Navigating a maze differently - auser study

Aryabrata Basu*

Emory Universityaryabrata.basu@emory.edu

Kyle Johnsen†

University of Georgiakjohnsen@uga.edu

September 24, 2018

Abstract

Navigating spaces is an embodied experience. Examples can vary from rescue workers trying to save peoplefrom natural disasters; a tourist finding their way to the nearest coffee shop, or a gamer solving a maze.Virtual reality allows these experiences to be simulated in a controlled virtual environment. However, virtualreality users remain anchored in the real world and the conventions by which the virtual environment isdeployed influence user performance. There is currently a need to evaluate the degree of influence imposedby extrinsic factors and virtual reality hardware on its users. Traditionally, virtual reality experiences havebeen deployed using Head-Mounted Displays with powerful computers rendering the graphical contentof the virtual environment; however, user input has been facilitated using an array of human interfacedevices including Keyboards, Mice, Trackballs, Touchscreens, Joysticks, Gamepads, Motion detecting camerasand Webcams. Some of these HIDs have also been introduced for non-immersive video games and generalcomputing. Due to this fact, a subset of virtual reality users has greater familiarity than others in usingthese HIDs. Virtual reality experiences that utilize gamepads (controllers) to navigate virtual environmentsmay introduce a bias towards usability among virtual reality users previously exposed to video-gaming.

This article presents an evaluative user study conducted using our ubiquitous virtual reality frameworkwith general audiences. Among our findings, we reveal a usability bias among virtual reality users whoare predominantly video gamers. Beyond this, we found a statistical difference in user behavior betweenuntethered immersive virtual reality experiences compared to untethered non-immersive virtual realityexperiences.

I. Introduction

Spatial navigation is one of the core abili-ties of human beings. It is a useful skillset that we employ on a daily basis to

navigate building interiors or busy streets toreach our destinations. The act of navigatinga physical environment requires the simulta-neous operation of both cognitive and motorfunctions [12]. The majority of virtual real-ity (VR) experiments are designed with spa-tial navigation in mind as the primary meansfor exploring virtual environments (VEs) [17].That said, the potential implications of VE inter-

*Corresponding author†Advisor

face and VR task model design on individualuser performance in VR have yet to be exploredadequately. As a result, understanding the un-derlying concepts of spatial navigation as amathematical construct and its relationship toimmersive VR user performance is particularlyimportant, especially if we were to adopt VRexercises designed to promote navigation skillset building in users with measurable gain inoutput.

In this work, we presume that the priorvideo-gaming experience, age and the genderof the users participating in a VR study im-pacts their respective performance in execut-ing immersive VR tasks. This study has beenextended from earlier work examining how

1

arX

iv:1

805.

0945

4v4

[cs

.HC

] 2

0 Se

p 20

18

Spatial navigation in immersive VR • May 2018 • Iteration. 1

Figure 1: Immersive (top) and non-immersive (bottom)perspective of all study participants at the startof the maze experience

physiological factors affect user performancein immersive VR [1]. In this article, we describeour virtual maze application as well as the de-tailed system design. We report an extendedanalysis conducted on the recorded user tra-jectory data. To that end, we defined a set ofmathematically derived features generalizedfor each trajectory such as distance traveled,decision points reached inside the maze, po-sitional curvature, head rotation amount, andcoverage of the maze.

II. Related Work

Our application and system is built upon thestrategies, technology, and research involvedin previous studies conducted on the topic ofspatial navigation in immersive VR. A numberof researchers have addressed issues relatedto spatial navigation and travel metaphors inimmersive VEs over the years [6, 14, 13, 18, 7].A subset of these works have also looked intothe affects of prior video gaming experience on

Figure 2: Key locations and decision nodes of the maze

immersive VR navigation performance [11].

In 1998, Bowman et al. proposed a formal-ized methodology and framework [6] for theevaluation of travel techniques in immersiveVEs. The basic construct of their frameworkwas a taxonomy of travel techniques. Their ex-perimental analysis revealed the need to gathermore information about user analytics insideVEs. In a second study, Bowman et al. stud-ied the effects of various travel techniques onthe spatial orientation [5] of users inside im-mersive VR environments. At the same time,Ruddle et al. introduced a formal study com-paring HMDs and desktop displays [10]. Theobjective of this study was to perform a base-line investigation that compared the two dif-ferent types of displays. They found that par-ticipants using the HMD navigated the virtualbuildings significantly more quickly, and de-veloped a significantly accurate sense of rela-tive straight-line distance. Behavioral analysesshowed that participants took advantage ofthe natural, head-tracked interface providedby the HMD in ways that include “lookingaround” more often while traveling throughthe VE and spending less time being stationarywhile choosing a direction in which to travel.

In 2009, Smith et al. conducted a researchstudy to look into the impact of previous com-

2

Spatial navigation in immersive VR • May 2018 • Iteration. 1

puter gaming experience, user perceived gam-ing ability, and actual gaming performance onnavigation tasks in a VE [11]. They found thatperceived gaming skill and progress in a linearfirst person shooter (FPS) game were foundto be the most consistent metrics. Both per-ceived gaming skill and progress in a linearFPS game bore a relationship to performancein trivial searches, primed searches, the num-ber of mistakes when performing an advancedtravel technique (jumping) and in travelingtime requiring high speed and accuracy. Smithet al. [11], stated ‘this may require the develop-ment of a gamer profile including both similarmetrics to those used in this paper, metrics overother 3D interface tasks such as selection andmanipulation and metrics for subjective condi-tions such as user disorientation, cybersicknessand presence.’ In 2010, Suma et al. reporteda user study comparing real walking withthree virtual travel techniques; namely, gaze-directed, pointing-directed, and torso-directedtravel [14]. Suma et al. found that real walkingis superior in terms of user performance as itallowed more cognitive capacity for processingand encoding stimuli than pointing-directedtravel metaphors. They also found that maleparticipants were slower and performed signif-icantly worse on the attention task when thespatial task became more difficult in contrast tofemale participants. In another related study,Suma et al. reported that for complex VEs withnumerous turns, virtual travel techniques areacceptable substitutes for real walking if thegoal of the application involves learning or rea-soning based on information presented in theenvironment [13].

Another component of studying spatial nav-igation is spatial trajectory analysis. In 2005,Zanbaka et al. described a between-subjects ex-periment that compared four different methodsof travel, their effect on cognition, and pathstaken in an immersive VE [18]. This studyused participants’ trajectory (position and headorientation) data in post-analysis by creatingoverlays that ultimately revealed further differ-ences in travel techniques. This study favored alarge tracked space over other travel techniques

in VR for applications where problem solvingand interpretation of material is important orwhere opportunity to train is minimal. Jeonget al. in 2005, reported on the differentiationof information-gathering ability in the real andthe VE [7]. An important finding of their studywas that the users path of finding informationwas similar, their information gathering abilitydiffers between the real and the VE.

This article does not address the cognitiveissues related to spatial exploration on anindividual basis. Rather our approach dis-tinguishes task performance (as a trend) be-tween two categories of users (gamers andnon-gamers) using VR systems. More thanone performance metric for solving a maze inVR has been deployed, which allowed for anuanced discussion of navigation abilities.

Prior work on the evaluation of ubiquitous,smartphone driven 3DUIs helped guide oursystem architecture [3, 2]. A light-weight VRsystem is logistically a sound design whenit comes to recruiting higher number of sub-jects for a research study. Other studies existthat looked into physical activity as a poten-tial marker for performances in immersive VRenvironments. One such study suggested thatphysical activities of users can be a predictorfor VR task performance [3]. In regards tophysical activity, virtual reality has also beenexplored as a means to motivate individualsto exercise, a concept better known as VR ex-ergaming [8]. While all these studies explorethe relationship between physical activity andVR performance, there has been a lack of con-crete connections between them. In contrastto quantifying physical fitness of users as sug-gested by Basu et al.[1, 8], our approach shiftsthe attention of the user performance predictorto the gaming profile.

III. Application

Our maze application requires the user to nav-igate a maze and find multiple target sites. Forconsistency, every player had a fixed startingpoint and fixed targets to reach in the maze.The context of the maze experience was similar

3

Spatial navigation in immersive VR • May 2018 • Iteration. 1

to a previous study [1], in which the subjectis required to find and rescue another human(avatar). The task included finding a clue (min-imap) before finding the human avatar andthen tracing back the path to the entrance ofthe maze. On average, the participants spent4.64 minutes in the immersive user interfacesetting and 6.18 minutes in the non-immersiveuser interface setting.

i. Game Design

The focal point of our VR experience was to tra-verse a maze environment, find multiple targetsites and trace back your path. To add to theexperience, we used an altruistic framework bysetting up a background story of a lost travelerinside the maze. Each participant was giventhe tasks of searching for and rescuing the lostcharacter in the maze. To better emphasizethis point, a restriction was placed upon theparticipant in the form of finding a map of themaze first. Once the user was successful infinding the map location, the user was thenasked to find the lost character in the maze.Upon meeting the character, which is an ani-mated humanoid avatar, the participants wererequired to trace back their path to the entranceof the maze to complete the experience. Themaze design also included strategically placedunique props (box, first-aid kit, etc.) to helpwith user localization and memory. Variouskey locations, including targets and props inour maze design, are illustrated in Figure 2.

ii. System Hardware

Our system setup included a mobile HMD anda wireless gaming controller. We opted forSamsung Galaxy Gear VR as our HMD solu-tion and Samsung Galaxy S6 Edge+ with 5.7inch screen size at 2560 pixels x 1440 pixels (518ppi) screen resolution as our primary smartphone display. For user controls in the maze,we opted for a Bluetooth compatible MadcatzC.T.R.L. gamepad controller which connectseasily with our display. A typical system setupdeployed for the study is illustrated in Figure 3.

Figure 3: A randomly selected participant partaking themaze experience

iii. System Software

The VR maze experience was developed us-ing the Unity game engine and deployed onthe HMD device as an Android app. Whileeach participant was busy solving the maze,the system software kept an active log of theirin-session trajectory data along with their headorientation data every 10 milliseconds. Fur-thermore, the system software also logged rel-evant event data such as participants quittingthe maze experience or participants findingthe targets successfully. A visual example ofparticipant tracking algorithm is illustrated inFigure 4.

Once the user data was collected, scripts de-veloped in C# programming language wereused to analyze the users’ trajectory datain Unity game engine. These scripts parsethrough user transformation (x,y,z) in real-timeand visualize their search patterns in the maze.This, in turn, allows to confirm visually emerg-ing differences in user performance in immer-sive VR. This sort of visual analysis can be gen-eralized to apply to other studies concerningspatial navigation in immersive VR.

IV. Study Design

Null hypothesis: There is not a significant im-pact on users’ ability to successfully navigate a 3Dmaze under the influence of technical factors suchas immersive, non-immersive user interface.

4

Spatial navigation in immersive VR • May 2018 • Iteration. 1

Figure 4: Visualization of a randomly selected partici-pant’s spatial trajectory data including gaze in-formation represented as highlighted in cyan.

Alternate hypothesis: There is a significantimpact on users’ ability to successfully navigate a3D maze under the influence of technical factorssuch as immersive, non-immersive user interface.

The primary goal behind the maze experi-ence is to understand how users with varyingdegree of exposure to video gaming solve acomplex three dimensional immersive naviga-tion problem. We wanted to quantify spatialdecision making in terms of user activity in themaze. To this end, we conducted a 2 x 1 studywith user type being one of our independentvariables. We recorded spatial user activity,including gaze activity, every 10 millisecondapart to help develop a model for spatial navi-gation performance. The experiment requiredusers to navigate a maze with a three-fold taskmodel. The following subsections describe thestudy design in details.

i. Population and Environment

Over a course of 8 months, 40 self-motivatedparticipants (mean-age = 35.93; SD = 11.11)volunteered for our study. Table 1 shows thebreakdown of the two study groups by popu-lation demographics. These participants com-

Gamer N-Gamer

Participants 21 19

GenderMale 14 4Female 7 15

Player Level

Novice 0 12Casual 6 6Pro 12 0Hardcore 2 0Experiential 1 1

Table 1: Study demographics, based on the question-naires; ‘Player Level’ is a derived scale reflect-ing participant’s involvement and approach tovideo-gaming; ‘Gamer’ vs. ‘Non-Gamer’ re-flects the self-assessment of being a gamer

prise a mix (by age, profession, background)of university staff and students at Emory Uni-versity's main campus. In compliance withthe Institutional Review Board guidelines ofboth University of Georgia and Emory Univer-sity, every participant's consent was obtainedand recorded on paper forms. Furthermore,the users were asked to fill out a series offorms including a general demographic ques-tionnaire, pre-simulator and post-simulatorsickness questionnaire and a presence question-naire. Additionally, these users were also askedto fill out a self-reported gaming profile assess-ment form. Each user was then tasked to solvethe same maze problem twice using an immer-sive user interface and then a non-immersiveuser interface. The order of deployment foreach user was random and can be classifiedbetween the following two categories:

Condition 1. Immersive first, and then non-immersive.

Condition 2. Non-immersive first, and thenimmersive.

Each participant was recruited using an on-line recruitment drive advertised through theinternal university mailing list. A specific labo-ratory was chosen to run the sessions with onlyone participant at a time. The choice of our

5

Spatial navigation in immersive VR • May 2018 • Iteration. 1

Features Definition

Distance traveled It is the totallength of thepath traveledbetween twopositions.

Coverage Total number ofunit cubes cov-ered (area).

Number of decision pointsreached

Total number ofdecision pointscovered in themaze.

Positional Curvature The signed an-gle of curvaturebetween twoconsecutiveposition vectors.

Head rotation amount The unsignedhead rotationangle betweentwo consecu-tive rotationtransforms.

Table 2: Features extracted from users’ trajectory datafor deep exploration and their correspondingdefinitions.

study space was particularly critical to removeany additional social anxiety within our partic-ipants. Our study investigator sat down withour participants and would start conversing toease them into our study. Upon agreeing to ourstudy requirements, a participant would starthis/her first round of maze solving followedby a five minute break before starting his/hersecond round. The order of deployment foreach user was random. The study investiga-tor made sure to capture an equal number ofparticipants for each category of deployment.Between the two sessions, the users were givenfive minute of break time to normalize theirfatigue level before starting their next session.

ii. Measures

Data automatically recorded during the courseof the maze navigation was used to measurethe performance of each participant. For eachsession, we recorded, every 10 millisecond, theparticipant’s spatial trajectory inside the mazealong with orientation of gaze information (Fig-ure 4). This allowed us to effectively play-back and calculate each participant’s spatio-temporal activity including, but not limited to,time spent at decision nodes in the maze andcumulative gaze rotation at decision nodes inthe maze.

In addition, the participants filled out back-ground survey questionnaires with empha-sis on video gaming activity profiling, pre-simulator and post-simulator sickness ques-tionnaire (SSQ) [9] and a presence question-naire [16] related to their immersive maze ex-perience.

The data contained information retrievedfrom 40 participants with the following at-tributes: subject id, date, player profile, gen-der, age, 20/20 vision, gaming hours, athleticscore, condition of deployment, time of com-pletion, relative simulator sickness score (low:16; high: 160), and presence score (low: -21,high: +21). Furthermore, we introduced a de-rived scale of attribute based on participant’sself reported gaming hours per week and theirvideo-gaming exposure profiling which we

6

Spatial navigation in immersive VR • May 2018 • Iteration. 1

have termed as ‘Player Level’ [see Table 1]. Ad-ditionally, for deep exploration of the trajec-tory data, we defined a set of mathematicallyderived features generalized for each trajec-tory such as distance traveled, decision points(nodes) reached inside the maze, positional cur-vature, head rotation amount, and coverage ofthe maze. Some of these trajectory features areillustrated in Figure 5. Positional curvature fea-ture refers to the curvature of the trajectory cal-culated per frame between successive positionvectors. Rotation amount feature refers to theunsigned angle calculated per frame betweensuccessive head rotation transforms stored asQuaternions. Coverage feature is simply thenumber of unit cubes covered (area) by theuser, calculated per frame. We conducted afollow up secondary analysis of the trajectorydata using the Dynamic Time Warping (DTW)algorithm [4] to compute alignment distancebetween two trajectory samples from our poolof participant data. We used R and SPSS to-gether to conduct the analysis of the trajectorydata.

Figure 5: Visualization of two trajectory features: posi-tional curvature and head rotation amount.

We obtained information from each ses-sions (immersive and non-immersive user in-terface). Pre-experience and post-experiencesurvey questions relevant to this analysis are

Survey Question Response type

Please indicate the number ofhours (per week) you playvideo games

Ordinal,1 (1-3 hours) . . .5 (10+ hours)

Please indicate your overall(perceived) athletic skills

Ordinal,1 (Very poor) . . .10 (Excellent)

Please indicate your responseto the questions related toplaying video games

Ordinal,Yes/No/Maybe

Please indicate and rate any ofthe symptoms listed in thetable below on a scale of 1 to10 (SSQ)

Ordinal,1 (Never felt) . . .10 (Str. felt)

Please rate your agreementwith the following statementsw.r.t. both the VR sessionswitnessed (Presence)

Likert,-3 (Str. Disagree). . .+3 (Str. Agree)

Briefly tell us about yourvirtual experiences(suggestions/comments)

Free Response

Extendedcomments/suggestions

Free Response

Table 3: Pre-experience and post-experience survey

shown in Table 3.

iii. Procedure and Tasks

At the start of the experiment, participantswere given elaborate verbal instructions ex-plaining the experimental procedure. Both thegamer and the non-gamer group were then in-terviewed in order to profile their backgroundand account for prior exposure to video gam-ing. The participants were given an overviewof the technology involved and were explainedhow to wear a HMD and control their in-gameavatar using the controller. They were also ex-plained their responsibility towards the study(finding the targets inside the maze in a spe-

7

Spatial navigation in immersive VR • May 2018 • Iteration. 1

Figure 6: Time of completion is significantly less for par-ticipants who self-reported their gaming profile(gprofile) as gamers

Figure 7: Presence score is higher for participants whoself-reported their gaming profile (gprofile) asnon-gamers

cific order, and if they were feeling nauseated,then they should immediately stop their VRexperience). The participants were explicitlytold that there were no time limits to their ses-sions and were advised to take their time tosolve the maze. The participants would theninitiate their two session maze experience witheither immersive or non-immersive user inter-face control schematics. During the course ofsolving the maze, the participants had the op-tion of verbally communicating with the studyinvestigator in case they needed any furtherinformation.

The maze task model consisted of three tasks,all of which involve wayfinding. These taskswere as follows:

1. Find your way to target 1, a mini map( 12b)

2. Find your way to target 2, the animatedhuman avatar ( 12a)

3. Find your way back to the entrance of themaze

The difference between immersive and non-immersive user interface lies in the way the par-ticipants controlled their gaze. Immersive userinterface control meant that the participantsused their natural head orientation to aligntheir view inside the maze environment (VE).Non-immersive user interface control meantthe participants had to control their view in-side the maze using the joystick analog controlon the gaming controller (gamepad) (Figure 1).The participants were required to be seatedwhile going through the study because thisconfiguration helped minimize simulator sick-ness. After completing their first session, theparticipants were given a break of 5 minutesbefore starting their next session to help nor-malize any fatigue conditions. In their secondsession, they were given the same maze onceagain but with a different user interface con-trol schematic from before. Upon completionof their second session, the post-experience sur-vey was administered, which included a pres-ence questionnaire [16] along with personalfeedback. Simulator sickness questionnaires

8

Spatial navigation in immersive VR • May 2018 • Iteration. 1

[9] were provided contextually in-between thesessions. At this point, all participant activityhave been logged, collated and uploaded forfurther analysis.

Figure 8: Repeated measures analysis for current userstudy; test of within-subjects effects.

Figure 9: Repeated measures analysis for current userstudy; test of within-subjects contrast.

V. Results

Only 19 participants out of 40 completed bothmaze sessions. We report that 2 participantsout of 40 were not appropriately logged dueto device I/O failure, resulting in completedata loss of 2 data points from the analysis.We investigate our dependent variable (timeof completion) against the set of independentvariables such as the participant’s gaming pro-file type (gamer or non-gamer), the type ofdeployment (immersive and non-immersive),presence score, simulator sickness score, age,and gender.

Figure 10: Visual comparison of user trajectories of arandomly selected gamer (left) and a non-gamer (right) while solving the maze in thenon-immersive interface.

We report a significant difference betweengroups of means of gaming profile type andtime of completion (F-value = 12.885, P =0.000914). The box plot of time of completionwith respect to gaming profile [see Figure 6]illustrates the difference between gamer andnon-gamer groups. We observe, a majority ofgamer participants took less time in solvingthe maze on average, whereas a majority ofnon-gamer participants took more time. Fur-thermore, figure 14 shows a positive correlation(scatterplot) between the time of completion ofthe maze (in seconds) and the age of the partic-ipants. We also report that female participantsin this study spent higher time completing themaze on average than the male participants inthis study [see Figure 15]. This suggests thatthere is a trend of age and gender togetherimpacting immersive VR user performance.

Additionally, we report a significant differ-ence between groups of means of gaming pro-file type and self-reported presence score ofthe successful participants (F-value = 9.565, P= 0.00699). The box plot of presence scorewith respect to gaming profile [see Figure 7]illustrates the difference between gamer andnon-gamer groups. Interestingly, we observedanother significant difference between groupsof means of condition of deployment and theself reported SSQ scores [see Figure 13]. It isindicative of the fact that immersive user in-terface control made participants nauseated toa higher degree than non-immersive user in-

9

Spatial navigation in immersive VR • May 2018 • Iteration. 1

Figure 11: Bi-modality test

terface control. All of these findings togetherare indicative of a trend, but we need inferen-tial statistics to make an argument conclusively.We then proceed to explore our recorded usertrajectory dataset.

Repeated measures analysis of the user tra-jectory dataset revealed that technical factors,such as immersive and non-immersive userinterfaces, do significantly impact immersiveVR user performance [see Figure 9 and Fig-ure 8]. This shows a significant main effect oftreatment (difference between immersive andnon-immersive user interface) and an interac-tive effect of treatment and condition (i.e. orderof deployment; immersive or non-immersivefirst, immersive or non-immersive second) com-bined. We find that condition is not a signifi-cant factor between subjects. Looking closelyat individual measures, we don’t find a differ-ence on any particular measure [see Figure 9].This suggests small but consistent differencesthat arise as a result of treatment (immersiveversus non-immersive user interface).

From our DTW analysis, we found that thealignment distance of the trajectories (vectorbased feature) between the gamer group andnon-gamer group is closer (tends to zero) inthe case of immersive user interface than non-immersive user interface. This fact is furtherbacked by a positive bi-modality test (see Fig-ure 11) between the two groups. This suggeststhat immersive user interface normalizes themaze solving experience between the gamerand non-gamer groups.

(a) 3D mannequin model

(b) 3D Minimap model

Figure 12: Targets inside the maze

VI. Discussion

We reject our null hypothesis and accept thealternate hypothesis. We found that prior ex-posure to video-gaming, age, and gender hasa measurable impact on immersive VR userperformance. While conceptualizing our ex-perimental design, there were concerns aboutchoosing the appropriate measurement param-eters as performance metric for VR tasks, weworried that the immersive and non-immersiveuser interface control paradigms would beconfusing to users, and that visual cues tounderstand decision making in mazes dur-ing analysis would be complicated and con-founding. During the course of the study, weobserved that users, who claimed to be self-perceived video-gamers, aren’t confused bythe immersive, non-immersive interface controlparadigm. The video-gaming group seemed totreat both immersive and non-immersive userinterface as the same, unlike the non-video-gaming group. The ability to replay user trajec-tory in the Unity game engine’s editor environ-

10

Spatial navigation in immersive VR • May 2018 • Iteration. 1

Figure 13: SSQ score is higher in immersive user inter-face control

Figure 14: Maze Study - Graph showing scatterplot oftime of completion of the maze as the y-axisand age of the participants as the x-axis.

ment proved to be very useful. These visualreplays helped us determine user trajectoryfeatures such as, curvature, and head-rotationamount [see Figure 5] to analyze.

Additional findings: The interface workedwell except twice when it failed to record thesession due to a device malfunction. Due to theminimalist nature of our system setup, we wereable to reduce points of failure in our systemso that our participants could focus better onthe study objectives. All user data logging wasdone in the background and were stored locallyon-board the HMD device for faster I/O. Themaze provided ample space-time interactivityfor the participants which in turn provided usa rich data set including high resolution usertrajectory data which can be parsed throughin real-time [see Figure 4]. The task model in

Figure 15: Maze Study - Boxplot showing time of com-pletion of the maze by group (gender).

our study also proved to be fairly difficult formost of the participants, which helped offsetthe balance between chance-based decisionsand skill-based decisions inside the maze. Theaverage time that the users spent in immersiveinterface was 268 seconds while the averagetime spent by the same users in non-immersiveinterface was 364 seconds. As suspected, therewere differences in performance between thegamer and the non-gamer groups [see Figure10]. In the context of spatial problem-solving,wayfinding, the combined finding of simulatorsickness and presence score could be indicativeof the fact that the two groups approached thestudy differently. From a pedagogical point ofview, if immersive VR applications were to bedeployed while making a meaningful impactto its user’s skill-building exercise, we wouldhave to account for the factor that differentgroups of users conceive and treat immersiveenvironments differently based on their priorexperience. Standardized immersive experi-ence, although logistically good, is not effectivein targeting individual specific needs.

VII. Conclusions

Immersive VR applications have been tradi-tionally deployed in a controlled laboratorysetting ever since the advent of HMDs [15].With the proliferation of highly accessible mo-bile communication devices such as the smart-phone, high pixel density display technology

11

Spatial navigation in immersive VR • May 2018 • Iteration. 1

became available as an alternative to dedicatedand tethered HMDs. The mobility of thesesmart devices inspired us to conceive a mo-bile VR platform that is untethered and canbe deployed universally. Our work representsanother step toward ubiquitous deploymentof immersive virtual experiences which willpotentially advance general VR usability, in-cite more research, and further the cause forpractical VR applications.

With the current study (a study of techni-cal factor), we showed that prior exposure tovideo-gaming, age, and gender is found tohave a measurable impact on user performancewhen it comes to spatial maze navigation inimmersive VR. The current study design in-volved a relatively complicated maze designthan Study III. The context of the maze expe-rience was similar to Study III, in which thesubject is required to find and rescue anotherhuman (avatar). The task included findinga clue (minimap) before finding the humanavatar and then tracing back the path to the en-trance of the maze. The hardware setup for thecurrent study was kept similar to Study III formaintaining consistency. The current study de-sign also looked at each participant partakingthe maze experience twice with the distinctionthat one of the session would require the userto use only the game-controller to navigate themaze while the other would involve the sameuser solving the same maze combining naturalhead gaze rotation and the game-controller tonavigate the maze. We found that the gamingprofile coupled with age and gender is prov-ing to be a strong indicator of success when itcomes to navigating immersive mazes in VR.After our preliminary finding, we focused onthe important issue of exploring the core im-pact on user task performance in an immer-sive VR setup deployed under an immersivesetting. We conducted extended analysis onrecorded user trajectory data. To that end, wedefined a set of mathematically derived fea-tures generalized for each trajectory such asdistance traveled, decision points reached in-side the maze, positional curvature, head rota-tion amount, and coverage of the maze. Upon

closer inspection of the variation in these trajec-tory features through repeated measure analy-sis, we found a small but consistent differencethat arises as a result of the treatment of ourfinal study design, that is immersive versusnon-immersive user interface. We find that im-mersive user interface helps users explore theVE better with natural head gaze control thanthe non-immersive user interface with analoggaze control. These findings are useful to VRdesigners in making appropriate trade-offs inVR level design for specific VR applications.

To summarize, in this work, we have showedthat prior exposure to video-gaming, age, andgender is found to have a measurable impacton user performance when it comes to spa-tial maze navigation in immersive VR. We con-ducted extended analysis on recorded user tra-jectory data. To that end, we defined a set ofmathematically derived features generalizedfor each trajectory such as distance traveled,decision points reached inside the maze, posi-tional curvature, head rotation amount, andcoverage of the maze. Upon closer inspec-tion of the variation in these trajectory featuresthrough repeated measure analysis, we founda small but consistent difference that arises as aresult of the treatment of our final study design,that is immersive versus non-immersive userinterface. We find that immersive user interfacehelps users explore the VE more with naturalhead gaze control than the non-immersive userinterface with analog gaze control.

References

[1] Basu, A., Ball, C., Manning, B., and

Johnsen, K. Effects of user physical fit-ness on performance in virtual reality. In2016 IEEE Symposium on 3D User Interfaces(3DUI) (March 2016), pp. 233–234.

[2] Basu, A., Johnsen, K., and Bogert, K.Poster: Evolution and usability of ubiqui-tous immersive 3d interfaces. In 2013 IEEESymposium on 3D User Interfaces (3DUI)(March 2013), pp. 131–132.

12

Spatial navigation in immersive VR • May 2018 • Iteration. 1

[3] Basu, A., Saupe, C., Refour, E., Raij, A.,and Johnsen, K. Immersive 3dui on onedollar a day. In 2012 IEEE Symposium on3D User Interfaces (3DUI) (March 2012),pp. 97–100.

[4] Bellman, R., and Kalaba, R. On adap-tive control processes. IRE Transactions onAutomatic Control 4, 2 (1959), 1–9.

[5] Bowman, D. A., Davis, E. T., Hodges, L. F.,and Badre, A. N. Maintaining spatial ori-entation during travel in an immersivevirtual environment. Presence: Teleopera-tors and Virtual Environments 8, 6 (1999),618–631.

[6] Bowman, D. A., Koller, D., and Hodges,L. F. A methodology for the evaluationof travel techniques for immersive virtualenvironments. Virtual Reality 3, 2 (Jun1998), 120–131.

[7] Jeong, D. H., Lee, C. S., Jeon, G. B., Song,C. G., Babu, S., and Hodges, L. Differ-entiation on information gathering abilityin real and virtual world. In Proceedingsof Pacific Graphics (2005), Citeseer, pp. 157–159.

[8] Johnsen, K., Ahn, S. J., Moore, J., Brown,S., Robertson, T. P., Marable, A., and

Basu, A. Mixed reality virtual pets to re-duce childhood obesity. IEEE Transactionson Visualization and Computer Graphics 20,4 (April 2014), 523–530.

[9] Kennedy, R. S., Lane, N. E., Berbaum,K. S., and Lilienthal, M. G. Simula-tor sickness questionnaire: An enhancedmethod for quantifying simulator sick-ness. The International Journal of AviationPsychology 3, 3 (1993), 203–220.

[10] Ruddle, R. A., Payne, S. J., and Jones,D. M. Navigating large-scale virtual envi-ronments: what differences occur betweenhelmet-mounted and desk-top displays?Presence: Teleoperators & Virtual Environ-ments 8, 2 (1999), 157–168.

[11] Smith, S. P., and Du’Mont, S. Measuringthe effect of gaming experience on virtualenvironment navigation tasks. In 3D UserInterfaces, 2009. 3DUI 2009. IEEE Sympo-sium on (2009), IEEE, pp. 3–10.

[12] Spiers, H. J., and Maguire, E. A.Thoughts, behaviour, and brain dynam-ics during navigation in the real world.NeuroImage 31, 4 (2006), 1826 – 1840.

[13] Suma, E., Finkelstein, S., Reid, M., Babu,S., Ulinski, A., and Hodges, L. F. Eval-uation of the cognitive effects of traveltechnique in complex real and virtual en-vironments. IEEE Transactions on Visual-ization and Computer Graphics 16, 4 (2010),690–702.

[14] Suma, E. A., Finkelstein, S. L., Clark,S., Goolkasian, P., and Hodges, L. F. Ef-fects of travel technique and gender on adivided attention task in a virtual environ-ment. In 3D User Interfaces (3DUI), 2010IEEE Symposium on (2010), IEEE, pp. 27–34.

[15] Sutherland, I. E. A head-mounted threedimensional display. In Proceedings ofthe December 9-11, 1968, Fall Joint Com-puter Conference, Part I (New York, NY,USA, 1968), AFIPS ’68 (Fall, part I), ACM,pp. 757–764.

[16] Usoh, M., Catena, E., Arman, S., and

Slater, M. Using presence questionnairesin reality. Presence: Teleoperators & VirtualEnvironments 9, 5 (2000), 497–503.

[17] Werkhoven, P., van Erp, J. B., and

Philippi, T. G. Navigating virtual mazes:The benefits of audiovisual landmarks.Displays 35, 3 (2014), 110 – 117.

[18] Zanbaka, C. A., Lok, B. C., Babu, S. V.,Ulinski, A. C., and Hodges, L. F. Com-parison of path visualizations and cogni-tive measures relative to travel techniquein a virtual environment. IEEE Transac-tions on Visualization and Computer Graphics11, 6 (2005), 694–705.

13