Evaluation of Guidance Systems in Public Infrastructures Using Eye Tracking in an Immersive Virtual Environment

Abstract

In this paper we present a novel method to evaluate guidance systems and navigation solutions for public infrastructures based on an immersive virtual environment in combination with a mobile eye tracking system. It opens new opportunities for way nding studies in public infrastructures already during the planning phase. Our approach embeds an interface for natural locomotion in the virtual environment o ering signi cant advantages related to the user's spatial perception as well as physical and cognitive demands. Accurate measurements of position, locomotion and gaze within the virtual environment enable a great simpli cation for the analysis of eye tracking data. We conducted a study with participants that included virtual and real world scenarios within a train station. First results exhibit similar behaviour of participants in the real and virtual environment, con rming the comparability and applicability of our method.

Keywords

• Way nding
• Immersive Virtual Environment
• Eye Tracking
• Gaze Analysis
• Visual Attention

-

Public transport infrastructures are becoming more multi-functional by serving as a short or long distance transportation hub and, for instance, as a shopping mall at the same time. For passengers and visitors, the increasing complexity of such infrastructures can lead to disorientation or subjective uncertainties. These aspects need to be fully taken into account already in the planning phase to prevent wrong design decisions which might result in higher costs in case they have to be changed later. To this end, it is important to understand how people interact with their environment and how di erent visual information sources in uence the orientation and navigation behaviour.

For such questions, eye tracking has become a popular method to investigate human way nding and to gain insights into perception and processing of visual information. Research on way nding has been conducted within lab and eld studies, which has provided evidence for comparable eye-movements in experimental and real world scenarios [ 2 ]. This study also revealed that the freedom of immersed participants to move their head and body is of high importance.

This paper presents our approach of combining an immersive virtual environment with an embedded interface for natural locomotion together with a mobile eye tracking system and demonstrates its applicability for way nding studies. 2

Immersive Virtual Environment

The immersive virtual environment DAVE [ 4 ], as shown in Figure 1, allows to e ciently perform way nding studies with enhanced comparability of the results due to controlling and reproducing the testing conditions (e.g. time, passers-by, train and bus schedule, etc.) for each participant. By extending the functionality of this test environment with a mobile eye tracking system from SMI, we are able to test di erent scenarios with alternating visual information within complex infrastructures.

Results in [ 7 ] show that body orientation and movement have a strong impact on the results of eye tracking studies in spatial navigation tasks. To enhance the awareness of movement and body perception within the virtual environment, it has been equipped with a Microsoft Kinect sensor for tracking the user's motion. The user has to walk in place to move forward and to turn the shoulders to invoke rotations in the virtual world. The outcome of [ 3 ] revealed the importance between head alignment and heading. In our observations, the head was consistently aligned with the current walking direction because the participants not only turned the shoulder but rather changed the complete body orientation into the walking direction.

In [ 6 ] a survey of empirical studies on the perception of egocentric distances in virtual environments is made and the egocentric distance is reported to be shorter with a mean about 74% of the modelled distances. The main in uencing factors to facilitate a veridical spatial perception are binocular disparity, high quality graphics, a rich environment and an enhanced user's sense of presence. Of great importance is the participants' feeling of their own body in the virtual environment. These requirements are ful lled by the immersive virtual environment DAVE.

The validity of our hands-free steering approach has been explored in [ 1 ] by conducting a case study with parallel test groups, exposing individuals to way nding exercises in the real world and the corresponding virtual world. The validation results showed that the perceived durations, egocentric distances and directions do not di er statistically signi cantly between the real and the virtual world.

Maybe the use of stereoscopic shutter glasses instead of head mounted devices providing the perception of the own body in the virtual scene, instead of a virtual avatar, provides an enhanced sense of presence and therefore a better metric in the virtual environment leading to better results in the estimation of egocentric distances and directions. But there is still much unknown about presence and a promising eld for further research on spatial perception in virtual environments. 3

Analysis Method

Within the virtual environment DAVE all activities, movements and user interactions are recorded to be available for detailed analysis. Figure 2 shows a participant performing an experiment in the virtual environment as well as the corresponding eye tracking video indicating the gaze point with an orange circle.

Given the highly accurate data on position, body alignment and viewing direction of the participant in the virtual environment together with the gaze point from the eye tracking system, it is possible to compute the intersection between the viewing ray and the triangle mesh of the 3D model for each frame. The overall accuracy is a combination of the eye tracking glasses (according to the manufacturer: 0:5 ) and the optical tracking. This enables to project an attention map onto the virtual 3D environment and furthermore to create saliency maps for visualizing the distribution of xation in the environment as well as the amount of time one particular object was xated. The aggregated intersections between the viewing rays of participants and the 3D mesh of the environment are shown in Figure 3.

The evaluations are carried out according to the three levels of metric from [ 5 ] which are de ned as follows: user task performance (time required, distance travelled, number of wrong decisions), physical behaviour (characteristics of locomotion, looking around, shape of trajectory) and decision making (viewed objects and signs identi ed via attention maps, thinking aloud, interviews). 4

Findings and Outlook

The applicability of the method crucially relies on the similarity between the spatial perception in the virtual and in the real environment. In order to demonstrate the comparability and applicability of the method, we conducted a way nding study with two scenarios within the virtual and real world setting of the main train station in Vienna. We recruited di erent public transport users with speci c demographic behaviour and experience. Each participant started on platform 12 and was assigned to walk rst to the departure location of the tram line D (scenario 1) and afterwards to the location of the local bus line 13A (scenario 2). Both scenarios are aided way nding scenarios, classi ed after the taxonomy in [ 8 ], using external aids like signage and maps. Aided way nding using signs is considered to be rather simple since it does not require considerable cognitive e ort from the user. Therefore it is important to provide all the relevant information at every decision point. Large public infrastructures such as the main train station in Vienna are very complex which makes the planning of the guidance system more di cult and error-prone.

The proposed method represents a new approach to support an e ective evaluation of guidance systems in a goal oriented manner. The main advantage of this method lies in evaluating a multitude of di erent scenarios already in the planning phase avoiding costly mistakes. It greatly simpli es the data analysis, also reducing time and costs to reveal xation, attention, saliency maps, objects of interest and route choices based on eye tracking measurements. Our approach provides a basis for an e cient evaluation of guidance systems and navigation solutions in public infrastructures and many additional application areas.

The analysis of the recorded data from the virtual and the real environment is still ongoing. First results of the eld studies already exhibit similar critical areas within the real and virtual world where participants showed similar behaviour facing the same challenges like insu cient signposting or misunderstanding of signs. The nal results will provide unique insights into task based visual attention and way nding behaviour and will reveal the full potential of the method.

Acknowledgments. This work is part of the project MOVING and is funded by the research program line ways2go within the framework of the Austrian strategic initiative IV2Splus Intelligent Transport Systems and Services plus under the project number 835733.

The study was performed using SMI Eye Tracking devices.

References

1. Bauer , D. , Schneckenburger , J. , Settgast , V. , Millonig , A. , Gartner , G.: Hands free steering in a virtual world for the evaluation of guidance systems in pedestrian infrastructures: design and validation . In: Transportation Research Board 92nd Annual Meeting . No. 13 - 1484 ( 2013 )

2. Foulsham , T. , Walker , E. , Kingstone , A. : The where, what and when of gaze allocation in the lab and the natural environment . Vision Research 51 ( 17 ), 1920 { 1931 ( 2011 )

3. Hollands , M.A. , Patla , A.E. , Vickers , J.N. : "look where you're going!": gaze behaviour associated with maintaining and changing the direction of locomotion . Experimental Brain Research 143 ( 2 ), 221 { 230 ( 2002 )

4. Lancelle , M. , Settgast , V. , Fellner , D. : De nitely a ordable virtual environment . Proceedings of IEEE virtual reality. IEEE 1 ( 2009 )

5. Lessels , Ruddle, R.A. : Three levels of metric for evaluating way nding . Presence Teleoperators Virtual Environments 15 ( 6 ), 637 { 654 ( 2006 )

6. Renner , R.S. , Velichkovsky , B.M. , Helmert , J.R.: The perception of egocentric distances in virtual environments - a review . ACM Comput. Surv . 46 ( 2 ), 23 : 123 : 40 ( 2013 )

7. Schwarzkopf , S. , von Stulpnagel, R., Buchner, S.J., Konieczny , L. : What lab eye tracking tells us about way nding a comparison of stationary and mobile eye tracking in a large building scenario . In: Eye Tracking for Spatial Research, Proceedings of the 1st International Workshop ( 2013 )

8. Wiener , J.M. , Bchner , S.J. , Hlscher , C. : Taxonomy of human way nding tasks: A knowledge-based approach . Spatial Cognition & Computation 9 ( 2 ), 152 { 165 ( 2009 )