflex alterations (Ebenholtz, 1992; Kennedy et al., 1990; Peli, 1995). Previous research examined various time delays between a participant's control input and subsequent visual scene update in a simulator. Uliano, Lambert, Kennedy, and Sheppard (1986) found no effect of increasing time delay on simulator sickness, whereas Frank, Casali, and Wierwille (1988) found increased participant discomfort when visual motion stimuli lagged inertial motion stimuli in a driving simulator. The Frank et al. finding is interesting because head-coupled VEs (the focus of this paper) necessarily require that the visual motion stimuli lag the (self-generated) inertial stimulation, which would imply increased sickness in these conditions. However, the particular task involved (control inputs via a joys tick or steering wheel) in these previous studies was not as intimately coupled to visual scene response as is self-generated head motion, so the generalization of these results to head-coupled environments is unclear.
Two experiments were conducted to identify the effects of specific simulation imperfections on simulator sickness. Experiment 1 investigated the effects of different image scale factors (through DFOV/GFOV manipulations) while holding time delay constant at its minimum achievable value for this configuration (48 ins). Because optic flow rate varies with image scale factor whereas vestibular stimulation remains constant, it was hypothesized that simulator sickness reports would rise for visual conditions in which the DFOV/GFOV ratio (i.e., image scale factor) differed from 1.0.
In Experiment 2, pure system time delay (i.e., transmission lag) was varied while image scale was held constant at 1.0. It was hypothesized that increasing time delay would increase reported levels of simulator sickness.
EXPERIMENT 1: IMAGE SCALE EXPERIMENT
Participants. The participants in this experiment were 11 adult volunteers (6 men and 5 women, mean age 28.5 years, range 19-39 years). All participants reported good health with no medical history of epilepsy or vestibular pathology. Participants were tested for normal or corrected visual acuity of 20/30 or better and voluntarily abstained from alcohol and drugs for 12 hr prior to participating in each session. Nine participants reported that they had experienced motion sickness in the past, and two participants reported having experienced simulator sickness in the past. Two participants claimed not to be susceptible to motion sickness, six participants reported slight susceptibility, and three reported moderate susceptibility. At the onset of testing, all participants were new to the VE used in this experiment and had not experienced a head-coupled virtual interface in the previous 30 days. Seven participants had never before been exposed to a head-coupled virtual interface.
Materials and apparatus. This experiment was conducted in the VR Effects Laboratory, a component of the Human Interface Technology Laboratory at the University of Washington. A Virtual i/O i-glasses! (Virtual i/O, Seattle, Washington) head-mounted display (HMD) displayed a full-color, bi-ocular virtual image to the participant with an angular resolution of 6.8 arc mm/color pixel. This HMD has a 25[degrees] horizontal x 19[degrees] vertical DFOV, with 100% visual overlap. Viewing software (WARP TV; WARP Ltd., Sausalito, California) generated a 3600 cylindrical image that was precomputed and prerendered into memory at program initiation (Figure 2). When a participant moved his or her head during the experiment, the portion of the image that corresponded to the new head position was drawn on the display directly from random access memory (shown as the square inset in Figure 2) via a memory address computed from head position sensor data. Thus the displayed image did not have to be rendered in real time, reducing total system latency. Only head rotations (pitch, yaw, and roll) were registered by this system.
WARP TV software allowed GFOV and system time delay to be easily manipulated. The minimum system time delay (from head tracker movement to visual scene update) for this configuration was 48 ms with an additional, selectable delay of up to 500 ms. WARP TV was run on a Pentium 166 MHz PC with an ATI Technologies Mach 64 Pro Turbo graphic accelerator with 4 MB video memory. The minimum system time latency (M = 48 ms) was fixed for this experiment, as was DFOV (25[degrees] horizontal x 19[degrees] vertical). Horizontal GFOV was systematically varied (12[degrees], 25[degrees], and 50[degrees]).
An InterSense IS-300 position tracking system (InterSense Inc., Burlington, Massachusetts) was used to track head movements at 200 Hz. The system latency was approximately 4 ms with an angular resolution of 0.02[degrees] RMS and a dynamic accuracy of 1[degrees] RMS. The tracker was mounted directly above and centered on the participant's head via an aluminum bar attached to the HMD.
Design. A single-factor, within-subjects design was employed to investigate three levels of image scale factor with the following DFOV/GFOV ratios: MIN (0.5), NEU (1.0), and MAG (2.0) for minification, neutral, and magnification, respectively. Each participant experienced a total of three sessions (i.e., one session per image scale factor). Each session was separated by a minimum of six days to minimize carryover effects. Sessions were counterbalanced across image scale factor.
Sickness reports were collected before, during, and after VE exposure. These dependent variables included spoken ratings of sickness level during exposure and simulator sickness questionnaire (SSQ) scores (Kennedy, Lane, Berbaum, & Lilienthal, 1993). The spoken ratings were collected at specified times during VE exposure (10, 20, and 30 min). Each report consisted of a number on a 4-point scale ranging from 0 to 3, in which 0 indicated no discomfort, 1 indicated slight but noticeable discomfort, 2 indicated moderate discomfort, and 3 indicated very strong discomfort such that the participant wished to take a break or end the experiment. SSQ data were collected pre-exposure, immediately postexposure, and 20 min postexposure. Postural stability data were also collected pre-exposure and 20 mm postexposure.
Procedure. Each experimental session began with a prebriefing, participant consent, a pre-exposure SSQ, and a test of visual acuity. Two pre-exposure balance trials were then conducted to obtain baseline values for use in assessing the participant's postexposure postural stability. In each balance trial, the participant stood on the lab floor in a Sharpened Romberg stance with eyes open for 30 s. This stance consists of placing one foot in front of the other, heel touching toe, weight evenly distributed between the legs, arms folded across the chest, and chin up (Hamilton, Kantor, & Magee, 1989). The number of stance breaks per trial was recorded and then averaged across trials for comparison with postexposure values.
After the balance trials, the participant was secured in a chair (with a five-point harness, and with Velcro straps for the feet), and the HMD was positioned on his or her head. The participant then experienced 30 min of active, task-driven interaction with the VE. Five different QuickTime VR (Apple, Cupertino, California) 360[degrees] cylindrical images were presented to the participant during the exposure period (approximately 6 min exposure to each image).
For each VE image, the first minute of exposure was designated as exploration time for the participant to memorize the spatial arrangement of objects within the virtual world. The participant performed several (mostly yaw) head rotations during this time in order to explore the entire visual scene. The remaining 5 min consisted of a series of visual search tasks. The participant began each search task looking straight ahead. The experimenter then called out a target, and the participant responded by finding the target, fixating on it, and orally identifying that the search was completed. The experimenter verified the response by monitoring a separate display that replicated the participant's visual scene. After spoken confirmation by the experimenter, the participant would again look straight ahead, and a new search target would be identified.
Targets were always located within approximately [+ or -] 100[degrees] azimuth of straight ahead. The purpose of the search task was to continually have the participant interact with the VE using active, unrestricted head movements. After the 5-min search period, the participant was presented with a new VE image, until all six images were experienced.
During the VE exposure period, no specific instructions were given to the participant regarding head movement characteristics. The participant moved his or her head as desired to explore the scene and successfully complete the search tasks. However, in order to gain knowledge on particular head movement patterns, head position epochs (yaw angle only) were collected. These epochs, 80 s in length, were collected at the beginning, halfway through, and near the end of the exposure period.
At 10, 20, and 30 mm, the participant was prompted to report his or her comfort level on a scale of 0 to 3, as described earlier. If a participant reported a 3, the display was immediately turned off and the participant was prompted to describe his or her symptoms. The participant was also prompted (and often encouraged) to end the experiment.
After the exposure period, the participant completed the first postexposure SSQ (POST 1) and a general posttest questionnaire. At 20 mm postexposure, the participant completed two balance trials in the same manner as in pre-exposure and filled out the second postexposure SSQ (POST2).
Sickness reports. Because the sickness data were positively skewed and ordinal, only non-parametric tests were used in the analyses. Eight of nine participants completed all experimental sessions. Because of attrition, three individuals (one for each level of the image scale factor) filled the remaining participant position. Repeated-measures statistical analyses were conducted only on the eight participants who completed all sessions.
SSQ Total Severity (TS) scores were derived in accordance with Kennedy et al. (1993). Mean and median SSQ TS scores over time are shown in Figure 3. A Friedman test revealed that simulator sickness was induced as a result of VE exposure (collapsed across image scale factor), [[chi].sup.2](2) = 14.25, p [less than] .001. Two-tailed Wilcoxon analyses revealed that POST1 SSQ scores (those collected immediately after VE exposure) were significantly higher than PRE SSQ scores, [T.sub..05, 8] = 2.52, p [less than] .02. POST2 scores were also significantly higher than PRE scores, [T.sub..05, 8] = 2.52, p [less than] .02, indicating that sickness effects, though reduced, were still present after 20 mm. POST2 scores were significantly lower then POST 1 scores, however, [T.sub..05, 8] = -2.24, p [less than] .03. Participant gender did not significantly influence sickness incidence.
Mean and median POST1 SSQ TS scores as a function of image scale factor are presented in Figure 4. A Friedman test indicated a significant effect of image scale factor, [[chi].sup.2](2) = 6.75, p [less than] .04. Wilcoxon tests (two-tailed) revealed that sickness magnitude was significantly lower in the NEU condition than in the MIN, [T.sub..05, 8] = -2.33, p [less than] .02, and MAG image scale conditions, [T.sup..05, 8] = -2.25, p [less than] .03. There was no statistical difference between the MIN and the MAG conditions.
Mean and median POST2 SSQ TS scores as a function of image scale factor are presented in Figure 5. A Friedman test revealed a significant effect of image scale factor, [[chi].sup.2] (2) = 6.81, p [less than] .04. Two-tailed Wilcoxon tests indicated that sickness reports were significantly lower in the NEU condition than in the MAG condition, [T.sub..05, 8] 8 = -2.20, p [less than] .03, and there was a trend for a similar difference between the NEU and MIN conditions, [T.sub..05, 8] = -1.83, p [less than] .0%).
A Friedman analysis of session order was not significant, [[chi].sup.2] (2) = 3.04, p [greater than] .22, indicating that POST1 SSQ scores were not affected by previous sessions of the experiment. POST2 SSQ data also did not differ significantly across sessions.
The three spoken ratings of sickness obtained during each exposure period (at 10, 20 and 30 min) were averaged prior to conducting a two-tailed Friedman test. Although averaged verbal sickness ratings were not significantly affected by image scale factor, there appeared to be a slight trend in that direction, [[chi].sup.2] = 4.56, p [less than] .11. Figure 6 presents mean and median sickness ratings during exposure, by image scale factor. When only the final sickness ratings were examined (i.e., those reported at 30 mm of exposure), the means for the MIN, NEU, and MAG conditions were 1.4, 0.2, and 0.9, respectively. Finally, there was a trend for sickness ratings to increase with exposure time, [[chi].sup.2](2)= 4.85, p [less than] .09.
Head motion (yaw). Given instabilities in the collection process, data analysis was conducted only for the five participants who had complete head-position data sets (three yaw position epochs per session, for three sessions). Each position epoch was analyzed for average angular velocity, maximum angular velocity, average angular acceleration, and maximum angular acceleration along the yaw axis (Table 1).
An analysis of variance (ANOVA) found a significant effect of image scale factor on average angular velocity, F(2, 8) = 5.43, p [less than].05. Paired two-tailed t-tests (Bonferroni correction) indicated a significant reduction in yaw-angle head velocity in the MAG condition as compared with the MIN condition, t(4) = 4.28, p [less than] .05. It should be noted that although optic flow rate varied with image scale factor, total angular distance traversed by the head for a given target did not. No other analyses revealed significance.
Postural stability. These data were not analyzed because of a complete lack of variance. The mean and standard deviation were both zero.
This experiment induced significant levels of simulator sickness as a result of a 30-mm exposure to a head-coupled virtual interface. Simulator sickness reports were shown to be related to VE image scale factor. Sickness reports in the NEU condition were approximately half the magnitude of those obtained in either the MIN or MAG condition. Note that the MIN and MAG conditions had image scale factors not equal to 1.0. The data were consistent whether the metric examined was SSQ TS score immediately postexposure, SSQ TS score 20 mm postexposure, or averaged spoken rating during exposure (though this last metric failed to reach statistical significance).
In order to apply additional meaning to SSQ scores, Stanney, Kennedy, and Drexler (1997) developed a classification system based upon thousands of reports from various simulators and virtual interfaces. This system uses median SSQ TS scores to associate the utilized simulation environment with one of six categories: no symptoms, negligible symptoms, minimal symptoms, significant symptoms, symptoms are a concern, and "a bad simulator." Using this system with POST1 SSQ median values, the MIN and MAG conditions would be classified as a bad simulator (the worst rating achievable), whereas the NEU condition would be classified as minimal symptoms. Using POST2 values, the MIN and MAG conditions would be rated as significant symptoms, whereas the NEU condition would be placed within the negligible symptoms category.
Simulator sickness was still present in the MIN and MAG image scale conditions 20 min postexposure, though its magnitude was decreased. Evidently the time constant for recovery is not much shorter than the time constant for sickness onset, at least for this virtual interface/task combination. Sickness stemming from the NEU condition, however, was nearly absent after 20 min (using the classification of Stanney et al., 1997). This result further indicates that GFOV angle is an important modulating factor in the occurrence of simulator sickness.
Head motion analyses indicated a modest trend for reduced head motion in the MAG condition as compared with the MIN condition, though the only metric to achieve statistical significance was average angular velocity. It is interesting to note that this reduction does not correspond with any significant increase in simulator sickness levels in the MAG condition over those obtained in the MIN condition. Therefore, increased optic flow associated with the MAG condition (rather than increased sickness incidence) is the most likely explanation for the reduction in head motion in that condition.
An additional observation of the complete data set (collapsed across image scale factor) found that neither head movement yaw-angle velocity nor acceleration declined appreciably over the 30-min exposure period. Thus it is reasonable to suggest that participants did not become overly fatigued as a result of the task.
It is important to note that some simulator sickness symptoms were reported in the NEU condition, only less so (by a factor of two) than in the other conditions. This result indicates that other provocative factors of simulator sickness were present but not controlled for in this experiment. A prime suspect, system time delay, is investigated in the next experiment. Other potential contributors include reduced resolution, display flicker, optical distortions, and form/fit of the HMD.
EXPERIMENT 2: TIME DELAY EXPERIMENT
Participants. The volunteers for Experiment 2 were 10 adults (6 men and 4 women, mean age 27.4 years, range 23-36 years), none of whom had participated in Experiment 1. Participant requirements were the same as in Experiment 1. Eight participants reported that they had experienced motion sickness in the past, and two of these reported experiencing simulator sickness in the past. Two participants claimed not to be susceptible to motion sickness, six reported slight susceptibility, one reported moderate susceptibility and one reported high susceptibility. Participants were unfamiliar with the VE prior to testing and had not experienced a head-coupled virtual interface in the previous 30 days (four participants had never before been exposed to a head-coupled virtual interface).
Design. A single-factor within-subjects design was used to assess two levels of step-onset time delay (125 ms, 250 ms) while image scale factor was held constant (at 1.0, the NEU condition of Experiment 1). Each participant experienced two sessions (one session per time delay condition). Each session was separated by a minimum of six days to minimize carryover effects. Sessions were counterbalanced across time delay. The dependent variables were the same as those in Experiment 1.
Materials and apparatus. The apparatus and experimental set-up were as described in Experiment 1, except that a Chattecx Balance Platform (Chattecx Corp., Chattanooga, Tennessee) was used to provide a more sensitive measure of postural stability, given the invariance in the balance data from Experiment 1. This system determined the participant's instantaneous center of gravity based on signals from force-sensing plates under each foot (Levine, Whittle, Beach, & Ollard, 1996). An overall dispersion score (cm) was derived by monitoring changes in the center-of-gravity estimates along the two horizontal directions (front/back and side to side) over the period of the test. The system provided center-of-gravity estimates at 100 Hz.
Additional time delay was achieved through the use of a delay buffer that held head position data for the required time before updating the virtual image. Thus it is considered "pure" time delay, or transmission lag, per Wickens (1992). System time delay was empirically verified and calibrated using a test system composed of a high-speed photo-sensor, digital oscilloscope, and mechanical tracker (see Draper, 1998).
The procedure was identical to that used in Experiment 1 except that the pre-exposure and postexposure postural stability test utilized a Chattecx Balance Platform. Participants stood in a Sharpened Romberg position (eyes open) while on this apparatus. Each participant performed two pre-exposure trials and two postex-posure trials. Each trial was 10 s long, and the average total dispersion score across both trials was recorded as the postural stability metric.
Sickness reports. Both sessions were successfully completed by eight participants. Because one participant was unable to attend the second session, another participant (matched by gender, age, and experience) took part in the remaining delay condition to complete the ninth participant data set. Nonparametric repeated-measures statistical tests considered only the eight participants who completed both sessions.
Mean and median SSQ TS scores across time are shown in Figure 7. A Friedman test of these data revealed that simulator sickness was induced as a result of VE exposure (collapsed across time delay), [[chi].sup.2](2) = 5.72, p = .05. Wilcoxon post hoc tests found POST1 SSQ scores to be significantly higher than PRE, [T.sub.05,8] = 2.21, p [less than] .03, or POST2 scores, [T.sub..05,8] = 2.20, p [less than] .03. POST2 SSQ scores were not significantly different from PRE values, indicating that the accumulated sickness had returned to pre-exposure levels within 20 min of ending the VE exposure. Participant gender did not significantly influence sickness incidence.
Mean and median sickness scores, by time delay, are presented in Figures 8 to 10. Two-tailed Wilcoxon analyses indicated no significant effect of time delay on POST1 SSQ TS scores (Figure 8), POST2 SSQ TS scores (Figure 9), or on the average verbal sickness rating during exposure (Figure 10).
A Friedman test of POST1 SSQ TS data across sessions was not significant, indicating that reports of sickness were not affected by prior exposures. POST2 SSQ data and sickness ratings during exposure also did not differ across sessions.
A Friedman test indicated a trend toward significance for sickness increasing with exposure time, [[chi].sup.2](2) = 6.0, p [less than] .06, when collapsed across time delay. Verbal ratings appeared to be higher at 20 and 30 min than at 10 min.
Head motion. Head position epochs for eight participants were analyzed as in Experiment 1 (Table 2). An ANOVA of average angular velocity revealed a significant main effect for exposure time, F(2, 14) = 10.80, p [less than] .01. Paired two-tailed t-tests (Bonferroni correction; [alpha] = .05) indicated significant increases in average yaw-angle head velocity at the midpoint and end of exposure as compared with the beginning of the exposure period. An ANOVA of average angular acceleration indicated main effects for time delay, F (1, 7) = 9.42, p [less than] .05, and exposure time, F(2, 14) = 1l.50, p [less than].001. There was a significant reduction in average yaw acceleration in the 250-ms delay condition as compared with the 125-ms delay condition. Paired t-tests (Bonferroni correction) of exposure times indicated a significant increase in average yaw-angle acceleration at the end of the exposure period as compared with the beginning and midpoint. No other analyses revealed significance.
Postural stability. Balance data were collected pre-exposure and approximately 20 mm postexposure. Wilcoxon analyses indicated no significant difference between PRE and POST data when collapsed across time delay, nor were there any significant effects of time delay condition on postural stability. Mean pre-exposure/postexposure dispersion scores were 5.5/5.1 at 125 ms and 4.9/5.7 at 250 ms.
Simulator sickness occurred as a result of exposure to the virtual interface, though symptom severity was significantly lower than that reported in the MIN and MAG conditions of Experiment 1. Simulator sickness magnitudes did not significantly change with increasing time delays from 125 ms to 250 ms, a finding that was consistent across all sickness metrics. However, it is interesting to note that the participants did perceive the larger delay, often making comments about the sluggishness of the system and the unpleasant motion.
Using the SSQ classification scheme (Stanney et al., 1997) on POST1 median SSQ TS scores, we classified both the 125-ms and 250-ms conditions as minimal symptoms. For POST2 scores, both conditions would be classified as negligible symptoms.
Head motion analyses indicated a minor but consistent trend toward a reduction in yaw rotational motion as time delay increased (though significance was achieved only for the average angular acceleration metric). This reduction is believed to be a result of the increased difficulty in compensating for the larger delay in controlling the visual scene. Participants' head movement velocity and acceleration appeared to increase significantly over the course of a session. Because increased simulator sickness levels are often associated with reduced head movements, the significant increase in head motion over the course of an exposure period offers further support to the finding that participants did not experience significant simulator sickness symptoms with increasing time delays. This increase is likely a result of the development of effective control strategies to compensate for the large time delays.
At the most rudimentary level, these experiments demonstrated the existence of simulator sickness as a result of exposure to a head-coupled virtual interface. However, this finding is far from novel. More interesting findings center on which factors induced simulator sickness and which did not, along with the resulting implications for existing theories.
The results of Experiment 1 suggest that image scale factor (i.e., DFOV/GFOV ratio) is a provocative factor in the inducement of simulator sickness. Experiment 2 failed to modulate sickness reports with increasing time delays. Thus results from Experiment 1 appear consistent with the sensory rearrangement theory, whereas the results of Experiment 2 do not.
One might contend that the range of time delays examined in Experiment 2 (125-250 ins) was not large enough to detect an effect. This range could be expanded downward to 48 ms if the results from the NEU condition of Experiment 1 were considered. The NEU experimental condition was equivalent to the two delay conditions used in Experiment 2 except for a reduced time delay (set at 48 ins) and a different participant group. Because few virtual interface systems have system time delays below 48 ms or above 250 ms, simulator sickness data were informally compared across these three time delays. Figure 11 displays the results of this comparison for each sickness metric. Though no statistical analyses were conducted, it appears that sickness magnitude does not change appreciably or systematically over the three time delays, spanning 48 to 250 ins.
Though not a statistically validated result, these expanded data are also not readily explained by the sensory rearrangement theory, which would seemingly forecast an increase in sickness with increasing time delay, at least to a point. However, increases in time delay did have noticeable effects in other areas. Motion perception varied with delay (i.e., the minimum time delay was not perceptible by the participant, whereas the 250-ms delay was very noticeable and deemed unpleasant), as did visual vestibulo-ocular reflex phase angles (Table 3; see Draper, 1998). Additionally, the increases in participant head motion over the course of each session in Experiment 2 are contrary to those expected in cases of rising sickness level.
Therefore, these data suggest (both within and across experiments) that fixed, pure time delays up to 250 ms are not especially provocative contributors to simulator sickness, a conclusion that is supported by earlier research (Uliano et al., 1986). Combined, Experiment 1 and Experiment 2 serve to illustrate the fallacy of attempting to utilize the sensory rearrangement theory to predict simulator sickness in virtual interfaces. The theory, though explanatory for many cases of simulator sickness, is incomplete (in its present form) for the purpose of making rigorous a priori design decisions to reduce sickness incidence in VEs.
Other theories of motion sickness also appear to have difficulty accounting for these data. As discussed earlier, the subjective vertical theory contends that only those sensory rearrangements that include significant otolith involvement are provocative. These experiments, however, elicited significant levels of simulator sickness even though the task involved predominantly yaw rotational head movements. Furthermore, the subjective vertical theory does not readily predict the differential sickness experienced across conditions in Experiment 1, given that the task was identical.
Finally, though these experiments do not specifically address the ecological theory of motion sickness, simulator sickness symptoms were elicited although the participants were never challenged posturally (they were firmly strapped into a chair with a five-point harness and with Velcro straps for the feet). However, it is possible that subtle instabilities in head-movement control may have existed, but went undetected, in those conditions that provoked simulator sickness.
Head Motion Analyses
Head motion characteristics were generally similar across both experiments, the exception being maximum angular acceleration (means: 520[degrees]/[s.sup.2] in Experiment I vs. 632[degrees]/[s.sup.2] in Experiment 2). The task involved was evidently the major determinant of head motion, and it remained consistent across conditions and experiments. There was no time pressure to find objects as quickly as possible, so participants often performed at a leisurely pace. Also, it is important to note that changing the image scale or time delay did not change the angular movement required to center each target in the display.
The significant variations of head motion across conditions were most likely the result of a decreased ability to control the visual scene in response to head movements. In Experiment 1 the MAO condition had an optic flow rate in response to head movements that was four times the rate associated with the MIN condition, because of the magnification differences. In Experiment 2 the 250-ms time delay was an obvious disruption to the smooth coupled relationship between head movement and visual scene response.
A common finding in simulator and motion sickness is that head movements are often curtailed when participants begin to experience simulator sickness. This would predict a reduced set of head movements in the MIN and MAG condition versus the NEU condition of Experiment 1. However, the analyses failed to support this prediction. The most likely explanation is that the power of head motion was concentrated at low frequencies because of task requirements, which resulted in a floor effect. That, combined with the small sample size for head motion analyses (n = 5), likely accounted for the apparent lack of correlation between head motion and sickness scores in Experiment 1. The significant increase in head-motion characteristics over the exposure period in Experiment 2 is in agreement with the lack of sickness symptoms reported by those participants, however.
Postural Stability Analyses
As expected, postural stability data failed to show a decrement as a result of exposure to the virtual interface. These tests were conducted primarily as a safety precaution prior to releasing the participants after the experiment, taking place approximately 20 min postexposure. Significant effects would have been more likely if these tests were administered immediately after the exposure period.
Though generalizing results from a few experiments is risky at best, these data at least suggest a few preliminary design guidelines for developing virtual interfaces that minimize the influence of simulator sickness. These guidelines were based primarily on the data obtained in these experiments, though outside research was leveraged as appropriate to solidify these recommendations. Further research is required to confirm and extend these initial guidelines.
1. Avoid image scale deviations from 1.0 magnification to reduce simulator sickness incidence. This can be accomplished by equating GFOV and DFOV (assuming that a spatially accurate model of the VE exists).
2. As a logical extension of this research, verify that the gain of the head-tracking system is accurately calibrated to minimize simulator sickness resulting from inaccurate optic flow rate in response to head rotations.
3. Minimize system time delays to improve user performance and enjoyment (Pausch et al., 1992; So & Griffin, 1995), but do not expect a significant reduction in sickness level. This guidance is currently limited to nearly constant, step-onset time delays and rotational head motion only.
4. Limit initial VE exposures to 10 MIN or less to minimize simulator sickness.
A preliminary report describing Experiment 1 was presented at the 1998 Annual Meeting of the Human Factors and Ergonomics Society. We wish to acknowledge Paul Schwartz and Zsolt Lorant (University of Washington) for software support and assistance in data collection. Chris Russell (Air Force Research Laboratory) developed the software analysis routines for the head position data. In addition, Todd Nelson, Robert Bolia, and Grant McMillan (Air Force Research Laboratory), as well as Heath Ruff (Sytronics Inc.), provided helpful comments on a draft of this manuscript. This research was funded in part by a grant from the Air Force Office of Scientific Research (F49620-93-0339). This work is not subject to United States copyright restrictions.
Mark H. Draper is a human factors engineer in the Human Effectiveness Directorate of the Air Force Research Laboratory at Wright-Patterson Air Force Base, Ohio. He received his Ph.D. in mechanical engineering from the University of Washington in 1998.
Erik S. Viirre is an adjunct assistant professor in the Division of Otolaryngology, University of California, San Diego School of Medicine. He received his Ph.D. in physiology from the University of Western Ontario in 1987.
Thomas A. Furness is a professor of industrial engineering and director of the Human Interface Technology Laboratory at the University of Washington, Seattle, Washington. He received his Ph.D. in engineering and applied science from the University of Southampton, England, in 1981.
Valerie J. Gawron is a principal engineer in the Flight Research Group at Veridian Engineering, Buffalo, New York. She received a Ph.D. in engineering psychology from the University of Illinois in 1980.
Bles, W., Bos, J. E., de Graf, B., Groen, E., & Wertheim, A. H. (1998), Motion sickness: Only one provocative conflict? Brain Research Bulletin, 47, 481-487.
Boyer, B. S., & Wickens, C. D. (1994). 3D weather displays far aircraft cockpits (Teachnical Report ARL-94-l1/NASA-94-4). Savoy: University of Illinois Institute of Aviation, Aviation Research Laboratory.
Danas, E. (1995). Mapping auditory space onto visual space. Unpublished master's thesis, University of Washington. Seattle.
Draper, M. H. (1998). The effects of image scale factor on vestibuloocular reflex adaptation and simulator sickness in head-coupled virtual environments. In Proceedings of the Human Factors and Ergonomics Society 42nd Annual Meeting (pp. 1481-1486). Santa Monica, CA: Human Factors and Ergonomics Society.
Durlach, N. I., & Mayor, A. 5. (Eds.). (1995). Virtual reality: Scientific and technological challenges. Washington, DC: National Academy Press.
Ebenholtz, S. M. (1992). Motion sickness and oculomotor systems in virtual environments. Presence, 1, 302-305.
Ebenholtz, S. M., Cohen, M. A., & Linder, B. J. (1994). The possible role of nystagmus in motion sickness: A hypothesis. Aviation, Space, and Environmental Medicine, 65, 1032-1035.
Foley, J. D., van Dam, A., Feiner, S. K., & Hughes. J. F. (1996). Computer graphics principle and practice (2nd ed.). Reading, MA: Addison-Wesley.
Frank, L. H., Casali, J. G., & Wierwille, W. (1988). Effects of visual display and motion system delays on operator performance and uneasiness in a driving simulator. Human Factors, 30, 201-217.
Griffin, M. J. (1990). Handbook of human vibration. London: Academic.
Hamilton, K. M., Kantor, L., & Magee, L. E. (1989). Limitations of postural equilibrium tests for examining simulator sickness. Aviation, Space, and Environmental Medicine, 60, 246-251.
Hendrix, C. M. (1994). Exploratory studies on the sense of presence in virtual environments as a function of visual and auditory display parameters. Unpublished master's thesis, University of Washington, Seattle.
Kennedy, R. S., Hettinger, L. J., & Lilienthal, M. G. (1990). Simulator sickness. In G. H. Crampton (Ed.), Motion and space sickness (pp. 317-342). Boca Raton, FL: CRC Press.
Kennedy, R. S., Lane. N. E., Berbaum, K. S., & Lilienthal, M. G. (1993). A simulator sickness questionnaire (SSQ): A new method for quantifying simulator sickness. International Journal of Aviation Psychology, 3, 203-220.
Levine, D., Whittle, M. W., Beach, J. A., & Ollard, P. G. (1996). Test-retest reliability of the Chattecx Balance System in the patient with hemiplegia. Journal of Rehabilitation Research and Development, 33, 36-44.
McCauley, M. E., & Sharkey, T. J. (1992). Cybersickness: Perception of self-motion in virtual environments. Presence, 1, 311-318.
Meyer, K., Applewhite, H. L., & Biocca, F. A. (1992). A survey of position trackers. Presence, 1, 173-200.
Pausch, R, Crea, T., & Conway, M. (1992). A literature survey for virtual environments: Military flight simulator visual systems and simulator sickness. Presence, 1, 344-363.
Peli, E. (1995, July). Real vision and virtual reality. Optics and Photonics News, 28-34.
Reason, J. T. (1978). Motion sickneas adaptation: A neural mismatch model. Journal of the Royal Society of Medicine, 71, 819-829.
Regan, C. (1995). An investigation into nausea and other side effects of head-coupled immersive virtual reality. Virtual Reality, 1, 17-32.
Riccio. G. E., & Stoffregen, T. A. (1991). An ecological theory of motion sickness and postural instability. Ecological Psychology, 3, 195-240.
Rushton. S., Mon-Williams, M., & Wann, J. P. (1994). Binocular vision in a bi-ocular world: New-generation head-mounted displays avoid causing visual deficit. Displays. 15, 255-260.
So, R. H., & Griffin, M. J. (1995). Head coupled virtual environment with display lag. In K. Carr & R. England (Eds.), Simulated and virtual realities: Elements of perception (pp. 103-112). London: Taylor & Francis.
Stanney, K. M., Kennedy, R. S., & Drexler, J. M. (1997). Cybersickness is not simulator sickness. In Proceedings of the Human Factors and Ergonomics Society 41st Annual Meeting (pp. 1138-1142). Santa Monica, CA: Humans Factors and Ergonomics Society.
Stoffregen, T. A., & Riccio, G. E. (1991). An ecological critique of the sensory conflict theory of motion sickness. Ecological Psychology, 3, 150-194.
Treisman, M, (1977). Motion sickness: An evolutionary hypothesis. Science, 197, 493-495.
Uliano, K. C., Lambert, E. Y., Kennedy, R. S., & Sheppard, D. J. (1986). The effects of asynchronous visual delays on simulator flight performance and tire development of simulator sickness symptomatology. (Technical Report NAVTRASYSCEN 85-D-0026-1, AD-A180 196). Orlando, FL: Naval Training Systems Center.
Viirre, E. S. (1994, July/August). A survey of medical issues and virtual reality technology. Virtual Reality World, 16-20.
Wicknes, C. D. (1992). Engineering psychology and human performance. New York: Harper Collins.
Wloka. M. M. (1995). Lag in multiprocessor virtual reality. Presence, 4, 50-63.
TABLE 1 Mean (and Standard Deviation) Yaw Angular Head Motion Summary Data Image Scale Average Maximum Average Experiment Velocity Velocity Acceleration (n=5) ([degrees]/s) ([degrees]/s) ([degrees]2/s) Group Mean (SD) 16.5 (3.2) 133.5 (27.6) 51.9 (11.6) Image Scale MIN 18.4 (4.5) 144.4 (54.7) 56.3 (17.0) NEU 16.6 (2.2) 146.6 (21.5) 52.3 (9.0) MAG 14.4 (3.7) 111.0 (26.3) 48.5 (11.8) Exposure Time First Image 16.5 (4.5) 133.8 (37.7) 52.5 (13.4) Midpoint 14.7 (3.2) 118.2 (33.8) 46.6 (10.0) Final Image 18.1 (4.1) 150.0 (35.8) 58.0 (14.9) Image Scale Maximum Experiment Acceleration (n=5) ([degrees]2/s) Group Mean (SD) 519.7 (132.5) Image Scale MIN 549.5 (204.1) NEU 566.3 (92.9) MAG 454.4 (110.0) Exposure Time First Image 521.2 (158.1) Midpoint 497.0 (148.9) Final Image 552.1 (136.3) TABLE 2 Mean (and Standard Deviation) Yaw Angular Head Motion Summary Data Time Delay Average Maximum Average Experiment Velocity Velocity Acceleration (n=8) ([degrees]/s) ([degrees]/s) ([degrees]2/s) Grand Mean (SD) 17.0 (3.1) 147.9 (37.6) 54.8 (14.1) Time Delay 125 ms 17.6 (3.4) 156.8 (36.8) 59.0 (14.5) 250 ms 16.5 (3.3) 138.9 (43.6) 50.6 (14.6) Exposure Time First Image 14.8 (3.3) 136.9 (37.9) 47.7 (14.9) Midpoint 17.1 (3.8) 147.9 (41.8) 52.8 (14.4) Final Image 19.3 (3.4) 158.8 (45.1) 64.0 (16.1) Time Delay Maximum Experiment Acceleration (n=8) ([degrees]2/s) Grand Mean (SD) 631.9 (164.7) Time Delay 125 ms 693.6 (190.0) 250 ms 570.1 (179.3) Exposure Time First Image 613.9 (197.3) Midpoint 613.3 (171.2) Final Image 668.4 (182.3) TABLE 3 Mean (and Standard Deviation) Visual Vestibulo-ocular Reflex Phase Recordings across the Various Time Delays in the Two Experiments Visual Condition WOR Phase (SD) Real World (No Delay) - 1.1[degrees] (2.0[degrees]) 48-ms Delay -14.0[degrees] (3.0[degrees]) 125-ms Delay -34.4[degrees] (7.8[degrees]) 250-ms Delay -73.6[degrees] (5.5[degrees]) Note: These values are "visual" VOR measurements caputured in lighted conditions with the participant actively focused on a fixed target within the visual scene and an induced inertial sinousodial oscillation at 0.8 Hz. Thus the response contains contributions from both the VOR and ocular fixation mechanisms. Real-world values were obtained by having the participant focus through the see-through HMD to a fixed target on the far wall of the laboratory (see Draper, 1998).
|Printer friendly Cite/link Email Feedback|
|Author:||Draper, Mark H.; Virre, Erik S.; Furness, Thomas A.; Gawron, Valerie J.|
|Article Type:||Statistical Data Included|
|Date:||Mar 22, 2001|
|Previous Article:||Effects of Hearing Protectors on Auditory Localization in Azimuth and Elevation.|
|Next Article:||Spatial Representations of Virtual Mazes: The Role of Visual Fidelity and Individual Differences.|